摘要:

Tutorial Review Liquid Chromatographic Determination Using Lanthanides as Time-Resolved Luminescence Probes for Drugs and Xenobiotics: Advantages and Limitations† A. Rieutord*a,b, P. Prognona, F. Brionb and G. Mahuziera a Laboratoire de Chimie Analytique II. Recherches en M�ethodologies Bioanalytiques, Facult�e de Pharmacie, 5 rue Jean-Baptiste Cl�ement, 92290 Ch�atenay-Malabry Cedex, France b H�opital Robert Debr�e, Service de Pharmacie et Laboratoire de Toxico-Pharmacologie, 48 Bd S�erurier, 75019 Paris, France Lanthanide sensitized luminescence is a very attractive alternative to UV detection and other luminescence techniques, i.e., fluorescence and phosphorescence, in separation science for the detection of drugs and xenobiotics because of the large Stokes shift, narrow emission bands and long lifetime.Some published applications of HPLC determination with lanthanide (Ln3+) sensitized luminescence detection are reviewed. Advantages and limitations of this technique are discussed.Normal-phase (NP) HPLC is not influenced by the quenching effect of water whereas reversed-phase (RP) HPLC is applicable to more compounds than NP-HPLC. However, pH adjustment and the quenching effect of water on Ln3+ luminescence are the main drawbacks of RP-HPLC. Elution properties and the need for pH adjustment are two arguments for selecting the mode of addition of Ln3+, i.e., pre- or post-column in the HPLC system. Sensitized Ln3+ luminescence detection is a much more specific method of detection than UV or fluorescence detection after HPLC separation but nevertheless, in some cases, does not always exhibit a significant increase in analytical performance when the donor itself is a strong fluorophore. The development of more powerful excitation sources could improve the limit of detection of the Ln3+ sensitized detection technique.This review suggests that it would be useful to obtain predicting factors about the drug to establish whether the latter is suitable to be measured using an HPLC–Ln3+ approach.Keywords: Lanthanides; energy transfer; time resolved luminescence; high-performance liquid chromatography; drugs; xenobiotics Luminescence techniques were extensively developed in the last two decades to improve the limits of detection for drug and xenobiotic measurements by high-performance liquid chromatography (HPLC). Even if these modes of detection are not as universal as UV/VIS spectrophotometry, they allow a gain in terms of specificity and sensitivity.Unfortunately, many organic compounds decay from the electronic excited state (S1) to the ground state (S0) without emission of light. Therefore, in order to extend the applications of luminescence techniques, the analyte may be derivatized1 to a fluorescent derivative, or environmental modifiers could be added to increase the native fluorescence.2,3 In the same way, progress made with the phosphorescence detection mode has allowed measurements at room temperature.4 Nevertheless, interferences due to fluorescence impurities, Raman and Rayleigh scattering and the fact that the phosphorescence detection requires deoxygenation of the sample can limit their use.5 To overcome these problems, determination of analytes using sensitized lanthanide ion luminescence detection appears to be a very attractive approach.Lanthanides (Ln3+) possess a large Stokes shift in their luminescence emission, narrow emission bands of 1–20 nm half-width and a long emission lifetime of the order of hundreds of microseconds.In particular, the last property allows time-resolved luminescence (TRL) measurements and therefore interferences can be overcome. Lanthanides are the 14 elements following lanthanum in the Periodic Table. The lanthanide elements are associated with the progressive filling of the 4f-electron subshell. The 4f electronic configuration plays a major role in the process called ‘energy transfer’ by which an excited sample transfers its excitation energy to an acceptor metal ion during the lifetime of the excited state.Since the first energy transfer was described between salicylate and Tb3+,6 numerous studies using lanthanide sensitized luminescence have been carried out. However, the first use of lanthanide ion as the detection chromophore in LC was † Presented at the VIIth International Symposium on Luminescence Spectrometry in Biomedical Analysis, Sophia Antipolis, Nice, France, April 17–19, 1996.Andr�e Rieutord is a hospital pharmacist at the Robert Debr�e Hospital in Paris and has carried out research at the Facult�e de Pharmacie (Ch�atenay-Malabry) in the Laboratoire of Chimie Analytique II of Professor G. Mahuzier. His Research interests involve improving the sensitivity of drug detection in biological fluids using lanthanide sensitized luminescence and other related luminescence techniques in order to apply these schemes to liquid chromatography or capillary electrophoresis separation methods.Analyst, May 1997, Vol. 122 (59R–66R) 59Rreported only in 1985.7 Regarding drug and xenobiotic determinations, several papers using HPLC and Ln3+ sensitized luminescence have been published on non-steroidal antiinflammatory agents (NSAIDs),8 ciprofloxacin,9 nucleic acids,10 bleomycin,11 tetracycline,12,13 steroids,14 orotate,15 theophylline,16 mycotoxines17 and thiol compounds.18–20 The objective of this review is to analyse constraints related to Ln3+ properties and HPLC and emphazise the advantages and limitations of such a technique.We shall discuss the use of normal-phase HPLC (NP-HPLC) and reversed phase HPLC (RP-HPLC) and the mode of addition of lanthanide salts either in pre- or post-column. Finally, we shall discuss the performance of the method. This review deals only with the use of Ln3+ sensitized luminescence as the detection mode in HPLC for drugs and other organic xenobiotics.Other major analytical aspects of the use of lanthanides related to fluorescence immunoassays and dynamic quenching of lanthanide luminescence for the measurement of inorganic anions, for instance, are not included. Generalities Lanthanide Characteristics Despite the poor TRL emission of free lanthanides in solution due to their low e and the efficient quenching of their luminescence by water molecules, they can be very useful for detection when sensitizers are added.Sensitized lanthanide luminescence is a well described method of detection. The Ln3+ complex in solution absorbs energy at a wavelength characteristic of the organic donor and emits radiation with a wavelength characteristic of Ln3+. This reaction known as ‘energy transfer’, has been extensively studied and very well described.21–24 Fig. 1 illustrates the theoretical diagram of energy transfer. Energy is absorbed by the organic ligand leading to an excited state singlet (S1).At this stage there are two possibilities: the molecule can return from S1 to S0 by a radiative transition (fluorescence) or a non-radiative process or can go over to an excited triplet state (T1). Once again, the organic compound has two possibilities: the molecule can go from T1 to S0 by a radiative process (phosphorescence) or a non-radiative process, but in the presence of lanthanide ion energy can be transferred to the metal ion, which may be excited and subsequently emit characteristic radiation.When Ln3+ is free, it will fluoresce when excited by intense radiation but its intensity emission is very weak. In fact, four lanthanide ions (Dy3+, Eu3+, Sm3+ and Tb3+) are mainly used owing to energy level considerations. Briefly, two types of energy transfer processes have been depicted. First, triply charged lanthanide cation can form a stable chelate with ligands when negatively charged donating groups are involved; this is called intramolecular energy transfer.In aqueous solutions, donor groups containing neutral oxygen or nitrogen atoms generally bind only when present in multidentate ligands that contain at least one or two other donor groups having negatively charged oxygen. Table 1 illustrates some structures required for compounds to chelate Ln3+. Second, the donating compound can transfer eneral mechanism to the metal; this is the intermolecular energy transfer.25 Drugs for which intermolecular transfer occur include theophylline16 and some a,b-unsaturated steroids.14 Table 2 lists the compounds involved in the latter process.Regarding the compounds listed in Tables 1 and 2, sensitized lanthanide luminescence so far has limited applications. To extend the applicability, a pre-column labelling reaction for thiols containing analytes was developed.18–20 The idea is to derivatize the analyte to a product that, first, is able to form a stable complex with Ln3+ and, second, sensitize the lanthanide luminescence.A commercially available derivatization reagent for thiol-containing analytes, i.e., 4-maleimidosalicylic acid (4-MSA), was chosen because the salicylate group has sensitizing properties for Tb3+. 4-MSA can be utilized in the reaction without any problems as unreacted 4-MSA reagent does not produce any fluorescence or, in the presence of Tb3+, Fig. 1 Energy transfer mechanism between organic donor and Tb3+.Table 1 List of drugs and xenobiotics for which intramolecular energy transfer has been described. Group chelating Ln3+ intermolecular transfer Compound Ref. a-Ketocarboxylate Carboxypsoralene 8 Citrinin 17 Fluoroquinolones 9 Orotate 15 a-Aminocarboxylate Flufenamic acid 8 Mefenamic acid 8 Niflumic acid 8 Ochratoxin A 17 b-Diketone like Ochratoxin A 17 (ketone and hydroxyl) Tetracyclines 12 Bleomycin 11 Phosphate Adenosine triphosphate 10 Amide and/or amino and ketone Tetracycline 12 Bleomycin 11 Hydroxyl Adenosine triphosphate 10 Bleomycin 11 Table 2 List of drugs for which intermolecular energy transfer has been described.Groups involved in intermolecular process Compound Ref. Carbonyl a,b-Unsaturated steroid 14 Theophylline 16 Amino Theophylline 16 60R Analyst, May 1997, Vol. 122any sensitized Tb3+ luminescence. Successful LC methods with sensitized Ln3+ luminescence have been developed for the determination of glutathione, N-acetylcysteine and l-cystein. 18 Owing to the long lifetime of sensitized lanthanide luminescence, time resolved luminescence (Fig. 2) is used. After the excited pulse, the period of measurement (gain time) starts after a delay time allowing discrimination of short-life luminescence such as background fluorescence. Parameters Influencing Energy Transfer Lanthanide salts are commercially available in chloride or nitrate forms. Chorides are widely used because they are water soluble and compatible with most organic solvents (methanol, acetonitrile, etc.). Greater sensitization has been observed for the chloride versus the nitrate salts in the case of an intermolecular transfer.7 This could be related to the greater ability of NO32 than Cl2 to form an inner sphere.26 Consequently, the likelihood of collisions leading to energy transfer should be larger for Cl2 than for NO32.Lanthanide luminescence is favoured by neutral or alkaline pH owing to ionization of electron donating groups.The ionization of molecules facilitate binding to Ln3+, allowing the formation of a strong chelate. Thus, optimum sensitivity has been reported at pH 8–10 for tetracycline,12–14 at pH 8–8.5 for orotate,15 at pH 6.8 for bleomycin,11 at pH 7.5–8.5 for fluoroquinolones9,27 and at pH 6.7 for an EDTA-like structure, e.g., ADR-925, the hydrolysis product of desrazoxane.28 Further alkalinization causes precipitation of lanthanide hydroxide. 11–15,27,28 Perry and Winefordner29,30 reported that TRIS buffer was likely to have chelating properties towards Ln3+ and thus prevent the quenching effect of water on lanthanide luminescence. However, in our experiments8,9 using diluted NaOH instead of TRIS buffer to adjust the pH to 7.4, we observed a similar enhancement of Tb3+ luminescence. We concluded that TRIS buffer would have no or a minor influence on the energy transfer process. Heat produces a quenching or enhancement effect on the luminescence depending on the type of energy transfer.As intermolecular collisional energy transfer occurs between naphthaldehyde and Eu3+, luminescence is enhanced with increase in temperature.7 In fact, higher temperature increases collisions between the analyte and lanthanide ion, favouring the energy transfer process. Conversly, Tb3+ luminescence sensitized by tetracycline or bleomycin is quenched with increase in temperature.11 In these cases, an increase in temperature increases collisional deactivation.Briefly, in an intramolecular energy transfer mechanism, the signal is decreased by increase in temperature and if a collisional process is involved the signal is increased. Oxygen is a matter of concern for intermolecular energy transfer. In that a particular type of energy transfer, sensitation, is only possible if the solvent is thoroughly deoxygenated since otherwise non-radiative decay of the excited ligand dominates.Similar quenching is not readily observed with chelates i.e intramolecular energy transfer. From our experience, with lanthanide chelates, we never observed any significant difference in luminescence when measuring samples deoxygenated or not.8,9 Another parameter influencing luminescence is the lanthanide concentration. Optimum sensitivity is observed when a certain lanthanide concentration is reached.8,9,15 There is evidence for an optimum [lanthanide]/[organic donor] ratio.This ratio varies considerably, from 1 to 250, depending on the compound.8,9,15,23,31 For the same compound it can vary whether micelles are added or not.32 Possible mechanisms controlling the effect of lanthanide concentration are the collisionnal quenching of free Ln3+, the quenching effect of Cl2 when using chloride salts of lanthanides and the direct excitation of lanthanide ions. In order to decrease radiationless energy transfer, a variety of molecules such as Triton X-100,8,9,13,24,33,34 Tween 2035 and sodium dodecyl sulfate (SDS)15 have been used to provide an insulating sheath around emitting species.By compartmentalizing the Ln3+ and the donor species in a discrete volume region, micelles facilitate the energy transfer process. In particular, micelle addition allows the removal of the efficient quenching effect of water in aqueous solutions. Micelle concentrations should be determined to obtain optimal energy transfer.8,9,14 Neutral molecules such as trioctylphosphine oxide (TOPO) have also been used.8,9,33,34 They have a good lone oxygen atom and a bulky saturated hydrocarbon tail that directs the solvent molecules away from the metal ion and hence contributes to the insulating sheath.36 Sensitized Lanthanide Luminescence in HPLC Drug measurement in complex biological matrices requires very sensitive and specific analytical techniques in order to monitor residue levels.In particular, for intracellular determinations low limits of detection (LOD) ( < 1028 mol l21) will be crucial in the future but at the moment this requirement seems difficult to fulfil systematically.In fact, sensitized lanthanide luminescence used in HPLC separations has been developed for several drugs such as NSAIDs,8 ciprofloxacin,9 bleomycin,11 orotate,15 tetracyclines,12 nucleic acids and nucleotides,10 steroids,14 theophylline,16 mycotoxines17 and thiol compounds, 18–20 with the aim of reaching this LOD and improving the selectivity of the analyte.The separation of organic compounds by HPLC depends on both the stationary phase and the mobile phase. Optimum elution is obtained when a certain equilibrium between the three parameters is reached. Moreover, using HPLC with sensitized lanthanide luminescence as a detection mode, the mobile phase should comply with the luminescence requirement of the lanthanide. We shall discuss the advantages and limitations of using this detection strategy in NP- and RP-HPLC.Reversed-phase HPLC This is the most commonly used method.8–12,15,17,20 Mobile phases are mainly prepared with water, methanol and acetonitrile, which are compatible with energy transfer requirements. A study8 was carried out on the luminescence of Tb3+–mefenamic acid in solvents with various polarity (Table 3). This emphasizes the need for a polar solvent to cause dissociation of the electron donating group, i.e., carboxylate for the NSAIDs, and Fig. 2 Schematic representation of operation of a pulsed-source timeresolved fluorimeter: A, excited pulse; B, fluorescence emission (short-life luminescence); and C, phosphorescence or ion fluorescence emission (longlife luminescence). td, Delay time; tg, gain time.Analyst, May 1997, Vol. 122 61Rfavours binding of Tb3+. NSAIDs are poor sensitizers of Tb3+ in apolar (hexane) or slightly polar solvents (acetonitrile, DMSO). Sensitization of lanthanide luminescence is facilitated in solvents containing hydroxide groups such as methanol and ethanol.Methanol gives better sensitivity on luminescence than acetonitrile and water. On the other hand, Vazquez et al.17 observed that luminescence of the Tb3+–ochratoxin A chelate was greater in butanol than in methanol. The higher viscosity of butanol is likely to limit the non-radioactive deactivation of the chelate owing to a decrease in the number of molecular collisions and to enhance the lanthanide luminescence.Conversely, water is a strong quencher of energy transfer between an organic donor and lanthanide ion, hence a high water content of an RP-HPLC mobile phase causes a dramatic quenching of Ln3+ luminescence. Two ways of limiting the quenching effect of water on luminescence have been described: decreasing the water content of mobile phase or using a displacing agent to remove water from the first coordination sphere of the lanthanide ion. The use of methanol instead of acetonitrile to prepare the mobile phase for the elution of ciprofloxacin allowed a decrease in the proportion of water9 because isoelution with methanol needed an increased methanol content compared with acetonitrile.Owing to the lower water content, better sensitization of Tb3+ was observed. Consequently, this small change improved the sensitivity but increased the back-pressure on-column. The other possibility is to displace water from the first coordination sphere of Ln3+.Acetate ion was the first agent used to discriminate the quenching effect of water on luminescence.7 An 11-fold gain in sensitivity was observed when acetate ion was added to the post-column reagent methanol–water (80 + 20). Hence the luminescence intensity was comparable to that obtained with a post-column reagent containing 100% methanol. The use of a micellar mobile phase was successfully employed to limit the quenching effect of water.14 Addition of SDS at the critical micelle concentration produced a slight increase of Tb3+ sensitized luminescence.In this case, the problem of terbium luminescence quenching in aqueous solutions was circumvented by compartmentalizing the Ln3+ acceptor and the organic donor in SDS micelles. The increased luminescence observed corresponds to the typical increase in lanthanide quantum yield on passing from water to an organic medium.37 An insulating sheath can be created around the Ln3+ ion if a synergistic agent having an oxygen atom for coordination can be selected, therefore excluding water.TOPO, with its lone oxygen atom and three octyl donating chains, protects not only rare earth ions but also the organic donor from collisional interferences. The enhancing effect of TOPO was observed on the sensitized luminescence of some chelates, e.g., Tb3+–ciprofloxacin9 and Tb3+–NSAIDs.8 However, an increased concentration of TOPO causes a decrease in the signal, probably due to the poor solubility of the TOPO molecules in a highly aqueous medium.8 The use of sensitized lanthanide luminescence in a detection scheme in RP-HPLC is the most commonly described approach.Nevertheless, the high water content of RP-HPLC mobile phases produces a strong quenching effect on the lanthanide luminescence. The second drawback is related to the fact that an alkaline pH favouring lanthanide chelate formation is not always compatible with the stationary phase in a conventional HPLC column.However, to overcome the latter problem, we report some solutions under Post-column addition below. Normal-phase HPLC In separation science, NP-HPLC is not used as often as RPHPLC. For polar compounds, however, NP-HPLC is often needed. Only two papers14,16 have described the use of sensitized lanthanide luminescence in NP-HPLC separation. In contrast to RP-HPLC, the absence of water in the mobile phase will prevent the quenching effect of water on luminescence and pH adjustment is not necessary.Addition of water to the mobile phase would favour the binding of Ln3+ to water molecules, preventing intermolecular transfer between an organic donor and Ln3+. Moreover, dissolved oxygen should be removed from the mobile phase to avoid oxygen quenching when energy transfer is involved in an intermolecular process.7 The use of sensitized lanthanide luminescence in NP-HPLC appears to be restricted to polar compounds, the groups of which involved in energy transfer do not require extensive dissociation as for theophylline and some steroids.14,16 Thus cyclohexane, hexane and ethyl acetate, common solvents in NP-HPLC, comply with the requirements of sensitized lanthanide luminescence only with respect to compounds involved in an intermolecular type of energy transfer.The composition of an optimum mobile phase is a compromise between elution properties, backpressure on-column and luminescence enhancement of lanthanide. 14,16 For instance, the use of cyclohexane resulted in an extremly high back-pressure and its replacement by hexane caused only a moderate increase of back-pressure and provided a similar enhancement value for the sensitized luminescence of terbium. Unfortunately, hexane was a relatively weak solvent for elution of theophylline.16 The addition of ethyl acetate to hexane resulted in an excellent separation and detection of theophylline.16 Cyclohexane has also been successfully employed as a mobile phase for Tb3+–steroids.Reverse micelles are used in NP-HPLC to enhance the luminescence of lanthanides sensitized by organic donors such as theophylline16 or steroids.14 Reverse or inverse micelles refer to surfactant aggregation in non-polar solvents. Mwalupindi and Warner16 showed the advantage of using a lanthanide counter ion as opposed to a micellar solution with the lanthanide salt. Fig. 3 illustrates the sensitivity gain obtained when using a mobile phase containing Tb3+ surfactant, i.e., Tb(EHS)3.The enhancement of sensitized terbium luminescence was attributed to the rigid environment of the lanthanide counter-ion in the synthesized surfactant compared with the lanthanide salt. The use of reverse micelles expands the scope of lanthanide sensitized luminescence as method of detection of many organic compounds. However, there is still a need to deoxygenate the sample solution prior to luminescence. Lanthanide Addition Detection of organic donors is linked to the addition of Ln3+ to the chromatographic system.This could be performed in either a pre- or post-column mode. Table 3 Luminescence intensity of Tb3+ (2.5 3 1024 mol l21) sensitized by mefenamic acid (2 31025 mol l21) in different solvents. All measures were recorded at 550 nm.8 Luminescence* Solvent (arbitrary units) Hexane 0.2 Chloroform 0.3 Acetone 1.0 Acetonitrile 1.0 DMF 0.8 DMSO 0.3 Ethanol 42 Methanol 43 Methanol–water (7 + 3) 0.01 TRIS (0.4 mol l21) (pH 7.4) 40 Water–NaOH (pH 7.4) 30 * The luminescence intensity of Tb alone in the different solvents was @0.01. 62R Analyst, May 1997, Vol. 122Pre-column Addition This is the simplest technique: lanthanide salts are added directly to the mobile phase and reaction between Ln3+ and the organic donor will occur before separation on the column. Neither dilution caused by post-column addition nor a noisy baseline caused by the use of a second pump is observed.Nevertheless, the high consumption of expensive Ln3+ reagent should be considered. As reported previously, alkaline pH is favourable for the observation of sensitized Ln3+ luminescence. Unfortunately, even a slightly alkaline pH (!8) is well known to be incompatible with the stability of classical reversed-phase supports owing to the risk of hydrolysis of the alkalyted silanols. The use of a polymer column is recommended and allows the use of a mobile phase having a pH of 8–10.12,15 In two cases, pre-column addition had no influence on retention,10,14 but in one case, Wenzel et al.11 observed a better resolution of Tb3+–bleomycin than bleomycin alone.Complexation with Tb3+, as described for copper, would reduce the number of sites on the bleomycin capable of simultaneously adsorbing on the stationary phase. Post-column Addition Although this mode of addition is slightly more complicated to set up, it extends the applications of sensitized Ln3+ luminescence in HPLC separations.A second pump is needed to draw post-column reagent (PCR) containing Ln3+. A residence time in a post-column reaction coil is not essential for binding to occur. In numerous cases, a simple mixing tee is sufficient if the PCR was conveniently optimized. 8,9 In somes cases, post-column addition is mandatory. Fig. 4 depicts chromatograms obtained for ciprofloxacin using the same mobile phase at different pH. A poor peak shape is observed when using a mobile phase of pH 5.8, whereas good resolution is achieved when the pH of the mobile phase is decreased to 2.5.However, at this pH ciprofloxacin is not ionized, preventing lanthanide binding. This illustrates that an alkaline pH favouring sensitized lanthanide luminescence is hardly compatible with an acidic mobile phase prepared for the separation of compounds involving acidic functions in RPHPLC. 8,9,12,13. Therefore, for compounds such as tetracyclines, fluoroquinolones and NSAIDs, sensitized lanthanide luminescence has been applied in HPLC using a post-column device to add the Ln3+ salts.A buffer mobile phase must be avoided8,9 or must not be too concentrated to be adequately adjusted to an alkaline pH using a PCR.12 The use of phosphate species for acidification is not recommended as there is a risk of precipitation of Ln(PO4)3 in the post-column mixing tee.31 Any lanthanide hydroxide or phosphate would clog the post-column mixing tee.The challenge here is that the composition of PCR must alkalinize the mobile phase to favour sensitized lanthanide luminescence without leading to precipitation of lanthanide hydroxide. For instance, the PCR is adjusted to give pH 7.6 in the mixing tee for optimum sensitivity regarding ciprofloxacin and some NSAIDs measurements8,9 Lanthanide hydroxide formation could be circumvented by adding EDTA to the PCR containing lanthanide.12 Alkalinization up to pH 12 is possible when EDTA is added owing to the formation of a stable and water-soluble chelate.As EDTA does not fully encapsulate Ln3+,38 the organic donor can bind and transfer its energy to the metal. For instance, Wenzel et al.12 were able to adjust the pH to 9 in the mixing tee for tetracycline measurement. Alkaline reagents used in the PCR are usually TRIS buffer,8,12 ammonia buffer,12 or triethylamine.9 The two buffers solutions show original properties that might explain the interest in using them.Ammonia–ammonium chloride buffer exhibits weaker complexation with lanthanide ion than water, increasing the luminescence of lanthanide ion. TRIS buffer also increases the signal obtained, probably because it penetrates the coordination sphere of the chelate, giving rise to a synergestic effect.27 As a general trend, sodium and potassium hydroxide must be avoided in order to prevent lanthanide hydroxide formation. TRIS buffer is particularly interesting owing to its compatibility with Ln3+ ions, which usually precipitate with other buffers (phosphate, carbonate).When lanthanides are added post-column, the detector flow cell, the outlet and inlet capillaries and the mixing tee can easily Fig. 3 Chromatograms for standard samples of theophylline using A, TbCl3 in ethyl hexyl sulfosuccinate (ethyl acetate–hexane, 6 + 4) and B, Tb(EHS)3 in ethyl acetate–hexane (6 + 4) as the mobile phase.16 Fig. 4 Effect of pH (adjusted with H3PO4) on the shape of the peak of ciprofloxacin (1 3 1025 mol l21 methanolic solution) on a C18 column.pH: (A) 5.8; (B) 4.0; (C) 3.5; (D) 2.9; and (E) 2.5. Eluent, methanol–water (50 + 50); flow rate, 0.9 ml min21; detection, direct fluorescence monitoring of the eluent, lex = 278 nm, lem = 440 nm.9 Analyst, May 1997, Vol. 122 63Rbe cleaned by rinsing regularly with dilute nitric acid (1 + 9) to remove lanthanide hydroxide. Post-column addition dilutes the column eluent and causes a decrease in sensitivity.Consequently, the flow rate of the second pump should be as low as possible. The pulsing of this pump will increase the noise and further raise the limit of detection. Hence a two-head plunger pump is recommended. As described for the mobile phase previously, the quenching effect of water must be considered when preparing the PCR. Performance The sensitized lanthanide luminescence detection strategy in HPLC is a very specific method.The likelihood of having a compound co-eluted with an interferent able to transfer its energy to the metal is very low. Fig. 5 illustrates chromatograms obtained when measuring two mycotoxins, i.e., citrinin and ochratoxin A, in cheese extracts.17 UV spectrometry [Fig. 5(a)] is obviously inapplicable, owing to its poor specificity. The selectivity observed when using sensitized lanthanide luminescence [Fig. 5(b)] is greater than with fluorescence detection [Fig. 5(c)].Similar observations have been reported for real samples, measuring N-acetylcysteine in urine39 (Fig. 6) and mefenamic acid in serum.8 (Fig. 7). Table 4 compares the limits of detection obtained by UV, fluorescence or TRL with sensitized lanthanide luminescence. Depending on the drug, the limits of detection observed using TRL with sensitized lanthanide luminescence are either better (orotate, ciprofloxacin) or worse than those using UV detection (NSAIDs). Regarding ciprofloxacin, fluorescence detection exhibits a threefold better sensitivity than TRL with sensitized lanthanide luminescence.There are two hypotheses to explain these observations. First, one could say that the use of a second pump to draw lanthanide reagent decreases the sensitivity. The second mechanism is suggested by Fig. 8. As illustrated, perfectly selective excitation of the donor is sometimes difficult to perform owing to the absorption profile of the free Tb3+ ions which can interfere at high concentration levels generating high background noise.Nevertheless, the excitation of the Tb3+– ciprofloxacin chelate at a wavelength more specific for the organic compound does not necessarily lead to a real increase in the signal-to-noise ratio, and finally a compromise must be found. Conclusion In this review, the use of sensitized lanthanide luminescence in HPLC for the detection of drugs and xenobiotics has been Fig. 5 (a) Chromatograms of cheese extracts spiked with 1, citrinin and 2, ochratoxin A, with UV detection (l = 331 nm).Extracts spiked with 3 3 1026 mol l21 citrinin and 5 3 1026 mol l21 ochratoxin A. (b) Chromatograms of cheese extracts spiked with 1, citrinin and 2, ochratoxin A, with fluorescence detection (lex = 331 nm, lem = 472 nm). Extracts spiked with 2 3 1023 mol l21 citrinin and 1 3 1026 mol l21 ochratoxin A. (c) Chromatograms of cheese extracts spiked with 1, citrinin and 2, ochratoxin A, with TRL detection (lex = 331 nm, lem = 545 nm).Extracts spiked with 4.5 3 1026 mol l21 citrinin and 1 3 1025 mol l21.17 A and B represent two different extraction methods. Fig. 6 Chromatogram of urine samples spiked with derivatiazed 3 3 1027 mol l21 N-acetylcysteine (with 4-MSA) detected by fluorescence (left) and by sensitized Tb3+ luminescence (right); the dotted lines represent the unspiked samples.39 Fig. 7 Chromatograms of serum samples spiked with 5 3 1026 mol l21 of 1, flufenamic acid and 2, niflumic acid obtained by (a) UV detection at 285 nm and (b) TRL detection using Tb3+ (lex = 340 nm, lem = 545 nm).8 64R Analyst, May 1997, Vol. 122outlined. Numerous compounds possess the structural ability for the energy transfer to occur or can be derivatized to make sensitized lanthanide luminescence detection possible. However, in 10 years only 10–15 papers have been published reporting such a detection strategy in HPLC. The method is highly specific compared with the usual absorbance detection and is better than fluorimetry.Nevertheless, as we described previously, the sensitized lanthanide luminescence technique in HPLC appears sometimes to be hardly more sensitive than UV spectrophotometry or fluorescence for compounds having a high e, e.g., NSAIDs, or having a high fluorescence quantum yield (FF) e.g., fluoroquinolones. This suggests that the sensitized lanthanide luminescence technique is particularly interesting for chromophores exhibiting weak native fluorescence and having a suitable structure for the energy transfer to occur.The use of more powerful energetic sources could decrease the limit of detection in some cases. However, the usual laser systems deliver only a fixed wavelength, e.g., N2 laser 337 nm and KrF laser 248 nm. These wavelengths do not often match the excitation wavelengths of most analytes. The optimum parametric oscillator which is a tunable frequency converter based on an Nd:YAG laser, might provide solutions in terms of the sensitivity of sensitized lanthanide luminescence.40 This review suggests that it would be useful to obtain predicting factors about the drug to establish whether the latter is suitable to be measured using an HPLC–Ln3+ approach.The perfect organic would have a high e, a high FP (phosphorescence quantum yield) to allow efficient energy transfer, a low FF to make energy transfer more interesting in terms of sensitivity and a high Ks (association constant) with the Ln3+ acceptor.Moreover, the energy of the excited triplet state of the organic compound concerned must be nearly equal to or lie above the resonance level of the Ln3+ ion. Studies to determine all these parameters to examine exactly their relevance in the energy transfer process towards Ln3+ ions are warranted. References 1 Farinotti, R., Th`ese de Doctorat d’Etat, Universit�e Paris Sud, 1983. 2 Hinze, W. L., Singh. H. N. Baba, Y., and Harvey, N.G., TrAC, Trends Anal. Chem. (Pers Ed.), 1984, 3, 193. 3 Hoshino, M., Imamura, M., Ikehara, K., and Hama, T., J. Phys. Chem., 1981, 85, 1820. 4 Donkerbroek, J. J., Gooijer, C., Velthorst, N. H., and Frei, R. W., Anal. Chem., 1982, 54, 891. 5 Rendell, D., Fluorescence and Phosphorescence Spectroscopy. Analytical Chemistry by Open Learning, Wiley, New York, 1987. 6 Weissman, S. I., J. Chem. Phys., 1942, 10, 214. 7 Dibella, E. E., Weisman, J. B., Joseph, M. J. Schultz, J.R., and Wenzel, T. J., J. Chromatogr., 1985, 328, 101. 8 Sargi, L., Th`ese de Doctorat d’Etat, Universit�e Paris Sud, 1992. 9 Rieutord, A., Vazquez, L., Soursac, M., Prognon, P., Blais J., Bourget, P., and Mahuzier, G., Anal. Chim. Acta, 1994, 290, 215. 10 Wenzel, T. J., and Colette, L. M., J. Chromatogr., 1988, 436, 299. 11 Wenzel, T. J., Zomlefer, K., Rapkin, S. B., and Keith, R. H., J. Liq. Chromatogr., 1995, 18, 1473. 12 Wenzel, T. J., Colette, L. M., Dahlen, D.T., Hendrickson, S. M., and Yarmaloff, L. W., J. Chromatogr., 1988, 433, 149. 13 Duggan, J. X., J. Liq. Chromatogr., 1991, 14, 2499. 14 Amin, M., Harrington, K., and Von Wandruska, R., Anal. Chem., 1993, 65, 2346. 15 Schreurs, M., Vissers, J. P. C., Gooijer, C., and Velthorst, N. H., Anal. Chim. Acta, 1992, 262, 201. 16 Mwalupindi, A. G., and Warner, I. M., Anal. Chim. Acta, 1995, 306, 49. 17 Vazquez, B. I., Fente, C., Franco, C., Cepeda, A., Prognon, P., and Mahuzier, G., J.Chromatogr. A, 1996, 727, 185. 18 Schreurs, M., Gooijer, C., and Velthorst, N. H., Anal. Chem., 1990, 62, 2053. 19 Schreurs, M., Gooijer, C., and Velthorst, N. H., Fresenius’ J. Anal. Chem., 1991, 339, 499. 20 Schreurs, M., Hellendoorn, L., Gooijer, C., and Velthorst, N. H., J. Chromatogr, 1991, 552, 625. 21 Hemmil�a, I., Scand. J. Clin. Lab. Invest., 1988, 48, 389. 22 Richardson, F. S., Chem. Rev., 1982, 82, 541. 23 Diamandis, E. P., Clin. Biochem., 1988, 21, 139. 24 Georges, J., Analyst, 1993, 118, 1481. 25 Heller, A., and Wasserman, E., J. Chem. Phys., 1965, 42, 949. 26 Choppin, G. R., and Bertha, S. L., J. Inorg. Nucl. Chem., 1973, 35, 1309. 27 Panadero, S., G`omez-Hens, A., and P�erez-Bendito, D., Anal. Chim. Acta, 1995, 303, 39. 28 Hasinoff, B. B., J. Chromatogr. B, 1994, 656, 451. 29 Perry, L. M., and Winefordner, J. D., Talanta, 1990, 10, 965. 30 Perry, L. M., and Winefordner, J. D., Anal. Chim. Acta, 1990, 237, 273. 31 Hirschy, L. M., Dose, E.V., and Wineforder, J. D., Anal. Chim. Acta, 1983, 147, 311. 32 Georges, J., and Ghazarian, S., Anal. Chim. Acta, 1993, 276, 401. 33 Moulin, C., Decambox, P., and Mauchien, P., Anal. Chim. Acta, 1991, 254, 145. 34 Zhu, G., Si, Z., Yang, J., and Ding, J., Anal. Chim. Acta, 1990, 231, 157. 35 Hemmil�a, I., Dahuku, S., Mukkala, V. M., Siitari, H., and L�ovgren, T., Anal. Biochem., 1984, 137, 335. 36 Halverson, F., Brinen, J. S., and Leto, J. R., J. Chem. Phys., 1964, 41, 157. Table 4 Comparison of detection limits (mol l21) obtained using UV spectrophotometry, fluorescence or sensitized lanthanide luminescence with timeresolved luminescence (TRL) in HPLC for several drugs and xenobiotics. Ln3+ Compound UV Ref.Fluorescence Ref. TRL Ref. used Orotate 1 3 1027 41 — 1 3 1028 15 Tb3+* Niflumic acid 2 3 1027 8 — 3 31026 8 Tb3+* Flufenamic acid 4 3 1027 8 — 1 31026 8 Tb3+* Mefenamic acid 4 3 1027 8 — 2.5 3 1026 8 Tb3+* Bleomycin A2 2 3 1027 11 — 3 3 1026 11 Tb3+† Ciprofloxacin 2 3 1027 42 3 3 1028 43 1 3 1027 9 Tb3+* Theophylline 4 3 1027 44 — 5 3 1028 16 Eu3+* * Post-column addition.† Pre-column addition. Fig. 8 Comparison of the peak height (limit of detection = 0.25 pmol injected) and the signal-to-background ratio of ciprofloxacin for excitation performed at (a) 320 and (b) 278 nm. In both instances the flow rate of the post-column pump was fixed at 0.2 ml min21.9 Analyst, May 1997, Vol. 122 65R37 Haas, Y., and Stein, G., J.Phys. Chem., 1971, 75, 3668. 38 Lind, M. D., Lee, B., and Hoard, J. L., J. Am. Chem. Soc., 1965, 87, 1611. 39 Gooijer, C., Schreurs, M., and Velthorst, N. H., in HPLC Detection: Newer Methods, ed. Patonay, G., VCH, New York, 1992, pp. 27– 55. 40 Van de Nesse, R. J., Velthorst, N. H. Brinkman, U. A. Th., and Gooijer, C., J. Chromatogr. A, 1995, 704, 1. 41 Ferrari, V., Giordano, G., Cracco, A. T., Dussini, N., Chiaandetti, L., and Zachelo, F., J. Chromatogr., 1989, 497, 101. 42 Vall�ee, F., Lebel, M., and Bergeron, M. G., Ther. Drug Monit., 1986, 8, 340. 43 Jehl, F., Gallion, C., Debs, J., Brogrard, J. M., Monteil, H., and Minck, R., J. Chromatogr, 1985, 339, 347. 44 Blanchard, J., Harvey, S., and Morgan, W. J., J. Chromatogr. Sci., 1990, 28, 203. Paper 6/07616E Received November 8, 1996 Accepted February 11, 1997 66R Analyst, May 1997, Vol. 122 Tutorial Review Liquid Chromatographic Determination Using Lanthanides as Time-Resolved Luminescence Probes for Drugs and Xenobiotics: Advantages and Limitations† A.Rieutord*a,b, P. Prognona, F. Brionb and G. Mahuziera a Laboratoire de Chimie Analytique II. Recherches en M�ethodologies Bioanalytiques, Facult�e de Pharmacie, 5 rue Jean-Baptiste Cl�ement, 92290 Ch�atenay-Malabry Cedex, France b H�opital Robert Debr�e, Service de Pharmacie et Laboratoire de Toxico-Pharmacologie, 48 Bd S�erurier, 75019 Paris, France Lanthanide sensitized luminescence is a very attractive alternative to UV detection and other luminescence techniques, i.e., fluorescence and phosphorescence, in separation science for the detection of drugs and xenobiotics because of the large Stokes shift, narrow emission bands and long lifetime.Some published applications of HPLC determination with lanthanide (Ln3+) sensitized luminescence detection are reviewed. Advantages and limitations of this technique are discussed. Normal-phase (NP) HPLC is not influenced by the quenching effect of water whereas reversed-phase (RP) HPLC is applicable to more compounds than NP-HPLC.However, pH adjustment and the quenching effect of water on Ln3+ luminescence are the main drawbacks of RP-HPLC. Elution properties and the need for pH adjustment are two arguments for selecting the mode of addition of Ln3+, ithe HPLC system. Sensitized Ln3+ luminescence detection is a much more specific method of detection than UV or fluorescence detection after HPLC separation but nevertheless, in some cases, does not always exhibit a significant increase in analytical performance when the donor itself is a strong fluorophore.The development of more powerful excitation sources could improve the limit of detection of the Ln3+ sensitized detection technique. This review suggests that it would be useful to obtain predicting factors about the drug to establish whether the latter is suitable to be measured using an HPLC–Ln3+ approach.Keywords: Lanthanides; energy transfer; time resolved luminescence; high-performance liquid chromatography; drugs; xenobiotics Luminescence techniques were extensively developed in the last two decades to improve the limits of detection for drug and xenobiotic measurements by high-performance liquid chromatography (HPLC). Even if these modes of detection are not as universal as UV/VIS spectrophotometry, they allow a gain in terms of specificity and sensitivity. Unfortunately, many organic compounds decay from the electronic excited state (S1) to the ground state (S0) without emission of light.Therefore, in order to extend the applications of luminescence techniques, the analyte may be derivatized1 to a fluorescent derivative, or environmental modifiers could be added to increase the native fluorescence.2,3 In the same way, progress made with the phosphorescence detection mode has allowed measurements at room temperature.4 Nevertheless, interferences due to fluorescence impurities, Raman and Rayleigh scattering and the fact that the phosphorescence detection requires deoxygenation of the sample can limit their use.5 To overcome these problems, determination of analytes using sensitized lanthanide ion luminescence detection appears to be a very attractive approach.Lanthanides (Ln3+) possess a large Stokes shift in their luminescence emission, narrow emission bands of 1–20 nm half-width and a long emission lifetime of the order of hundreds of microseconds.In particular, the last property allows time-resolved luminescence (TRL) measurements and therefore interferences can be overcome. Lanthanides are the 14 elements following lanthanum in the Periodic Table. The lanthanide elements are associated with the progressive filling of the 4f-electron subshell. The 4f electronic configuration plays a major role in the process called ‘energy transfer’ by which an excited sample transfers its excitation energy to an acceptor metal ion during the lifetime of the excited state.Since the first energy transfer was described between salicylate and Tb3+,6 numerous studies using lanthanide sensitized luminescence have been carried out. However, the first use of lanthanide ion as the detection chromophore in LC was † Presented at the VIIth International Symposium on Luminescence Spectrometry in Biomedical Analysis, Sophia Antipolis, Nice, France, April 17–19, 1996. Andr�e Rieutord is a hospital pharmacist at the Robert Debr�e Hospital in Paris and has carried out research at the Facult�e de Pharmacie (Ch�atenay-Malabry) in the Laboratoire of Chimie Analytique II of Professor G.Mahuzier. His Research interests involve improving the sensitivity of drug detection in biological fluids using lanthanide sensitized luminescence and other related luminescence techniques in order to apply these schemes to liquid chromatography or capillary electrophoresis separation methods.Analyst, May 1997, Vol. 122 (59R–66R) 59Rreported only in 1985.7 Regarding drug and xenobiotic determinations, several papers using HPLC and Ln3+ sensitized luminescence have been published on non-steroidal antiinflammatory agents (NSAIDs),8 ciprofloxacin,9 nucleic acids,10 bleomycin,11 tetracycline,12,13 steroids,14 orotate,15 theophylline,16 mycotoxines17 and thiol compounds.18–20 The objective of this review is to analyse constraints related to Ln3+ properties and HPLC and emphazise the advantages and limitations of such a technique. We shall discuss the use of normal-phase HPLC (NP-HPLC) and reversed phase HPLC (RP-HPLC) and the mode of addition of lanthanide salts either in pre- or post-column. Finally, we shall discuss the performance of the method.This review deals only with the use of Ln3+ sensitized luminescence as the detection mode in HPLC for drugs and other organic xenobiotics. Other major analytical aspects of the use of lanthanides related to fluorescence immunoassays and dynamic quenching of lanthanide luminescence for the measurement of inorganic anions, for instance, are not included.Generalities Lanthanide Characteristics Despite the poor TRL emission of free lanthanides in solution due to their low e and the efficient quenching of their luminescence by water molecules, they can be very useful for detection when sensitizers are added. Sensitized lanthanide luminescence is a well described method of detection.The Ln3+ complex in solution absorbs energy at a wavelength characteristic of the organic donor and emits radiation with a wavelength characteristic of Ln3+. This reaction known as ‘energy transfer’, has been extensively studied and very well described.21–24 Fig. 1 illustrates the theoretical diagram of energy transfer. Energy is absorbed by the organic ligand leading to an excited state singlet (S1). At this stage there are two possibilities: the molecule can return from S1 to S0 by a radiative transition (fluorescence) or a non-radiative process or can go over to an excited triplet state (T1).Once again, the organic compound has two possibilities: the molecule can go from T1 to S0 by a radiative process (phosphorescence) or a non-radiative process, but in the presence of lanthanide ion energy can be transferred to the metal ion, which may be excited and subsequently emit characteristic radiation. When Ln3+ is free, it will fluoresce when excited by intense radiation but its intensity emission is very weak.In fact, four lanthanide ions (Dy3+, Eu3+, Sm3+ and Tb3+) are mainly used owing to energy level considerations. Briefly, two types of energy transfer processes have been depicted. First, triply charged lanthanide cation can form a stable chelate with ligands when negatively charged donating groups are involved; this is called intramolecular energy transfer. In aqueous solutions, donor groups containing neutral oxygen or nitrogen atoms generally bind only when present in multidentate ligands that contain at least one or two other donor groups having negatively charged oxygen.Table 1 illustrates some structures required for compounds to chelate Ln3+. Second, the donating compound can transfer energy only by a collisional mechanism to the metal; this is the intermolecular energy transfer.25 Drugs for which intermolecular transfer occur include theophylline16 and some a,b-unsaturated steroids.14 Table 2 lists the compounds involved in the latter process.Regarding the compounds listed in Tables 1 and 2, sensitized lanthanide luminescence so far has limited applications. To extend the applicability, a pre-column labelling reaction for thiols containing analytes was developed.18–20 The idea is to derivatize the analyte to a product that, first, is able to form a stable complex with Ln3+ and, second, sensitize the lanthanide luminescence.A commercially available derivatization reagent for thiol-containing analytes, i.e., 4-maleimidosalicylic acid (4-MSA), was chosen because the salicylate group has sensitizing properties for Tb3+. 4-MSA can be utilized in the reaction without any problems as unreacted 4-MSA reagent does not produce any fluorescence or, in the presence of Tb3+, Fig. 1 Energy transfer mechanism between organic donor and Tb3+. Table 1 List of drugs and xenobiotics for which intramolecular energy transfer has been described.Group chelating Ln3+ intermolecular transfer Compound Ref. a-Ketocarboxylate Carboxypsoralene 8 Citrinin 17 Fluoroquinolones 9 Orotate 15 a-Aminocarboxylate Flufenamic acid 8 Mefenamic acid 8 Niflumic acid 8 Ochratoxin A 17 b-Diketone like Ochratoxin A 17 (ketoneroxyl) Tetracyclines 12 Bleomycin 11 Phosphate Adenosine triphosphate 10 Amide and/or amino and ketone Tetracycline 12 Bleomycin 11 Hydroxyl Adenosine triphosphate 10 Bleomycin 11 Table 2 List of drugs for which intermolecular energy transfer has been described.Groups involved in intermolecular process Compound Ref. Carbonyl a,b-Unsaturated steroid 14 Theophylline 16 Amino Theophylline 16 60R Analyst, May 1997, Vol. 122any sensitized Tb3+ luminescence. Successful LC methods with sensitized Ln3+ luminescence have been developed for the determination of glutathione, N-acetylcysteine and l-cystein. 18 Owing to the long lifetime of sensitized lanthanide luminescence, time resolved luminescence (Fig. 2) is used. After the excited pulse, the period of measurement (gain time) starts after a delay time allowing discrimination of short-life luminescence such as background fluorescence. Parameters Influencing Energy Transfer Lanthanide salts are commercially available in chloride or nitrate forms. Chorides are widely used because they are water soluble and compatible with most organic solvents (methanol, acetonitrile, etc.).Greater sensitization has been observed for the chloride versus the nitrate salts in the case of an intermolecular transfer.7 This could be related to the greater ability of NO32 than Cl2 to form an inner sphere.26 Consequently, the likelihood of collisions leading to energy transfer should be larger for Cl2 than for NO32. Lanthanide luminescence is favoured by neutral or alkaline pH owing to ionization of electron donating groups. The ionization of molecules facilitate binding to Ln3+, allowing the formation of a strong chelate.Thus, optimum sensitivity has been reported at pH 8–10 for tetracycline,12–14 at pH 8–8.5 for orotate,15 at pH 6.8 for bleomycin,11 at pH 7.5–8.5 for fluoroquinolones9,27 and at pH 6.7 for an EDTA-like structure, e.g., ADR-925, the hydrolysis product of desrazoxane.28 Further alkalinization causes precipitation of lanthanide hydroxide. 11–15,27,28 Perry and Winefordner29,30 reported that TRIS buffer was likely to have chelating properties towards Ln3+ and thus prevent the quenching effect of water on lanthanide luminescence.However, in our experiments8,9 using diluted NaOH instead of TRIS buffer to adjust the pH to 7.4, we observed a similar enhancement of Tb3+ luminescence. We concluded that TRIS buffer would have no or a minor influence on the energy transfer process. Heat produces a quenching or enhancement effect on the luminescence depending on the type of energy transfer. As intermolecular collisional energy transfer occurs between naphthaldehyde and Eu3+, luminescence is enhanced with increase in temperature.7 In fact, higher temperature increases collisions between the analyte and lanthanide ion, favouring the energy transfer process.Conversly, Tb3+ luminescence sensitized by tetracycline or bleomycin is quenched with increase in temperature.11 In these cases, an increase in temperature increases collisional deactivation. Briefly, in an intramolecular energy transfer mechanism, the signal is decreased by increase in temperature and if a collisional process is involved the signal is increased.Oxygen is a matter of concern for intermolecular energy transfer. In that a particular type of energy transfer, sensitation, is only possible if the solvent is thoroughly deoxygenated since otherwise non-radiative decay of the excited ligand dominates. Similar quenching is not readily observed with chelates i.e intramolecular energy transfer.From our experience, with lanthanide chelates, we never observed any significant difference in luminescence when measuring samples deoxygenated or not.8,9 Another parameter influencing luminescence is the lanthanide concentration. Optimum sensitivity is observed when a certain lanthanide concentration is reached.8,9,15 There is evidence for an optimum [lanthanide]/[organic donor] ratio. This ratio varies considerably, from 1 to 250, depending on the compound.8,9,15,23,31 For the same compound it can vary whether micelles are added or not.32 Possible mechanisms controlling the effect of lanthanide concentration are the collisionnal quenching of free Ln3+, the quenching effect of Cl2 when using chloride salts of lanthanides and the direct excitation of lanthanide ions.In order to decrease radiationless energy transfer, a variety of molecules such as Triton X-100,8,9,13,24,33,34 Tween 2035 and sodium dodecyl sulfate (SDS)15 have been used to provide an insulating sheath around emitting species.By compartmentalizing the Ln3+ and the donor species in a discrete volume region, micelles facilitate the energy transfer process. In particular, micelle addition allows the removal of the efficient quenching effect of water in aqueous solutions. Micelle concentrations should be determined to obtain optimal energy transfer.8,9,14 Neutral molecules such as trioctylphosphine oxide (TOPO) have also been used.8,9,33,34 They have a good lone oxygen atom and a bulky saturated hydrocarbon tail that directs the solvent molecules away from the metal ion and hence contributes to the insulating sheath.36 Sensitized Lanthanide Luminescence in HPLC Drug measurement in complex biological matrices requires very sensitive and specific analytical techniques in order to monitor residue levels.In particular, for intracellular determinations low limits of detection (LOD) ( < 1028 mol l21) will be crucial in the future but at the moment this requirement seems difficult to fulfil systematically.In fact, sensitized lanthanide luminescence used in HPLC separations has been developed for several drugs such as NSAIDs,8 ciprofloxacin,9 bleomycin,11 orotate,15 tetracyclines,12 nucleic acids and nucleotides,10 steroids,14 theophylline,16 mycotoxines17 and thiol compounds, 18–20 with the aim of reaching this LOD and improving the selectivity of the analyte. The separation of organic compounds by HPLC depends on both the stationary phase and the mobile phase.Optimum elution is obtained when a certain equilibrium between the three parameters is reached. Moreover, using HPLC with sensitized lanthanide luminescence as a detection mode, the mobile phase should comply with the luminescence requirement of the lanthanide. We shall discuss the advantages and limitations of using this detection strategy in NP- and RP-HPLC. Reversed-phase HPLC This is the most commonly used method.8–12,15,17,20 Mobile phases are mainly prepared with water, methanol and acetonitrile, which are compatible with energy transfer requirements.A study8 was carried out on the luminescence of Tb3+–mefenamic acid in solvents with various polarity (Table 3). This emphasizes the need for a polar solvent to cause dissociation of the electron donating group, i.e., carboxylate for the NSAIDs, and Fig. 2 Schematic representation of operation of a pulsed-source timeresolved fluorimeter: A, excited pulse; B, fluorescence emission (short-life luminescence); and C, phosphorescence or ion fluorescence emission (longlife luminescence).td, Delay time; tg, gain time. Analyst, May 1997, Vol. 122 61Rfavours binding of Tb3+. NSAIDs are poor sensitizers of Tb3+ in apolar (hexane) or slightly polar solvents (acetonitrile, DMSO). Sensitization of lanthanide luminescence is facilitated in solvents containing hydroxide groups such as methanol and ethanol.Methanol gives better sensitivity on luminescence than acetonitrile and water. On the other hand, Vazquez et al.17 observed that luminescence of the Tb3+–ochratoxin A chelate was greater in butanol than in methanol. The higher viscosity of butanol is likely to limit the non-radioactive deactivation of the chelate owing to a decrease in the number of molecular collisions and to enhance the lanthanide luminescence. Conversely, water is a strong quencher of energy transfer between an organic donor and lanthanide ion, hence a high water content of an RP-HPLC mobile phase causes a dramatic quenching of Ln3+ luminescence.Two ways of limiting the quenching effect of water on luminescence have been described: decreasing the water content of mobile phase or using a displacing agent to remove water from the first coordination sphere of the lanthanide ion. The use of methanol instead of acetonitrile to prepare the mobile phase for the elution of ciprofloxacin allowed a decrease in the proportion of water9 because isoelution with methanol needed an increased methanol content compared with acetonitrile.Owing to the lower water content, better sensitization of Tb3+ was observed. Consequently, this small change improved the sensitivity but increased the back-pressure on-column. The other possibility is to displace water from the first coordination sphere of Ln3+. Acetate ion was the first agent used to discriminate the quenching effect of water on luminescence.7 An 11-fold gain in sensitivity was observed when acetate ion was added to the post-column reagent methanol–water (80 + 20).Hence the luminescence intensity was comparable to that obtained with a post-column reagent containing 100% methanol. The use of a micellar mobile phase was successfully employed to limit the quenching effect of water.14 Addition of SDS at the critical micelle concentration produced a slight increase of Tb3+ sensitized luminescence.In this case, the problem of terbium luminescence quenching in aqueous solutions was circumvented by compartmentalizing the Ln3+ acceptor and the organic donor in SDS micelles. The increased luminescence observed corresponds to the typical increase in lanthanide quantum yield on passing from water to an organic medium.37 An insulating sheath can be created around the Ln3+ ion if a synergistic agent having an oxygen atom for coordination can be selected, therefore excluding water.TOPO, with its lone oxygen atom and three octyl donating chains, protects not only rare earth ions but also the organic donor from collisional interferences.The enhancing effect of TOPO was observed on the sensitized luminescence of some chelates, e.g., Tb3+–ciprofloxacin9 and Tb3+–NSAIDs.8 However, an increased concentration of TOPO causes a decrease in the signal, probably due to the poor solubility of the TOPO molecules in a highly aqueous medium.8 The use of sensitized lanthanide luminescence in a detection scheme in RP-HPLC is the most commonly described approach.Nevertheless, the high water content of RP-HPLC mobile phases produces a strong quenching effect on the lanthanide luminescence. The second drawback is related to the fact that an alkaline pH favouring lanthanide chelate formation is not always compatible with the stationary phase in a conventional HPLC column. However, to overcome the latter problem, we report some solutions under Post-column addition below.Normal-phase HPLC In separation science, NP-HPLC is not used as often as RPHPLC. For polar compounds, however, NP-HPLC is often needed. Only two papers14,16 have described the use of sensitized lanthanide luminescence in NP-HPLC separation. In contrast to RP-HPLC, the absence of water in the mobile phase will prevent the quenching effect of water on luminescence and pH adjustment is not necessary. Addition of water to the mobile phase would favour the binding of Ln3+ to water molecules, preventing intermolecular transfer between an organic donor and Ln3+.Moreover, dissolved oxygen should be removed from the mobile phase to avoid oxygen quenching when energy transfer is involved in an intermolecular process.7 The use of sensitized lanthanide luminescence in NP-HPLC appears to be restricted to polar compounds, the groups of which involved in energy transfer do not require extensive dissociation as for theophylline and some steroids.14,16 Thus cyclohexane, hexane and ethyl acetate, common solvents in NP-HPLC, comply with the requirements of sensitized lanthanide luminescence only with respect to compounds involved in an intermolecular type of energy transfer.The composition of an optimum mobile phase is a compromise between elution properties, backpressure on-column and luminescence enhancement of lanthanide. 14,16 For instance, the use of cyclohexane resulted in an extremly high back-pressure and its replacement by hexane caused only a moderate increase of back-pressure and provided a similar enhancement value for the sensitized luminescence of terbium.Unfortunately, hexane was a relatively weak solvent for elution of theophylline.16 The addition of ethyl acetate to hexane resulted in an excellent separation and detection of theophylline.16 Cyclohexane has also been successfully employed as a mobile phase for Tb3+–steroids. Reverse micelles are used in NP-HPLC to enhance the luminescence of lanthanides sensitized by organic donors such as theophylline16 or steroids.14 Reverse or inverse micelles refer to surfactant aggregation in non-polar solvents.Mwalupindi and Warner16 showed the advantage of using a lanthanide counter ion as opposed to a micellar solution with the lanthanide salt. Fig. 3 illustrates the sensitivity gain obtained when using a mobile phase containing Tb3+ surfactant, i.e., Tb(EHS)3. The enhancement of sensitized terbium luminescence was attributed to the rigid environment of the lanthanide counter-ion in the synthesized surfactant compared with the lanthanide salt.The use of reverse micelles expands the scope of lanthanide sensitized luminescence as method of detection of many organic compounds. However, there is still a need to deoxygenate the sample solution prior to luminescence. Lanthanide Addition Detection of organic donors is linked to the addition of Ln3+ to the chromatographic system.This could be performed in either a pre- or post-column mode. Table 3 Luminescence intensity of Tb3+ (2.5 3 1024 mol l21) sensitized by mefenamic acid (2 31025 mol l21) in different solvents. All measures were recorded at 550 nm.8 Luminescence* Solvent (arbitrary units) Hexane 0.2 Chloroform 0.3 Acetone 1.0 Acetonitrile 1.0 DMF 0.8 DMSO 0.3 Ethanol 42 Methanol 43 Methanol–water (7 + 3) 0.01 TRIS (0.4 mol l21) (pH 7.4) 40 Water–NaOH (pH 7.4) 30 * The luminescence intensity of Tb alone in the different solvents was @0.01. 62R Analyst, May 1997, Vol. 122Pre-column Addition This is the simplest technique: lanthanide salts are added directly to the mobile phase and reaction between Ln3+ and the organic donor will occur before separation on the column. Neither dilution caused by post-column addition nor a noisy baseline caused by the use of a second pump is observed. Nevertheless, the high consumption of expensive Ln3+ reagent should be considered.As reported previously, alkaline pH is favourable for the observation of sensitized Ln3+ luminescence. Unfortunately, even a slightly alkaline pH (!8) is well known to be incompatible with the stability of classical reversed-phase supports owing to the risk of hydrolysis of the alkalyted silanols. The use of a polymer column is recommended and allows the use of a mobile phase having a pH of 8–10.12,15 In two cases, pre-column addition had no influence on retention,10,14 but in one case, Wenzel et al.11 observed a better resolution of Tb3+–bleomycin than bleomycin alone.Complexation with Tb3+, as described for copper, would reduce the number of sites on the bleomycin capable of simultaneously adsorbing on the stationary phase. Post-column Addition Although this mode of addition is slightly more complicated to set up, it extends the applications of sensitized Ln3+ luminescence in HPLC separations. A second pump is needed to draw post-column reagent (PCR) containing Ln3+.A residence time in a post-column reaction coil is not essential for binding to occur. In numerous cases, a simple mixing tee is sufficient if the PCR was conveniently optimized. 8,9 In somes cases, post-column addition is mandatory. Fig. 4 depicts chromatograms obtained for ciprofloxacin using the same mobile phase at different pH. A poor peak shape is observed when using a mobile phase of pH 5.8, whereas good resolution is achieved when the pH of the mobile phase is decreased to 2.5.However, at this pH ciprofloxacin is not ionized, preventing lanthanide binding. This illustrates that an alkaline pH favouring sensitized lanthanide luminescence is hardly compatible with an acidic mobile phase prepared for the separation of compounds involving acidic functions in RPHPLC. 8,9,12,13. Therefore, for compounds such as tetracyclines, fluoroquinolones and NSAIDs, sensitized lanthanide luminescence has been applied in HPLC using a post-column device to add the Ln3+ salts.A buffer mobile phase must be avoided8,9 or must not be too concentrated to be adequately adjusted to an alkaline pH using a PCR.12 The use of phosphate species for acidification is not recommended as there is a risk of precipitation of Ln(PO4)3 in the post-column mixing tee.31 Any lanthanide hydroxide or phosphate would clog the post-column mixing tee. The challenge here is that the composition of PCR must alkalinize the mobile phase to favour sensitized lanthanide luminescence without leading to precipitation of lanthanide hydroxide.For instance, the PCR is adjusted to give pH 7.6 in the mixing tee for optimum sensitivity regarding ciprofloxacin and some NSAIDs measurements8,9 Lanthanide hydroxide formation could be circumvented by adding EDTA to the PCR containing lanthanide.12 Alkalinization up to pH 12 is possible when EDTA is added owing to the formation of a stable and water-soluble chelate.As EDTA does not fully encapsulate Ln3+,38 the organic donor can bind and transfer its energy to the metal. For instance, Wenzel et al.12 were able to adjust the pH to 9 in the mixing tee for tetracycline measurement. Alkaline reagents used in the PCR are usually TRIS buffer,8,12 ammonia buffer,12 or triethylamine.9 The two buffers solutions show original properties that might explain the interest in using them. Ammonia–ammonium chloride buffer exhibits weaker complexation with lanthanide ion than water, increasing the luminescence of lanthanide ion.TRIS buffer also increases the signal obtained, probably because it penetrates the coordination sphere of the chelate, giving rise to a synergestic effect.27 As a general trend, sodium and potassium hydroxide must be avoided in order to prevent lanthanide hydroxide formation. TRIS buffer is particularly interesting owing to its compatibility with Ln3+ ions, which usually precipitate with other buffers (phosphate, carbonate).When lanthanides are added post-column, the detector flow cell, the outlet and inlet capillaries and the mixing tee can easily Fig. 3 Chromatograms for standard samples of theophylline using A, TbCl3 in ethyl hexyl sulfosuccinate (ethyl acetate–hexane, 6 + 4) and B, Tb(EHS)3 in ethyl acetate–hexane (6 + 4) as the mobile phase.16 Fig. 4 Effect of pH (adjusted with H3PO4) on the shape of the peak of ciprofloxacin (1 3 1025 mol l21 methanolic solution) on a C18 column.pH: (A) 5.8; (B) 4.0; (C) 3.5; (D) 2.9; and (E) 2.5. Eluent, methanol–water (50 + 50); flow rate, 0.9 ml min21; detection, direct fluorescence monitoring of the eluent, lex = 278 nm, lem = 440 nm.9 Analyst, May 1997, Vol. 122 63Rbe cleaned by rinsing regularly with dilute nitric acid (1 + 9) to remove lanthanide hydroxide. Post-column addition dilutes the column eluent and causes a decrease in sensitivity. Consequently, the flow rate of the second pump should be as low as possible.The pulsing of this pump will increase the noise and further raise the limit of detection. Hence a two-head plunger pump is recommended. As described for the mobile phase previously, the quenching effect of water must be considered when preparing the PCR. Performance The sensitized lanthanide luminescence detection strategy in HPLC is a very specific method. The likelihood of having a compound co-eluted with an interferent able to transfer its energy to the metal is very low.Fig. 5 illustrates chromatograms obtained when measuring two mycotoxins, i.e., citrinin and ochratoxin A, in cheese extracts.17 UV spectrometry [Fig. 5(a)] is obviously inapplicable, owing to its poor specificity. The selectivity observed when using sensitized lanthanide luminescence [Fig. 5(b)] is greater than with fluorescence detection [Fig. 5(c)]. Similar observations have been reported for real samples, measuring N-acetylcysteine in urine39 (Fig. 6) and mefenamic acid in serum.8 (Fig. 7). Table 4 compares the limits of detection obtained by UV, fluorescence or TRL with sensitized lanthanide luminescence. Depending on the drug, the limits of detection observed using TRL with sensitized lanthanide luminescence are either better (orotate, ciprofloxacin) or worse than those using UV detection (NSAIDs). Regarding ciprofloxacin, fluorescence detection exhibits a threefold better sensitivity than TRL with sensitized lanthanide luminescence. There are two hypotheses to explain these observations.First, one could say that the use of a second pump to draw lanthanide reagent decreases the sensitivity. The second mechanism is suggested by Fig. 8. As illustrated, perfectly selective excitation of the donor is sometimes difficult to perform owing to the absorption profile of the free Tb3+ ions which can interfere at high concentration levels generating high background noise. Nevertheless, the excitation of the Tb3+– ciprofloxacin chelate at a wavelength more specific for the organic compound does not necessarily lead to a real increase in the signal-to-noise ratio, and finally a compromise must be found.Conclusion In this review, the use of sensitized lanthanide luminescence in HPLC for the detection of drugs and xenobiotics has been Fig. 5 (a) Chromatograms of cheese extracts spiked with 1, citrinin and 2, ochratoxin A, with UV detection (l = 331 nm).Extracts spiked with 3 3 1026 mol l21 citrinin and 5 3 1026 mol l21 ochratoxin A. (b) Chromatograms of cheese extracts spiked with 1, citrinin and 2, ochratoxin A, with fluorescence detection (lex = 331 nm, lem = 472 nm). Extracts spiked with 2 3 1023 mol l21 citrinin and 1 3 1026 mol l21 ochratoxin A. (c) Chromatograms of cheese extracts spiked with 1, citrinin and 2, ochratoxin A, with TRL detection (lex = 331 nm, lem = 545 nm). Extracts spiked with 4.5 3 1026 mol l21 citrinin and 1 3 1025 mol l21.17 A and B represent two different extraction methods. Fig. 6 Chromatogram of urine samples spiked with derivatiazed 3 3 1027 mol l21 N-acetylcysteine (with 4-MSA) detected by fluorescence (left) and by sensitized Tb3+ luminescence (right); the dotted lines represent the unspiked samples.39 Fig. 7 Chromatograms of serum samples spiked with 5 3 1026 mol l21 of 1, flufenamic acid and 2, niflumic acid obtained by (a) UV detection at 285 nm and (b) TRL detection using Tb3+ (lex = 340 nm, lem = 545 nm).8 64R Analyst, May 1997, Vol. 122outlined. Numerous compounds possess the structural ability for the energy transfer to occur or can be derivatized to make sensitized lanthanide luminescence detection possible. However, in 10 years only 10–15 papers have been published reporting such a detection strategy in HPLC. The method is highly specific compared with the usual absorbance detection and is better than fluorimetry. Nevertheless, as we described previously, the sensitized lanthanide luminescence technique in HPLC appears sometimes to be hardly more sensitive than UV spectrophotometry or fluorescence for compounds having a high e, e.g., NSAIDs, or having a high fluorescence quantum yield (FF) e.g., fluoroquinolones. This suggests that the sensitized lanthanide luminescence technique is particularly interesting for chromophores exhibiting weak native fluorescence and having a suitable structure for the energy transfer to occur.The use of more powerful energetic sources could decrease the limit of detection in some cases. However, the usual laser systems deliver only a fixed wavelength, e.g., N2 laser 337 nm and KrF laser 248 nm. These wavelengths do not often match the excitation wavelengths of most analytes. The optimum parametric oscillator which is a tunable frequency converter based on an Nd:YAG laser, might provide solutions in terms of the sensitivity of sensitized lanthanide luminescence.40 This review suggests that it would be useful to obtain predicting factors about the drug to establish whether the latter is suitable to be measured using an HPLC–Ln3+ approach.The perfect organic would have a high e, a high FP (phosphorescence quantum yield) to allow efficient energy transfer, a low FF to make energy transfer more interesting in terms of sensitivity and a high Ks (association constant) with the Ln3+ acceptor. Moreover, the energy of the excited triplet state of the organic compound concerned must be nearly equal to or lie above the resonance level of the Ln3+ ion.Studies to determine all these parameters to examine exactly their relevance in the energy transfer process towards Ln3+ ions are warranted. References 1 Farinotti, R., Th`ese de Doctorat d’Etat, Universit�e Paris Sud, 1983. 2 Hinze, W. L., Singh. H. N. Baba, Y., and Harvey, N. G., TrAC, Trends Anal. Chem. (Pers Ed.), 1984, 3, 193. 3 Hoshino, M., Imamura, M., Ikehara, K., and Hama, T., J.Phys. Chem., 1981, 85, 1820. 4 Donkerbroek, J. J., Gooijer, C., Velthorst, N. H., and Frei, R. W., Anal. Chem., 1982, 54, 891. 5 Rendell, D., Fluorescence and Phosphorescence Spectroscopy. Analytical Chemistry by Open Learning, Wiley, New York, 1987. 6 Weissman, S. I., J. Chem. Phys., 1942, 10, 214. 7 Dibella, E. E., Weisman, J. B., Joseph, M. J. Schultz, J. R., and Wenzel, T. J., J. Chromatogr., 1985, 328, 101. 8 Sargi, L., Th`ese de Doctorat d’Etat, Universit�e Paris Sud, 1992. 9 Rieutord, A., Vazquez, L., Soursac, M., Prognon, P., Blais J., Bourget, P., and Mahuzier, G., Anal. Chim. Acta, 1994, 290, 215. 10 Wenzel, T. J., and Colette, L. M., J. Chromatogr., 1988, 436, 299. 11 Wenzel, T. J., Zomlefer, K., Rapkin, S. B., and Keith, R. H., J. Liq. Chromatogr., 1995, 18, 1473. 12 Wenzel, T. J., Colette, L. M., Dahlen, D. T., Hendrickson, S. M., and Yarmaloff, L. W., J. Chromatogr., 1988, 433, 149. 13 Duggan, J. X., J. Liq. Chromatogr., 1991, 14, 2499. 14 Amin, M., Harrington, K., and Von Wandruska, R., Anal. Chem., 1993, 65, 2346. 15 Schreurs, M., Vissers, J. P. C., Gooijer, C., and Velthorst, N. H., Anal. Chim. Acta, 1992, 262, 201. 16 Mwalupindi, A. G., and Warner, I. M., Anal. Chim. Acta, 1995, 306, 49. 17 Vazquez, B. I., Fente, C., Franco, C., Cepeda, A., Prognon, P., and Mahuzier, G., J. Chromatogr. A, 1996, 727, 185. 18 Schreurs, M., Gooijer, C., and Velthorst, N. H., Anal. Chem., 1990, 62, 2053. 19 Schreurs, M., Gooijer, C., and Velthorst, N. H., Fresenius’ J. Anal. Chem., 1991, 339, 499. 20 Schreurs, M., Hellendoorn, L., Gooijer, C., and Velthorst, N. H., J. Chromatogr, 1991, 552, 625. 21 Hemmil�a, I., Scand. J. Clin. Lab. Invest., 1988, 48, 389. 22 Richardson, F. S., Chem. Rev., 1982, 82, 541. 23 Diamandis, E. P., Clin. Biochem., 1988, 21, 139. 24 Georges, J., Analyst, 1993, 118, 1481. 25 Heller, A., and Wasserman, E., J. Chem. Phys., 1965, 42, 949. 26 Choppin, G. R., and Bertha, S. L., J. Inorg. Nucl. Chem., 1973, 35, 1309. 27 Panadero, S., G`omez-Hens, A., and P�erez-Bendito, D., Anal. Chim. Acta, 1995, 303, 39. 28 Hasinoff, B. B., J. Chromatogr. B, 1994, 656, 451. 29 Perry, L. M., and Winefordner, J. D., Talanta, 1990, 10, 965. 30 Perry, L. M., and Winefordner, J. D., Anal. Chim. Acta, 1990, 237, 273. 31 Hirschy, L. M., Dose, E. V., and Wineforder, J. D., Anal. Chim. Acta, 1983, 147, 311. 32 Georges, J., and Ghazarian, S., Anal. Chim. Acta, 1993, 276, 401. 33 Moulin, C., Decambox, P., and Mauchien, P., Anal. Chim. Acta, 1991, 254, 145. 34 Zhu, G., Si, Z., Yang, J., and Ding, J., Anal. Chim. Acta, 1990, 231, 157. 35 Hemmil�a, I., Dahuku, S., Mukkala, V. M., Siitari, H., and L�ovgren, T., Anal. Biochem., 1984, 137, 335. 36 Halverson, F., Brinen, J. S., and Leto, J. R., J. Chem. Phys., 1964, 41, 157. Table 4 Comparison of detection limits (mol l21) obtained using UV spectrophotometry, fluorescence or sensitized lanthanide luminescence with timeresolved luminescence (TRL) in HPLC for several drugs and xenobiotics. Ln3+ Compound UV Ref. Fluorescence Ref. TRL Ref. used Orotate 1 3 1027 41 — 1 3 1028 15 Tb3+* Niflumic acid 2 3 1027 8 — 3 31026 8 Tb3+* Flufenamic acid 4 3 1027 8 — 1 31026 8 Tb3+* Mefenamic acid 4 3 1027 8 — 2.5 3 1026 8 Tb3+* Bleomycin A2 2 3 1027 11 — 3 3 1026 11 Tb3+† Ciprofloxacin 2 3 1027 42 3 3 1028 43 1 3 1027 9 Tb3+* Theophylline 4 3 1027 44 — 5 3 1028 16 Eu3+* * Post-column addition. † Pre-column addition. Fig. 8 Comparison of the peak height (limit of detection = 0.25 pmol injected) and the signal-to-background ratio of ciprofloxacin for excitation performed at (a) 320 and (b) 278 nm. In both instances the flow rate of the post-column pump was fixed at 0.2 ml min21.9 Analyst, May 1997, Vol. 122 65R37 Haas, Y., and Stein, G., J. Phys. Chem., 1971, 75, 3668. 38 Lind, M. D., Lee, B., and Hoard, J. L., J. Am. Chem. Soc., 1965, 87, 1611. 39 Gooijer, C., Schreurs, M., and Velthorst, N. H., in HPLC Detection: Newer Methods, ed. Patonay, G., VCH, New York, 1992, pp. 27– 55. 40 Van de Nesse, R. J., Velthorst, N. H. Brinkman, U. A. Th., and Gooijer, C., J. Chromatogr. A, 1995, 704, 1. 41 Ferrari, V., Giordano, G., Cracco, A. T., Dussini, N., Chiaandetti, L., and Zachelo, F., J. Chromatogr., 1989, 497, 101. 42 Vall�ee, F., Lebel, M., and Bergeron, M. G., Ther. Drug Monit., 1986, 8, 340. 43 Jehl, F., Gallion, C., Debs, J., Brogrard, J. M., Monteil, H., and Minck, R., J. Chromatogr, 1985, 339, 347. 44 Blanchard, J., Harvey, S., and Morgan, W. J., J. Chromatogr. Sci., 1990, 28, 203. Paper 6/07616E Received November 8, 1996 Accepted February 11, 1997 66R Analyst, May

ISSN:0003-2654    

DOI:10.1039/a607616e

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Analytical Methods for the Determination of Platinum in Biological and Environmental Materials A Review Maria Balcerzak Department of Analytical Chemistry, Warsaw University of Technology, Noakowskiego 3, 00-664 Warsaw, Poland Summary of Contents Introduction Pre-treatment Procedures Analytical Methods Separation Methods Quantification of Platinum Available Information on Platinum in Biological and Environmental Materials Conclusions References Keywords: Noble metal determination; platinum determination; clinical analysis; environmental analysis; review Introduction Improvements in detection power, reliability, precision, selectivity and economy of analytical methods are still a challenge to the analyst.The determination of trace amounts of essential and toxic elements at very low concentrations (ng and sub-ng levels) and their speciation in various biological and environmental materials are examples of problems whose solution requires improved analytical procedures.The determination of noble metals presents even more difficult problems. Owing to the great variety of complex matrices (geological, industrial and biological), low concentrations to be determined (ng and pg levels in geological and biological samples) and heterogeneity of examined materials, special pre-treatment of the samples is usually necessary. Quantitative transformation of the platinum metals into suitable complexes, isolation of individual elements from the interfering matrix and their preconcentration up to the level detectable by an appropriate analytical technique are generally required.1–3 The low reactivity of the noble metals towards single chemical reagents, their great chemical similarities (especially between the pairs Ru and Os, Rh and Ir, Pt and Pd4,5), the formation of compounds of similar composition and properties, numerous oxidation states, ability to form many species in a given oxidation state, tendency for hydrolysis and also the formation of polynuclear complexes make their determination (especially at low concentration levels) very difficult even for experienced analysts.Complete dissolution of the noble metals, quantitative conversion into suitable complexes and their separation and preconcentration before final detection are critical steps in almost all analytical procedures. The choice of an appropriate method of determination is a difficult task, especially for a nonexpert.All noble metal species in solutions are extremely coordinated and complex. The great inertness of many complexes means that any ligand substitution requires strong reaction conditions, such as heating, high reagent concentrations and long reaction times. The complexity of the chemical properties of the noble metals and the kinetics of their reactions in solutions pose serious problems, especially to those who deal with their separation and preconcentration by liquid chromatography and their determination by spectrophotometric methods.A standard solution containing the platinum metal in a definite oxidation state and chemical form is required for calibration of the analytical method to be used. Until recently, mainly geochemists involved in exploration for the noble metals and analysts dealing with the analysis of various industrial products have been interested in developing techniques for the determination of very low concentrations of these metals in various matrices.The discovery of the anticancer activity of cisplatin by Rosenberg and co-workers6–8 was the beginning of worldwide interest and intensive studies on the chemistry and biochemistry of platinum complexes acting as cytotoxic agents and on the determination of ultratrace amounts of platinum in biological matrices. The monitoring of cisplatin drugs, introduced into chemotherapy in 1978, has become a very important problem. Cisplatin and carboplatin (the ‘second-generation’ compound) are effective antitumor agents, especially in cases of small cell lung cancers, tumors of the head and neck, carcinomas of the prostate, etc.9–11 However, treatment with cisplatin is complicated by the occurrence of severe sideeffects, among which the high nephrotoxicity is a basic dosage limiting factor.Nausea and vomiting may at times be severe and protracted in patients receiving cisplatin. Sensory neuropathy and high-frequency hearing loss after several cycles of therapy Maria Balcerzak is an Associate Professor in the Department of Analytical Chemistry at Warsaw University of Technology, Poland.She completed her Ph.D. (1978) and Dr.Sc. (habilitation; 1991) degrees in analytical chemistry of the noble metals at the Faculty of Chemistry, Warsaw University of Technology. Separation, preconcentration and determination of trace amounts of noble metals are her main research interests. She has been twice awarded the special prize of Scientific Secretary of the Polish Academy of Sciences for team achievements on noble metals.Her current research activity focuses on speciation analysis of metals and metalloids in biological as well as environmental materials. She lectures on analytical chemistry at the undergraduate level and analytical chemistry of the noble metals at postgraduate level. She is a member of the Trace Inorganic Analysis Committee of the Polish Academy of Sciences. Analyst, May 1997, Vol. 122 (67R–74R) 67Rare not uncommon. Carboplatin shows less nephrotoxicity and is less ototoxic, but is more myelosuppressive.9 It has a range of activity similar to cisplatin. Numerous toxicities of the drugs have led to interest in the determination of trace amounts of platinum absorbed by the human body and of physiological effects of various platinum complexes on living organisms.12–34 Can platinum be accumulated and what is the baseline of platinum in the human body are questions which must be quickly and adequately answered. The interest in the determination of ultra-trace amounts of platinum arose with the introduction in 1975 of the noble metals (Pt, Pd and Rh) as components of automotive catalysts in order to decrease the emission of carbon monoxide, unburnt hydrocarbons and nitrogen oxides in exhaust gases.Soon it turned out that these catalysts were mobile sources of noble metals in the environment. Platinum as the major noble metal in the catalysts (typically containing 0.08% Pt, 0.04% Pd and 0.005–0.007% Rh) is of the greatest concern.There are already data that particles containing nanocrystalline platinum attached to alumina are present in the environment, probably as a consequence of their release from catalysts. The fact that platinum might have entered the food chain was the reason for the rapid and detailed evaluation of the platinum content in various environmental samples and of its toxicological effect on biological systems. Airborne particles,35,36 dust,37–40 soils,41–43 urban gullypots,44 water,45–47 plants,39,48–53 biotic materials48,52,54 and sediments44,45,47 were examined. Relatively high levels of platinum in dust samples from San Diego (0.7 mg g21) have been reported.55 California was the first area where cars were fitted with platinum metal catalysts.The early generations of catalysts emitted much more Pt per km than the present types. An increase in platinum content in various environmental matrices can be noted.The platinum concentration in Swedish road sediments increased from 3.0 ng g21 (1984) to 8.9 ng g21 (1991).44 According to recent data from the UK, platinum concentrations were in the range 0.3–8 ng g21 for road side samples and 0.42–29.8 ng g21 for dust.42 Increased uptake of platinum by plants, especially those treated with some watersoluble compounds, e.g., [Pt(NH3)4](NO3)2, has been reported. 51 Because of difficulties with the determination of platinum at ppb and sub-ppb concentration levels and the lack of suitable certified reference materials, data are scarce, especially for the content of platinum in various complex environmental matrices.The results obtained in various laboratories differ significantly. 32,42 Some attempts have been made to establish the accuracy of the methods by using geological reference materials. 38,45,56 However, most of these materials contain platinum at higher concentration (mg g21) levels.Interlaboratory comparisons and confirmation of the results by various methods are of great value for improving the quality of the determination of platinum, especially in environmental samples. The sample preparation procedures are critical analytical steps and influence the quality of the final results. The representativeness of the sample examined, a digestion procedure which ensures the quantitative conversion of platinum into a suitable complex and elimination of matrix interferences are essential, especially in the analysis of environmental materials.Special procedures for the purification of reagents and vessels are required in order to reduce the blanks below the limits of detection of the most sensitive detection techniques, such as ICP-MS or ETAAS. Analytical methodologies for the quantification of platinum in various biological and environmental samples are critically reviewed in this paper.Pre-treatment Procedures The kind of digestion procedure to be used depends on the kind of matrix, the preconcentration step and the detection method. Relatively simple sample preparation steps can be used for biological materials, especially human body fluids and tissues, containing platinum at concentrations high enough to be detected directly without any sample pre-treatment.13,14,16,23,40 It is generally sufficient, especially for ICP-MS, to dilute the sample with 1–2% nitric acid in order to lower the concentration of dissolved solids, usually below 0.2%.In some cases, however, digestion of samples is necessary. Decomposition of blood, plasma or urine with nitric, hydrochloric and perchloric acid in a high-pressure asher (HPA) at gradually increasing temperature up to 320 °C, followed by adsorptive voltammetric (AV) detection has been reported.19 Dry ashing of blood, spiked blood and a saline standard with concentrated nitric acid at programmed temperatures (1 h at 200 °C, 30 min at 250 °C, 1 h at 350 °C, 30 min at 425 °C and 3 h at 800 °C) followed by dissolution of the residue in aqua regia was also used prior to the determination of platinum by an AV method.15 Dissolution of the samples in aqua regia or in nitric acid of various concentrations (1, 3 and 11 mol l21) after dry ashing at 500 °C for 24 h was applied to the determination of platinum in serum and urine by ETAAS.22 Wet ashing procedures based on the treatment of the samples with concentrated nitric acid followed by boiling with 3% hydrogen peroxide for 3–5 min22 or perchloric acid15 were used to digest various tissues and blood samples before the determination of platinum by ETAAS or ICP-MS methods, respectively.Microwave digestion was applied to mineralize human blood samples prior to the detection of platinum by ICP-AES12 and ICPMS. 31 Wet ashing digestion can be carried out in open or closed systems under pressure. Losses of platinum have been observed during dry ashing procedures despite the complete recovery of the platinum added to the samples.22 The losses of platinum may be due to high volatility of the organometallic compounds of the element present in various environmental and biological samples.Unfortunately, there are insufficient data on the speciation of platinum, especially as far as the presence of organometallic compounds in various matrices is concerned. Attempts to reduce losses of platinum chlorides by their conversion into platinum nitrates have been reported recently. 44 The quantitative digestion of a large variety of environmental samples with complex matrices (not always known) is a very difficult analytical task. Because of the high heterogeneity of the materials examined, large samples as sufficiently representative as possible should be used. Special sample pretreatment and isolation of platinum from the interfering matrix is required.Low concentrations (ng or sub-ng levels) of platinum in environmental materials are difficult to determine even by the most sensitive techniques, such as ETAAS or ICPMS. High pressure ashing with nitric and hydrochloric acid using a suitable temperature program ensuring complete decomposition of organic matrices was used for some plants and biotic materials.48,51,52 Dissolution in aqua regia was the most suitable method for converting platinum from dust,40 sediments44,45,47 and airborne particles35 into soluble complexes.Preliminary treatment of the samples with HNO3, HF or HClO4 may be necessary for complete digestion of more complex matrices. Careful removal of the excess of nitric acid from the sample before any further preconcentration step, especially by liquid chromatography, is necessary. Nitroso complexes of platinum, which may be present in solutions after treatment of the sample with aqua regia, may behave differently to chloride complexes.Hydrofluoric acid should be completely volatilized from the sample because it can damage the samplers. The concentration level of platinum in the reagents used should be carefully controlled in order to avoid high blank values. It turned out that some re-distilled nitric acids contained 68R Analyst, May 1997, Vol. 122more than 1 ng of platinum in 50 ml (the volume used in a single analysis).45 Acids of Suprapur grade (Merck, Darmstadt, Germany) are sufficiently pure for the determination of platinum at ppb levels.Also, the purity of nitric and hydrochloric acids purified by sub-boiling distillation was found to be sufficient.40,43,48,51 The vessels used for the decomposition of the samples can be a source of severe contamination. Memory effects resulting in high and unpredictable blanks can be observed when the same vessels are used for the decomposition of samples with different platinum contents.37 A three-step cleaning process, vapour cleaning with hydrochloric acid, cleaning with nitric and hydrochloric acid in the high pressure asher and vapor cleaning with nitric acid, reduced blank values of the procedure to below the limit of detection of ETAAS.48 Containers made of synthetic, high-purity quartz glass (Suprasil, Heraeus Quarzschmelze, Hanau, Germany) cleaned with nitric acid for about 4–5 h have been considered to ensure undetectable blank levels.37 Water purified with a Millipore Milli-Q water purification system was generally used in all analytical procedures. In view of great difficulties in collecting representative samples and maintaining their integrity during handling and storage, pre-treatment steps should be kept to the minimum.Analytical Methods Reliable and efficient analytical methods are required for the determination of platinum at ng and sub-ng levels in a wide variety of biological and environmental matrices. ETAAS, AV, ICP-AES, ICP-MS and neutron activation analysis (NAA) are the most sensitive techniques and find the widest application.However, the direct use of these techniques is considerably restricted owing to interferences caused by matrix elements and the lack of certified reference materials for extreme trace levels of platinum in different materials. The ETAAS technique has some drawbacks for the determination of platinum. The atomization of platinum requires a relatively high temperature. Platinum is also prone to carbide formation.The technique is sensitive to drift and matrix interferences. Generally, the measuring range does not cover the concentration ranges of interest, especially in the case of environmental materials. Separation and preconcentration steps prior to the final detection are necessary. Adsorptive voltammetric determination is most often based on the measurement of the catalytic reduction of protons by the complex of platinum with formazone (a condensation product of formaldehyde and hydrazine) accumulated on the surface of a hanging mercury drop electrode.The electrochemically active complex lowers the hydrogen overpotential at the mercury electrode, thus producing a very sensitive catalytic current, which is measured in the differential pulse mode. The method is highly sensitive for platinum but is negatively affected by organic matrices, especially surface-active agents. The organic matrix is usually destroyed before the determination.Dry ashing is believed to be the most reliable digestion procedure for the destruction of different matrices with high concentrations of organic substances. ICP-MS is the most promising technique for the determination of ultra-trace levels of elements owing to its high speed, excellent detection limits, wide dynamic range, possibility of accurate multi-element analysis and unique capability of measuring element isotopic ratios. However, ICP-MS signals suffer from some types of matrix and spectral interferences.57,58 High salt concentrations can lead to suppression or enhancement of the measured signal.These effects are considered as matrix effects. Spectral interferences occur because of the limited resolution of the generally applied quadrupole mass spectrometers. Signals from ions, polyatomic groups and analytes with masses that differ by less than 0.5 u can overlap. Sample dilution, chemical separation, chemical modification, alternative sample introduction, mathematical correction, calibration procedures such as standard addition and isotope dilution, which allow one to eliminate or correct interferences in ICP-MS methods, have been discussed in detail.58 The detection of platinum by ICP-MS, especially in complex environmental matrices, is generally preceded by a separation/preconcentration step.NAA offers very low detection limits, but requires access to a nuclear reactor. It is not suitable for routine analysis, but it is particularly useful for checking other methods.Separation Methods Ion-exchange chromatography is most frequently used for matrix elimination and preconcentration of platinum from various samples. All platinum metals can be separated from base metals (including hafnium) using cation or anion exchangers. 2,59–62 Ion-exchange methods have been developed mostly for chloro complexes of platinum metals owing to their common use in analytical procedures. Anionic chloro complexes are strongly retained on basic anion exchange resins from dilute acids or simply pass through cation-exchange columns.Basic metals which exist as cations are retained on cationic exchangers. Strong anionic exchangers find the widest applications in analytical separation procedures for platinum. 27,45,47,52,59,61–63 However, it must be stressed that satisfactory results can be obtained in media containing platinum in the form of a complex of definite composition.A review of separation methods of the platinum metals based on ionexchange chromatography and solvent extraction developed in 1950–83 has been presented by Al-Bazi and Chow.59 Because of the complex nature of the solution chemistry of platinum metals and difficulties in developing reliable separation methods, especially for samples containing several platinum metals, the interest of analysts has been focused on the wider application of various sorbents and complexes for the effective separation and preconcentration of these metals.Enhancement of selectivity may be achieved owing to differences in the tendency of individual metals to form complexes in various media. An on-line sorbent extraction preconcentration system has been developed54 for the determination of platinum and rhodium by ETAAS and ICP-AES after accumulation of their bis(carboxymethyl)dithiocarbamate (CMDTC) chelates on a microcolumn packed with XAD-4 resin. The method was applied to the determination of platinum in some polluted biotic materials.Methods of separation of the platinum metals based on liquid chromatography after their extraction in the form of chelates with 1-(2-pyridylazo)-2-naphthol (PAN) have been presented.60,64 The separation and determination of PtII, OsIV, IrIV, RuIII, CoII and NiII with 2-(6-methyl-2-benzothiazolylazo)- 5-diethylaminophenol (MBTAE) has been described.65 Platinum( iv), PdII, RhIII, RuIII and IrIII can be separated by sorption on an ion-exchange resin containing thiosemicarbazide as chelating agent followed by selective elution.66 Separation of PtII, PdII and RuIII quinolin-8-olates on Silasorb-600 with chloroform– propan-2-ol (98 + 2) has been examined.67 Extraction of the platinum–dithizonate complex in the presence of tin(ii) chloride as labilizing agent,68 followed by ETAAS detection has also been used.48 Activated charcoal46 and modified silica gel49 have been used to preconcentrate platinum from fresh waters and plants before its determination by ICP-MS and ICP-AES, respectively.Coprecipitation with selenium as a collector for platinum, gold, palladium and rhodium from natural waters before their determination by ETAAS and total reflection X-ray fluorescence was studied.69 It should be noted that the methods involving separation steps are time consuming. Close attention must be paid to the Analyst, May 1997, Vol. 122 69Rchemical behavior of the analyte.Direct methods that allow the determination of very low levels of individual elements in complex matrices are needed. Quantification of Platinum The concentration of platinum in some clinical samples is high enough to be detected by ETAAS,13,22,24 AV,15,19,21,24 ICPAES12,18,23 and ICP-MS.14–16 ICP-MS is the most often used technique for the determination of ultra-trace amounts of platinum in environmental matrices.35,40,42,43,46,47,51,56 Relatively large amounts of base metals present in samples may cause severe matrix interferences.High concentrations of sodium perchlorate and sodium chloride decrease the sensitivity of the platinum determination whereas the presence of sulfides and natural organic (fulvic) acids increases the sensitivity by a factor of up to 4.70 Sulfates decrease the sensitivity. Calibration procedures, such as standard addition or isotope dilution, are recommended especially when the nature of the matrix is unknown or cannot be easily matched with the standards.Gold,15,16,21 europium14 and indium25 were used as internal standards to compensate for the matrix effects in the ICP-MS determination of platinum in some clinical samples. Oxides of hafnium, ytterbium and tungsten cause spectral interferences in the determination of platinum.35 Hafnium, which occurs in the earth’s crust at concentrations over 300 times higher than that of platinum, enhances the platinum signal. The HfO signal overlaps the signals of the three most abundant platinum isotopes, 194Pt (32.9%), 195Pt (33.8%) and 196Pt (25.2%).Application of the stable isotope dilution (ID) technique to the evaluation of platinum concentration by ICP-MS can be an effective approach to compensate for the suppression of sensitivity by interferents.17,43,47 ID–ICP-MS measurement is relatively free from interferences that may be caused by complex matrices.71,72 It is based on the addition of a known amount of an enriched isotope (called the ‘spike’) to the sample and calculation of the analyte concentration by measuring the altered isotopic ratio instead of the absolute ion intensity.The isotopic ratio defines the element concentration and does not change with element losses. The enriched isotope should be added to the sample at an early stage of the analytical procedure. An advantage of the isotope dilution technique is that the chemical separation of the element to be determined need not be quantitative.The method requires calibration of the isotopic abundances and concentration of the spike. Isotope ratios are less affected by instrumental drift than are the ion sensitivities. ID–ICP-MS was used for the simultaneous determination of Pt, Re and Ir in natural waters and sediments after anion-exchange separation of the chloro complexes of Pt and Ir and perrhenate ion (ReO42).47 Platinum in urine was determined by ID–GCMS using bis(trifluoroethyl)dithiocarbamate as chelating agent and enriched 192Pt.17 The use of ID–ICP-MS for the determination of ultra-trace amounts of platinum in geological samples has recently been reported.73,74 Attempts to improve the detection limits of ICP-MS for platinum by improving the nebulization rates and transmission of the analyte into the plasma have recently been reported. 30,40,43,75 The capability of ICP-MS with electrothermal vaporization (ETV) of determining platinum in various matrices without preconcentration or matrix separation steps40 and in fresh waters after preconcentration of the analyte on activated charcoal46 has been investigated.The coupling of a thermospray system for sample introduction to the plasma resulted in an increase in the platinum signal intensity by a factor of 9 compared with the signal obtained using a conventional pneumatic nebulizer.75 Direct electrospray ionization of eluent into the plasma, recently developed for the successful coupling of LC to ICP-MS, has been used for separation of the anticancer drug cisplatin, its trans isomer and their cytotoxic hydrated complexes.30 The use of a high resolution (HR) magnetic sector field instrument instead of a quadrupole (Q) in ICP-MS represents an interesting approach to the determination of ultra-trace amounts of platinum in complex matrices.Using HR-ICP-MS the sensitivity of the determination of physiological levels of Pt, Pd, Rh and Au in human blood was increased by about two orders of magnitude.31 HR-ICP-MS has recently been used for the determination of platinum in grass samples.51 Radiochemical neutron activation analysis (RNAA) based on the production of 199Au as indicator radionuclide [198Pt(n,g)199Pt]?199Au] and its separation using different procedures has also been used for the determination of platinum in some biological and environmental materials.26,27,32,76–79 Platinum contents in some tissues (kidney, liver and lung) of laboratory mice with transplanted L1210 leukemia and treated with cisplatin were determined using extraction of 199Au with diethyldithiocarbamate (DDTC).32 However, the platinum content determined (31 ng g21) in a standard reference material (Bowen’s kale) in this work differed significantly from previous data (198 ng g21).76 An RNAA method using 199Au was also used for the determination of platinum in three BCR Certified Reference Materials (No. 184, bovine muscle; 185, bovine liver; 186, pig kidney).79 Electrolytic separation of gold on a niobium cathode ensured a very high radiochemical purity and eliminated the interferences from calcium and other elements.Unfortunately, only the bovine liver was found to contain platinum above the detection limit (2 ng g21). The selective separation of gold based on its adsorption on polyurethane foam from hydrochloric acid medium was employed in the determination of platinum in street dust and corn plant.38 Extraction with dithizone supplemented the analytical procedure for samples containing high antimony (!5 mg g21) and low platinum ( < 50 ng g21) concentrations.The results obtained for dust samples were in good agreement with the platinum content evaluated with the aid of other methods (ETAAS, ICP-MS, ID– ICP-MS and AV). However, large differences were observed for the corn plant and geological standard reference material SU-1. According to the authors, this might have been due to insufficient homogeneity of the samples.Results for the content of platinum in liver, kidney, placenta and brain of mice obtained by NAA were better correlated with the drug treatment compared with determination by ETAAS.27 Detailed studies on the effects of neutron activation conditions, such as short irradiation, cyclic and long irradiation, with both thermal and epithermal neutrons, on the determination of platinum and other noble metals in vegetation were carried out.53 Available Information on Platinum in Biological and Environmental Materials The contents of platinum in various biological and environmental materials evaluated by different analytical techniques are presented in Tables 1 and 2, respectively. It can be seen that most analytical investigations have focused on the evaluation of the content of platinum in body fluids and tissues after administration of platinum-containing anticancer drugs.14,18,23,25,29,32 The natural levels of platinum in human body may be two to three orders of magnitude lower than those in patients treated with cisplatin.Owing to difficulties in measuring extremely low platinum concentrations, only a few data on the baseline of platinum in human body have been published.15,19,21 A comparison of published values obtained in different laboratories for the same materials reveals great differences. Natural levels of platinum determined in body fluids differ significantly.Values of 0.8–6.9 ng l21 and 0.56 mg l21 for blood and 0.5–15 ng l21 and 0.18 mg l21 for urine have been reported.19,21 70R Analyst, May 1997, Vol. 122The relatively higher (up to 100 times) platinum contents in the urine and blood of industrial workers show that large amounts of platinum are emitted during catalyst manufacture and recycling. Platinum concentrations of 32–180 ng l21 in blood, 95–280 ng l21 in plasma and 21–2900 ng l21 in urine were reported.19 Platinum compounds, especially the soluble salts, are toxic and chronic industrial exposure to them is responsible for a syndrome known as platinosis.Chloroplati- Table 1 Analytical data for the content of platinum in biological samples Concentration Material Sample digestion Separation technique Detection technique range (DL) Ref. Human blood HNO3 ICP-AES (20 mg l21) 12 PTFE bomb, MW l = 214.423 nm Human plasma Dilution with ETAAS 0.84–242 mmol l21 13 Plasma ultrafiltrate 0.15 mol l21 NaCl 0.24–242 mmol l21 Saliva and 0.2 mol l21 0.12–3.64 mmol l21 Urine HCl 2.02–12.1 (mmol– mmol) l21 Plasma Dilution with ICP-MS 100–1000 mg l21 14 chemotherapy 1% HNO3 europium as treatment internal standard Human blood Dry ashing AV 0.1–2.8 mg l21 15 natural level HNO3, HCl ICP-MS Wet ashing gold as internal HNO3, HClO4 standard Serum Dilution with ICP-MS 170–2800 mg l21 16 1% HNO3 gold as internal standard ETAAS Urine HNO3,H3PO4, GC MS 125 ± 6 mg l21 17 SRM 2670 H2O2 bis(trifluoroethyl) dithiocarbamate chelate Urine ICP-AES 0.6–21.7 mg l21 18 chemotherapy l = 214.423 nm (0.05 mg l21) treatment Urine High pressure AV 0.5–15 ng l21 19 ashing (21–2900 ng l21) Blood HNO3, HCl, @0.8–6.9 ng l21 HClO4 (32–180 ng l21) Blood plasma @0.8–6.9 ng l21 baseline (exposed (95–280 ng l21) people) (0.2 ng l21) Urine Dilution with Extraction with DDTC AV 0.1–100 mg ml21 20 0.15 mol l21 NaCl Blood AV 0.56 mg l21 21 Urine 0.18 mg l21 baseline Serum, urine, tissues Dry ashing ETAAS 56 pg 22 HCl, HNO3 Wet ashing HNO3, H2O2 Plasma Dilution with DCP-AES 2.53 mg ml21 23 chemotherapy 0.2% HNO3 l = 265.95 nm (0.2 mg ml21) treatment Urine (0.04 mg ml21) Ultrafiltrate (0.02 mg ml21) Urine Dilution with AV (0.4 mg ml21) 24 HNO3 ETAAS Blood HNO3, HCl ICP-MS (0.1 mg l21) 25 Urine indium as internal Tissues standard chemotherapy ETAAS (10 mg l21) treatment Animal tissues HNO3 Ion exchange RNAA 0.6–82.5 mg g21 27 Plants 0.33–810 ng g21 Plasma ultrafiltrate Dilution with HPLC ICP-MS 250–4600 ng ml21 29 chemotherapy acetonitrile + water Hypersil Phenyl (0.6 ng ml21) treatment 5 mm bonded silica Plants, animal tissues Extraction with DDTC RNAA 31 ng g21 32 chemotherapy treatment Urine ICP-MS 1.2–35 ng l21 40 Urine HNO3 ID–ICP-MS 7.77 ng g21 57 Animal tissues Separation of Au on RNAA (2 ng g21) 79 Nb cathode Analyst, May 1997, Vol. 122 71Rnates irritate the skin and mucous membranes.They can cause allergic reactions, resulting in high rates of occupational asthma and dermatitis.11 In contrast, platinum in the metallic state is non-toxic and non-allergenic. The transformation mechanism of metallic platinum into bioavailable forms is still unknown and has to be investigated.Increasing trends in the content of platinum in environmentally relevant matrices should be noted. A threefold increase in the platinum content in Swedish road sediments between 1984 and 1991 has been reported.44 An increased uptake of platinum by plants growing on polluted soils has been found.50,51 Reported platinum concentrations in plants range from 3.76 ng g2143 (ref. 43) to as high as 2070 mg g21.48 Large differences in data obtained by different laboratories for the same environmental samples examined should be noted.42 A lack of suitable reference materials with certified values for platinum does not allow the reliability of the published results to be confirmed. The evaluation of the final hazard of platinum to living organisms requires speciation studies of the element in various biological and environmental matrices.As was well established earlier,80 the bioavailability of a particular essential or toxic element depends on its chemical form. Investigations of platinum species in various matrices have become a great challenge to the analysts. Plants as one of the main components of the human food chain may be of greatest interest.50,51,81 However, so far the analytical data have mostly been limited to results obtained by a combination of gel permeation chromatography with AV.50 Some platinum species have been isolated from grass samples treated with tetraammineplatinum(ii) nitrate, [Pt(NH3)4](NO3)2.It was found that 90% and < 10% of platinum is bound to low molecular mass species (about 1 kDa) and to species from 19 up to > 1000 kDa, respectively. Speciation studies are of great value for the explanation of the chemical behavior and the reaction mechanisms of anticancer agents.It is known that only about 5% of the total platinum from the applied drugs undergoes biotransformation processes with the formation of products with anticancer activity. Most of the platinum is excreted from the body or converted into unreactive compounds with proteins. The requirements as to the sensitivity and reliability of analytical techniques increase in the case of speciation studies carried out in order to elucidate the chemical Table 2 Analytical data for the content of platinum in environmental materials Concentration Material Sample digestion Separation technique Detection technique range (DL) Ref.Airborne HF, HClO4 Cation exchange ICP-MS 0.014–0.184 mg g21 35 particulate matter aqua regia, HCl (0.005 mg g21) Airborne dust HPA AV 0.6–130 mg kg21 36 HNO3, HCl, HClO4, H2SO4 Filter dust HPA AV 148–187 ng 37 Corn H2SO4, HCl 80 mg kg21 Ryegrass 320–1500 ng kg21 Street dust HNO3, HCl, HF Adsorption of Au RNAA 14.5 ng g21 38 Corn plants HNO3, HCl on polyurethane foam 642 ng g21 Plants HNO3, HCl ETAAS 0.050–2.6 mg g21 39 Dust PTFE bomb (0.3 ng) Dust HPA ICP-MS 47.2–67.9 mg kg21 40 HNO3, HClO4, HF aqua regia Soils Nickel sulfide fire INAA 330 ± 223 ng g21 41 assay Soils ICP-MS < 0.3–8 ng g21 42 Road dust 0.42–29.8 ng g21 Plants HNO3, HClO4 Sorption of ID–ICP-MS 3.76 ng g21 43 Soils HF, aqua regia, bis(carboxymethyl) 79.4 ng g21 HNO3 dithiocarbamate Urban gullypots Dry ashing AV 1.7–3.8 ng l21 44 Sediments HNO3 (0.5 ng g21) HCl + HNO3 HCl Marine waters HNO3, HClO4, Anion exchange ETAAS 90–150 pg l21 45 Sediments HNO3 + HCl 0.7–21.9 ng g21 Waters Activated charcoal ICP-MS (0.3–0.8 ng l21) 46 Waters HCl, HF, HNO3 Anion exchange ID–ICP-MS (14 pg) 47 Sediments MW HNO3, HCl Plants HPA Extraction ETAAS 1.45–2070 mg kg21 48 HNO3, HCl SnCl2, dithizone Plants Silica gel Separon AES 2–20 mg in 25 ml 49 SGX C18 gold as internal matrix solution standard Grass HPA HR-ICP-MS 2 ng g21 (highways) 51 HNO3 + HCl 0.2 ng g21 (rural areas) Plants HPA Anion exchange UV 0.07–1.52 mg kg21 52 aqua regia (40 ng g21) Plants Dry ashing Ion exchange RNAA 0.1–0.83 mg g21 53 Biotic materials HPA Sorption of ETAAS 0.07–1.5 ng g21 54 HNO3, HCl bis(carboxymethyl) ICP-AES dithiocarbamate 72R Analyst, May 1997, Vol. 122transformation of platinum in the human body and reaction mechanism. Conclusions The development of analytical methods allowing the determination of low platinum concentrations (mg l21, ng g21, ng l21 and pg g21) in a wide variety of biological and environmental materials is a challenge to the analyst.Extremely sensitive and selective methods are required for the determination of platinum, in particular in complex environmental matrices. The quality of the results obtained is limited by the representativeness of the sample examined, the kind of digestion procedure applied to the quantitative conversion of platinum into a suitable complex, the preconcentration step and elimination of matrix interferences.A fundamental problem to be resolved is the lack of suitable reference materials containing platinum at concentration levels of interest. An adequate knowledge of the bioavailability, bioaccumulation and toxicity of platinum to living organisms requires accurate analytical data on the total platinum content in various biological and environmental samples and on the formation of species of various biological activity.However, the proper evaluation of the concentrations of various platinum forms is seriously hampered by the insufficient detection limits of available analytical techniques. Improvements in analytical methods applied to the isolation and preconcentration of particular platinum species prior to final detection are of great interest. This work was supported by the State Committee for Scientific Research (project No. 3 T09A 042 08, sponsored 1995–97).References 1 Beamish, F. E., and Van Loon, J. C., Recent Advances in the Analytical Chemistry of the Noble Metals, Pergamon Press, Oxford, 1972. 2 Ginsburg, S. I., Ezerskaya, N. A., Prokof’eva, I. V., Fedorenko, N. F., Shlenskaya, V. I., and Belskii, N. K., Analytical Chemistry of Platinum Metals, Nauka, Moscow, 1972. 3 Van Loon, J. C., and Barefoot, R. R., Determination of the Precious Metals, Selected Instrumental Methods, Wiley, Chichester, 1991. 4 Griffith, W.P., The Chemistry of the Rarer Platinum Metals (Os, Ru, Ir and Rh), Wiley, London, 1967. 5 Livingstone, S. E., The Chemistry of Ruthenium, Rhodium, Palladium, Osmium, Iridium and Platinum, Pergamon Press, Oxford, 1973. 6 Rosenberg, B., Van Camp, L., and Krigas, T., Nature (London), 1965, 205, 698. 7 Rosenberg, B., Van Camp, L., Grimely, E., and Thomson, A., J. Biol. Chem., 1967, 242, 347. 8 Rosenberg, B., Van Camp, L., Trosko, J. E., and Mansour, V. H., Nature (London), 1969, 222, 385. 9 Chemistry of the Platinum Group Metals, ed., Hartley, F.R., Elsevier, Amsterdam, 1991. 10 Metal Complexes in Cancer Chemotherapy, ed., Keppler, B. K., VCH, Weinheim, 1993. 11 Handbook on Metals in Clinical and Analytical Chemistry, ed., Seiler, H. G., Sigel, A., and Sigel, H., Marcel Dekker, New York, 1994. 12 Di Noto, V., Ni, D., Dalla Via, L., Scomazzon, F., and Vidali, M., Analyst, 1995, 120, 1669. 13 Van Warmerdam, L. J. C., Van Tellingen, O., Maes, R.A. A., and Beijnen, J. H., Fresenius’ J. Anal. Chem., 1995, 351, 777. 14 Allain, P., Berre, S., Mauras, A., and Le Bouil, A., Biol. Mass Spectrom., 1992, 21, 141. 15 Nygren, O., Vaughan, G. T., Florence, T. M., Morrison, G. M. P., Warner, I. M., and Dale, L. S., Anal. Chem., 1990, 62, 1637. 16 Casetta, B., Roncadin, M., Montanari, G., and Furlanut, M., At. Spectrosc., 1991, 12, 81. 17 Aggarwal, S. K., Kinter, M., and Herold, D. A., J. Am. Soc. Mass Spectrom., 1991, 2, 85. 18 Lo, F.B., Aral, D. K., and Nazar, M. A., J. Anal. Toxicol., 1987, 11, 242. 19 Messerschmidt, J., Alt, F., T�olg, G., Angerer, J., and Schaller, K. H., Fresenius’ J. Anal. Chem., 1992, 343, 391. 20 Schmid, G. M., and Atherton, D. R., Anal. Chem., 1986, 58, 1956. 21 Vaughan, G. T., and Florence, T. M., Sci. Total Environ., 1992, 111, 47. 22 McGahan, M. Ch., and Tyczkowska, K., Spectrochim. Acta, Part B, 1987, 42, 665. 23 McLoughlin, S., Bowdler, D., and Roberts, N. B., J. Anal.At. Spectrom., 1988, 3, 273. 24 Shearan, P., and Smyth, M. R., Analyst, 1988, 113, 609. 25 Tothill, P., Matheson, L. M., Smyth, J. F., and McKay, K., J. Anal. At. Spectrom., 1990, 5, 619. 26 Haeberlin, S. T., Lux, F., Karl, J., Spruss, T., and Sch�onenberger, H., J. Radioanal. Nucl. Chem., 1987, 113, 461. 27 Esposito, M., Collecchi, P., Oddone, M., and Meloni, S., J. Radioanal. Nucl. Chem., 1987, 113, 437. 28 Ivanov, A. V., Zheligovskaya, N. N., and Nesterenko, P. N., Zh. Anal.Khim., 1995, 50, 190. 29 Cairns, W. R. L., Ebdon, L., and Hill, S. J., Fresenius&rs J. Anal. Chem., 1996, 355, 202. 30 Ehrsson, H. C., Wallin, I. B., Andersson, A. S., and Edlund, P. O., Anal. Chem., 1995, 67, 3608. 31 Begerow, J., and Dunemann, L., J. Anal. At. Spectrom., 1996, 11, 303. 32 Taskaev, E., Karaivanova, M., and Grigorov, T., J. Radioanal. Nucl. Chem., 1988, 120, 75. 33 Alimonti, A., Dominici, C., Petrucci, F., La Torre, F., and Caroli, S., Anal. Chim. Hung., 1991, 128, 527. 34 Barefoot, R. R., and Van Loon, J. C., Anal. Chim. Acta, 1996, 334, 5. 35 Mukai, H., Ambe, Y., and Morita, M., J. Anal. At. Spectrom., 1990, 5, 75. 36 Alt, F., Bambauer, A., Hoppstock, K., Mergler, B., and T�olg, G., Fresenius’ J. Anal. Chem., 1993, 346, 693. 37 Hoppstock, K., Alt, F., Cammann, K., and Weber, G., Fresenius’ Z. Anal. Chem., 1989, 335, 813. 38 Wildhagen, D., and Krivan, V., Anal. Chim. Acta, 1993, 274, 257. 39 Beinrohr, E., Lee, M. L., Tsch�opel, P., and T�olg, G., Fresenius’ J.Anal. Chem., 1993, 346, 689. 40 Schramel, P., Wendler, I., and Lustig, S., Fresenius’ J. Anal. Chem., 1995, 353, 115. 41 Heinrich, E., Schmidt, G., and Kratz, K. L., Fresenius’ J. Anal. Chem., 1996, 354, 883. 42 Farago, M. E., Kavanagh, P., Blanks, R., Kelly, J., Kazantzis, G., Thornton, I., Simpson, P. R., Cook, J. M., Parry, S., and Hall, G. M., Fresenius’ J. Anal. Chem., 1996, 354, 660. 43 Parent, M., Vanhoe, H., Moens, L., and Dams, R., Fresenius’ J.Anal. Chem., 1996, 354, 664. 44 Wei, Ch., and Morrison, G. M., Anal. Chim. Acta, 1994, 284, 587. 45 Hodge, V., Stallard, M., Koide, M., and Goldberg, E. D., Anal. Chem., 1986, 58, 616. 46 Hall, G. E. M., and Pelchat, J. C., J. Anal. At. Spectrom., 1993, 8, 1059. 47 Colodner, D. C., Boyle, E. A., and Edmond, J. M., Anal. Chem., 1993, 65, 1419. 48 Alt, F., Jerono, U., Messerschmidt, J., and T�olg, G., Mikrochim. Acta, 1988, III, 299. 49 Otruba, V., Strnadov�a, M., and Skaln�ýkov�a, B., Talanta, 1993, 40, 221. 50 Messerschmidt, J., Alt, F., and T�olg, G., Anal. Chim. Acta, 1994, 291, 161. 51 Vanhaecke, F., Verstraete, D., Moens, L., and Dams, R., paper presented at the Winter Conference on Plasma Spectrochemistry, Florida, 1996, FP54. 52 Jerono, U., Alt, F., Messerschmidt, J., and T�olg, G., Mikrochim. Acta, 1992, 108, 221. 53 Valente, I. M., Minski, M. J., and Peterson, P. J., J. Radioanal. Chem., 1982, 71, 115. 54 Lee, M. L., T�olg, G., Beinrohr, E., and Tsch�opel, P., Anal.Chim. Acta, 1993, 272, 193. 55 Hodge, V. F., and Stallard, M. O., Environ. Sci. Technol., 1986, 20, 1058. 56 Perry, B. J., Barefoot, R. R., and Van Loon, J. C., Trends Anal. Chem., 1995, 14, 388. Analyst, May 1997, Vol. 122 73R57 Olivares, J. A., and Houk, R. S., Anal. Chem., 1986, 58, 20. 58 Dams, R. F. J., Goossens, J., and Moens, L., Mikrochim. Acta, 1995, 119, 277. 59 Al-Bazi, S. J., and Chow, A., Talanta, 1984, 31, 815. 60 Alimarin, I. P., Ivanov, V.M., Bol’shova, T. A., and Basova, E. M., Fresenius’ J. Anal. Chem., 1989, 335, 63. 61 Alimarin, I. P., Basova, E. M., Bol’shova, T. A., and Ivanov, V. M., Zh. Anal. Khim., 1986, 41, 5. 62 Rocklin, R. D., Anal. Chem., 1984, 56, 1959. 63 Brooks, R. R., and Ahrens, L. H., Spectrochim. Acta, 1960, 16, 783. 64 Basova, E. M., Bol’shova, T. A., Ivanov, V. M., and Morozova, N. B., Zh. Anal. Khim., 1989, 44, 680. 65 Liu, W. P., and Liu, Q. P., Fresenius’ J. Anal. Chem., 1994, 350, 671. 66 Siddhanta, S., and Das, H. R., Talanta, 1985, 32, 457. 67 Malykhin, A. Yu., Bol’shova, T. A., Ermolenko, I. N., Bobrovich, I. B., Nikishkin, I. A., and Polyakova, L. A., Zh. Anal. Khim., 1989, 44, 886. 68 Balcerzak, M., Analysis, 1994, 22, 353. 69 Eller, R., Alt, F., T�olg, G., and Tobschall, H. J., Fresenius’ Z. Anal. Chem., 1989, 334, 723. 70 Wood, S. A., Vlassopoulos, D., and Mucci, A., Anal. Chim. Acta, 1990, 229, 227. 71 Fassett, J. D., and Paulsen, P. J., Anal. Chem., 1989, 61, 643A. 72 McLaren, J. W., Beauchemin, D., and Berman, S. S., Anal. Chem., 1987, 59, 610. 73 Enzweiler, J., Potts, P. J., and Jarvis, K. E., Analyst, 1995, 120, 1391. 74 Yi, Y.V., and Masuda, A., Anal. Chem., 1996, 68, 1444. 75 Parent, M., Vanhoe, H., Moens, L., and Dams, R., Anal. Chim. Acta, 1996, 320, 1. 76 Nadkarni, R. A., and Morrison, G. H., J. Radioanal. Chem., 1977, 38, 435. 77 Zeisler, R., and Greenberg, R. R., J. Radioanal. Chem., 1982, 75, 27. 78 Sykes, T.R., Stephens-Newsham, L. G., Apps, M. J., and Noujaim, A. A., J. Radioanal. Chem., 1982, 69, 441. 79 Rietz, B., and Heydorn, K., J. Radioanal. Nucl. Chem., 1993, 174, 49. 80 Cornelis, R., Borguet, F., and De Kompe, J., Anal. Chim. Acta, 1993, 283, 183. 81 Bettmer, J., Buscher, W., and Cammann, K., Fresenius’ J. Anal. Chem., 1996, 354, 521. Paper 6/08153C Received December 3, 1996 Accepted February 10, 1997 74R Analyst, May 1997, Vol. 122 Analytical Methods for the Determination of Platinum in Biological and Environmental Materials A Review Maria Balcerzak Department of Analytical Chemistry, Warsaw University of Technology, Noakowskiego 3, 00-664 Warsaw, Poland Summary of Contents Introduction Pre-treatment Procedures Analytical Methods Separation Methods Quantification of Platinum Available Information on Platinum in Biological and Environmental Materials Conclusions References Keywords: Noble metal determination; platinum determination; clinical analysis; environmental analysis; review Introduction Improvements in detection power, reliability, precision, selectivity and economy of analytical methods are still a challenge to the analyst.The determination of trace amounts of essential and toxic elements at very low concentrations (ng and sub-ng levels) and their speciation in various biological and environmental materials are examples of problems whose solution requires improved analytical procedures. The determination of noble metals presents even more difficult problems.Owing to the great variety of complex matrices (geological, industrial and biological), low concentrations to be determined (ng and pg levels in geological and biological samples) and heterogeneity of examined materials, special pre-treatment of the samples is usually necessary. Quantitative transformation of the platinum metals into suitable complexes, isolation of individual elements from the interfering matrix and their preconcentration up to the level detectable by an appropriate analytical technique are generally required.1–3 The low reactivity of the noble metals towards single chemical reagents, their great chemical similarities (especially between the pairs Ru and Os, Rh and Ir, Pt and Pd4,5), the formation of compounds of similar composition and properties, numerous oxidation states, ability to form many species in a given oxidation state, tendency for hydrolysis and also the formation of polynuclear complexes make their determination (especially at low concentration levels) very difficult even for experienced analysts.Complete dissolution of the noble metals, quantitative conversion into suitable complexes and their separation and preconcentration before final detection are critical steps in almost all analytical procedures. The choice of an appropriate method of determination is a difficult task, especially for a nonexpert. All noble metal species in solutions are extremely coordinated and complex.The great inertness of many complexes means that any ligand substitution requires strong reaction conditions, such as heating, high reagent concentrations and long reaction times. The complexity of the chemical properties of the noble metals and the kinetics of their reactions in solutions pose serious problems, especially to those who deal with their separation and preconcentration by liquid chromatography and their determination by spectrophotometric methods.A standard solution containing the platinum metal in a definite oxidation state and chemical form is required for calibration of the analytical method to be used. Until recently, mainly geochemists involved in exploration for the noble metals and analysts dealing with the analysis of various industrial products have been interested in developing techniques for the determination of very low concentrations of thesices.The discovery of the anticancer activity of cisplatin by Rosenberg and co-workers6–8 was the beginning of worldwide interest and intensive studies on the chemistry and biochemistry of platinum complexes acting as cytotoxic agents and on the determination of ultratrace amounts of platinum in biological matrices. The monitoring of cisplatin drugs, introduced into chemotherapy in 1978, has become a very important problem. Cisplatin and carboplatin (the ‘second-generation’ compound) are effective antitumor agents, especially in cases of small cell lung cancers, tumors of the head and neck, carcinomas of the prostate, etc.9–11 However, treatment with cisplatin is complicated by the occurrence of severe sideeffects, among which the high nephrotoxicity is a basic dosage limiting factor.Nausea and vomiting may at times be severe and protracted in patients receiving cisplatin. Sensory neuropathy and high-frequency hearing loss after several cycles of therapy Maria Balcerzak is an Associate Professor in the Department of Analytical Chemistry at Warsaw University of Technology, Poland.She completed her Ph.D. (1978) and Dr.Sc. (habilitation; 1991) degrees in analytical chemistry of the noble metals at the Faculty of Chemistry, Warsaw University of Technology. Separation, preconcentration and determination of trace amounts of noble metals are her main research interests. She has been twice awarded the special prize of Scientific Secretary of the Polish Academy of Sciences for team achievements on noble metals.Her current research activity focuses on speciation analysis of metals and metalloids in biological as well as environmental materials. She lectures on analytical chemistry at the undergraduate level and analytical chemistry of the noble metals at postgraduate level. She is a member of the Trace Inorganic Analysis Committee of the Polish Academy of Sciences. Analyst, May 1997, Vol. 122 (67R–74R) 67Rare not uncommon.Carboplatin shows less nephrotoxicity and is less ototoxic, but is more myelosuppressive.9 It has a range of activity similar to cisplatin. Numerous toxicities of the drugs have led to interest in the determination of trace amounts of platinum absorbed by the human body and of physiological effects of various platinum complexes on living organisms.12–34 Can platinum be accumulated and what is the baseline of platinum in the human body are questions which must be quickly and adequately answered. The interest in the determination of ultra-trace amounts of platinum arose with the introduction in 1975 of the noble metals (Pt, Pd and Rh) as components of automotive catalysts in order to decrease the emission of carbon monoxide, unburnt hydrocarbons and nitrogen oxides in exhaust gases.Soon it turned out that these catalysts were mobile sources of noble metals in the environment. Platinum as the major noble metal in the catalysts (typically containing 0.08% Pt, 0.04% Pd and 0.005–0.007% Rh) is of the greatest concern.There are already data that particles containing nanocrystalline platinum attached to alumina are present in the environment, probably as a consequence of their release from catalysts. The fact that platinum might have entered the food chain was the reason for the rapid and detailed evaluation of the platinum content in various environmental samples and of its toxicological effect on biological systems. Airborne particles,35,36 dust,37–40 soils,41–43 urban gullypots,44 water,45–47 plants,39,48–53 biotic materials48,52,54 and sediments44,45,47 were examined. Relatively high levels of platinum in dust samples from San Diego (0.7 mg g21) have been reported.55 California was the first area where cars were fitted with platinum metal catalysts.The early generations of catalysts emitted much more Pt per km than the present types. An increase in platinum content in various environmental matrices can be noted.The platinum concentration in Swedish road sediments increased from 3.0 ng g21 (1984) to 8.9 ng g21 (1991).44 According to recent data from the UK, platinum concentrations were in the range 0.3–8 ng g21 for road side samples and 0.42–29.8 ng g21 for dust.42 Increased uptake of platinum by plants, especially those treated with some watersoluble compounds, e.g., [Pt(NH3)4](NO3)2, has been reported. 51 Because of difficulties with the determination of platinum at ppb and sub-ppb concentration levels and the lack of suitable certified reference materials, data are scarce, especially for the content of platinum in various complex environmental matrices. The results obtained in various laboratories differ significantly. 32,42 Some attempts have been made to establish the accuracy of the methods by using geological reference materials. 38,45,56 However, most of these materials contain platinum at higher concentration (mg g21) levels.Interlaboratory comparisons and confirmation of the results by various methods are of great value for improving the quality of the determination of platinum, especially in environmental samples. The sample preparation procedures are critical analytical steps and influence the quality of the final results. The representativeness of the sample examined, a digestion procedure which ensures the quantitative conversion of platinum into a suitable complex and elimination of matrix interferences are essential, especially in the analysis of environmental materials.Special procedures for the purification of reagents and vessels are required in order to reduce the blanks below the limits of detection of the most sensitive detection techniques, such as ICP-MS or ETAAS. Analytical methodologies for the quantification of platinum in various biological and environmental samples are critically reviewed in this paper. Pre-treatment Procedures The kind of digestion procedure to be used depends on the kind of matrix, the preconcentration step and the detection method.Relatively simple sample preparation steps can be used for biological materials, especially human body fluids and tissues, containing platinum at concentrations high enough to be detected directly without any sample pre-treatment.13,14,16,23,40 It is generally sufficient, especially for ICP-MS, to dilute the sample with 1–2% nitric acid in order to lower the concentration of dissolved solids, usually below 0.2%.In some cases, however, digestion of samples is necessary. Decomposition of blood, plasma or urine with nitric, hydrochloric and perchloric acid in a high-pressure asher (HPA) at gradually increasing temperature up to 320 °C, followed by adsorptive voltammetric (AV) detection has been reported.19 Dry ashing of blood, spiked blood and a saline standard with concentrated nitric acid at programmed temperatures (1 h at 200 °C, 30 min at 250 °C, 1 h at 350 °C, 30 min at 425 °C and 3 h at 800 °C) followed by dissolution of the residue in aqua regia was also used prior to the determination of platinum by an AV method.15 Dissolution of the samples in aqua regia or in nitric acid of various concentrations (1, 3 and 11 mol l21) after dry ashing at 500 °C for 24 h was applied to the determination of platinum in serum and urine by ETAAS.22 Wet ashing procedures based on the treatment of the samples with concentrated nitric acid followed by boiling with 3% hydrogen peroxide for 3–5 min22 or perchloric acid15 were used to digest various tissues and blood samples before the determination of platinum by ETAAS or ICP-MS methods, respectively.Microwave digestion was applied to mineralize human blood samples prior to the detection of platinum by ICP-AES12 and ICPMS. 31 Wet ashing digestion can be carried out in open or closed systems under pressure. Losses of platinum have been observed during dry ashing procedures despite the complete recovery of the platinum added to the samples.22 The losses of platinum may be due to high volatility of the organometallic compounds of the element present in various environmental and biological samples.Unfortunately, there are insufficient data on the speciation of platinum, especially as far as the presence of organometallic compounds in various matrices is concerned. Attempts to reduce losses of platinum chlorides by their conversion into platinum nitrates have been reported recently. 44 The quantitative digestion of a large variety of environmental samples with complex matrices (not always known) is a very difficult analytical task. Because of the high heterogeneity of the materials examined, large samples as sufficiently representative as possible should be used. Special sample pretreatment and isolation of platinum from the interfering matrix is required. Low concentrations (ng or sub-ng levels) of platinum in environmental materials are difficult to determine even by the most sensitive techniques, such as ETAAS or ICPMS.High pressure ashing with nitric and hydrochloric acid using a suitable temperature program ensuring complete decomposition of organic matrices was used for some plants and biotic materials.48,51,52 Dissolution in aqua regia was the most suitable method for converting platinum from dust,40 sediments44,45,47 and airborne particles35 into soluble complexes.Preliminary treatment of the samples with HNO3, HF or HClO4 may be necessary for complete digestion of more complex matrices. Careful removal of the excess of nitric acid from the sample before any further preconcentration step, especially by liquid chromatography, is necessary. Nitroso complexes of platinum, which may be present in solutions after treatment of the sample with aqua regia, may behave differently to chloride complexes. Hydrofluoric acid should be completely volatilized from the sample because it can damage the samplers. The concentration level of platinum in the reagents used should be carefully controlled in order to avoid high blank values.It turned out that some re-distilled nitric acids contained 68R Analyst, May 1997, Vol. 122more than 1 ng of platinum in 50 ml (the volume used in a single analysis).45 Acids of Suprapur grade (Merck, Darmstadt, Germany) are sufficiently pure for the determination of platinum at ppb levels.Also, the purity of nitric and hydrochloric acids purified by sub-boiling distillation was found to be sufficient.40,43,48,51 The vessels used for the decomposition of the samples can be a source of severe contamination. Memory effects resulting in high and unpredictable blanks can be observed when the same vessels are used for the decomposition of samples with different platinum contents.37 A three-step cleaning process, vapour cleaning with hydrochloric acid, cleaning with nitric and hydrochloric acid in the high pressure asher and vapor cleaning with nitric acid, reduced blank values of the procedure to below the limit of detection of ETAAS.48 Containers made of synthetic, high-purity quartz glass (Suprasil, Heraeus Quarzschmelze, Hanau, Germany) cleaned with nitric acid for about 4–5 h have been considered to ensure undetectable blank levels.37 Water purified with a Millipore Milli-Q water purification system was generally used in all analytical procedures. In view of great difficulties in collecting representative samples and maintaining their integrity during handling and storage, pre-treatment steps should be kept to the minimum.Analytical Methods Reliable and efficient analytical methods are required for the determination of platinum at ng and sub-ng levels in a wide variety of biological and environmental matrices. ETAAS, AV, ICP-AES, ICP-MS and neutron activation analysis (NAA) are the most sensitive techniques and find the widest application.However, the direct use of these techniques is considerably restricted owing to interferences caused by matrix elements and the lack of certified reference materials for extreme trace levels of platinum in different materials. The ETAAS technique has some drawbacks for the determination of platinum. The atomization of platinum requires a relatively high temperature. Platinum is also prone to carbide formation.The technique is sensitive to drift and matrix interferences. Generally, the measuring range does not cover the concentration ranges of interest, especially in the case of environmental materials. Separation and preconcentration steps prior to the final detection are necessary. Adsorptive voltammetric determination is most often based on the measurement of the catalytic reduction of protons by the complex of platinum with formazone (a condensation product of formaldehyde and hydrazine) accumulated on the surface of a hanging mercury drop electrode.The electrochemically active complex lowers the hydrogen overpotential at the mercury electrode, thus producing a very sensitive catalytic current, which is measured in the differential pulse mode. The method is highly sensitive for platinum but is negatively affected by organic matrices, especially surface-active agents. The organic matrix is usually destroyed before the determination. Dry ashing is believed to be the most reliable digestion procedure for the destruction of different matrices with high concentrations of organic substances.ICP-MS is the most promising technique for the determination of ultra-trace levels of elements owing to its high speed, excellent detection limits, wide dynamic range, possibility of accurate multi-element analysis and unique capability of measuring element isotopic ratios. However, ICP-MS signals suffer from some types of matrix and spectral interferences.57,58 High salt concentrations can lead to suppression or enhancement of the measured signal.These effects are considered as matrix effects. Spectral interferences occur because of the limited resolution of the generally applied quadrupole mass spectrometers. Signals from ions, polyatomic groups and analytes with masses that differ by less than 0.5 u can overlap. Sample dilution, chemical separation, chemical modification, alternative sample introduction, mathematical correction, calibration procedures such as standard addition and isotope dilution, which allow one to eliminate or correct interferences in ICP-MS methods, have been discussed in detail.58 The detection of platinum by ICP-MS, especially in complex environmental matrices, is generally preceded by a separation/preconcentration step.NAA offers very low detection limits, but requires access to a nuclear reactor. It is not suitable for routine analysis, but it is particularly useful for checking other methods.Separation Methods Ion-exchange chromatography is most frequently used for matrix elimination and preconcentration of platinum from various samples. All platinum metals can be separated from base metals (including hafnium) using cation or anion exchangers. 2,59–62 Ion-exchange methods have been developed mostly for chloro complexes of platinum metals owing to their common use in analytical procedures. Anionic chloro complexes are strongly retained on basic anion exchange resins from dilute acids or simply pass through cation-exchange columns.Basic metals which exist as cations are retained on cationic exchangers. Strong anionic exchangers find the widest applications in analytical separation procedures for platinum. 27,45,47,52,59,61–63 However, it must be stressed that satisfactory results can be obtained in media containing platinum in the form of a complex of definite composition. A review of separation methods of the platinum metals based on ionexchange chromatography and solvent extraction developed in 1950–83 has been presented by Al-Bazi and Chow.59 Because of the complex nature of the solution chemistry of platinum metals and difficulties in developing reliable separation methods, especially for samples containing several platinum metals, the interest of analysts has been focused on the wider application of various sorbents and complexes for the effective separation and preconcentration of these metals.Enhancement of selectivity may be achieved owing to differences in the tendency of individual metals to form complexes in various media. An on-line sorbent extraction preconcentration system has been developed54 for the determination of platinum and rhodium by ETAAS and ICP-AES after accumulation of their bis(carboxymethyl)dithiocarbamate (CMDTC) chelates on a microcolumn packed with XAD-4 resin. The method was applied to the determination of platinum in some polluted biotic materials.Methods of separation of the platinum metals based on liquid chromatography after their extraction in the form of chelates with 1-(2-pyridylazo)-2-naphthol (PAN) have been presented.60,64 The separation and determination of PtII, OsIV, IrIV, RuIII, CoII and NiII with 2-(6-methyl-2-benzothiazolylazo)- 5-diethylaminophenol (MBTAE) has been described.65 Platinum( iv), PdII, RhIII, RuIII and IrIII can be separated by sorption on an ion-exchange resin containing thiosemicarbazide as chelating agent followed by selective elution.66 Separation of PtII, PdII and RuIII quinolin-8-olates on Silasorb-600 with chloroform– propan-2-ol (98 + 2) has been examined.67 Extraction of the platinum–dithizonate complex in the presence of tin(ii) chloride as labilizing agent,68 followed by ETAAS detection has also been used.48 Activated charcoal46 and modified silica gel49 have been used to preconcentrate platinum from fresh waters and plants before its determination by ICP-MS and ICP-AES, respectively.Coprecipitation with selenium as a collector for platinum, gold, palladium and rhodium from natural waters before their determination by ETAAS and total reflection X-ray fluorescence was studied.69 It should be noted that the methods involving separation steps are time consuming. Close attention must be paid to the Analyst, May 1997, Vol. 122 69Rchemical behavior of the analyte. Direct methods that allow the determination of very low levels of individual elements in complex matrices are needed.Quantification of Platinum The concentration of platinum in some clinical samples is high enough to be detected by ETAAS,13,22,24 AV,15,19,21,24 ICPAES12,18,23 and ICP-MS.14–16 ICP-MS is the most often used technique for the determination of ultra-trace amounts of platinum in environmental matrices.35,40,42,43,46,47,51,56 Relatively large amounts of base metals present in samples may cause severe matrix interferences.High concentrations of sodium perchlorate and sodium chloride decrease the sensitivity of the platinum determination whereas the presence of sulfides and natural organic (fulvic) acids increases the sensitivity by a factor of up to 4.70 Sulfates decrease the sensitivity. Calibration procedures, such as standard addition or isotope dilution, are recommended especially when the nature of the matrix is unknown or cannot be easily matched with the standards.Gold,15,16,21 europium14 and indium25 were used as internal standards to compensate for the matrix effects in the ICP-MS determination of platinum in some clinical samples. Oxides of hafnium, ytterbium and tungsten cause spectral interferences in the determination of platinum.35 Hafnium, which occurs in the earth’s crust at concentrations over 300 times higher than that of platinum, enhances the platinum signal. The HfO signal overlaps the signals of the three most abundant platinum isotopes, 194Pt (32.9%), 195Pt (33.8%) and 196Pt (25.2%).Application of the stable isotope dilution (ID) technique to the evaluation of platinum concentration by ICP-MS can be an effective approach to compensate for the suppression of sensitivity by interferents.17,43,47 ID–ICP-MS measurement is relatively free from interferences that may be caused by complex matrices.71,72 It is based on the addition of a known amount of an enriched isotope (called the ‘spike’) to the sample and calculation of the analyte concentration by measuring the altered isotopic ratio instead of the absolute ion intensity.The isotopic ratio defines the element concentration and does not change with element losses. The enriched isotope should be added to the sample at an early stage of the analytical procedure. An advantage of the isotope dilution technique is that the chemical separation of the element to be determined need not be quantitative.The method requires calibration of the isotopic abundances and concentration of the spike. Isotope ratios are less affected by instrumental drift than are the ion sensitivities. ID–ICP-MS was used for the simultaneous determination of Pt, Re and Ir in natural waters and sediments after anion-exchange separation of the chloro complexes of Pt and Ir and perrhenate ion (ReO42).47 Platinum in urine was determined by ID–GCMS using bis(trifluoroethyl)dithiocarbamate as chelating agent and enriched 192Pt.17 The use of ID–ICP-MS for the determination of ultra-trace amounts of platinum in geological samples has recently been reported.73,74 Attempts to improve the detection limits of ICP-MS for platinum by improving the nebulization rates and transmission of the analyte into the plasma have recently been reported. 30,40,43,75 The capability of ICP-MS with electrothermal vaporization (ETV) of determining platinum in various matrices without preconcentration or matrix separation steps40 and in fresh waters after preconcentration of the analyte on activated charcoal46 has been investigated.The coupling of a thermospray system for sample introduction to the plasma resulted in an increase in the platinum signal intensity by a factor of 9 compared with the signal obtained using a conventional pneumatic nebulizer.75 Direct electrospray ionization of eluent into the plasma, recently developed for the successful coupling of LC to ICP-MS, has been used for separation of the anticancer drug cisplatin, its trans isomer and their cytotoxic hydrated complexes.30 The use of a high resolution (HR) magnetic sector field instrument instead of a quadrupole (Q) in ICP-MS represents an interesting approach to the determination of ultra-trace amounts of platinum in complex matrices.Using HR-ICP-MS the sensitivity of the determination of physiological levels of Pt, Pd, Rh and Au in human blood was increased by about two orders of magnitude.31 HR-ICP-MS has recently been used for the determination of platinum in grass samples.51 Radiochemical neutron activation analysis (RNAA) based on the production of 199Au as indicator radionuclide [198Pt(n,g)199Pt]?199Au] and its separation using different procedures has also been used for the determination of platinum in some biological and environmental materials.26,27,32,76–79 Platinum contents in some tissues (kidney, liver and lung) of laboratory mice with transplanted L1210 leukemia and treated with cisplatin were determined using extraction of 199Au with diethyldithiocarbamate (DDTC).32 However, the platinum content determined (31 ng g21) in a standard reference material (Bowen’s kale) in this work differed significantly from previous data (198 ng g21).76 An RNAA method using 199Au was also used for the determination of platinum in three BCR Certified Reference Materials (No. 184, bovine muscle; 185, bovine liver; 186, pig kidney).79 Electrolytic separation of gold on a niobium cathode ensured a very high radiochemical purity and eliminated the interferences from calcium and other elements.Unfortunately, only the bovine liver was found to contain platinum above the detection limit (2 ng g21). The selective separation of gold based on its adsorption on polyurethane foam from hydrochloric acid medium was employed in the determination of platinum in street dust and corn plant.38 Extraction with dithizone supplemented the analytical procedure for samples containing high antimony (!5 mg g21) and low platinum ( < 50 ng g21) concentrations.The results obtained for dust samples were in good agreement with the platinum content evaluated with the aid of other methods (ETAAS, ICP-MS, ID– ICP-MS and AV). However, large differences were observed for the corn plant and geological standard reference material SU-1. According to the authors, this might have been due to insufficient homogeneity of the samples.Results for the content of platinum in liver, kidney, placenta and brain of mice obtained by NAA were better correlated with the drug treatment compared with determination by ETAAS.27 Detailed studies on the effects of neutron activation conditions, such as short irradiation, cyclic and long irradiation, with both thermal and epithermal neutrons, on the determination of platinum and other noble metals in vegetation were carried out.53 Available Information on Platinum in Biological and Environmental Materials The contents of platinum in various biological and environmental materials evaluated by different analytical techniques are presented in Tables 1 and 2, respectively.It can be seen that most analytical investigations have focused on the evaluation of the content of platinum in body fluids and tissues after administration of platinum-containing anticancer drugs.14,18,23,25,29,32 The natural levels of platinum in human body may be two to three orders of magnitude lower than those in patients treated with cisplatin.Owing to difficulties in measuring extremely low platinum concentrations, only a few data on the baseline of platinum in human body have been published.15,19,21 A comparison of published values obtained in different laboratories for the same materials reveals great differences. Natural levels of platinum determined in body fluids differ significantly. Values of 0.8–6.9 ng l21 and 0.56 mg l21 for blood and 0.5–15 ng l21 and 0.18 mg l21 for urine have been reported.19,21 70R Analyst, May 1997, Vol. 122The relatively higher (up to 100 times) platinum contents in the urine and blood of industrial workers show that large amounts of platinum are emitted during catalyst manufacture and recycling. Platinum concentrations of 32–180 ng l21 in blood, 95–280 ng l21 in plasma and 21–2900 ng l21 in urine were reported.19 Platinum compounds, especially the soluble salts, are toxic and chronic industrial exposure to them is responsible for a syndrome known as platinosis.Chloroplati- Table 1 Analytical data for the content of platinum in biological samples Concentration Material Sample digestion Separation technique Detection technique range (DL) Ref. Human blood HNO3 ICP-AES (20 mg l21) 12 PTFE bomb, MW l = 214.423 nm Human plasma Dilution with ETAAS 0.84–242 mmol l21 13 Plasma ultrafiltrate 0.15 mol l21 NaCl 0.24–242 mmol l21 Saliva and 0.2 mol l21 0.12–3.64 mmol l21 Urine HCl 2.02–12.1 (mmol– mmol) l21 Plasma Dilution with ICP-MS 100–1000 mg l21 14 chemotherapy 1% HNO3 europium as treatment internal standard Human blood Dry ashing AV 0.1–2.8 mg l21 15 natural level HNO3, HCl ICP-MS Wet ashing gold as internal HNO3, HClO4 standard Serum Dilution with ICP-MS 170–2800 mg l21 16 1% HNO3 gold as internal standard ETAAS Urine HNO3,H3PO4, GC MS 125 ± 6 mg l21 17 SRM 2670 H2O2 bis(trifluoroethyl) dithiocarbamate chelate Urine ICP-AES 0.6–21.7 mg l21 18 chemotherapy l = 214.423 nm (0.05 mg l21) treatment Urine High pressure AV 0.5–15 ng l21 19 ashing (21–2900 ng l21) Blood HNO3, HCl, @0.8–6.9 ng l21 HClO4 (32–180 ng l21) Blood plasma @0.8–6.9 ng l21 baseline (exposed (95–280 ng l21) people) (0.2 ng l21) Urine Dilution with Extraction with DDTC AV 0.1–100 mg ml21 20 0.15 mol l21 NaCl Blood AV 0.56 mg l21 21 Urine 0.18 mg l21 baseline Serum, urine, tissues Dry ashing ETAAS 56 pg 22 HCl, HNO3 Wet ashing HNO3, H2O2 Plasma Dilution with DCP-AES 2.53 mg ml21 23 chemotherapy 0.2% HNO3 l = 265.95 nm (0.2 mg ml21) treatment Urine (0.04 mg ml21) Ultrafiltrate (0.02 mg ml21) Urine Dilution with AV (0.4 mg ml21) 24 HNO3 ETAAS Blood HNO3, HCl ICP-MS (0.1 mg l21) 25 Urine indium as internal Tissues standard chemotherapy ETAAS (10 mg l21) treatment Animal tissues HNO3 Ion exchange RNAA 0.6–82.5 mg g21 27 Plants 0.33–810 ng g21 Plasma ultrafiltrate Dilution with HPLC ICP-MS 250–4600 ng ml21 29 chemotherapy acetonitrile + water Hypersil Phenyl (0.6 ng ml21) treatment 5 mm bonded silica Plants, animal tissues Extraction with DDTC RNAA 31 ng g21 32 chemotherapy treatment Urine ICP-MS 1.2–35 ng l21 40 Urine HNO3 ID–ICP-MS 7.77 ng g21 57 Animal tissues Separation of Au on RNAA (2 ng g21) 79 Nb cathode Analyst, May 1997, Vol. 122 71Rnates irritate the skin and mucous membranes.They can cause allergic reactions, resulting in high rates of occupational asthma and dermatitis.11 In contrast, platinum in the metallic state is non-toxic and non-allergenic. The transformation mechanism of metallic platinum into bioavailable forms is still unknown and has to be investigated.Increasing trends in the content of platinum in environmentally relevant matrices should be noted. A threefold increase in the platinum content in Swedish road sediments between 1984 and 1991 has been reported.44 An increased uptake of platinum by plants growing on polluted soils has been found.50,51 Reported platinum concentrations in plants range from 3.76 ng g2143 (ref. 43) to as high as 2070 mg g21.48 Large differences in data obtained by different laboratories for the same environmental samples examined should be noted.42 A lack of suitable reference materials with certified values for platinum does not allow the reliability of the published results to be confirmed. The evaluation of the final hazard of platinum to living organisms requires speciation studies of the element in various biological and environmental matrices.As was well established earlier,80 the bioavailability of a particular essential or toxic element depends on its chemical form. Investigations of platinum species in various matrices have become a great challenge to the analysts. Plants as one of the main components of the human food chain may be of greatest interest.50,51,81 However, so far the analytical data have mostly been limited to results obtained by a combination of gel permeation chromatography with AV.50 Some platinum species have been isolated from grass samples treated with tetraammineplatinum(ii) nitrate, [Pt(NH3)4](NO3)2.It was found that 90% and < 10% of platinum is bound to low molecular mass species (about 1 kDa) and to species from 19 up to > 1000 kDa, respectively. Speciation studies are of great value for the explanation of the chemical behavior and the reaction mechanisms of anticancer agents.It is known that only about 5% of the total platinum from the applied drugs undergoes biotransformation processes with the formation of products with anticancer activity. Most of the platinum is excreted from the body or converted into unreactive compounds with proteins. The requirements as to the sensitivity and reliability of analytical techniques increase in the case of speciation studies carried out in order to elucidate the chemical Table 2 Analytical data for the content of platinum in environmental materials Concentration Material Sample digestion Separation technique Detection technique range (DL) Ref.Airborne HF, HClO4 Cation exchange ICP-MS 0.014–0.184 mg g21 35 particulate matter aqua regia, HCl (0.005 mg g21) Airborne dust HPA AV 0.6–130 mg kg21 36 HNO3, HCl, HClO4, H2SO4 Filter dust HPA AV 148–187 ng 37 Corn H2SO4, HCl 80 mg kg21 Ryegrass 320–1500 ng kg21 Street dust HNO3, HCl, HF Adsorption of Au RNAA 14.5 ng g21 38 Corn plants HNO3, HCl on polyurethane foam 642 ng g21 Plants HNO3, HCl ETAAS 0.050–2.6 mg g21 39 Dust PTFE bomb (0.3 ng) Dust HPA ICP-MS 47.2–67.9 mg kg21 40 HNO3, HClO4, HF aqua regia Soils Nickel sulfide fire INAA 330 ± 223 ng g21 41 assay Soils ICP-MS < 0.3–8 ng g21 42 Road dust 0.42–29.8 ng g21 Plants HNO3, HClO4 Sorption of ID–ICP-MS 3.76 ng g21 43 Soils HF, aqua regia, bis(carboxymethyl) 79.4 ng g21 HNO3 dithiocarbamate Urban gullypots Dry ashing AV 1.7–3.8 ng l21 44 Sediments HNO3 (0.5 ng g21) HCl + HNO3 HCl Marine waters HNO3, HClO4, Anion exchange ETAAS 90–150 pg l21 45 Sediments HNO3 + HCl 0.7–21.9 ng g21 Waters Activated charcoal ICP-MS (0.3–0.8 ng l21) 46 Waters HCl, HF, HNO3 Anion exchange ID–ICP-MS (14 pg) 47 Sediments MW HNO3, HCl Plants HPA Extraction ETAAS 1.45–2070 mg kg21 48 HNO3, HCl SnCl2, dithizone Plants Silica gel Separon AES 2–20 mg in 25 ml 49 SGX C18 gold as internal matrix solution standard Grass HPA HR-ICP-MS 2 ng g21 (highways) 51 HNO3 + HCl 0.2 ng g21 (rural areas) Plants HPA Anion exchange UV 0.07–1.52 mg kg21 52 aqua regia (40 ng g21) Plants Dry ashing Ion exchange RNAA 0.1–0.83 mg g21 53 Biotic materials HPA Sorption of ETAAS 0.07–1.5 ng g21 54 HNO3, HCl bis(carboxymethyl) ICP-AES dithiocarbamate 72R Analyst, May 1997, Vol. 122transformation of platinum in the human body and reaction mechanism. Conclusions The development of analytical methods allowing the determination of low platinum concentrations (mg l21, ng g21, ng l21 and pg g21) in a wide variety of biological and environmental materials is a challenge to the analyst.Extremely sensitive and selective methods are required for the determination of platinum, in particular in complex environmental matrices. The quality of the results obtained is limited by the representativeness of the sample examined, the kind of digestion procedure applied to the quantitative conversion of platinum into a suitable complex, the preconcentration step and elimination of matrix interferences.A fundamental problem to be resolved is the lack of suitable reference materials containing platinum at concentration levels of interest. An adequate knowledge of the bioavailability, bioaccumulation and toxicity of platinum to living organisms requires accurate analytical data on the total platinum content in various biological and environmental samples and on the formation of species of various biological activity.However, the proper evaluation of the concentrations of various platinum forms is seriously hampered by the insufficient detection limits of available analytical techniques. Improvements in analytical methods applied to the isolation and preconcentration of particular platinum species prior to final detection are of great interest. This work was supported by the State Committee for Scientific Research (project No. 3 T09A 042 08, sponsored 1995–97). References 1 Beamish, F.E., and Van Loon, J. C., Recent Advances in the Analytical Chemistry of the Noble Metals, Pergamon Press, Oxford, 1972. 2 Ginsburg, S. I., Ezerskaya, N. A., Prokof’eva, I. V., Fedorenko, N. F., Shlenskaya, V. I., and Belskii, N. K., Analytical Chemistry of Platinum Metals, Nauka, Moscow, 1972. 3 Van Loon, J. C., and Barefoot, R. R., Determination of the Precious Metals, Selected Instrumental Methods, Wiley, Chichester, 1991. 4 Griffith, W. P., The Chemistry of the Rarer Platinum Metals (Os, Ru, Ir and Rh), Wiley, London, 1967. 5 Livingstone, S. E., The Chemistry of Ruthenium, Rhodium, Palladium, Osmium, Iridium and Platinum, Pergamon Press, Oxford, 1973. 6 Rosenberg, B., Van Camp, L., and Krigas, T., Nature (London), 1965, 205, 698. 7 Rosenberg, B., Van Camp, L., Grimely, E., and Thomson, A., J. Biol. Chem., 1967, 242, 347. 8 Rosenberg, B., Van Camp, L., Trosko, J. E., and Mansour, V. H., Nature (London), 1969, 222, 385. 9 Chemistry of the Platinum Group Metals, ed., Hartley, F.R., Elsevier, Amsterdam, 1991. 10 Metal Complexes in Cancer Chemotherapy, ed., Keppler, B. K., VCH, Weinheim, 1993. 11 Handbook on Metals in Clinical and Analytical Chemistry, ed., Seiler, H. G., Sigel, A., and Sigel, H., Marcel Dekker, New York, 1994. 12 Di Noto, V., Ni, D., Dalla Via, L., Scomazzon, F., and Vidali, M., Analyst, 1995, 120, 1669. 13 Van Warmerdam, L. J. C., Van Tellingen, O., Maes, R. A. A., and Beijnen, J. H., Fresenius’ J. Anal.Chem., 1995, 351, 777. 14 Allain, P., Berre, S., Mauras, A., and Le Bouil, A., Biol. Mass Spectrom., 1992, 21, 141. 15 Nygren, O., Vaughan, G. T., Florence, T. M., Morrison, G. M. P., Warner, I. M., and Dale, L. S., Anal. Chem., 1990, 62, 1637. 16 Casetta, B., Roncadin, M., Montanari, G., and Furlanut, M., At. Spectrosc., 1991, 12, 81. 17 Aggarwal, S. K., Kinter, M., and Herold, D. A., J. Am. Soc. Mass Spectrom., 1991, 2, 85. 18 Lo, F. B., Aral, D. K., and Nazar, M. A., J.Anal. Toxicol., 1987, 11, 242. 19 Messerschmidt, J., Alt, F., T�olg, G., Angerer, J., and Schaller, K. H., Fresenius’ J. Anal. Chem., 1992, 343, 391. 20 Schmid, G. M., and Atherton, D. R., Anal. Chem., 1986, 58, 1956. 21 Vaughan, G. T., and Florence, T. M., Sci. Total Environ., 1992, 111, 47. 22 McGahan, M. Ch., and Tyczkowska, K., Spectrochim. Acta, Part B, 1987, 42, 665. 23 McLoughlin, S., Bowdler, D., and Roberts, N. B., J. Anal. At. Spectrom., 1988, 3, 273. 24 Shearan, P., and Smyth, M.R., Analyst, 1988, 113, 609. 25 Tothill, P., Matheson, L. M., Smyth, J. F., and McKay, K., J. Anal. At. Spectrom., 1990, 5, 619. 26 Haeberlin, S. T., Lux, F., Karl, J., Spruss, T., and Sch�onenberger, H., J. Radioanal. Nucl. Chem., 1987, 113, 461. 27 Esposito, M., Collecchi, P., Oddone, M., and Meloni, S., J. Radioanal. Nucl. Chem., 1987, 113, 437. 28 Ivanov, A. V., Zheligovskaya, N. N., and Nesterenko, P. N., Zh. Anal. Khim., 1995, 50, 190. 29 Cairns, W. R. L., Ebdon, L., and Hill, S.J., Fresenius’ J. Anal. Chem., 1996, 355, 202. 30 Ehrsson, H. C., Wallin, I. B., Andersson, A. S., and Edlund, P. O., Anal. Chem., 1995, 67, 3608. 31 Begerow, J., and Dunemann, L., J. Anal. At. Spectrom., 1996, 11, 303. 32 Taskaev, E., Karaivanova, M., and Grigorov, T., J. Radioanal. Nucl. Chem., 1988, 120, 75. 33 Alimonti, A., Dominici, C., Petrucci, F., La Torre, F., and Caroli, S., Anal. Chim. Hung., 1991, 128, 527. 34 Barefoot, R. R., and Van Loon, J.C., Anal. Chim. Acta, 1996, 334, 5. 35 Mukai, H., Ambe, Y., and Morita, M., J. Anal. At. Spectrom., 1990, 5, 75. 36 Alt, F., Bambauer, A., Hoppstock, K., Mergler, B., and T�olg, G., Fresenius’ J. Anal. Chem., 1993, 346, 693. 37 Hoppstock, K., Alt, F., Cammann, K., and Weber, G., Fresenius’ Z. Anal. Chem., 1989, 335, 813. 38 Wildhagen, D., and Krivan, V., Anal. Chim. Acta, 1993, 274, 257. 39 Beinrohr, E., Lee, M. L., Tsch�opel, P., and T�olg, G., Fresenius’ J. Anal. Chem., 1993, 346, 689. 40 Schramel, P., Wendler, I., and Lustig, S., Fresenius’ J. Anal. Chem., 1995, 353, 115. 41 Heinrich, E., Schmidt, G., and Kratz, K. L., Fresenius’ J. Anal. Chem., 1996, 354, 883. 42 Farago, M. E., Kavanagh, P., Blanks, R., Kelly, J., Kazantzis, G., Thornton, I., Simpson, P. R., Cook, J. M., Parry, S., and Hall, G. M., Fresenius’ J. Anal. Chem., 1996, 354, 660. 43 Parent, M., Vanhoe, H., Moens, L., and Dams, R., Fresenius’ J. Anal. Chem., 1996, 354, 664. 44 Wei, Ch., and Morrison, G. M., Anal. Chim. Acta, 1994, 284, 587. 45 Hodge, V., Stallard, M., Koide, M., and Goldberg, E. D., Anal. Chem., 1986, 58, 616. 46 Hall, G. E. M., and Pelchat, J. C., J. Anal. At. Spectrom., 1993, 8, 1059. 47 Colodner, D. C., Boyle, E. A., and Edmond, J. M., Anal. Chem., 1993, 65, 1419. 48 Alt, F., Jerono, U., Messerschmidt, J., and T�olg, G., Mikrochim. Acta, 1988, III, 299. 49 Otruba, V., Strnadov�a, M., and Skaln�ýkov�a, B., Talanta, 1993, 40, 221. 50 Messerschmidt, J., Alt, F., and T�olg, G., Anal. Chim. Acta, 1994, 291, 161. 51 Vanhaecke, F., Verstraete, D., Moens, L., and Dams, R., paper presented at the Winter Conference on Plasma Spectrochemistry, Florida, 1996, FP54. 52 Jerono, U., Alt, F., Messerschmidt, J., and T�olg, G., Mikrochim. Acta, 1992, 108, 221. 53 Valente, I. M., Minski, M. J., and Peterson, P. J., J. Radioanal. Chem., 1982, 71, 115. 54 Lee, M. L., T�olg, G., Beinrohr, E., and Tsch�opel, P., Anal. Chim. Acta, 1993, 272, 193. 55 Hodge, V. F., and Stallard, M. O., Environ. Sci. Technol., 1986, 20, 1058. 56 Perry, B. J., Barefoot, R. R., and Van Loon, J. C., Trends Anal. Chem., 1995, 14, 388. Analyst, May 1997, Vol. 122 73R57 Olivares, J. A., and Houk, R. S., Anal. Chem., 1986, 58, 20. 58 Dams, R. F. J., Goossens, J., and Moens, L., Mikrochim. Acta, 1995, 119, 277. 59 Al-Bazi, S. J., and Chow, A., Talanta, 1984, 31, 815. 60 Alimarin, I. P., Ivanov, V. M., Bol’shova, T. A., and Basova, E. M., Fresenius’ J. Anal. Chem., 1989, 335, 63. 61 Alimarin, I. P., Basova, E. M., Bol’shova, T. A., and Ivanov, V. M., Zh. Anal. Khim., 1986, 41, 5. 62 Rocklin, R. D., Anal. Chem., 1984, 56, 1959. 63 Brooks, R. R., and Ahrens, L. H., Spectrochim. Acta, 1960, 16, 783. 64 Basova, E. M., Bol’shova, T. A., Ivanov, V. M., and Morozova, N. B., Zh. Anal. Khim., 1989, 44, 680. 65 Liu, W. P., and Liu, Q. P., Fresenius’ J. Anal. Chem., 1994, , and Das, H. R., Talanta, 1985, 32, 457. 67 Malykhin, A. Yu., Bol’shova, T. A., Ermolenko, I. N., Bobrovich, I. B., Nikishkin, I. A., and Polyakova, L. A., Zh. Anal. Khim., 1989, 44, 886. 68 Balcerzak, M., Analysis, 1994, 22, 353. 69 Eller, R., Alt, F., T�olg, G., and Tobschall, H. J., Fresenius’ Z. Anal. Chem., 1989, 334, 723. 70 Wood, S. A., Vlassopoulos, D., and Mucci, A., Anal. Chim. Acta, 1990, 229, 227. 71 Fassett, J. D., and Paulsen, P. J., Anal. Chem., 1989, 61, 643A. 72 McLaren, J. W., Beauchemin, D., and Berman, S. S., Anal. Chem., 1987, 59, 610. 73 Enzweiler, J., Potts, P. J., and Jarvis, K. E., Analyst, 1995, 120, 1391. 74 Yi, Y.V., and Masuda, A., Anal. Chem., 1996, 68, 1444. 75 Parent, M., Vanhoe, H., Moens, L., and Dams, R., Anal. Chim. Acta, 1996, 320, 1. 76 Nadkarni, R. A., and Morrison, G. H., J. Radioanal. Chem., 1977, 38, 435. 77 Zeisler, R., and Greenberg, R. R., J. Radioanal. Chem., 1982, 75, 27. 78 Sykes, T. R., Stephens-Newsham, L. G., Apps, M. J., and Noujaim, A. A., J. Radioanal. Chem., 1982, 69, 441. 79 Rietz, B., and Heydorn, K., J. Radioanal. Nucl. Chem., 1993, 174, 49. 80 Cornelis, R., Borguet, F., and De Kompe, J., Anal. Chim. Acta, 1993, 283, 183. 81 Bettmer, J., Buscher, W., and Cammann, K., Fresenius’ J. Anal. Chem., 1996, 354, 521. Paper 6/08153C Received December 3, 1996 Accepted February 10, 1997 74R Analyst, May 1997, Vol. 1

ISSN:0003-2654    

DOI:10.1039/a608153c

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Sampling Intercomparisons for Aldehydes in Simulated Workplace Air E. Goelen, M. Lambrechts and F. Geyskens VITO, Environment Division, Boeretang 200, 2400 Mol, Belgium Thirty one laboratories of various EU Member States have participated in two interlaboratory comparisons in order to assess errors of personal sampling methods associated with both the sampling and the analytical steps. In contrast to conventional quality control schemes, this project particularly focuses attention on the sampling and identification step; it is executed by means of sampling exercises and has included discussions on potential sources of error.In a sampling exercise, participants come to a central facility and perform measurements on synthetic workplace air in a laboratory installation. Concentration levels of formaldehyde, acrolein, glutaraldehyde and acetaldehyde between 0.1 and 2 times the limit value for workplace air were prepared at various humidity levels and with acetone, occasionally, as interferent.Sampling times varied from 1–4 h. The related analytical work is performed at the analyst’s own laboratory. The intention is for each participant to determine the observed value of the delivered standard atmosphere using the sampling method of his own choice. Trueness (bias), precision and relative overall uncertainty of each method–laboratory combination is calculated and verified towards compliance with EN 482, which outlines minimum performance criteria.The first challenge involved the precise gas phase generation of the selected analytes in high air flows (up to 300 l min21) and calculating the true value only by direct reference to primary standards. This was accomplished by modifying the capillary dosage injection technique so that reactive compounds, like low molecular mass aldehydes, could be dosed with the same accuracy and precision as unreactive solvents. A permeation tube with high emission rate was developed for formaldehyde. Up to ten different sampling techniques were evaluated.The measurement methods used by the majority of the participants were based on pumped sampling on silica cartridges (or tubes) and glass fiber filters, coated with 2,4-dinitrophenylhydrazine. It was observed that for formaldehyde, and in some cases for acetaldehyde and glutaraldehyde, the majority of the method–laboratory combinations complied with an overall uncertainty of 30%. The results for acrolein, however, indicated a systematic negative bias, often larger than minus 50% of the true value, caused by the decomposition of the acrolein DNPH derivative in the presence of excess acid and excess DNPH.Keywords: Sampling intercomparisons; aldehydes; measurement method evaluation; bias and precision of methods; compliance with EN 482 minimum performance criteria From an occupational point of view, four aldehydes are of special interest, namely, formaldehyde, acrolein, glutaraldehyde and acetaldehyde.A large number of methods for their determination have been published. Literature references to the methods used are incorporated in Table 1. The most recent use reagent coated sorbents or filters in combination with pumped or diffusive sampling. A stable derivative is formed in situ on the adsorbent. The derivative is solvent desorbed and determined quantitatively by a sensitive technique such as GC or HPLC. Pre-coated samplers for pumped and diffusive sampling are commercially available and widely used.Minimum performance criteria for workplace air methods were published by the European Standardization Committee (CEN) in EN 482 (July 1994).14 In order to evaluate and compare the performance of currently used methods and to gather data on their overall uncertainty, the Standards, Measurements and Testing programme of the European Commission funded a quality assurance project that included both the Table 1 Summary and references of aldehyde sampling methods Methods and Ex. 1 Ex. 2 references 24 labs 26 labs Remark (ex. 2) Al, Solid A1-SG-X*1–4 14 20 home-made: 3; Waters: 13 sorbent SKC tube; Supelco A1-XAD-X5 1 2 A1-XAD-Y6,7 1 4 home-made: 2 A2, Filter A2-GFF-X†8 9 5 25 mm: 2; 13 mm: 3 A2-PAP-X 2 37 mm: 1 A3, Impinger A3-X3,9 4 3 Diffusive D1-GFF/PAP-X10 2 5 GMD badge samplers D2-GFF-Z11 1 1 3M badge D3-GFF-X12 1 1 37 mm GFF, badge D4-FL-X13 — 1 Radiello * Various home made types or purchased from (three different) commercial suppliers.† 13 mm, 25 mm or 37 mm filter holders. A, active sampling; A1, coated solid sorbent; A2, impregnated filter; A3, impinger; X, 2,4-dinitrophenylhydrazine; Y, 2-(hydroxymethyl)piperidine; Z, bisulfite; SG, on silica gel; XAD, on Amberlite XAD-2; GFF, on glass-fibre filter; PAP, on cellulose filter; FL, on florisil; D, diffusive sampling; D1, diffusive sampler with a poly(propylene) housing of 60 3 30 3 5 mm, which includes an impregnated filter of 20 3 45 mm; D2, 3M Type 3750 formaldehyde monitor; D3, standard 37 mm filter holder including an impregnated glass-fiber filter; D4, radial diffusive sampler with an adsorbing cartridge of 60 mm length and 5.8 mm od; 100-mesh hole size and 0.1 mm wire diameter, sealed at the end by two net cups.Analyst, May 1997, Vol. 122 (411–419) 411sampling and analytical step with emphasis on the sampling step. In a sampling exercise, participants come to a central facility and perform measurements on synthetic workplace air in a laboratory installation simulating, as closely as possible, daily occupational hygiene situations.The related analytical work is done by each participant in his own laboratory. The intention is to determine the observed value of the delivered standard atmosphere. State-of-the-art bias, within- and between-laboratory relative standard deviations as well as overall uncertainties for method–laboratory combinations are calculated by reference to the theoretical value.The output and conclusions of two aldehyde sampling exercises, both held in 1995, are summarized here. Participation has been on a voluntary basis and limited to 30 laboratories per intercomparison. The sampling exercises were organised at the coordinators’ institute (VITO, Mol) where a purpose-built facility is available. A technical discussion succeeded each sampling intercomparison. Experimental Design Essential parts of the dedicated facility at the coordinators’ institute consisted of dosing devices for the dynamic gas-phase generation of aldehydes and the sampling manifold (length 46 m).Very precisely controlled standard atmospheres are distributed through a glass manifold of 40 mm glass tubing, downstream of which there are 33 sampling points for pumped sampling and three exposure chambers for diffusive sampling.15 The design is such that up to thirty participants can sample simultaneously from the same aldehyde calibration mixture applying pumped and/or diffusive sampling.Carrier gas flows amount up to 300 l min21. Concentration levels are between 0.1 and 2 times the limit value and are accomplished using home made capillary dosage devices. These devices were originally developed for unreactive solvents16 but modified successfully for low molecular mass aldehydes17 which resulted in an original concept not previously described in the literature.18 Formaldehyde is generated by decomposition of paraformaldehyde in a permeation tube at 90 °C, resulting in a permeation rate of up to 150 mg min21.The dosing devices for acrolein, acetaldehyde and glutaraldehyde basically all consist of a stainless-steel container, a reservoir for the substance under investigation, a glass capillary and an evaporation section which includes flash heaters and a nebuliser. A pressure difference across the capillary forces the liquid into the evaporation section and a nitrogen carrier gas stream of up to 10 l min21 allows its injection as vapour into the sampling manifold.Mass measurements of the injected amount in combination with the total carrier gas flow rate result in a theoretical concentration value, with which participants’ results are evaluated and compared. Both the mass loss of analyte and the volumetric flow of air were determined using techniques traceable to gravimetric standards. Homogeneity and stability measurements during intercomparisons were performed for the various analytes as follows: (a) formaldehyde: monitor with photo-acoustic detector for stability and an automated colorimetric analyser (interference free) for homogeneity along the manifold; (b) acrolein and glutaraldehyde: on-line GC with photo-ionisation detector (10.2 eV); and (c) acetaldehyde and acetone: on-line GC with flame ionisation detector.All homogeneity and stability measurements have shown a relative standard deviation of between 0.5 and 3% for periods of up to 8 h. Sampling Intercomparisons: Set-up, Participants and Methods A sampling exercise lasts 2–3 d and is typically composed of a number of runs (up to 9); a run is a period of time (between 2 and 4 h) in which the concentration level of the selected analytes remains constant.Variables in the different runs also included (besides the number of analytes and the concentration levels) the duration of each run, the water vapour content of the carrier gas [typically 20–80% relative humidity (RH) under laboratory conditions] and an analytical interfering compound, such as acetone.Details of both sampling exercises are presented as the horizontal axes of Figs. 1 and 2. The decision regarding the priority aldehydes was taken in a preliminary workshop19 and at the same time a reference material certification project20 was initiated for the same aldehydes. After each sampling exercise, the participants’ observed values and the applied method(s) are centralised at the coordinators’ institute.These data are presented in anonymous form and distributed as tables and figures showing bias versus theoretical value and within- and between-laboratory relative standard deviations for the different method–laboratory combinations. A technical discussion meeting after each exercise has Fig. 1 Percentage of results within 30% relative overall uncertainty (bars), a within-laboratory RSD of < 5% (2) and a bias of > 25% («).All method–laboratory combinations of aldehyde sampling exercise 1. Fig. 2 As Fig. 1 but aldehyde sampling exercise 2. 412 Analyst, May 1997, Vol. 122been the forum to discuss errors associated to either methods or laboratories. Thirty one laboratories from 11 Member States have participated in the first (22–24 March 1995; code 1) and/or in the second (10–11 October 1995; code 2) intercomparison. The list is given in the Appendix.Each laboratory received information regarding the analytes involved, the composition of each run and an indication of the order of magnitude of the concentration level(s) prior to the sampling exercise. There was no limitation on the type of sampler, the number of samples (minimum of six, simultaneously or successively) per run nor the type of sampling technique (pumped and/or diffusive). The way of sampling and the sorbent-coating combination used is summarised in Table 1.Some trends (based on intercomparison 2) are observed from Table 1. Pumped sampling (82%) is used mainly with preference for DNPH coated on silica (56% of pumped samplers). The D1 type of diffusive sampler is by far the most popular (63%) among the diffusive samplers. Analyses are performed using HPLC (83%) and GC (13%), with spectrophotometry being performed by one participant. Results and Discussion Overall Performance The ability of a representative number of European laboratories to measure these aldehydes is illustrated in Figs. 1–4.A potential influence of different sampling situations on the overall uncertainty and performance is assessed and illustrated for various runs. Figs. 1 and 2 show the performance of laboratories per compound, per run and per sampling exercise in relation to: (a) the relative overall uncertainty (ROU) as a percentage of results within 30% relative overall uncertainty; (b) within-laboratory relative standard deviation smaller than 5%; and (c) the percentage of results with a bias above 25%.The horizontal axis contains, grouped per run, the accepted theoretical value of the analytes, the run number and run time, the humidity level and the presence of acetone (as an interfering compound). The vertical line (percentage of results) indicates the percentage of mean values for all method–laboratories fulfilling the selected criteria, the mean value of a laboratory for a compound in a run is the result of (a minimum of) six replicates, sampled simultaneously or successively.Fig. 3 presents potential differences between the most frequently used pumped sampling techniques while Fig. 4 shows differences between derivatives (DNPH versus HMP) as well as selected pumped versus selected diffusive sampling techniques. Horizontal and vertical axes have the same format as Figs. 1 and 2. Conclusions regarding the state-of-the-art reveal that performance differences are mainly influenced by the type of analyte (aldehyde) determined.Of all factors that might presumably affect bias and precision, it is indeed a fact that for formaldehyde an average of 67% method–laboratory combinations comply with the minimum performance requirement of 30% overall uncertainty, while for acrolein and glutaraldehyde the figures are 29 and 39%, respectively. Differences due to a variation in concentration levels, sampling time or humidity level are less pronounced than the analyte involved.Separation problems in liquid chromatography seem to have a negligible effect since sampled analytes on their own or as a mixture do not alter the ROU significantly. Only breakthrough due to high acetaldehyde concentration levels (e.g., run 2, exercise 2) has lowered the accuracy of most results substantially. The high number of outliers ( > 25%) for acrolein and glutaraldehyde may even question the accuracy of the method as it is currently used. In particular for glutaraldehyde, the phenomenon may also be caused by a lack of experience of some of the participating laboratories.Any substantial differences between various types of pumped sampling were assessed. Fig. 3 differentiates between sampling techniques for the DNPH method. Except for acrolein in runs 1 and 8, no significant differences between solid sorbent, filter and impinger methods with respect to uncertainty were found. Pumped sampling using a DNPH coating versus HMP and diffusive sampling leads to the observation that the overall uncertainty of diffusive samplers is even slightly lower than that obtained with pumped sampling.This is most likely due to the greater experience of aldehyde measurements possessed by the laboratories applying diffusive sampling. Diffusive sampling was not used frequently and only a few laboratories determined other aldehydes besides formaldehyde. HMP coating for acrolein has lead to systematically better results than the DNPH coating.Compliance of Aldehyde Personal Sampling Methods to EN 482 Compliance with minimum overall uncertainty requirements are detailed per analyte without differentiation between Fig. 3 Percentage of method–laboratory combinations per sampling technique with a relative overall uncertainty smaller than 30%. Aldehyde sampling exercise 1. Analyst, May 1997, Vol. 122 413 Formaldehyde Being the most prevalent aldehyde, it was of upmost importance to explain detected bias and attempt to improve the method(s) for detection.Figs. 1–5 show the current state-of-the-art. Typically, between 60 and 80% of the method–laboratory combinations had an ROU < 30%; between 25 and 50% had a bias within 10% of the true value. These data are worse than results in similar exercises for aromatic and chlorinated hydrocarbons, but acceptable and adequately in accordance with minimum performance requirements. Table 2 illustrates, for the selected performance indicators, the significant improvements through participation in successive sampling intercomparisons.The percentage of results for all method– laboratory combinations regarding overall uncertainty, bias within 10% and within-laboratory relative standard deviation improved by 9, 2 and 11%, respectively. The bias and withinlaboratory standard deviation for all individual method– laboratory combinations is detailed in Figs. 6 and 8 for two selected runs containing formaldehyde.Fourteen laboratories using the A1-SG-X method in exercise 1 often underestimated the accepted true value (Fig. 8). Inaccurate sampling volume was mentioned and found to be a potential cause. More consensus was obtained following the hypothesis that the recovery of formaldehyde as 2,4-dinitrophenylhydrazone was not 100%, as the majority of the participants had assumed. Fourteen laboratories later determined the recovery and found a value between 89 and 95% which would explain the underestimation shown in Fig. 8. These data are confirmed in the literature.21 Laboratory 49 stated that higher results (at least 15% difference) were obtained when 10 ml instead of 3 ml of desorption liquid were used. A change of solvent–sorbent ratio can indeed affect desorption efficiency. This phenomenon is known for charcoal, silica gel and porous polymers, but has not been described for reagent coated sorbents and their derivatives; 22 it nevertheless improved the results of laboratory 49. The desorption efficiency for formaldehyde was also found to be related to the loading on the cartridge (laboratory 16).Results obtained through the glass fiber filter method (A2- GFF-X; laboratories 16, 41, 42, 43, 47, 38) were quite scattered around the reference value. A critical factor here is breakthrough which is caused by either insufficient DNPH coating, mechanical damage of filters, drying out effects or the presence of interfering compounds ( > 1 l in the presence of acetone).Breakthrough is also found by the A1-SG-X method in the presence of acetone and above 1 l of sampling volume; up to 2 l in the presence of acetaldehyde is no problem. A 2 mg amount of coating is sufficient for 13 mm filters, but it should be increased for 25 and 37 mm filters. Drying out effects can be avoided by adding glycerine to the coating solution. The impinger method (laboratories 25, 27, 46 and 48) was very reliable.Formaldehyde collection in this way and direct injection of the sample into the GC or HPLC seems to avoid a number of steps which frequently introduce errors. The low results of laboratory 48 are explained by the sampling volume, which amounts to up to twice the value of other laboratories applying the same method. Disadvantages of the impinger collector, e.g., it is cumbersome to use and unsuitable for extended collection times due to solvent evaporation, seem to be particularly compensated by its accuracy.Reasons for bias applicable to the formaldehyde measurement methods were: (a) high concentrations of interfering compounds like acetaldehyde Table 2 Summary of formaldehyde measurement method improvements for selected performance indicators ROU Bias within Within-lab % of results* < 30% 10% RSD < 5% Exercise 1 64 40 49 Exercise 2 73 42 60 * Mean value of seven runs for exercise 1 and four runs for exercise 2.Fig. 6 As Fig. 8 except formaldehyde 0.308 ± 0.009 ppm (v/v). Run 2. Exercise 2. Fig. 7 Bias and within-laboratory RSD for all method–laboratory combinations. Acrolein 0.127 ± 0.004 ppm (v/v). Run 6. Exercise 1. Fig. 8 Bias and within-laboratory RSD for all method–laboratory combinations for selected formaldehyde runs. Formaldehyde 1.03 ± 0.03 ppm (v/v). Run 1. Exercise 1. Analyst, May 1997, Vol. 122 415and acetone cause a negative bias; it is assumed that this impedes the complete fixation of other compounds; (b) care should be taken with commercial formaldehyde hydrazone standards since laboratory 7 reported a 12% discrepancy with the BCR formaldehyde hydrazone standard (CRM 546);23 (c) no relationship was found between the overall uncertainty of formaldehyde measurements and the relative humidity level, which is confirmed in literature;24 (d) no relationship could be demonstrated between the bias of laboratories applying the DNPH method and the storage time; storage at 218 °C (even up to 8 weeks) did not result in a negative bias.Background levels of hydrazone derivatives on cartridges increase slightly with time and temperature. Therefore in exercise 2 an increased number of laboratories have stored at 218 °C and only for one to three weeks. Taking these potential sources of error into consideration resulted in an improved data set illustrated in Fig. 6. Fig. 9 illustrates that throughout the various runs, the between-laboratory relative standard deviation for the various methods was quite consistent and is between 20 and 60%.This is significantly higher than those values obtained for aromatic and chlorinated hydrocarbons (typically between 5 and 20%).28 An RSD between 10 and 30% was obtained in exercise II. This improvement is even visible in Fig. 6, where the results are less scattered; for the A1-SG-X method the results are also better centered around the reference value.Acrolein Sampling and analysis of acrolein is not yet common practice. All results for acrolein obtained through methods based on DNPH showed a significant negative bias; participants typically reported a value between 20 and 50% of the reference concentration (Fig. 7). This leads to an extremely low number of laboratories that comply with the EN 482 minimum performance criteria. Neither humidity, concentration levels, sampling time, interfering compounds nor other variables in sampling or analysis played an important role.The phenomenon was fully explained by the decomposition of the acrolein hydrazone on the samplers in the presence of excess of both DNPH and acid. Participants reported besides the main acrolein peak in the chromatogram (at the retention time of the acrolein hydrazone in desorption solution) a number of other peaks only appearing when acrolein was present in the atmosphere to be sampled (Fig. 10). The peak areas obtained vary in an unpredictable manner and both the number as well as peak heights were different between various method–laboratory combinations.The main peak is found to decrease with time, the additional peaks (between 3 and 6) increase rapidly as the sample deteriorates (Fig. 11). The decrease on a cartridge between 30 min and 24 h after sampling is shown in Fig. 11. Peaks 1, 2 and 3 originate from the decomposition of the acrolein hydrazone. This phenomenon leads to underestimation of the reference value since the main peak no longer represents the complete acrolein hydrazone concentration.The original acrolein hydrazone is partly rearranged on the sampler into a number of decomposition products. One of these products could be identified by laboratory 47 and consisted of three DNPH and two acrolein molecules. Once the sample is desorbed from the cartridge, tube or filter, no further degradation takes place, presumably because the acid is not present in excess in the solution.The decomposition products interfere and may overlap with other carbonyl compound peaks in the chromatogram which may result in false overestimation of these compounds (e.g., acetaldehyde, acetone, glutaraldehyde). Some of the observations described above are mentioned in a few publications. The decomposition of the acrolein hydrazone at room temperature is such that only 30% is left after one week. This was one of the explanations for incomplete recovery during acrolein sampling Fig. 9 Comparison of between-laboratory RSD (%) for selected sampling techniques and runs in exercise 1. Fig. 10 HPLC chromatogram with acrolein hydrazone decomposition products eluting between 3.50 and 5.30 min. 1, DNPH; 2, formaldehyde hydrazone; 3–7, origin from acrolein hydrazone. HPLC: UV at 375 mm, C18 column (35°C); AcN–H2O (75 + 25) at 1 ml min21; acrolein DNPH standard solution at 3.88 min. Fig. 11 Decomposition (degradation) of an acrolein standard solution (0.5 ml of 55 mg ml21) on cartridges coated with DNPH. 416 Analyst, May 1997, Vol. 122of automobile exhaust; storage at 0 °C would retard decomposition significantly.21,25 This last observation was not confirmed in the sampling project. In another study, decomposition was considered to be caused by ozone and an ozone scrubber was recommended before a cartridge to avoid decomposition of hydrazone derivatives.26,27 Due to these fundamental problems associated with the DNPH method, the overall uncertainty could only be improved by, for instance, taking the sum of all decomposition products into consideration to calculate the acrolein concentration or by passing immediately after sampling the solution through an ion exchange sorbent (e.g., Bond Elut SCX).Based on the results of this project, it can concluded that no currently available DNPH method has sufficient accuracy for acrolein measurements. The alternative method, which applies 2-(hydroxymethyl)- piperidine as derivative, was only applied by laboratories 16 and 34 but proved to form a stable acrolein derivative.The GC method applied by these participants showed very accurate results (Fig 7, laboratory 16). Glutaraldehyde The results obtained for glutaraldehyde were accurate for a number of laboratories but the picture for the total group and a majority of the participants was not encouraging. The obtained bias is explained by a number of parameters detailed below.As a whole, the results improved only slightly in the second sampling exercise. A number of parameters relevant for the performance of method–laboratory combinations are derived from Fig. 1 to 5 and are summarized in Table 3. Comparing the data for glutaraldehyde with the same set for toluene or other compounds where no sampling or significant analytical problems occur, illustrates that some aspects of this method are still subject to improvements. The values for toluene have been published previously.28 The bias and within-laboratory standard deviations for all method–laboratory combinations for selected runs with glutaraldehyde are illustrated in Fig. 12 (high humidity level) and Fig. 13 (low humidity level). In both (and all other) cases a quite scattered picture with a large overall between-laboratory relative standard deviation is obtained. Comparing both pictures illustrates a shift of the data from substantially above the reference value (Fig. 12) to 220–30% below (Fig. 13). This phenomenon is fully explained by the humidity level. Most laboratories obtain higher results when the humidity level of the reference gas mixture is high. This effect was confirmed when evaluating in detail the results of laboratory 23. These data were affected by the humidity level. Results close to the reference value were found in cases of a high humidity level (75%–80%) and 251% and 224% underestimation in case of 38 and 51% relative humidity, respectively. The detailed picture is shown in Fig. 14. One of the other factors that affected the performance of laboratories was the poor storage stability of the glutaraldehyde hydrazone on the sampling media. Storage at 220 °C for up to 4 weeks is no problem. However, if storage takes place in a refrigerator at 4–6 °C only 55–60% of the original amount is left on the sampler. Table 3 Comparison of glutaraldehyde measurement method performance indicators Within-lab Between- % of ROU Bias within RSD lab results* < 30% 30% < 5% RSD† Exercise 1 41 33 30 37 Exercise 2 43 24 44 39 * Mean value of seven runs for exercise I and three runs for exercise II.† Mean RSD of all active sampling methods, not percentage of results. Fig. 12 Bias and within-laboratory RSD for all method–laboratory combinations for selected glutaraldehyde runs. Glutaraldehyde 0.262 ± 0.008 ppm (v/v). Run 2. Exercise 1. Fig. 13 As Fig. 12 except glutaraldehyde 0.149 ± 0.004 ppm (v/v).Run 6. Exercise 1. Fig. 14 Correlation between the bias of glutaraldehyde measurements for one selected laboratory and the relative humidity level. 8, % relative humidity; X, % of reference concentration for the selected laboratory; O % of reference concentration for the mean of all laboratories. Analyst, May 1997, Vol. 122 417Although no separation problems occur in the various HPLC chromatograms between glutaraldehyde and the other compounds, some decomposition products of the acrolein hydrazone on the sampler can overlap with the glutaraldehyde hydrazone.In addition, the glutaraldehyde hydrazone appears as two isomers in the chromatogram, so the use of only one peak again leads to errors. All these factors that cause underestimation or overestimation of the reference value result in large variations in the percentage of method–laboratory combinations that fulfil the overall uncertainty criterion of 30% (Figs. 1 and 2). Typically only 20–50% of the method–laboratory combinations complied to EN 482.There appears to be no significant differences between methods. Respectively, 50–60% and 15–25% of the method– laboratory combinations use A1-SG-X and A2-GFF-X. Only one or two apply the methods A3-X, A1-XAD-X and D1-GFFX. Although the A1-SG-X method has systematically the lowest between laboratory RSD (Fig. 9), the smallest overall bias is obtained by laboratories using A2-GFF-X. The bias for the diffusive sampling methods was typically within 15%, but these laboratories obtained comparable results with their pumped method (A1-SG-X or A1-GFF-X).The results of laboratory 47 with the diffusive sampler even suggest that the sampling medium (glass fiber filter or paper) for the D1 method has affected the results. The glass fiber filter resulted in 10% higher results. These differences were not systematically detected for formaldehyde. Taking the above factors into account, the DNPH methods have the potential to produce accurate results for glutaraldehyde.This is proven, e.g., in run 5 of exercise 1, where the best results were obtained so that 52% of the method–laboratory combinations had a bias within 10% of the reference value; 25% had a within-laboratory relative standard deviation below 5%. Acetaldehyde and acetone Due to the high workplace air limit values for acetaldehyde (100 ppm) compared to the other aldehydes studied, this compound was only included in run 5 of exercise 1 and run 2 of exercise 2.The bias and within-laboratory relative standard deviation for all method–laboratory combinations is detailed in Fig. 15 for run 5 (1). Although a number of method–laboratory combinations succeeded in obtaining very accurate results (40% has a bias within 20%), a substantial number of outliers (60% with a bias > 25%) were found. The outlying results were caused by breakthrough on the samplers since the sampling volume was optimised for ppb levels of the other aldehydes.Besides breakthrough, care should be taken to avoid interference in the HPLC chromatogram of the large acetaldehyde peak with the small surrounding peaks especially of acrolein. The introduction of acetone in the programme was intended to detect possible interferences in the sampling and analysis of especially acrolein. It transpired that interference may indeed occur in the HPLC chromatogram but that acrolein measurements suffer especially from the instability of the acrolein hydrazone, a phenomenon which is detailed above.The results of run 7 (exercise 1) and run 5 (exercise 2) were not significantly affected by the presence of acetone. It is clear, however, that the presence of acetone may hamper the measurement of carbonyl compounds in air. Conclusion During the past decade, great progress has been made and at the same time a number of new methods have been developed and described in literature for the measurement of airborne aldehydes.The work presented here evaluated the most frequently used methods for determining aldehydes. A workplace air atmosphere containing known amounts of aldehydes was dynamically prepared and distributed through a manifold which allowed to evaluate bias and precision of the currently used methods. Up to 30 participants could sample on the spot at the coordinators’ laboratory from a gravimetrically determined reference concentration.The existing capillary dosage technique was therefore thoroughly reviewed so that gas phase generation of reactive compounds, like acrolein, glutaraldehyde and acetaldehyde, became feasible over a wide concentration range and in high air flows so that pumped and diffusive sampling techniques could be assessed at the same time. The results obtained by a selected number of European laboratories in two sampling exercises suggest that the current methods for formaldehyde comply in general and for an acceptable number of laboratories with EN 482 minimum performance criteria. The DNPH method produced poor results for acrolein which has led to the conclusion to abandon or significantly alter the method due the instability of the acrolein hydrazone on the sampler leading to decomposition of the derivative. Methods applied for glutaraldehyde and acetaldehyde seem principally and practically all right but their limited (although currently strongly increasing) use in daily occupational hygiene practice leads to errors which are usually anticipated by a group of more experienced users but are often not taken into account by the majority of the participants.Besides the overall results presented here and the discussion of potential sources of error, it is evident that this project also gained information on parameters like breakthrough volumes, uptake rates, analytical instrument optimisation, desorption efficiency tests, sampling volumes, active versus diffusive sampling within the same laboratory and other practical aspects which are not discussed fully here.These details are available in technical reports and can be obtained free through the various participants listed in the Appendix. This project was partly funded by the Standards, Measurements and Testing programme of the European Commission. The valuable advice of Dr. S. Vandendriessche in the design and management of the project is gratefully acknowledged.The author is indebted to all the participants in the project for their contribution and cooperation. The list of participants per country code is detailed below in the Appendix. Appendix B Mr. R. Grosjean and Mr. R. Mesmacque: Ministry of Labour; Dr. Roosels, Mrs. M. Noel, Mr. J.-M. Bosiroy, Mr. J. Boulanger: Funds for Occupational diseases; Mr. E. Fig. 15 Bias and within-laboratory RSD for all method–laboratory combinations. Acetaldehyde 31.3 ± 0.9 ppm v/v.Run. 5. Exercise 1. 418 Analyst, May 1997, Vol. 122Goelen (project coordinator), Mrs. F. Geyskens, Mr. M. Lambrechts, Mr. R. Bormans, Mr. T. De Ceuster, Mr. R. Mannaerts: VITO. CH Dr. C. K. Huynh, Mr. Vernay, Mr. Boiteux: Institut Universitaire Romande de Sant�e au Travail. G Dr. J. Auffarth, Dr. R. Hebisch, Mr. K. Rentel: Bundesanstalt fur Arbeitsschutz; Dr. W. Kramer, Mr. R. Schmitt: BASF; Dr. B. Strieffler, Mr. K. Schneider: Niedersachsisches Landesamt fur Okologie; Dr.J. Schnelle: GSF - Forschungszentrum fur Umwelt und Gesundheit; Dr. H. Fricke: Institut fur Gefahrstoff-Forschung; Dr. M. Weigl, Mr. Fehlauer: Berufsgenossenschaft Nahrungsmittel und Gaststatten Potsdam; Dr. B. Andrejs, Mrs. Kiefer, Mrs. C. Schuh: Berufsgenossenschaft Nahrungsmittel und Gaststatten Mannheim; Dr. Wensing, Dr. Schwarzer, Mr. M. Berbig, Mr. C. Kahre: TUV Nord; Dr. T. Wiesmuller, Mr. J. Schymonski, Mr. P. Zellner: Ecoplan; Mr. M. Bruckschlegel: IKPO Universitat Stuttgart; Dr. D.Ullrich, Mrs. C. Scheller: BGA, Institut fur Wasser-, Boden-, Lufthygiene. DK Dr. K. Egmose, Mrs. B. Grau-Hansen: Miljoe-Kemi. ES Dr. X. Guardino, Mrs. C. Santolaya, Mr. A. Marti: INSHT Barcelona; Mr. J. L. Perez Alvarez: Grupo Interlab; Mr. J. Comino, Mr. R. Manuel Ruiz, Mr. E. Arnaiz: Incohinsa. F Mr. R. Dujardien, Mr. H. Adrien, Mr. B. Brouart: INERIS. I Dr. V. Cocheo, Dr. F. Quaglio: Fondazione Savatore Maugeri. NL Mr. H. P. Bos, Mrs. S. Linders: RIVM; Mr. R.Peeters, Mr. M. Houtzager: TNO. SF Mrs. M.-L. Henricks-Eckermann: Turku Regional Institute of Occupational Health; Dr. E. Priha, Mrs. A. Jalkanen: Tampere Regional Institute of Occupational Health; Mrs. T. Tirkkonen, Mr. P. Tapanimaki, Mrs. R. Ketola: VTT - Chemical Technology. S Ass. Prof. J. Levin, Mr. R. Lindahl: National Institute for Working Life; Dr. L. Johnson, Mr. L. Rosell, Mrs. I. Isaksson: National Testing and Research Institute: Mr. R. Nordlinder, Mr.G. Ljungkvist: Sahlgrenska University Hospital. UK Mr. J. Cuthbert: Health and Safety Executive; Mr. G. Bebbington: Severn Trent Laboratories. B, Belgium; CH, Switzerland, G, Germany, DK, Denmark; ES, Spain; F, France; I, Italy, NL, Netherlands; SF, Finland; S, Sweden; UK, United Kingdom. References 1 Selim, S., J. Chromatogr., 1977, 136, 271. 2 Beasley, R. K., Hoffman, C. E., Rueppel, M. L., and Worley, J. W., Anal. Chem., 1980, 52, 1110. 3 Grosjean, D., and Kok, G. L., Anal.Chem., 1982, 54, 1221. 4 Tejada, S. B., Int. J. Environ. Anal. Chem., 1986, 26, 167. 5 Andersson, K., Andersson, G., Nilsson, C., and Levin, J. O., Chemosphere, 1979, 8, 812. 6 Kennedy, E. R., and Hill, R. H., Anal. Chem., 1982, 54, 1739. 7 Kennedy, E. R., O’Connor, P. F., and Gagnon, Y. T., Anal. Chem., 1984, 56, 2120. 8 Levin, J. O. Andersson, K., Lindahl, R., and Nilsson, C. A., Anal. Chem., 1985, 57, 1032. 9 Kuwata, K., Uebori, M., and Yamasaki, Y., J. Chromatogr., 1979, 17, 264. 10 Levin, J. O. Lindahl, R., and Andersson, K., Environ. Sci. Technol., 1986, 20, 1272. 11 Determination of Formaldehyde Vapours in Air, 3M Occupational Health and Environmental Safety Division, St. Paul, MN, USA. 12 Pfaffli, P., Virtanen, H., Riutta, O., and Hayri, L., in Clean Air at Work: New Trends in Assessment and Measurement for the 1990s, ed. Brown, R. H., Curtis, M. Saunders, K. J., and Vandendriessche, S., Royal Society of Chemistry, 1993, pp. 198–200. 13 Cocheo, V., Boaretto, C., and Sacco, P., Am.Ind. Hyg. Assoc. J., 1996, 57, 897. 14 European Committee for Standardization, Assessment of Workplace Exposure—General Requirements for the Performance of Procedures for the Measurement of Chemical Agents, CEN/TC137/ European Standard EN 482, Brussels, July 1994. 15 Goelen, E., Lambrechts, M., Geyskens, F., and De Fr�e, R., paper presented at the 2nd European Workshop on Mass Spectrometry in Occupational Health, Les Diablerets, Switzerland, June 16–18, 1993. 16 Goelen, E., Geyshens, F., Lambrechts, M., and Rymen, T., Int.J. Environ. Anal. Chem., 1992, 47, 217. 17 Goelen, E., Geyskens, F., and Lambrechts, M., Facility for the Generation and Distribution of Aldehyde Atmospheres in Air at Occupational Hygiene Levels, internal VITO report (MIE-DI-95-16), 1995. 18 Nelson, G. O., Gas Mixtures—Preparation and Control, Lewis Publishers, Chelsea, 1992. 19 Workshop on Possible New Projects in the Field of Workplace Air Monitoring, Brussels, March 25–27, 1991. 20 Levin, J. O., Lindahl, R., and Heeremans, C.E.M., EUR report, Brussels, 1996. 21 Lipari, F., and Swarin, S. J., J. Chromatogr., 1982, 247, 297. 22 Dommer, R. A., and Melcher, R. G., Am. Ind. Hyg. Assoc. J., 1978, 39, 240. 23 European Commission, BCR reference materials, Institute for Reference Materials and Measurements (Joint Research Centre), Geel, 1996. 24 Kuwata, K., Uebori, M., Yamasaki, H., and Kuge, T., Anal. Chem., 1983, 55, 2013. 25 Vainiotalo, S., Matveinen, K., in Clean Air at Work: New Trends in Assessment and Measurement for the 1990s, ed. Brown, R. H., Curtis, M., Saunders, K. J., and Vandendriessche, S., Royal Society of Chemistry, Cambridge, 1993, pp. 204–205. 26 Arnts, R. R., and Tejada, S. B., Environ. Sci. Technol., 1989, 23, 1428. 27 Barone, J. P., Walter, T. H., in Clean Air at Work: New Trends in Assessment and Measurement for the 1990s, ed. Brown, R. H., Curtis, M., Saunders, K. J., and Vandendriessche, S., Royal Society of Chemistry, Cambridge, 1993, pp. 162–168. 28 Goelen, E., Geyskens, F., and Lambrechts, M., Ann. Occup. Hyg., 1997, in the press. Paper 6/07047G Received October 16, 1996 Accepted January 13, 1997 Analyst, May 1997, Vol. 122 419 Sampling Intercomparisons for Aldehydes in Simulated Workplace Air E. Goelen, M. Lambrechts and F. Geyskens VITO, Environment Division, Boeretang 200, 2400 Mol, Belgium Thirty one laboratories of various EU Member States have participated in two interlaboratory comparisons in order to assess errors of personal sampling methods associated with both the sampling and the analytical steps.ontrast to conventional quality control schemes, this project particularly focuses attention on the sampling and identification step; it is executed by means of sampling exercises and has included discussions on potential sources of error. In a sampling exercise, participants come to a central facility and perform measurements on synthetic workplace air in a laboratory installation.Concentration levels of formaldehyde, acrolein, glutaraldehyde and acetaldehyde between 0.1 and 2 times the limit value for workplace air were prepared at various humidity levels and with acetone, occasionally, as interferent. Sampling times varied from 1–4 h. The related analytical work is performed at the analyst’s own laboratory. The intention is for each participant to determine the observed value of the delivered standard atmosphere using the sampling method of his own choice. Trueness (bias), precision and relative overall uncertainty of each method–laboratory combination is calculated and verified towards compliance with EN 482, which outlines minimum performance criteria.The first challenge involved the precise gas phase generation of the selected analytes in high air flows (up to 300 l min21) and calculating the true value only by direct reference to primary standards.This was accomplished by modifying the capillary dosage injection technique so that reactive compounds, like low molecular mass aldehydes, could be dosed with the same accuracy and precision as unreactive solvents. A permeation tube with high emission rate was developed for formaldehyde. Up to ten different sampling techniques were evaluated. The measurement methods used by the majority of the participants were based on pumped sampling on silica cartridges (or tubes) and glass fiber filters, coated with 2,4-dinitrophenylhydrazine. It was observed that for formaldehyde, and in some cases for acetaldehyde and glutaraldehyde, the majority of the method–laboratory combinations complied with an overall uncertainty of 30%.The results for acrolein, however, indicated a systematic negative bias, often larger than minus 50% of the true value, caused by the decomposition of the acrolein DNPH derivative in the presence of excess acid and excess DNPH.Keywords: Sampling intercomparisons; aldehydes; measurement method evaluation; bias and precision of methods; compliance with EN 482 minimum performance criteria From an occupational point of view, four aldehydes are of special interest, namely, formaldehyde, acrolein, glutaraldehyde and acetaldehyde. A large number of methods for their determination have been published. Literature references to the methods used are incorporated in Table 1. The most recent use reagent coated sorbents or filters in combination with pumped or diffusive sampling.A stable derivative is formed in situ on the adsorbent. The derivative is solvent desorbed and determined quantitatively by a sensitive technique such as GC or HPLC. Pre-coated samplers for pumped and diffusive sampling are commercially available and widely used. Minimum performance criteria for workplace air methods were published by the European Standardization Committee (CEN) in EN 482 (July 1994).14 In order to evaluate and compare the performance of currently used methods and to gather data on their overall uncertainty, the Standards, Measurements and Testing programme of the European Commission funded a quality assurance project that included both the Table 1 Summary and references of aldehyde sampling methods Methods and Ex. 1 Ex. 2 references 24 labs 26 labs Remark (ex. 2) Al, Solid A1-SG-X*1–4 14 20 home-made: 3; Waters: 13 sorbent SKC tube; Supelco A1-XAD-X5 1 2 A1-XAD-Y6,7 1 4 home-made: 2 A2, Filter A2-GFF-X†8 9 5 25 mm: 2; 13 mm: 3 A2-PAP-X 2 37 mm: 1 A3, Impinger A3-X3,9 4 3 Diffusive D1-GFF/PAP-X10 2 5 GMD badge samplers D2-GFF-Z11 1 1 3M badge D3-GFF-X12 1 1 37 mm GFF, badge D4-FL-X13 — 1 Radiello * Various home made types or purchased from (three different) commercial suppliers.† 13 mm, 25 mm or 37 mm filter holders. A, active sampling; A1, coated solid sorbent; A2, impregnated filter; A3, impinger; X, 2,4-dinitrophenylhydrazine; Y, 2-(hydroxymethyl)piperidine; Z, bisulfite; SG, on silica gel; XAD, on Amberlite XAD-2; GFF, on glass-fibre filter; PAP, on cellulose filter; FL, on florisil; D, diffusive sampling; D1, diffusive sampler with a poly(propylene) housing of 60 3 30 3 5 mm, which includes an impregnated filter of 20 3 45 mm; D2, 3M Type 3750 formaldehyde monitor; D3, standard 37 mm filter holder including an impregnated glass-fiber filter; D4, radial diffusive sampler with an adsorbing cartridge of 60 mm length and 5.8 mm od; 100-mesh hole size and 0.1 mm wire diameter, sealed at the end by two net cups.Analyst, May 1997, Vol. 122 (411–419) 411sampling and analytical step with emphasis on the sampling step. In a sampling exercise, participants come to a central facility and perform measurements on synthetic workplace air in a laboratory installation simulating, as closely as possible, daily occupational hygiene situations. The related analytical work is done by each participant in his own laboratory. The intention is to determine the observed value of the delivered standard atmosphere. State-of-the-art bias, within- and between-laboratory relative standard deviations as well as overall uncertainties for method–laboratory combinations are calculated by reference to the theoretical value.The output and conclusions of two aldehyde sampling exercises, both held in 1995, are summarized here. Participation has been on a voluntary basis and limited to 30 laboratories per intercomparison.The sampling exercises were organised at the coordinators’ institute (VITO, Mol) where a purpose-built facility is available. A technical discussion succeeded each sampling intercomparison. Experimental Design Essential parts of the dedicated facility at the coordinators’ institute consisted of dosing devices for the dynamic gas-phase generation of aldehydes and the sampling manifold (length 46 m). Very precisely controlled standard atmospheres are distributed through a glass manifold of 40 mm glass tubing, downstream of which there are 33 sampling points for pumped sampling and three exposure chambers for diffusive sampling.15 The design is such that up to thirty participants can sample simultaneously from the same aldehyde calibration mixture applying pumped and/or diffusive sampling.Carrier gas flows amount up to 300 l min21. Concentration levels are between 0.1 and 2 times the limit value and are accomplished using home made capillary dosage devices.These devices were originally developed for unreactive solvents16 but modified successfully for low molecular mass aldehydes17 which resulted in an original concept not previously described in the literature.18 Formaldehyde is generated by decomposition of paraformaldehyde in a permeation tube at 90 °C, resulting in a permeation rate of up to 150 mg min21. The dosing devices for acrolein, acetaldehyde and glutaraldehyde basically all consist of a stainless-steel container, a reservoir for the substance under investigation, a glass capillary and an evaporation section which includes flash heaters and a nebuliser.A pressure difference across the capillary forces the liquid into the evaporation section and a nitrogen carrier gas stream of up to 10 l min21 allows its injection as vapour into the sampling manifold. Mass measurements of the injected amount in combination with the total carrier gas flow rate result in a theoretical concentration value, with which participants’ results are evaluated and compared.Both the mass loss of analyte and the volumetric flow of air were determined using techniques traceable to gravimetric standards. Homogeneity and stability measurements during intercomparisons were performed for the various analytes as follows: (a) formaldehyde: monitor with photo-acoustic detector for stability and an automated colorimetric analyser (interference free) for homogeneity along the manifold; (b) acrolein and glutaraldehyde: on-line GC with photo-ionisation detector (10.2 eV); and (c) acetaldehyde and acetone: on-line GC with flame ionisation detector.All homogeneity and stability measurements have shown a relative standard deviation of between 0.5 and 3% for periods of up to 8 h. Sampling Intercomparisons: Set-up, Participants and Methods A sampling exercise lasts 2–3 d and is typically composed of a number of runs (up to 9); a run is a period of time (between 2 and 4 h) in which the concentration level of the selected analytes remains constant.Variables in the different runs also included (besides the number of analytes and the concentration levels) the duration of each run, the water vapour content of the carrier gas [typically 20–80% relative humidity (RH) under laboratory conditions] and an analytical interfering compound, such as acetone. Details of both sampling exercises are presented as the horizontal axes of Figs. 1 and 2. The decision regarding the priority aldehydes was taken in a preliminary workshop19 and at the same time a reference material certification project20 was initiated for the same aldehydes. After each sampling exercise, the participants’ observed values and the applied method(s) are centralised at the coordinators’ institute. These data are presented in anonymous form and distributed as tables and figures showing bias versus theoretical value and within- and between-laboratory relative standard deviations for the different method–laboratory combinations.A technical discussion meeting after each exercise has Fig. 1 Percentage of results within 30% relative overall uncertainty (bars), a within-laboratory RSD of < 5% (2) and a bias of > 25% («). All method–laboratory combinations of aldehyde sampling exercise 1. Fig. 2 As Fig. 1 but aldehyde sampling exercise 2. 412 Analyst, May 1997, Vol. 122been the forum to discuss errors associated to either methods or laboratories. Thirty one laboratories from 11 Member States have participated in the first (22–24 March 1995; code 1) and/or in the second (10–11 October 1995; code 2) intercomparison.The list is given in the Appendix. Each laboratory received information regarding the analytes involved, the composition of each run and an indication of the order of magnitude of the concentration level(s) prior to the sampling exercise. There was no limitation on the type of sampler, the number of samples (minimum of six, simultaneously or successively) per run nor the type of sampling technique (pumped and/or diffusive).The way of sampling and the sorbent-coating combination used is summarised in Table 1. Some trends (based on intercomparison 2) are observed from Table 1. Pumped sampling (82%) is used mainly with preference for DNPH coated on silica (56% of pumped samplers). The D1 type of diffusive sampler is by far the most popular (63%) among the diffusive samplers.Analyses are performed using HPLC (83%) and GC (13%), with spectrophotometry being performed by one participant. Results and Discussion Overall Performance The ability of a representative number of European laboratories to measure these aldehydes is illustrated in Figs. 1–4. A potential influence of different sampling situations on the overall uncertainty and performance is assessed and illustrated for various runs. Figs. 1 and 2 show the performance of laboratories per compound, per run and per sampling exercise in relation to: (a) the relative overall uncertainty (ROU) as a percentage of results within 30% relative overall uncertainty; (b) within-laboratory relative standard deviation smaller than 5%; and (c) the percentage of results with a bias above 25%. The horizontal axis contains, grouped per run, the accepted theoretical value of the analytes, the run number and run time, the humidity level and the presence of acetone (as an interfering compound).The vertical line (percentage of results) indicates the percentage of mean values for all method–laboratories fulfilling the selected criteria, the mean value of a laboratory for a compound in a run is the result of (a minimum of) six replicates, sampled simultaneously or successively. Fig. 3 presents potential differences between the most frequently used pumped sampling techniques while Fig. 4 shows differences between derivatives (DNPH versus HMP) as well as selected pumped versus selected diffusive sampling techniques.Horizontal and vertical axes have the same format as Figs. 1 and 2. Conclusions regarding the state-of-the-art reveal that performance differences are mainly influenced by the type of analyte (aldehyde) determined. Of all factors that might presumably affect bias and precision, it is indeed a fact that for formaldehyde an average of 67% method–laboratory combinations comply with the minimum performance requirement of 30% overall uncertainty, while for acrolein and glutaraldehyde the figures are 29 and 39%, respectively.Differences due to a variation in concentration levels, sampling time or humidity level are less pronounced than the analyte involved. Separation problems in liquid chromatography seem to have a negligible effect since sampled analytes on their own or as a mixture do not alter the ROU significantly. Only breakthrough due to high acetaldehyde concentration levels (e.g., run 2, exercise 2) has lowered the accuracy of most results substantially. The high number of outliers ( > 25%) for acrolein and glutaraldehyde may even question the accuracy of the method as it is currently used.In particular for glutaraldehyde, the phenomenon may also be caused by a lack of experience of some of the participating laboratories. Any substantial differences between various types of pumped sampling were assessed.Fig. 3 differentiates between sampling techniques for the DNPH method. Except for acrolein in runs 1 and 8, no significant differences between solid sorbent, filter and impinger methods with respect to uncertainty were found. Pumped sampling using a DNPH coating versus HMP and diffusive sampling leads to the observation that the overall uncertainty of diffusive samplers is even slightly lower than that obtained with pumped sampling. This is most likely due to the greater experience of aldehyde measurements possessed by the laboratories applying diffusive sampling. Diffusive sampling was not used frequently and only a few laboratories determined other aldehydes besides formaldehyde.HMP coating for acrolein has lead to systematically better results than the DNPH coating. Compliance of Aldehyde Personal Sampling Methods to EN 482 Compliance with minimum overall uncertainty requirements are detailed per analyte without differentiation between Fig. 3 Percentage of method–laboratory combinations per sampling technique with a relative overall uncertainty smaller than 30%.Aldehyde sampling exercise 1. Analyst, May 1997, Vol. 122 413 Formaldehyde Being the most prevalent aldehyde, it was of upmost importance to explain detected bias and attempt to improve the method(s) for detection. Figs. 1–5 show the current state-of-the-art. Typically, between 60 and 80% of the method–laboratory combinations had an ROU < 30%; between 25 and 50% had a bias within 10% of the true value. These data are worse than results in similar exercises for aromatic and chlorinated hydrocarbons, but acceptable and adequately in accordance with minimum performance requirements.Table 2 illustrates, for the selected performance indicators, the significant improvements through participation in successive sampling intercomparisons. The percentage of results for all method– laboratory combinations regarding overall uncertainty, bias within 10% and within-laboratory relative standard deviation improved by 9, 2 and 11%, respectively.The bias and withinlaboratory standard deviation for all individual method– laboratory combinations is detailed in Figs. 6 and 8 for two selected runs containing formaldehyde. Fourteen laboratories using the A1-SG-X method in exercise 1 often underestimated the accepted true value (Fig. 8). Inaccurate sampling volume was mentioned and found to be a potential cause.More consensus was obtained following the hypothesis that the recovery of formaldehyde as 2,4-dinitrophenylhydrazone was not 100%, as the majority of the participants had assumed. Fourteen laboratories later determined the recovery and found a value between 89 and 95% which would explain the underestimation shown in Fig. 8. These data are confirmed in the literature.21 Laboratory 49 stated that higher results (at least 15% difference) were obtained when 10 ml instead of 3 ml of desorption liquid were used.A change of solvent–sorbent ratio can indeed affect desorption efficiency. This phenomenon is known for charcoal, silica gel and porous polymers, but has not been described for reagent coated sorbents and their derivatives; 22 it nevertheless improved the results of laboratory 49. The desorption efficiency for formaldehyde was also found to be related to the loading on the cartridge (laboratory 16). Results obtained through the glass fiber filter method (A2- GFF-X; laboratories 16, 41, 42, 43, 47, 38) were quite scattered around the reference value.A critical factor here is breakthrough which is caused by either insufficient DNPH coating, mechanical damage of filters, drying out effects or the presence of interfering compounds ( > 1 l in the presence of acetone). Breakthrough is also found by the A1-SG-X method in the presence of acetone and above 1 l of sampling volume; up to 2 l in the presence of acetaldehyde is no problem.A 2 mg amount of coating is sufficient for 13 mm filters, but it should be increased for 25 and 37 mm filters. Drying out effects can be avoided by adding glycerine to the coating solution. The impinger method (laboratories 25, 27, 46 and 48) was very reliable. Formaldehyde collection in this way and direct injection of the sample into the GC or HPLC seems to avoid a number of steps which frequently introduce errors. The low results of laboratory 48 are explained by the sampling volume, which amounts to up to twice the value of other laboratories applying the same method.Disadvantages of the impinger collector, e.g., it is cumbersome to use and unsuitable for extended collection times due to solvent evaporation, seem to be particularly compensated by its accuracy. Reasons for bias applicable to the formaldehyde measurement methods were: (a) high concentrations of interfering compounds like acetaldehyde Table 2 Summary of formaldehyde measurement method improvements for selected performance indicators ROU Bias within Within-lab % of results* < 30% 10% RSD < 5% Exercise 1 64 40 49 Exercise 2 73 42 60 * Mean value of seven runs for exercise 1 and four runs for exercise 2.Fig. 6 As Fig. 8 except formaldehyde 0.308 ± 0.009 ppm (v/v). Run 2. Exercise 2. Fig. 7 Bias and within-laboratory RSD for all method–laboratory combinations. Acrolein 0.127 ± 0.004 ppm (v/v). Run 6. Exercise 1. Fig. 8 Bias and within-laboratory RSD for all method–laboratory combinations for selected formaldehyde runs.Formaldehyde 1.03 ± 0.03 ppm (v/v). Run 1. Exercise 1. Analyst, May 1997, Vol. 122 415and acetone cause a negative bias; it is assumed that this impedes the complete fixation of other compounds; (b) care should be taken with commercial formaldehyde hydrazone standards since laboratory 7 reported a 12% discrepancy with the BCR formaldehyde hydrazone standard (CRM 546);23 (c) no relationship was found between the overall uncertainty of formaldehyde measurements and the relative humidity level, which is confirmed in literature;24 (d) no relationship could be demonstrated between the bias of laboratories applying the DNPH method and the storage time; storage at 218 °C (even up to 8 weeks) did not result in a negative bias.Background levels of hydrazone derivatives on cartridges increase slightly with time and temperature. Therefore in exercise 2 an increased number of laboratories have stored at 218 °C and only for one to three weeks.Taking these potential sources of error into consideration resulted in an improved data set illustrated in Fig. 6. Fig. 9 illustrates that throughout the various runs, the between-laboratory relative standard deviation for the various methods was quite consistent and is between 20 and 60%. This is significantly higher than those values obtained for aromatic and chlorinated hydrocarbons (typically between 5 and 20%).28 An RSD between 10 and 30% was obtained in exercise II.This improvement is even visible in Fig. 6, where the results are less scattered; for the A1-SG-X method the results are also better centered around the reference value. Acrolein Sampling and analysis of acrolein is not yet common practice. All results for acrolein obtained through methods based on DNPH showed a significant negative bias; participants typically reported a value between 20 and 50% of the reference concentration (Fig. 7). This leads to an extremely low number of laboratories that comply with the EN 482 minimum performance criteria. Neither humidity, concentration levels, sampling time, interfering compounds nor other variables in sampling or analysis played an important role. The phenomenon was fully explained by the decomposition of the acrolein hydrazone on the samplers in the presence of excess of both DNPH and acid. Participants reported besides the main acrolein peak in the chromatogram (at the retention time of the acrolein hydrazone in desorption solution) a number of other peaks only appearing when acrolein was present in the atmosphere to be sampled (Fig. 10). The peak areas obtained vary in an unpredictable manner and both the number as well as peak heights were different between various method–laboratory combinations. The main peak is found to decrease with time, the additional peaks (between 3 and 6) increase rapidly as the sample deteriorates (Fig. 11).The decrease on a cartridge between 30 min and 24 h after sampling is shown in Fig. 11. Peaks 1, 2 and 3 originate from the decomposition of the acrolein hydrazone. This phenomenon leads to underestimation of the reference value since the main peak no longer represents the complete acrolein hydrazone concentration. The original acrolein hydrazone is partly rearranged on the sampler into a number of decomposition products. One of these products could be identified by laboratory 47 and consisted of three DNPH and two acrolein molecules. Once the sample is desorbed from the cartridge, tube or filter, no further degradation takes place, presumably because the acid is not present in excess in the solution.The decomposition products interfere and may overlap with other carbonyl compound peaks in the chromatogram which may result in false overestimation of these compounds (e.g., acetaldehyde, acetone, glutaraldehyde).Some of the observations described above are mentioned in a few publications. The decomposition of the acrolein hydrazone at room temperature is such that only 30% is left after one week. This was one of the explanations for incomplete recovery during acrolein sampling Fig. 9 Comparison of between-laboratory RSD (%) for selected sampling techniques and runs in exercise 1. Fig. 10 HPLC chromatogram with acrolein hydrazone decomposition products eluting between 3.50 and 5.30 min. 1, DNPH; 2, formaldehyde hydrazone; 3–7, origin from acrolein hydrazone. HPLC: UV at 375 mm, C18 column (35°C); AcN–H2O (75 + 25) at 1 ml min21; acrolein DNPH standard solution at 3.88 min. Fig. 11 Decomposition (degradation) of an acrolein standard solution (0.5 ml of 55 mg ml21) on cartridges coated with DNPH. 416 Analyst, May 1997, Vol. 122of automobile exhaust; storage at 0 °C would retard decomposition significantly.21,25 This last observation was not confirmed in the sampling project.In another study, decomposition was considered to be caused by ozone and an ozone scrubber was recommended before a cartridge to avoid decomposition of hydrazone derivatives.26,27 Due to these fundamental problems associated with the DNPH method, the overall uncertainty could only be improved by, for instance, taking the sum of all decomposition products into consideration to calculate the acrolein concentration or by passing immediately after sampling the solution through an ion exchange sorbent (e.g., Bond Elut SCX).Based on the results of this project, it can concluded that no currently available DNPH method has sufficient accuracy for acrolein measurements. The alternative method, which applies 2-(hydroxymethyl)- piperidine as derivative, was only applied by laboratories 16 and 34 but proved to form a stable acrolein derivative. The GC method applied by these participants showed very accurate results (Fig 7, laboratory 16).Glutaraldehyde The results obtained for glutaraldehyde were accurate for a number of laboratories but the picture for the total group and a majority of the participants was not encouraging. The obtained bias is explained by a number of parameters detailed below. As a whole, the results improved only slightly in the second sampling exercise. A number of parameters relevant for the performance of method–laboratory combinations are derived from Fig. 1 to 5 and are summarized in Table 3.Comparing the data for glutaraldehyde with the same set for toluene or other compounds where no sampling or significant analytical problems occur, illustrates that some aspects of this method are still subject to improvements. The values for toluene have been published previously.28 The bias and within-laboratory standard deviations for all method–laboratory combinations for selected runs with glutaraldehyde are illustrated in Fig. 12 (high humidity level) and Fig. 13 (low humidity level).In both (and all other) cases a quite scattered picture with a large overall between-laboratory relative standard deviation is obtained. Comparing both pictures illustrates a shift of the data from substantially above the reference value (Fig. 12) to 220–30% below (Fig. 13). This phenomenon is fully explained by the humidity level. Most laboratories obtain higher results when the humidity level of the reference gas mixture is high. This effect was confirmed when evaluating in detail the results of laboratory 23.These data were affected by the humidity level. Results close to the reference value were found in cases of a high humidity level (75%–80%) and 251% and 224% underestimation in case of 38 and 51% relative humidity, respectively. The detailed picture is shown in Fig. 14. One of the other factors that affected the performance of laboratories was the poor storage stability of the glutaraldehyde hydrazone on the sampling media.Storage at 220 °C for up to 4 weeks is no problem. However, if storage takes place in a refrigerator at 4–6 °C only 55–60% of the original amount is left on the sampler. Table 3 Comparison of glutaraldehyde measurement method performance indicators Within-lab Between- % of ROU Bias within RSD lab results* < 30% 30% < 5% RSD† Exercise 1 41 33 30 37 Exercise 2 43 24 44 39 * Mean value of seven runs for exercise I and three runs for exercise II. † Mean RSD of all active sampling methods, not percentage of results.Fig. 12 Bias and within-laboratory RSD for all method–laboratory combinations for selected glutaraldehyde runs. Glutaraldehyde 0.262 ± 0.008 ppm (v/v). Run 2. Exercise 1. Fig. 13 As Fig. 12 except glutaraldehyde 0.149 ± 0.004 ppm (v/v). Run 6. Exercise 1. Fig. 14 Correlation between the bias of glutaraldehyde measurements for one selected laboratory and the relative humidity level. 8, % relative humidity; X, % of reference concentration for the selected laboratory; O % of reference concentration for the mean of all laboratories. Analyst, May 1997, Vol. 122 417Although no separation problems occur in the various HPLC chromatograms between glutaraldehyde and the other compounds, some decomposition products of the acrolein hydrazone on the sampler can overlap with the glutaraldehyde hydrazone. In addition, the glutaraldehyde hydrazone appears as two isomers in the chromatogram, so the use of only one peak again leads to errors.All these factors that cause underestimation or overestimation of the reference value result in large variations in the percentage of method–laboratory combinations that fulfil the overall uncertainty criterion of 30% (Figs. 1 and 2). Typically only 20–50% of the method–laboratory combinations complied to EN 482. There appears to be no significant differences between methods. Respectively, 50–60% and 15–25% of the method– laboratory combinations use A1-SG-X and A2-GFF-X.Only one or two apply the methods A3-X, A1-XAD-X and D1-GFFX. Although the A1-SG-X method has systematically the lowest between laboratory RSD (Fig. 9), the smallest overall bias is obtained by laboratories using A2-GFF-X. The bias for the diffusive sampling methods was typically within 15%, but these laboratories obtained comparable results with their pumped method (A1-SG-X or A1-GFF-X). The results of laboratory 47 with the diffusive sampler even suggest that the sampling medium (glass fiber filter or paper) for the D1 method has affected the results.The glass fiber filter resulted in 10% higher results. These differences were not systematically detected for formaldehyde. Taking the above factors into account, the DNPH methods have the potential to produce accurate results for glutaraldehyde. This is proven, e.g., in run 5 of exercise 1, where the best results were obtained so that 52% of the method–laboratory combinations had a bias within 10% of the reference value; 25% had a within-laboratory relative standard deviation below 5%.Acetaldehyde and acetone Due to the high workplace air limit values for acetaldehyde (100 ppm) compared to the other aldehydes studied, this compound was only included in run 5 of exercise 1 and run 2 of exercise 2. The bias and within-laboratory relative standard deviation for all method–laboratory combinations is detailed in Fig. 15 for run 5 (1). Although a number of method–laboratory combinations succeeded in obtaining very accurate results (40% has a bias within 20%), a substantial number of outliers (60% with a bias > 25%) were found.The outlying results were caused by breakthrough on the samplers since the sampling volume was optimised for ppb levels of the other aldehydes. Besides breakthrough, care should be taken to avoid interference in the HPLC chromatogram of the large acetaldehyde peak with the small surrounding peaks especially of acrolein. The introduction of acetone in the programme was intended to detect possible interferences in the sampling and analysis of especially acrolein.It transpired that interference may indeed occur in the HPLC chromatogram but that acrolein measurements suffer especially from the instability of the acrolein hydrazone, a phenomenon which is detailed above. The results of run 7 (exercise 1) and run 5 (exercise 2) were not significantly affected by the presence of acetone. It is clear, however, that the presence of acetone may hamper the measurement of carbonyl compounds in air.Conclusion During the past decade, great progress has been made and at the same time a number of new methods have been developed and described in literature for the measurement of airborne aldehydes. The work presented here evaluated the most frequently used methods for determining aldehydes. A workplace air atmosphere containing known amounts of aldehydes was dynamically prepared and distributed through a manifold which allowed to evaluate bias and precision of the currently used methods.Up to 30 participants could sample on the spot at the coordinators’ laboratory from a gravimetrically determined reference concentration. The existing capillary dosage technique was therefore thoroughly reviewed so that gas phase generation of reactive compounds, like acrolein, glutaraldehyde and acetaldehyde, became feasible over a wide concentration range and in high air flows so that pumped and diffusive sampling techniques could be assessed at the same time.The results obtained by a selected number of European laboratories in two sampling exercises suggest that the current methods for formaldehyde comply in general and for an acceptable number of laboratories with EN 482 minimum performance criteria. The DNPH method produced poor results for acrolein which has led to the conclusion to abandon or significantly alter the method due the instability of the acrolein hydrazone on the sampler leading to decomposition of the derivative.Methods applied for glutaraldehyde and acetaldehyde seem principally and practically all right but their limited (although currently strongly increasing) use in daily occupational hygiene practice leads to errors which are usually anticipated by a group of more experienced users but are often not taken into account by the majority of the participants. Besides the overall results presented here and the discussion of potential sources of error, it is evident that this project also gained information on parameters like breakthrough volumes, uptake rates, analytical instrument optimisation, desorption efficiency tests, sampling volumes, active versus diffusive sampling within the same laboratory and other practical aspects which are not discussed fully here.These details are available in technical reports and can be obtained free through the various participants listed in the Appendix.This project was partly funded by the Standards, Measurements and Testing programme of the European Commission. The valuable advice of Dr. S. Vandendriessche in the design and management of the project is gratefully acknowledged. The author is indebted to all the participants in the project for their contribution and cooperation. The list of participants per country code is detailed below in the Appendix. Appendix B Mr.R. Grosjean and Mr. R. Mesmacque: Ministry of Labour; Dr. Roosels, Mrs. M. Noel, Mr. J.-M. Bosiroy, Mr. J. Boulanger: Funds for Occupational diseases; Mr. E. Fig. 15 Bias and within-laboratory RSD for all method–laboratory combinations. Acetaldehyde 31.3 ± 0.9 ppm v/v. Run. 5. Exercise 1. 418 Analyst, May 1997, Vol. 122Goelen (project coordinator), Mrs. F. Geyskens, Mr. M. Lambrechts, Mr. R. Bormans, Mr. T. De Ceuster, Mr. R. Mannaerts: VITO. CH Dr. C. K. Huynh, Mr. Vernay, Mr.Boiteux: Institut Universitaire Romande de Sant�e au Travail. G Dr. J. Auffarth, Dr. R. Hebisch, Mr. K. Rentel: Bundesanstalt fur Arbeitsschutz; Dr. W. Kramer, Mr. R. Schmitt: BASF; Dr. B. Strieffler, Mr. K. Schneider: Niedersachsisches Landesamt fur Okologie; Dr. J. Schnelle: GSF - Forschungszentrum fur Umwelt und Gesundheit; Dr. H. Fricke: Institut fur Gefahrstoff-Forschung; Dr. M. Weigl, Mr. Fehlauer: Berufsgenossenschaft Nahrungsmittel und Gaststatten Potsdam; Dr.B. Andrejs, Mrs. Kiefer, Mrs. C. Schuh: Berufsgenossenschaft Nahrungsmittel und Gaststatten Mannheim; Dr. Wensing, Dr. Schwarzer, Mr. M. Berbig, Mr. C. Kahre: TUV Nord; Dr. T. Wiesmuller, Mr. J. Schymonski, Mr. P. Zellner: Ecoplan; Mr. M. Bruckschlegel: IKPO Universitat Stuttgart; Dr. D. Ullrich, Mrs. C. Scheller: BGA, Institut fur Wasser-, Boden-, Lufthygiene. DK Dr. K. Egmose, Mrs. B. Grau-Hansen: Miljoe-Kemi. ES Dr. X. Guardino, Mrs. C. Santolaya, Mr. A. Marti: INSHT Barcelona; Mr.J. L. Perez Alvarez: Grupo Interlab; Mr. J. Comino, Mr. R. Manuel Ruiz, Mr. E. Arnaiz: Incohinsa. F Mr. R. Dujardien, Mr. H. Adrien, Mr. B. Brouart: INERIS. I Dr. V. Cocheo, Dr. F. Quaglio: Fondazione Savatore Maugeri. NL Mr. H. P. Bos, Mrs. S. Linders: RIVM; Mr. R. Peeters, Mr. M. Houtzager: TNO. SF Mrs. M.-L. Henricks-Eckermann: Turku Regional Institute of Occupational Health; Dr. E. Priha, Mrs. A. Jalkanen: Tampere Regional Institute of Occupational Health; Mrs.T. Tirkkonen, Mr. P. Tapanimaki, Mrs. R. Ketola: VTT - Chemical Technology. S Ass. Prof. J. Levin, Mr. R. Lindahl: National Institute for Working Life; Dr. L. Johnson, Mr. L. Rosell, Mrs. I. Isaksson: National Testing and Research Institute: Mr. R. Nordlinder, Mr. G. Ljungkvist: Sahlgrenska University Hospital. UK Mr. J. Cuthbert: Health and Safety Executive; Mr. G. Bebbington: Severn Trent Laboratories. B, Belgium; CH, Switzerland, G, Germany, DK, Denmark; ES, Spain; F, France; I, Italy, NL, Netherlands; SF, Finland; S, Sweden; UK, United Kingdom. References 1 Selim, S., J. Chromatogr., 1977, 136, 271. 2 Beasley, R. K., Hoffman, C. E., Rueppel, M. L., and Worley, J. W., Anal. Chem., 1980, 52, 1110. 3 Grosjean, D., and Kok, G. L., Anal. Chem., 1982, 54, 1221. 4 Tejada, S. B., Int. J. Environ. Anal. Chem., 1986, 26, 167. 5 Andersson, K., Andersson, G., Nilsson, C., and Levin, J. O., Chemosphere, 1979, 8, 812. 6 Kennedy, E. R., and Hill, R. H., Anal. Chem., 1982, 54, 1739. 7 Kennedy, E. R., O’Connor, P. F., and Gagnon, Y. T., Anal. Chem., 1984, 56, 2120. 8 Levin, J. O. Andersson, K., Lindahl, R., and Nilsson, C. A., Anal. Chem., 1985, 57, 1032. 9 Kuwata, K., Uebori, M., and Yamasaki, Y., J. Chromatogr., 1979, 17, 264. 10 Levin, J. O. Lindahl, R., and Andersson, K., Environ. Sci. Technol., 1986, 20, 1272. 11 Determination of Formaldehyde Vapours in Air, 3M Occupational Health and Environmental Safety Division, St. Paul, MN, USA. 12 Pfaffli, P., Virtanen, H., Riutta, O., and Hayri, L., in Clean Air at Work: New Trends in Assessment and Measurement for the 1990s, ed. Brown, R. H., Curtis, M. Saunders, K. J., and Vandendriessche, S., Royal Society of Chemistry, 1993, pp. 198–200. 13 Cocheo, V., Boaretto, C., and Sacco, P., Am. Ind. Hyg. Assoc. J., 1996, 57, 897. 14 European Committee for Standardization, Assessment of Workplace Exposure—General Requirements for the Performance of Procedures for the Measurement of Chemical Agents, CEN/TC137/ European Standard EN 482, Brussels, July 1994. 15 Goelen, E., Lambrechts, M., Geyskens, F., and De Fr�e, R., paper presented at the 2nd European Workshop on Mass Spectrometry in Occupational Health, Les Diablerets, Switzerland, June 16–18, 1993. 16 Goelen, E., Geyshens, F., Lambrechts, M., and Rymen, T., Int. J. Environ. Anal. Chem., 1992, 47, 217. 17 Goelen, E., Geyskens, F., and Lambrechts, M., Facility for the Generation and Distribution of Aldehyde Atmospheres in Air at Occupational Hygiene Levels, internal VITO report (MIE-DI-95-16), 1995. 18 Nelson, G. O., Gas Mixtures—Preparation and Control, Lewis Publishers, Chelsea, 1992. 19 Workshop on Possible New Projects in the Field of Workplace Air Monitoring, Brussels, March 25–27, 1991. 20 Levin, J. O., Lindahl, R., and Heeremans, C.E.M., EUR report, Brussels, 1996. 21 Lipari, F., and Swarin, S. J., J. Chromatogr., 1982, 247, 297. 22 Dommer, R. A., and Melcher, R. G., Am. Ind. Hyg. Assoc. J., 1978, 39, 240. 23 European Commission, BCR reference materials, Institute for Reference Materials and Measurements (Joint Research Centre), Geel, 1996. 24 Kuwata, K., Uebori, M., Yamasaki, H., and Kuge, T., Anal. Chem., 1983, 55, 2013. 25 Vainiotalo, S., Matveinen, K., in Clean Air at Work: New Trends in Assessment and Measurement for the 1990s, ed. Brown, R. H., Curtis, M., Saunders, K. J., and Vandendriessche, S., Royal Society of Chemistry, Cambridge, 1993, pp. 204–205. 26 Arnts, R. R., and Tejada, S. B., Environ. Sci. Technol., 1989, 23, 1428. 27 Barone, J. P., Walter, T. H., in Clean Air at Work: New Trends in Assessment and Measurement for the 1990s, ed. Brown, R. H., Curtis, M., Saunders, K. J., and Vandendriessche, S., Royal Society of Chemistry, Cambridge, 1993, pp. 162–168. 28 Goelen, E., Geyskens, F., and Lambrechts, M., Ann. Occup. Hyg., 1997, in the press. Paper 6/07047G Received October 16, 1996 Accepted January 13, 1997 Analyst, May 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a607047g

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Sampling and Analytical Quality Control of the Determination of Aluminium in Soybean Leaves Deming Donga, Michael H. Ramseyb and Iain Thorntonb a Department of Environmental Science, Jilin University, Changchun 130023, China b Environmental Geochemistry Research Group, Centre for Environmental Technology and Department of Geology, Imperial College of Science, Technology and Medicine, London, UK SW7 2BP A pot trial was set up with nine soil types, each replicated five times, in randomized blocks, to investigate the effect of soil properties on the uptake of Al by soybean plants.The variance of the measurements was estimated using five replicates of each soil type. This then enabled the statistical significance of the difference between results for each soil type to be established. The analytical variance was determined by two analyses per pot. Systematic errors during the analytical procedure were estimated by the analysis of reagent blanks and CRMs. Analysis of variance (ANOVA) was used to estimate the proportions of total variance contributed by analysis, sampling and geochemistry. Robust statistics were required because of the high sensitivity of classical ANOVA to a small number of outlying values, due to either technical or geochemical causes.In the case study described here, traditional analytical quality control apparently gave a satisfactory result. However, the results from robust ANOVA showed that the sampling error was the limiting factor for this application.It is suggested that the environmental analyst must consider errors from procedures of both sampling and analysis of environmental samples. Keywords: Bias; variance; aluminium; quality control; sampling Although Al is the most abundant metal and the third most common element in the Earth’s crust, its content in plant materials is at the trace level. Reliable determination of Al concentration in plant materials is a major problem caused by secondary Al contamination from soil/dust particles adhering to the surface of plant materials during storage and laboratory manipulations.1,2 Therefore, considerable efforts have been made to reduce Al contamination during the analytical procedure. 1,3–5 Analytical quality control is an accepted part of environmental chemical analysis.It is a characteristic feature of environmental analytical chemistry that large batches of samples, usually of soil/sediments and plants, are required to be analysed with the shortest possible turnaround time.The two parameters used to assess analytical quality in environmental analytical chemistry are bias and precision. The bias and precision normally qualify the systematic and random errors in environmental analytical chemistry. Systematic errors in sample preparation and analysis can be partially estimated by the analysis of reagent blanks and CRMs. The random error in the analytical process (i.e., analytical precision) can be estimated by the analysis of duplicate portions of test materials.6 However, analytical quality control does not estimate the random error introduced in the process of sampling.The random sampling error can be expressed as the sampling variance, but cannot be estimated directly from the analyses of duplicate samples because the sampling variance is overprinted by the analytical variance. The environmental analyst may attempt to take a sufficiently large sample to provide sample representativity, but rarely is any attempt made to estimate the measurement uncertainty that remains due to the sampling process.Ramsey et al.7 reported that the sampling variance might be estimated using a Sampling and Analytical Quality Control Scheme (SAQCS = ‘SAX’). The problem of quantitative estimation of sampling variance was solved empirically. The case study described here is used to show that: (1) estimations of both sampling and analytical errors are important in environmental sample analysis; (2) the analytical precision required must be judged by comparison with both the sampling precision and the geochemical variance.Experimental A greenhouse experiment was designed in order to investigate the environmental factors that control the uptake of Al from soil into soybean plants. For this purpose, it was necessary to measure reliably the differences between Al concentrations in soybean plants grown on different soil types.A pot trial was therefore set up with nine soil types, each replicated five times, in randomized blocks. This resulted in a total of 45 pots. The pots were watered with de-ionized water to keep the soil moisture approximately equal to that found in the field, (i.e., 20–25% m/m). After 2 months, whole plants were harvested at 5 cm above soil level to avoid soil contamination.1 The plant materials were rinsed thoroughly with de-ionized water to reduce surface contamination, dried at 30 °C, weighed, then ground prior to chemical analysis.Care was taken at each stage to minimize any Al contamination from either the laboratory procedures used or the laboratory atmosphere. The Al concentration in the plant leaves, after acid digestion, was then determined by ICP-AES at 308.2 nm and reported on a dry mass basis. The relevant instrumental conditions have been described previously by Ramsey et al.2 The sampling variance of the measurements was estimated using five replicates of each soil type.This then enabled the statistical significance of the difference between results for each soil type to be established. The analytical variance was determined by two chemical analyses per pot. Systematic errors during the analytical procedure were estimated by the analysis of reagent blanks and CRMs. House reference materials (HRMs) were used to control the betweenbatch bias and the overall bias of results throughout the period of the analysis.HRMs of appropriate overall composition were selected and inserted into each batch of samples at random positions. This was done conveniently at the time of weighing. The concentrations of the analyte in the HRMs were at background and elevated levels, respectively, in order to monitor the accuracy across the whole range of concentrations determined. For each HRM, the mean and standard deviation of the analysis results was obtained. The analytical bias was obtained by a comparison between the measured and accepted values of Al concentrations.The CRMs were analysed Analyst, May 1997, Vol. 122 (421–424) 421occasionally to estimate the bias of the method. The CRMs used were NIES (National Institute for Environmental Studies, Japan) No. 7 Tea Leaves, and NIST SRMs 1573 Tomato Leaves and 1572 Citrus Leaves. Reagent blanks were included as 10% of the samples at irregular intervals within each batch to ascertain the level of the analyte in the reagents and laboratory environment. They could therefore identify any contamination, either from the addition of extraneous material to the sample or due to carry over when anomalous and background samples were analysed together.The statistical significance of the bias between the mean value of element concentrations measured for the analyte and the certified value (for the reference materials) or zero (for the reagent blanks) was assessed using the t-test at the 95% confidence level.The analysis of a sample in duplicate and comparison of the two results has long been used as a method by which analysts have gauged the repeatability of their analysis. In properly designed experiments, duplicated analysis can be used as an estimate of the analytical precision within each batch of analysis. In each batch, 10% of the samples were randomly selected and analysed in duplicate, thereby representing the full range of analyte concentrations and matrices. Precision can be rapidly tested against an empirical standard of precision using a control chart.The use of these charts is based on the methodology of Thompson and Howarth6 and involves plotting the absolute difference between pairs of duplicate analyses against the mean value of the pairs. If the data are distributed on the charts such that 50% of the points lie below the 50% line, (or 90% below the 90% line, and/or 99% below the 99% line), a precision of 10% has then been achieved at a 95% confidence limit.A Sampling and Analytical Quality Control Scheme (SAQCS = ‘SAX’), recommended by Ramsey et al.7 and Ramsey,8 was applied to quantify both the analytical and sampling variance. The statistical technique used to separate the analytical variance from the sampling variance was robust analysis of variance (ANOVA). Results and Discussion Estimation of Systematic Errors Using Reagent Blanks Reagent blanks were used to ascertain the background level of the analyte and identify any contamination introduced in the entire digestion/analysis procedure.The statistical significance of the bias between the mean value of the Al concentrations in reagent blanks and zero was assessed using the t-test at the 95% confidence level. The t values calculated from reagent blanks for the analytical method used (Table 1) were larger than the tabulated t value (t = 2.57 at the 95% confidence level when n = 6). These results show that the bias is significantly greater than zero, and suggest that the reagent blanks for the method have been contaminated, possibly by dust particles from the laboratory environment.1,2 Therefore, all determinations using the method have to be corrected, by the subtraction of the mean value of the Al contents of the reagent blanks.Estimates of Analytical Bias Using Reference Materials The analysis of CRMs and HRMs was used to determine the analytical bias. The overall bias for each reference material was obtained by a comparison between the measured and accepted values of the Al concentrations. The statistical significance of the bias was also assessed.The measured Al concentrations in the CRMs and HRMs and the bias from the accepted values are shown in Table 2. All of the HRMs and CRMs used that had significant bias, had bias values of less than 10%, which is considered adequate, with the exception of SRM 1572 Citrus Leaves, for which a large negative significant bias (216.3%) was found.The reason for the large bias needs to be studied further. Although the results of the t-test showed that there were significant difference between the measured and accepted values for SRM 1572 Citrus Leaves, HRM 12 and HRM 13, this was not of sufficient magnitude to present a problem. The high recovery of Al from the CRMs and HRMs suggests that the trueness of the analytical method is satisfactory. Precision of the Analytical Method The analytical precision was evaluated by analysing 10% of the test materials in duplicate. Test materials were selected at random and therefore, on average, represented the full range of analyte concentrations across the samples.The use of precision charts is a simple way to test the precision of an analytical method.6 The precision chart for the determination of Al in soybean leaves is given in Fig. 1. This result showed that analyses of the soybean leaves for Al achieved a precision of less than ±10% (with a 95% confidence limit).Therefore, the precision of all batches of analyses was considered acceptable. Assessment of the Proportions of Analytical and Sampling Variance in the Experimental Design The SAX procedure requires: (1) Duplication of both sampling and analysis in a balanced design (i.e., analytical duplicates taken on both sample duplicates). (2) The separation of both the sampling and the analytical variance. The sum of the variance of the two measurement processes of sampling (s2 s) and analysis (s2 a) can be called the measurement variance (s2 meas), where Table 1 Al concentrations in the reagent blanks for the determination of Al in soybean leaves.n = 6 in all cases Calculated t Batch No. Mean/mg g21 s/mg g21 value 1 3.54 1.05 8.26 2 3.67 1.14 7.89 3 3.60 1.14 7.74 4 5.35 1.15 11.40 5 3.37 0.78 10.58 6 2.20 0.74 7.28 7 2.53 0.72 8.61 8 1.83 0.50 8.97 Table 2 Overall accuracy estimate of Al from CRMs and HRMs Accepted Measured Reference value/ value/ s/ Bias material mg g21 mg g21 mg g21 n (%) Citrus leaves (NIST SRM 1572) 92 77 3.8 15 216.3* Tomato Leaves (NIST SRM 1573) 1200† 1197 28.5 15 20.2 Tea Leaves (NIES No. 7) 775 724 21.2 10 26.0 HRM 11 120 115 5.5 4 24.2 HRM 12 101 97 4.8 26 24.0* HRM 13 500 512 25.4 26 2.4* HRM 14 780 764 24.7 4 22.1 * Significant bias at the 95% confidence level. † Provisional value. 422 Analyst, May 1997, Vol. 122Mean of duplicate results Difference between results 1 10 100 1000 100 10 1 0.1 99% 90% 50% s2 meas = s2 s + s2 a (1) (3) The comparison of these ‘measurement’ variances with the ‘true’ variability of the element in the samples excluding these errors, i.e., geochemical variance (s2 g).If the sources of variation are independent, the total variance (s2 total) is given by eqn (2). The three-component variances can be estimated using ANOVA s2 total = s2 g + s2 s + s2 a (2) The acceptable proportions of the measurement and analytical variances can be approximately related to the total variance as follows: 1% s2 total < s2 meas < 20% s2 total (3) 1% s2 meas < s2 a < 20% s2 meas (4) If the measurement variance contributes more than 20% of the total variance, then significant geochemical information will become progressively more obscured.In the case where the analytical variance contributes more than 20% to the measurement variance, analysis dominates the measurement variance. The analytical precision can then usefully be improved.If, however, the measurement variance is below 1% of the total variance, any reduction will not be cost-effective. This is similarly the case if the analytical variance is below 1% of the measurement variance. These limits are set, therefore, in a proportion to the geochemical variability rather than an arbitrary fixed precision as applied above (e.g., 10%). Analysis of variance, as a statistical technique, can be used to separate the three sources of variance and give sg, ss and sa which are estimates of sg, ss and sa, respectively.The advantage of variance as a measure of random error (i.e., precision) is that variances (s2) are additive whereas standard deviations (s) are not.7 In this work, SAX was applied to the determination of Al in soybean leaves. Estimated mean and standard deviations by using classical ANOVA of all the determinations of Al and robust ANOVA from the SAX data are both shown in Table 3.The analytical data quality was assessed as being generally satisfactory, as discussed above. The classical ANOVA of the soybean leaves suggested a mean Al concentration of 94.4 mg g21 with a total standard deviation (stotal) of 131.5 mg g21. However, the results given by the classical ANOVA may be erroneous because of four factors. These are: (1) the random errors were not normally distributed; (2) few samples had very high levels of Al; (3) there were a few poor duplicate analyses; and (4) the analytical error was not homogeneous (i.e., sa increases with concentration). These observations are inconsistent with the assumptions on which the classical ANOVA depends.7 In practice, it is difficult to make geochemical measurements that conform fully to the assumptions.For example, it is well established that analytical variance does vary as a function of concentration. This invalidates the assumption of homogeneity, for most geochemical measurements, although the application of the classical ANOVA to geochemical data has been widely reported.7 Robust ANOVA, however, has been shown to give more accurate estimates of variance in simulation studies.9,10 Robust statistics rely on the accommodation of outlying values rather than their rejection, and have been shown to be particularly appropriate for the description of analytical data.8,9 Briefly, the robust estimates of the mean and standard deviation are calculated by an iterative process, in which outlying observations are assigned reduced weights when they occur beyond a certain distance (e.g., 1.5s) from the estimated mean.Robust ANOVA should, therefore, provide more accurate estimates of the component variances.8 An important objective in this application of ANOVA was to check that the measurement variance originates in the processes of measurement (sampling and analysis). In this application the measurement variance contributed 20.7% to the total variance, which was approximately equal to the limit of 20%.Reliable interpretation of the geochemistry can be made, therefore, in the interpretation of the differences between the Al uptake by the soybean plants in the various soil types. The analytical method only contributed 0.1% to the total variance, which indicates that the sampling variance was the limiting factor for this application. This domination of the measurement variance by the sampling variance is probably caused by a few samples that were contaminated by soil/dust particles,1 in which the concentrations of Al were very high.Expressed in a more traditional way, the estimated analytical standard deviation (sa) was 2.2% of the overall mean, which is less than the usual arbitrary target of 5% (1sa) or 10% (2sa). Furthermore, the analytical variance contributed only 0.5% to the measurement variance. It can be concluded that the analytical precision was fit-for-purpose and better than required.Conclusions Analysts often expect the highest accuracy and greatest precision for chemical analysis. For certain applications, this is, Fig. 1 Ten per cent precision control chart for the determination of Al in soybean leaves. Table 3 SAX results for concentrations of Al in soybean leaves Standard deviation/mg g21 Mean/ Analysis Sampling Measurement Geochemical Total Method mg g21 (sa) (ss ) (sm ) (sg ) (stotal) Classical 94.4 1.92 31.99 32.05 127.51 131.5 Robust 46.9 1.03 14.38 14.42 28.24 31.7 Analyst, May 1997, Vol. 122 423no doubt, very important. For environmental analysis, however, analytical precision needs to be related to an estimate of the variance caused by the sampling process. A Sampling and Analytical Quality Control Scheme (SAX) extends traditional analytical control into the field environment to measure and also to provide targets for both analytical and sampling precision.This method utilizes replicate samples that are further duplicated when they are analysed, in addition to the other quality control materials, such as reference materials and reagent blanks. The case study described here showed that the systematic errors during the analytical procedure were acceptable with recoveries of Al of > 94% measured in reference materials after subtracting background values measured in reagent blanks. The application of SAX showed that the measurement variance contributed 20.7% to the total variance, which was approximately equal to the suggested limit of 20%.In this case, a reliable interpretation of the geochemistry can be made, for example, that there are the differences between the Al uptake by the soybean plants in the various soil types. The analytical method only contributed 0.1% to the total variance, which indicates that the sampling variance was the limiting factor for this application. This result suggests that the estimation of errors in the sampling process is very important in the chemical analysis of environmental materials.References 1 Dong, D., PhD Thesis, University of London, 1993. 2 Ramsey, M. H., Dong, D., Thornton, I., Watt, J., and Giddens, R., Environ. Geochem. Health, 1991, 13, 114. 3 Delves, H. T., Suchak, B., and Fellows, C. S., in Aluminium in Food and the Environment, ed. Massey, R. C., and Taylor, D., Royal Society of Chemistry, London, 1989, pp. 52–67. 4 Skelly, E. M., and Distefano, F. T., Appl. Spectrosc., 1988, 42, 1302. 5 Dong, D., Ramsey, M. H., and Thornton, I., J. Geochem. Explor., 1995, 55, 223. 6 Thompson, M., and Howarth, R. J., Analyst, 1976, 101, 690. 7 Ramsey, M. H., Thomopson, M., and Hale, M., J. Geochem. Explor. 1992, 44, 23. 8 Ramsey, M. H., Appl. Geochem., 1993, Suppl. Issue, 2, 149. 9 Analytical Methods Committee, Analyst, 1989, 114, 1693. 10 Analytical Methods Committee, Analyst, 1989, 114, 1699. Paper 6/07636J Received November 11, 1996 Accepted February 3, 1997 424 Analyst, May 1997, Vol. 122 Sampling and Analytical Quality Control of the Determination of Aluminium in Soybean Leaves Deming Donga, Michael H. Ramseyb and Iain Thorntonb a Department of Environmental Science, Jilin University, Changchun 130023, China b Environmental Geochemistry Research Group, Centre for Environmental Technology and Department of Geology, Imperial College of Science, Technology and Medicine, London, UK SW7 2BP A pot trial was set up with nine soil types, each replicated five times, in randomized blocks, to investigate the effect of soil properties on the uptake of Al by soybean plants.The variance of the measurements was estimated using five replicates of each soil type. This then enabled the statistical significance of the difference between results for each soil type to be established. The analytical variance was determined by two analyses per pot. Systematic errors during the analytical procedure were estimated by the analysis of reagent blanks and CRMs.Analysis of variance (ANOVA) was used to estimate the proportions of total variance contributed by analysis, sampling and geochemistry. Robust statistics were required because of the high sensitivity of classical ANOVA to a small number of outlying values, due to either technical or geochemical causes. In the case study described here, traditional analytical quality control apparently gave a satisfactory result.However, the results from robust ANOVA showed that the sampling error was the limiting factor for this application. It is suggested that the environmental analyst must consider errors from procedures of both sampling and analysis of environmental samples. Keywords: Bias; variance; aluminium; quality control; sampling Although Al is the most abundant metal and the third most common element in the Earth’s crust, its content in plant materials is at the trace level.Reliable determination of Al concentration in plant materials is a major problem caused by secondary Al contamination from soil/dust particles adhering to the surface of plant materials during storage and laboratory manipulations.1,2 Therefore, considerable efforts have been made to reduce Al contamination during the analytical procedure. 1,3–5 Analytical quality control is an accepted part of environmental chemical analysis. It is a characteristic feature of environmental analytical chemistry that large batches of samples, usually of soil/sediments and plants, are required to be analysed with the shortest possible turnaround time.The two parameters used to assess analytical quality in environmental analytical chemistry are bias and precision. The bias and precision normally qualify the systematic and random errors in environmental analytical chemistry. Systematic errors in sample preparation and analysis can be partially estimated by the analysis of reagent blanks and CRMs.The random error in the analytical process (i.e., analytical precision) can be estimated by the analysis of duplicate portions of test materials.6 However, analytical quality control does not estimate the random error introduced in the process of sampling. The random sampling error can be expressed as the sampling variance, but cannot be estimated directly from the analyses of duplicate samples because the sampling variance is overprinted by the analytical variance.The environmental analyst may attempt to take a sufficiently large sample to provide sample representativity, but rarely is any attempt made to estimate the measurement uncertainty that remains due to the sampling process. Ramsey et al.7 reported that the sampling variance might be estimated using a Sampling and Analytical Quality Control Scheme (SAQCS = ‘SAX’). The problem of quantitative estimation of sampling variance was solved empirically.The case study described here is used to show that: (1) estimations of both sampling and analytical errors are important in environmental sample analysis; (2) the analytical precision required must be judged by comparison with both the sampling precision and the geochemical variance. Experimental A greenhouse experiment was designed in order to investigate the environmental factors that control the uptake of Al from soil into soybean plants. For this purpose, it was necessary to measure reliably the differences between Al concentrations in soybean plants grown on different soil types.A pot trial was therefore set up with nine soil types, each replicated five times, in randomized blocks. This resulted in a total of 45 pots. The pots were watered with de-ionized water to keep the soil moisture approximately equal to that found in the field, (i.e., 20–25% m/m). After 2 months, whole plants were harvested at 5 cm above soil level to avoid soil contamination.1 The plant materials were rinsed thoroughly with de-ionized water to reduce surface contamination, dried at 30 °C, weighed, then ground prior to chemical analysis.Care was taken at each stage to minimize any Al contamination from either the laboratory procedures used or the laboratory atmosphere. The Al concentration in the plant leaves, after acid digestion, was then determined by ICP-AES at 308.2 nm and reported on a dry mass basis. The relevant instrumental conditions have been described previously by Ramsey et al.2 The sampling variance of the measurements was estimated using five replicates of each soil type. This then enabled the statistical significance of the difference between results for each soil type to be established.The analytical variance was determined by two chemical analyses per pot. Systematic errors during the analytical procedure were estimated by the analysis of reagent blanks and CRMs. House reference materials (HRMs) were used to control the betweenbatch bias and the overall bias of results throughout the period of the analysis.HRMs of appropriate overall composition were selected and inserted into each batch of samples at random positions. This was done conveniently at the time of weighing. The concentrations of the analyte in the HRMs were at background and elevated levels, respectively, in order to monitor the accuracy across the whole range of concentrations determined. For each HRM, the mean and standard deviation of the analysis results was obtained.The analytical bias was obtained by a comparison between the measured and accepted values of Al concentrations. The CRMs were analysed Analyst, May 1997, Vol. 122 (421–424) 421occasionally to estimate the bias of the method. The CRMs used were NIES (National Institute for Environmental Studies, Japan) No. 7 Tea Leaves, and NIST SRMs 1573 Tomato Leaves and 1572 Citrus Leaves. Reagent blanks were included as 10% of the samples at irregular intervals within each batch to ascertain the level of the analyte in the reagents and laboratory environment.They could therefore identify any contamination, either from the addition of extraneous material to the sample or due to carry over when anomalous and background samples were analysed together. The statistical significance of the bias between the mean value of element concentrations measured for the analyte and the certified value (for the reference materials) or zero (for the reagent blanks) was assessed using the t-test at the 95% confidence level.The analysis of a sample in duplicate and comparison of the two results has long been used as a method by which analysts have gauged the repeatability of their analysis. In properly designed experiments, duplicated analysis can be used as an estimate of the analytical precision within each batch of analysis. In each batch, 10% of the samples were randomly selected and analysed in duplicate, thereby representing the full range of analyte concentrations and matrices.Precision can be rapidly tested against an empirical standard of precision using a control chart. The use of these charts is based on the methodology of Thompson and Howarth6 and involves plotting the absolute difference between pairs of duplicate analyses against the mean value of the pairs. If the data are distributed on the charts such that 50% of the points lie below the 50% line, (or 90% below the 90% line, and/or 99% below the 99% line), a precision of 10% has then been achieved at a 95% confidence limit. A Sampling and Analytical Quality Control Scheme (SAQCS = ‘SAX’), recommended by Ramsey et al.7 and Ramsey,8 was applied to quantify both the analytical and sampling variance.The statistical technique used to separate the analytical variance from the sampling variance was robust analysis of variance (ANOVA). Results and Discussion Estimation of Systematic Errors Using Reagent Blanks Reagent blanks were used to ascertain the background level of the analyte and identify any contamination introduced in the entire digestion/analysis procedure. The statistical significance of the bias between the mean value of the Al concentrations in reagent blanks and zero was assessed using the t-test at the 95% confidence level.The t values calculated from reagent blanks for the analytical method used (Table 1) were larger than the tabulated t value (t = 2.57 at the 95% confidence level when n = 6).These results show that the bias is significantly greater than zero, and suggest that the reagent blanks for the method have been contaminated, possibly by dust particles from the laboratory environment.1,2 Therefore, all determinations using the method have to be corrected, by the subtraction of the mean value of the Al contents of the reagent blanks. Estimates of Analytical Bias Using Reference Materials The analysis of CRMs and HRMs was used to determine the analytical bias.The overall bias for each reference material was obtained by a comparison between the measured and accepted values of the Al concentrations. The statistical significance of the bias was also assessed. The measured Al concentrations in the CRMs and HRMs and the bias from the accepted values are shown in Table 2. All of the HRMs and CRMs used that had significant bias, had bias values of less than 10%, which is considered adequate, with the exception of SRM 1572 Citrus Leaves, for which a large negative significant bias (216.3%) was found.The reason for the large bias needs to be studied further. Although the results of the t-test showed that there were significant difference between the measured and accepted values for SRM 1572 Citrus Leaves, HRM 12 and HRM 13, this was not of sufficient magnitude to present a problem. The high recovery of Al from the CRMs and HRMs suggests that the trueness of the analytical method is satisfactory.Precision of the Analytical Method The analytical precision was evaluated by analysing 10% of the test materials in duplicate. Test materials were selected at random and therefore, on average, represented the full range of analyte concentrations across the samples. The use of precision charts is a simple way to test the precision of an analytical method.6 The precision chart for the determination of Al in soybean leaves is given in Fig. 1. This result showed that analyses of the soybean leaves for Al achieved a precision of less than ±10% (with a 95% confidence limit). Therefore, the precision of all batches of analyses was considered acceptable. Assessment of the Proportions of Analytical and Sampling Variance in the Experimental Design The SAX procedure requires: (1) Duplication of both sampling and analysis in a balanced design (i.e., analytical duplicates taken on both sample duplicates).(2) The separation of both the sampling and the analytical variance. The sum of the variance of the two measurement processes of sampling (s2 s) and analysis (s2 a) can be called the measurement variance (s2 meas), where Table 1 Al concentrations in the reagent blanks for the determination of Al in soybean leaves. n = 6 in all cases Calculated t Batch No. Mean/mg g21 s/mg g21 value 1 3.54 1.05 8.26 2 3.67 1.14 7.89 3 3.60 1.14 7.74 4 5.35 1.15 11.40 5 3.37 0.78 10.58 6 2.20 0.74 7.28 7 2.53 0.72 8.61 8 1.83 0.50 8.97 Table 2 Overall accuracy estimate of Al from CRMs and HRMs Accepted Measured Reference value/ value/ s/ Bias material mg g21 mg g21 mg g21 n (%) Citrus leaves (NIST SRM 1572) 92 77 3.8 15 216.3* Tomato Leaves (NIST SRM 1573) 1200† 1197 28.5 15 20.2 Tea Leaves (NIES No. 7) 775 724 21.2 10 26.0 HRM 11 120 115 5.5 4 24.2 HRM 12 101 97 4.8 26 24.0* HRM 13 500 512 25.4 26 2.4* HRM 14 780 764 24.7 4 22.1 * Significant bias at the 95% confidence level.† Provisional value. 422 Analyst, May 1997, Vol. 122Mean of duplicate results Difference between results 1 10 100 1000 100 10 1 0.1 99% 90% 50% s2 meas = s2 s + s2 a (1) (3) The comparison of these ‘measurement’ variances with the ‘true’ variability of the element in the samples excluding these errors, i.e., geochemical variance (s2 g). If the sources of variation are independent, the total variance (s2 total) is given by eqn (2). The three-component variances can be estimated using ANOVA s2 total = s2 g + s2 s + s2 a (2) The acceptable proportions of the measurement and analytical variances can be approximately related to the total variance as follows: 1% s2 total < s2 meas < 20% s2 total (3) 1% s2 meas < s2 a < 20% s2 meas (4) If the measurement variance contributes more than 20% of the total variance, then significant geochemical information will become progressively more obscured.In the case where the analytical variance contributes more than 20% to the measurement variance, analysis dominates the measurement variance.The analytical precision can then usefully be improved. If, however, the measurement variance is below 1% of the total variance, any reduction will not be cost-effective. This is similarly the case if the analytical variance is below 1% of the measurement variance. These limits are set, therefore, in a proportion to the geochemical variability rather than an arbitrary fixed precision as applied above (e.g., 10%).Analysis of variance, as a statistical technique, can be used to separate the three sources of variance and give sg, ss and sa which are estimates of sg, ss and sa, respectively. The advantage of variance as a measure of random error (i.e., precision) is that variances (s2) are additive whereas standard deviations (s) are not.7 In this work, SAX was applied to the determination of Al in soybean leaves. Estimated mean and standard deviations by using classical ANOVA of all the determinations of Al and robust ANOVA from the SAX data are both shown in Table 3.The analytical data quality was assessed as being generally satisfactory, as discussed above. The classical ANOVA of the soybean leaves suggested a mean Al concentration of 94.4 mg g21 with a total standard deviation (stotal) of 131.5 mg g21. However, the results given by the classical ANOVA may be erroneous because of four factors. These are: (1) the random errors were not normally distributed; (2) few samples had very high levels of Al; (3) there were a few poor duplicate analyses; and (4) the analytical error was not homogeneous (i.e., sa increases with concentration).These observations are inconsistent with the assumptions on which the classical ANOVA depends.7 In practice, it is difficult to make geochemical measurements that conform fully to the assumptions. For example, it is well established that analytical variance does vary as a function of concentration.This invalidates the assumption of homogeneity, for most geochemical measurements, although the application of the classical ANOVA to geochemical data has been widely reported.7 Robust ANOVA, however, has been shown to give more accurate estimates of variance in simulation studies.9,10 Robust statistics rely on the accommodation of outlying values rather than their rejection, and have been shown to be particularly appropriate for the description of analytical data.8,9 Briefly, the robust estimates of the mean and standard deviation are calculated by an iterative process, in which outlying observations are assigned reduced weights when they occur beyond a certain distance (e.g., 1.5s) from the estimated mean.Robust ANOVA should, therefore, provide more accurate estimates of the component variances.8 An important objective in this application of ANOVA was to check that the measurement variance originates in the processes of measurement (sampling and analysis).In this application the measurement variance contributed 20.7% to the total variance, which was approximately equal to the limit of 20%. Reliable interpretation of the geochemistry can be made, therefore, in the interpretation of the differences between the Al uptake by the soybean plants in the various soil types. The analytical method only contributed 0.1% to the total variance, which indicates that the sampling variance was the limiting factor for this application.This domination of the measurement variance by the sampling variance is probably caused by a few samples that were contaminated by soil/dust particles,1 in which the concentrations of Al were very high. Expressed in a more traditional way, the estimated analytical standard deviation (sa) was 2.2% of the overall mean, which is less than the usual arbitrary target of 5% (1sa) or 10% (2sa). Furthermore, the analytical variance contributed only 0.5% to the measurement variance. It can be concluded that the analytical precision was fit-for-purpose and better than required.Conclusions Analysts often expect the highest accuracy and greatest precision for chemical analysis. For certain applications, this is, Fig. 1 Ten per cent precision control chart for the determination of Al in soybean leaves. Table 3 SAX results for concentrations of Al in soybean leaves Standard deviation/mg g21 Mean/ Analysis Sampling Measurement Geochemical Total Method mg g21 (sa) (ss ) (sm ) (sg ) (stotal) Classical 94.4 1.92 31.99 32.05 127.51 131.5 Robust 46.9 1.03 14.38 14.42 28.24 31.7 Analyst, May 1997, Vol. 122 423no doubt, very important.For environmental analysis, however, analytical precision needs to be related to an estimate of the variance caused by the sampling process. A Sampling and Analytical Quality Control Scheme (SAX) extends traditional analytical control into the field environment to measure and also to provide targets for both analytical and sampling precision. This method utilizes replicate samples that are further duplicated when they are analysed, in addition to the other quality control materials, such as reference materials and reagent blanks. The case study described here showed that the systematic errors during the analytical procedure were acceptable with recoveries of Al of > 94% measured in reference materials after subtracting background values measured in reagent blanks. The application of SAX showed that the measurement variance contributed 20.7% to the total variance, which was approximately equal to the suggested limit of 20%. In this case, a reliable interpretation of the geochemistry can be made, for example, that there are the differences between the Al uptake by the soybean plants in the various soil types. The analytical method only contributed 0.1% to the total variance, which indicates that the sampling variance was the limiting factor for this application. This result suggests that the estimation of errors in the sampling process is very important in the chemical analysis of environmental materials. References 1 Dong, D., PhD Thesis, University of London, 1993. 2 Ramsey, M. H., Dong, D., Thornton, I., Watt, J., and Giddens, R., Environ. Geochem. Health, 1991, 13, 114. 3 Delves, H. T., Suchak, B., and Fellows, C. S., in Aluminium in Food and the Environment, ed. Massey, R. C., and Taylor, D., Royal Society of Chemistry, London, 1989, pp. 52–67. 4 Skelly, E. M., and Distefano, F. T., Appl. Spectrosc., 1988, 42, 1302. 5 Dong, D., Ramsey, M. H., and Thornton, I., J. Geochem. Explor., 1995, 55, 223. 6 Thompson, M., and Howarth, R. J., Analyst, 1976, 101, 690. 7 Ramsey, M. H., Thomopson, M., and Hale, M., J. Geochem. Explor. 1992, 44, 23. 8 Ramsey, M. H., Appl. Geochem., 1993, Suppl. Issue, 2, 149. 9 Analytical Methods Committee, Analyst, 1989, 114, 1693. 10 Analytical Methods Committee, Analyst, 1989, 114, 1699. Paper 6/07636J Received November 11, 1996 Accepted February 3, 1997 424 Analyst, May 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a607636j

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Solid-phase Extraction of Phenols and Pesticides in Water With a Modified Polymeric Resin N. Masqu�e , M. Gali`a, R. M. Marc�e and F. Borrull Departament de Qu�ýmica Anal�ýtica i Qu�ýmica Org`anica, Universitat Rovira i Virgili, Pça. Imperial Tarraco 1, 43005 Tarragona, Spain A comparative study of a chemically modified polymeric resin which has a benzoyl group and several commercial sorbents, such as PLRP-S, Envi-chrom P and LIChrolut EN, for the solid-phase extraction of various phenolic compounds and pesticides was carried out.Selectivity and breakthrough volumes for these compounds with different sorbents were studied by coupling an on-line solid-phase extraction system with a modified liquid chromatograph equipped with a UV detector. After determining the chromatographic conditions, linearity range and detection limits, the method was applied to the determination of these compounds in drinking and surface water. For 25 ml of sample, the recovery was > 70% for all the compounds, except for 4-nitrophenol (62%), and detection limits were between 0.2 and 0.8 mg l21.Keywords: Modified polymeric resin; solid-phase extraction; phenols; pesticides; water Phenolic compounds and pesticides are two very important groups of compounds which are of interest for the environment and their determination is receiving increasing attention because of their toxicity. Phenolic compounds are usually determined by reversed-phase liquid chromatography (RPLC) using different detection systems such as UV (diode-array detection),1–4 electrochemical5,6 or fluorescence.7,8 On the other hand, pesticides are usually determined by GC or RPLC, with a variety of detection systems.9–13 Unfortunately, the use of these detectors usually does not enable the detection limits required by legislation to be reached and hence it is necessary for the sample to be enriched prior to the chromatographic analysis.In recent years, solid-phase extraction (SPE) has become a very important technique for sample preparation in the environmental field, because of its advantages over liquid–liquid extraction.14,15 Two methods, on-line and off-line SPE, have been applied to determine phenolic compounds and pesticides in environmental waters. 12,16 The advantages of on-line trace enrichment procedures are better sensitivity, lower consumption of organic solvents, higher automation potential and simplicity of the analysis compared with off-line procedures.17,18 Several types of sorbents have been developed, such as C18 and C8 bonded to silica for apolar compounds, and carbon black and polymeric resins for more polar compounds.19 While most of the 11 phenolic compounds controlled by the EPA and some pesticides have high breakthrough volumes for commonly used sorbents, phenol, some nitrophenols, oxamyl and methomyl have low breakthrough volumes,9,20 except for some highly cross-linked styrene–divinylbenzenes (Envi-chrom P and LIChrolut EN)21,22 or carbon material.23 Different sorbents, mainly based on the use of highly crosslinked copolymers or chemically modified polymeric resins, have been developed.7,24,25 However, using sorbents in the precolumn which have very different characteristics from the stationary phase in the analytical column (such as a highly cross-linked sorbent and a C18 column, respectively) may lead to significant peak broadening. This can be solved by eluting the compounds retained on the pre-column with only the organic solvent of the mobile phase.18,23 In this paper, a chemically modified polymeric sorbent with a benzoyl group was synthesized and tested for the on-line SPE of some pesticides and phenolic compounds in order to obtain better breakthrough volumes for the determination of polar compounds in environmental waters.The results obtained with the synthesized sorbent, viz., breakthrough volumes, capacity and selectivity, were compared with those obtained for different commercial polymeric sorbents.Experimental Equipment The chromatographic experiments were performed with two Shimadzu (Tokyo, Japan) LC-10AD pumps, with a Shimadzu SPD-10A UV spectrophotometric detector. The temperature of the column was controlled by a Shimadzu CTO-10A oven and the chromatographic data were collected and recorded using a Hewlett-Packard (Avondale, PA, USA) HP-3365 Series II Chemstation which was controlled by Windows 3.1 (Microsoft). The analytical column was a 250 3 4 mm id stainlesssteel column packed with Spherisorb ODS 2, 5 mm (Teknokroma, Barcelona, Spain).To check the response of the instrument, standard solutions were injected through a Rheodyne (Cotati, CA, USA) valve with a 20 ml loop, and an automatic Must column-switching device (Spark Holland, Emmen, The Netherlands) was used with on-line SPE. The on-line trace enrichment process was carried out using steel pre-columns of 10 33 mm id laboratorypacked with the different sorbents studied.A Waters (Milford, MA, USA) M45 pump was used to deliver the sample. Reagents and Standards Various phenolic compounds and pesticides were studied. The phenolic compounds were: phenol (Ph), 4-nitrophenol (4-NP), 2,4-dinitrophenol (2,4-DNP) and 2-chlorophenol (2-CP); they were obtained from Aldrich-Chemie (Steinheim, Germany). The pesticides studied included triazines (simazine and atrazine), carbamates (methomyl and oxamyl), (4-chloro-2-methylphenoxy) acetic acid (MCPA) and bentazone.Except for bentazone, which was obtained from Dr. Ehrenstorfer (Augsburg, Germany), the remainder of the compounds were from Riedel-de Ha�en (Seelze, Germany). Standard solutions of 2000 mg l21 of each compound were prepared in methanol. A mixture of all compounds used was prepared weekly by diluting the standard solutions with Milli-Q water (Millipore, Bedford, MA, USA), and more diluted working solutions were prepared daily by diluting with Milli-Q or river water.All solutions were stored at 4 °C in a refrigerator. HPLC-gradient-grade methanol (Scharlau, Barcelona, Spain) and Milli-Q water were used to prepare the mobile phase. Hydrochloric acid (Probus, Badalona, Spain) was used to adjust Analyst, May 1997, Vol. 122 (425–428) 425the pH of the mobile phase and the sample before SPE to 3 and 2.5, respectively. Synthetic Procedure The chemically modified resin was obtained from porous crosslinked polystyrene–divinylbenzene (PS–DVB) beads.Amberchrome GC-161 (Tosohaas, Montgomeryville, PA, USA) is a spherical resin with an average particle size of 50–100 mm and an average pore size of 110–175 Å. The benzoyl derivative was prepared as follows: a portion (2.5 g) of Amberchrome GC-161 was stirred in 15 ml of nitrobenzene at room temperature for 12 h. Aluminium chloride (4.3 g) was added slowly and under mechanical stirring and the reaction mixture was cooled to 0 °C. Then 4.9 g of benzoyl chloride were added dropwise and the reaction mixture was stirred for 3 h.The reaction mixture was quenched by adding acetone with 1% hydrochloric acid, and excess liquid was decanted. The polymer was washed twice with methanol and dried under vacuum at 60 °C. The recovery of the beads was essentially quantitative. The polymer was characterized by IR spectrometry and the extent of the modification (60%) was established from the elemental analyses. Chromatographic Conditions The eluents for the chromatographic separation were Milli-Q water adjusted to pH 3 with sulfuric acid (solvent A) and methanol (solvent B).The flow rate was 1 ml min21 and the temperature of the column oven was set at 65 °C. The gradient profile was 20% solvent B initially, 50% solvent B after 25 min and 100% solvent B after 28 min (held for 2 min), after which the mobile phase returned to the initial conditions in 2 min.The detection was carried out at different wavelengths (at 240 nm for oxamyl, methomyl, bentazone, simazine and atrazine; at 280 nm for Ph, 4-NP, 2,4-DNP and 2-CP; and at 230 nm for MCPA) with a programme of wavelengths: first at 240 nm, then at 280 nm after 7 min, at 240 nm again at 13.5 min, then at 230 nm after 21.2 minutes and finally at 240 nm after 23 min. On-line Trace E-line trace enrichment was carried out with several polymeric sorbents so as to compare them with the sorbent synthesized here.The sorbents chosen were PLRP-S of 100 Å and 20 mm (Polymer Laboratories, Amherst, MA, USA), Envi-chrom P (Supelco, Bellefonte, PA, USA) and LIChrolut EN of 40–120 mm (Merck, Darmstadt, Germany). Prior to the preconcentration step, the pH of the samples was adjusted to 2.5 with hydrochloric acid. A modification of the common elution design was used to desorb the analytes from the sorbents.18 A Must column-switching device with two switching valves was used to clean the tubes, activate the pre-column and control the sample volume to be preconcentrated.Before use, all pre-columns were conditioned by flushing with methanol for 1 min at 2 ml min21 and activated with 2 ml of Milli-Q water adjusted to pH 2.5 with hydrochloric acid. Then, different sample volumes were preconcentrated at 4 ml min21. In the next step, the analytes trapped on the pre-column were desorbed in the back-flush mode only by the organic solvent (methanol) of the mobile phase so as to prevent broadening of peaks due to the different nature of the sorbent on the precolumn and the analytical column.This enables sorbents with high retention for the compounds to be used. Real samples were filtered through 0.45 mm nylon membranes (Supelco) before preconcentration. Results and Discussion This study discusses a sorbent obtained by chemically modifying a PS–DVB resin, which can be used in the SPE of some phenolic compounds and pesticides.The electrophilic substitution of the PS–DVB resin with benzoyl chloride was carried out by the Friedel–Crafts reaction. Nitrobenzene was used as solvent in the presence of aluminium chloride as catalyst at low temperatures. In order to prevent the beads from degrading, suitable stirring conditions were established. The polymer was characterized by IR spectrometry and the presence of the modified material was proved by the appearance of a band assignable to the carbonyl group (1670 cm21 ).Several reaction times (3, 5, 7 and 24 h) were tested in order to study the length of the reaction, and similar degrees of substitution (60%) were obtained. Hence, it was concluded that the reaction time does not affect the degree of modification. After the sorbent had been synthesized and characterized, it was initially tested for the on-line SPE of Ph, which is a polar compound with a low breakthrough volume for most of the commercial sorbents.6,9,12 The different breakthrough curves for Ph obtained for the synthesized sorbent and the commercial sorbents (PLRP-S, Envi-chrom P and LIChrolut EN) are shown in Fig. 1. To obtain these curves, a standard solution of 10 mg l21 of Ph dissolved in Milli-Q water at pH 2.5 (with hydrochloric acid) was directly introduced into the detector, by-passing the Rheodyne valve with the pre-column; when a stable response was obtained, the Rheodyne valve was moved so that the sample was passed through the precolumn at 1 ml min21.The signal was measured with a UV detector at 280 nm. If the breakthrough volume is considered as the volume at which the detector signal reaches 10% of the total signal, the breakthrough volumes obtained for PLRP-S, Envi-Chrom P, the synthesized sorbent and LIChrolut EN were 4, 7, 13 and 30 ml, respectively. Fig. 1 confirms that the synthesized sorbent is better than PLRP-S and Envi-chrom P but no better than LIChrolut EN for Ph recovery.A gradient separation was optimized to separate the ten compounds studied. The chromatogram obtained under the optimum conditions is shown in Fig. 2. A programme of wavelengths in the UV spectrophotometric detector enabled each compound to be detected at its maximum absorbance. Good linearity of response by direct injection of between 0.25–0.1 and 40 mg l21 was obtained for all compounds with good regression coefficients (R2) ranging from 0.9993 for atrazine to 0.9999 for simazine.Detection limits26 attained ranged from 10 mg l21 for oxamyl, methomyl, 2,4-DNP and simazine to 50 mg l21 for Ph, 2-CP and MCPA. In order to calculate the recoveries for each compound, different sample volumes (10, 25, 50 and 100 ml) of the analyte mixtures studied were preconcentrated on the synthesized sorbent. In Table 1 the recoveries of these compounds are shown. Good recoveries were obtained for all compounds when Fig. 1 Breakthrough curves for a standard solution of 10 mg l21 of Ph, adjusted to pH 2.5, obtained for the different sorbents used. 426 Analyst, May 1997, Vol. 122100 ml of a standard solution at the 20 mg l21 level were analysed with the synthesized sorbent. Only Ph has a recovery near 40%; the remainder of the compounds have values higher than 70% with RSDs lower than 3% for three replicate analyses. In this study, to obtain a good recovery of Ph, 25 ml of sample were selected for further studies. A comparative study of the different sorbents, preconcentrating 25 ml of a standard solution spiked at 8 mg l21 adjusted to pH 2.5 with HCl, was carried out.Table 2 shows the recoveries found. The values obtained for Ph with the synthesized sorbent are better than those obtained using commercial sorbents such as PLRP-S or Envi-chrom P, and similar to those obtained with LIChrolut EN. For the other compounds similar results were obtained for all sorbents when this volume of sample was analysed. When 100 ml of sample spiked at the 2 mg l21 level were preconcentrated on PLRP-S, the recoveries of oxamyl and methomyl were lower (23 and 15%, respectively) and Ph was eluted; however, when the same volume was preconcentrated on the synthesized sorbent, the recoveries of oxamyl and methomyl were 77 and 70%, respectively, and Ph had a recovery of nearly 40%.Table 3 shows that the synthesized sorbent is suitable for determining these compounds. When Ebro river water was analysed, similar recoveries were obtained.The results obtained for oxamyl and methomyl are higher than those obtained for carbon sorbents which are recommended for polar compounds.23 Linearity of the response for the total analytical system, including the preconcentration step, was checked for a sample volume of 25 ml of Ebro river water spiked at different concentrations. Good linearity was obtained from 0.8–4 to 50 mg l21. R2 values were between 0.9996 and 0.9999 and the detection limits26 were between 0.2 and 0.8 mg l21.The results obtained are shown in Table 4. In order to compare the selectivity of the synthesized sorbent with that of the other commercially available sorbents which are recommended for the determination of selected compounds, 25 ml of Ebro river water spiked with 8 mg l21 of each compound were analysed with the different sorbents chosen. With LIChrolut EN, additional peaks appeared and the peak at 14 min Fig. 2 Chromatogram corresponding to a standard solution of 10 mg l21 of each compound studied under optimum conditions. 1, Oxamyl; 2, methomyl; 3, Ph; 4, 4-NP; 5, 2,4-DNP; 6, 2-CP; 7, bentazone; 8, simazine; 9, MCPA; and 10, atrazine. Table 1 Recoveries obtained with the synthesized sorbent using different volumes of 20 mg l 21 standard solution. Values are the means of three determinations and are expressed as a percentage. For conditions, see text Recovery (%)* Compound 10 ml 25 ml 50 ml 100 ml Oxamyl 84 80 77 77 Methomyl 82 76 73 70 Ph 91 70 58 40 4-NP 81 78 78 78 2,4-DNP 81 81 80 85 2-CP 80 72 75 79 Bentazone 85 85 82 88 Simazine 80 80 79 83 MCPA 80 73 73 78 Atrazine 86 76 75 79 * % RSDs are lower than 9% in all instances.Table 2 Recoveries obtained with PLRP-S , Envi-chrom P, the synthesized sorbent and LIChrolut EN using 25 ml of standard solution spiked at the 8 mg l21 level. Values are the means of three determinations and are expressed as a percentage. For conditions, see text Recovery (%)* Envi-chrom Synthesized Lichrolut Compound PLRP-S P sorbent EN Oxamyl 75 65 74 70 Methomyl 68 64 72 74 Ph 33 47 71 78 4-NP 73 71 62 81 2,4-DNP 77 73 72 77 2-CP 75 71 80 83 Bentazone 86 78 79 83 Simazine 73 77 76 84 MCPA 81 79 74 83 Atrazine 77 75 77 83 * % RSDs are lower than 10% in all instances. Table 3 Mean recoveries and RSDs (n = 3) of SPE with the synthesized sorbent and PLRP-S for 100 ml of standard solution spiked with 2 mg l21 of each compound in Milli-Q and Ebro river water.All values in % Synthesized sorbent PLRP-S PLRP-S (Milli-Q water) (Milli-Q water) (River water) Re- Re- Re- Compound covery RSD covery RSD covery RSD Oxamyl 77 2 23 3 21 5 Methomyl 69 2 15 1 16 2 Ph 40 3 — — — — 4-NP 78 2 32 4 40 8 2,4-DNP 85 2 79 4 81 3 2-CP 79 3 71 1 69 1 Bentazone 88 1 86 6 80 1 Simazine 83 3 65 4 57 8 MCPA 78 1 75 1 89 2 Atrazine 78 3 75 3 80 1 Table 4 Study of the linearity range and detection limits of the method Linearity range/ Detection limit/ Compound mg l21 mg l21 Oxamyl 0.8–50 0.2 Methomyl 1–50 0.4 Ph 4–50 0.8 4-NP 1–50 0.4 2,4-DNP 0.8–50 0.2 2-CP 4–50 0.8 Bentazone 0.8–50 0.2 Simazine 0.8–50 0.2 MCPA 1–50 0.4 Atrazine 0.8–50 0.2 Analyst, May 1997, Vol. 122 427was larger with this sorbent than with the synthesized sorbent. Fig. 3 shows the chromatograms of 25 ml of sample and the same sample spiked with a standard solution of 4 mg l21 of each compound in tap and Ebro river water (3c and 3d, and 3a and 3b, respectively).In the river water chromatograms, four peaks with the same retention times as bentazone, oxamyl, methomyl and Ph appear; however, for the first three compounds, the signal was near the detection limits, whereas for bentazone, a concentration of 1.5 mg l21 was found. This pesticide has already been found in samples of the same origin in our laboratory using MS the detection technique in the chromatographic system. Conclusions It has been shown that the synthesized sorbent has higher recoveries for the determination of some phenolic compounds and pesticides in surface and tap water than other commercially available sorbents such as PLRP-S or Envi-chrom P.Compared with these commercial sorbents, the synthesized sorbent gives better results for oxamyl, methomyl and Ph, and similar results for the less polar compounds. The matrix effect was similar to that obtained when Envi-chrom P or LIChrolut EN was used.The authors thank the Direcci�o General de Recerca de la Generalitat de Catalunya for supporting this study. References 1 Elvira-Cozal, C., Cano-Faura, P., P�erez-Arribas, L.V., Le�on-Gonz`alez, M. E., and Polo-Diez, L. M., Chromatographia, 1995, 40, 91. 2 Ruana, J., Urbe, I., and Borrull, F., J. Chromatogr. A, 1993, 655, 217. 3 Puig, D., and Barcel�o, D., Chromatographia, 1995, 40, 435. 4 Fr�ebortov�a, J., and Tatarkovicov�a, V., Analyst, 1994, 119, 1519. 5 Galcer�an, M. T., and J�auregui, O., Anal.Chim. Acta, 1995, 304, 75. 6 Pocurull, E., S�anchez, G., Borrull, F., and Marc�e, R. M., J. Chromatogr. A, 1995, 696, 31. 7 Masqu�e, N., Gali`a, M., Marc�e, R. M., and Borrull, F., J. Chromatogr. A, in the press. 8 Lamprecht, G., and Huber, J. F. K., J. Chromatogr. A, 1994, 667, 47. 9 Marc�e, R. M., Prosen, H., Crespo, C., Calull, M., Borrull, F., and Brinkman, U. A. Th., J. Chromatogr. A, 1995, 696, 63. 10 Bagheri, H., Slobodnik, J., Marc�e, R. M., Ghijsen, R.T., and Brinkman, U. A. Th., Chromatographia, 1993, 37, 159. 11 Molina, C., Honning, M., and Barcel�o, D., Anal. Chem., 1994, 66, 4444. 12 Aguilar, C., Borrull, F., and Marc�e, R. M., J. Chromatogr. A, 1996, 754, 77. 13 Liska, I., Brouwer, E. R., Ostheimer, A. G. L., Lingeman, H., and Brinkman, U. A. Th., Int. J. Environ. Anal. Chem., 1992, 47, 267. 14 Hennion, M. C., Trends Anal. Chem., 1991, 10, 317. 15 Font, G., Ma�nes, J., Molt�o, J. C., and Pic�o, Y., J. Chromatogr., 1993, 642, 135. 16 Brinkman, U. A. Th., J. Chromatogr. A, 1994, 665, 217. 17 Pocurull, E., Marc�e, R. M., and Borrull, F., Chromatographia, 1995, 41, 521. 18 Pocurull, E., Marc�e, R. M., and Borrull, F., J. Chromatogr. A, 1996, 738, 1. 19 Liska, I., Kuthan, A., and Krupik, J., J. Chromatogr., 1990, 509, 123. 20 Pocurull, E., Calull, M., Marc�e, R. M., and Borrull, F., Chromatographia, 1994, 38, 579. 21 Pocurull, E., Calull, M., Marc�e, R. M., and Borrull, F., J. Chromatogr. A, 1996, 719, 105. 22 Puig, D., and Barcel�o, D., J. Chromatogr. A., 1996, 733, 371. 23 Slobodnik, J., � Oztezkizan, � O., Lingeman, H., and Brinkman, U. A. Th., J. Chromatogr., 1996, 750, 227. 24 Sun, J. J., and Fritz, J. S., J. Chromatogr., 1992, 590, 197. 25 Tsyurupa, M. P., Llyin, M. M., Andreeva, A. I., and Davankov, V. A., Fresenius’ J. Anal. Chem., 1995, 352, 672. 26 Boqu�e, R., and Rius, X., J. Chem. Educ., 1993, 70, 230. Paper 6/07504E Received November 4, 1996 Accepted February 6, 1997 Fig. 3 Chromatograms obtained by on-line trace enrichment with the synthesized sorbent of a 25 ml sample. (a) Ebro river water, (b) Ebro river water spiked with 4 mg l21 of each compound, (c) tap water and (d) tap water spiked with 4 mg l21 of each compound. For peak designation, see Fig. 2. 428 Analyst, May 1997, Vol. 122 Solid-phase Extraction of Phenols and Pesticides in Water With a Modified Polymeric Resin N. Masqu�e , M. Gali`a, R. M. Marc�e and F. Borrull Departament de Qu�ýmica Anal�ýtica i Qu�ýmica Org`anica, Universitat Rovira i Virgili, Pça.Imperial Tarraco 1, 43005 Tarragona, Spain A comparative study of a chemically modified polymeric resin which has a benzoyl group and several commercial sorbents, such as PLRP-S, Envi-chrom P and LIChrolut EN, for the solid-phase extraction of various phenolic compounds and pesticides was carried out. Selectivity and breakthrough volumes for these compounds with different sorbents were studied by coupling an on-line solid-phase extraction system with a modified liquid chromatograph equipped with a UV detector.After determining the chromatographic conditions, linearity range and detection limits, the method was applied to the determination of these compounds in drinking and surface water. For 25 ml of sample, the recovery was > 70% for all the compounds, except for 4-nitrophenol (62%), and detection limits were between 0.2 and 0.8 mg l21. Keywords: Modified polymeric resin; solid-phase extraction; phenols; pesticides; water Phenolic compounds and pesticides are two very important groups of compounds which are of interest for the environment and their determination is receiving increasing attention because of their toxicity.Phenolic compounds are usually determined by reversed-phase liquid chromatography (RPLC) using different detection systems such as UV (diode-array detection),1–4 electrochemical5,6 or fluorescence.7,8 On the other hand, pesticides are usually determined by GC or RPLC, with a variety of detection systems.9–13 Unfortunately, the use of these detectors usually does not enable the detection limits required by legislation to be reached and hence it is necessary for the sample to be enriched prior to the chromatographic analysis.In recent years, solid-phase extraction (SPE) has become a very important technique for sample preparation in the environmental field, because of its advantages over liquid–liquid extraction.14,15 Two methods, on-line and off-line SPE, have been applied to determine phenolic compounds and pesticides in environmental waters. 12,16 The advantages of on-line trace enrichment procedures are better sensitivity, lower consumption of organic solvents, higher automation potential and simplicity of the analysis compared with off-line procedures.17,18 Several types of sorbents have been developed, such as C18 and C8 bonded to silica for apolar compounds, and carbon black and polymeric resins for more polar compounds.19 While most of the 11 phenolic compounds controlled by the EPA and some pesticides have high breakthrough volumes for commonly used sorbents, phenol, some nitrophenols, oxamyl and methomyl have low breakthrough volumes,9,20 except for some highly cross-linked styrene–divinylbenzenes (Envi-chrom P and LIChrolut EN)21,22 or carbon material.23 Different sorbents, mainly based on the use of highly crosslinked copolymers or chemically modified polymeric resins, have been developed.7,24,25 However, using sorbents in the precolumn which have very different characteristics from the stationaorbent and a C18 column, respectively) may lead to significant peak broadening.This can be solved by eluting the compounds retained on the pre-column with only the organic solvent of the mobile phase.18,23 In this paper, a chemically modified polymeric sorbent with a benzoyl group was synthesized and tested for the on-line SPE of some pesticides and phenolic compounds in order to obtain better breakthrough volumes for the determination of polar compounds in environmental waters.The results obtained with the synthesized sorbent, viz., breakthrough volumes, capacity and selectivity, were compared with those obtained for different commercial polymeric sorbents. Experimental Equipment The chromatographic experiments were performed with two Shimadzu (Tokyo, Japan) LC-10AD pumps, with a Shimadzu SPD-10A UV spectrophotometric detector. The temperature of the column was controlled by a Shimadzu CTO-10A oven and the chromatographic data were collected and recorded using a Hewlett-Packard (Avondale, PA, USA) HP-3365 Series II Chemstation which was controlled by Windows 3.1 (Microsoft).The analytical column was a 250 3 4 mm id stainlesssteel column packed with Spherisorb ODS 2, 5 mm (Teknokroma, Barcelona, Spain).To check the response of the instrument, standard solutions were injected through a Rheodyne (Cotati, CA, USA) valve with a 20 ml loop, and an automatic Must column-switching device (Spark Holland, Emmen, The Netherlands) was used with on-line SPE. The on-line trace enrichment process was carried out using steel pre-columns of 10 33 mm id laboratorypacked with the different sorbents studied. A Waters (Milford, MA, USA) M45 pump was used to deliver the sample.Reagents and Standards Various phenolic compounds and pesticides were studied. The phenolic compounds were: phenol (Ph), 4-nitrophenol (4-NP), 2,4-dinitrophenol (2,4-DNP) and 2-chlorophenol (2-CP); they were obtained from Aldrich-Chemie (Steinheim, Germany). The pesticides studied included triazines (simazine and atrazine), carbamates (methomyl and oxamyl), (4-chloro-2-methylphenoxy) acetic acid (MCPA) and bentazone. Except for bentazone, which was obtained from Dr.Ehrenstorfer (Augsburg, Germany), the remainder of the compounds were from Riedel-de Ha�en (Seelze, Germany). Standard solutions of 2000 mg l21 of each compound were prepared in methanol. A mixture of all compounds used was prepared weekly by diluting the standard solutions with Milli-Q water (Millipore, Bedford, MA, USA), and more diluted working solutions were prepared daily by diluting with Milli-Q or river water. All solutions were stored at 4 °C in a refrigerator. HPLC-gradient-grade methanol (Scharlau, Barcelona, Spain) and Milli-Q water were used to prepare the mobile phase.Hydrochloric acid (Probus, Badalona, Spain) was used to adjust Analyst, May 1997, Vol. 122 (425–428) 425the pH of the mobile phase and the sample before SPE to 3 and 2.5, respectively. Synthetic Procedure The chemically modified resin was obtained from porous crosslinked polystyrene–divinylbenzene (PS–DVB) beads. Amberchrome GC-161 (Tosohaas, Montgomeryville, PA, USA) is a spherical resin with an average particle size of 50–100 mm and an average pore size of 110–175 Å.The benzoyl derivative was prepared as follows: a portion (2.5 g) of Amberchrome GC-161 was stirred in 15 ml of nitrobenzene at room temperature for 12 h. Aluminium chloride (4.3 g) was added slowly and under mechanical stirring and the reaction mixture was cooled to 0 °C. Then 4.9 g of benzoyl chloride were added dropwise and the reaction mixture was stirred for 3 h.The reaction mixture was quenched by adding acetone with 1% hydrochloric acid, and excess liquid was decanted. The polymer was washed twice with methanol and dried under vacuum at 60 °C. The recovery of the beads was essentially quantitative. The polymer was characterized by IR spectrometry and the extent of the modification (60%) was established from the elemental analyses. Chromatographic Conditions The eluents for the chromatographic separation were Milli-Q water adjusted to pH 3 with sulfuric acid (solvent A) and methanol (solvent B).The flow rate was 1 ml min21 and the temperature of the column oven was set at 65 °C. The gradient profile was 20% solvent B initially, 50% solvent B after 25 min and 100% solvent B after 28 min (held for 2 min), after which the mobile phase returned to the initial conditions in 2 min. The detection was carried out at different wavelengths (at 240 nm for oxamyl, methomyl, bentazone, simazine and atrazine; at 280 nm for Ph, 4-NP, 2,4-DNP and 2-CP; and at 230 nm for MCPA) with a programme of wavelengths: first at 240 nm, then at 280 nm after 7 min, at 240 nm again at 13.5 min, then at 230 nm after 21.2 minutes and finally at 240 nm after 23 min.On-line Trace Enrichment On-line trace enrichment was carried out with several polymeric sorbents so as to compare them with the sorbent synthesized here. The sorbents chosen were PLRP-S of 100 Å and 20 mm (Polymer Laboratories, Amherst, MA, USA), Envi-chrom P (Supelco, Bellefonte, PA, USA) and LIChrolut EN of 40–120 mm (Merck, Darmstadt, Germany).Prior to the preconcentration step, the pH of the samples was adjusted to 2.5 with hydrochloric acid. A modification of the common elution design was used to desorb the analytes from the sorbents.18 A Must column-switching device with two switching valves was used to clean the tubes, activate the pre-column and control the sample volume to be preconcentrated. Before use, all pre-columns were conditioned by flushing with methanol for 1 min at 2 ml min21 and activated with 2 ml of Milli-Q water adjusted to pH 2.5 with hydrochloric acid.Then, different sample volumes were preconcentrated at 4 ml min21. In the next step, the analytes trapped on the pre-column were desorbed in the back-flush mode only by the organic solvent (methanol) of the mobile phase so as to prevent broadening of peaks due to the different nature of the sorbent on the precolumn and the analytical column.This enables sorbents with high retention for the compounds to be used. Real samples were filtered through 0.45 mm nylon membranes (Supelco) before preconcentration. Results and Discussion This study discusses a sorbent obtained by chemically modifying a PS–DVB resin, which can be used in the SPE of some phenolic compounds and pesticides. The electrophilic substitution of the PS–DVB resin with benzoyl chloride was carried out by the Friedel–Crafts reaction.Nitrobenzene was used as solvent in the presence of aluminium chloride as catalyst at low temperatures. In order to prevent the beads from degrading, suitable stirring conditions were established. The polymer was characterized by IR spectrometry and the presence of the modified material was proved by the appearance of a band assignable to the carbonyl group (1670 cm21 ). Several reaction times (3, 5, 7 and 24 h) were tested in order to study the length of the reaction, and similar degrees of substitution (60%) were obtained.Hence, it was concluded that the reaction time does not affect the degree of modification. After the sorbent had been synthesized and characterized, it was initially tested for the on-line SPE of Ph, which is a polar compound with a low breakthrough volume for most of the commercial sorbents.6,9,12 The different breakthrough curves for Ph obtained for the synthesized sorbent and the commercial sorbents (PLRP-S, Envi-chrom P and LIChrolut EN) are shown in Fig. 1. To obtain these curves, a standard solution of 10 mg l21 of Ph dissolved in Milli-Q water at pH 2.5 (with hydrochloric acid) was directly introduced into the detector, by-passing the Rheodyne valve with the pre-column; when a stable response was obtained, the Rheodyne valve was moved so that the sample was passed through the precolumn at 1 ml min21. The signal was measured with a UV detector at 280 nm. If the breakthrough volume is considered as the volume at which the detector signal reaches 10% of the total signal, the breakthrough volumes obtained for PLRP-S, Envi-Chrom P, the synthesized sorbent and LIChrolut EN were 4, 7, 1and 30 ml, respectively.Fig. 1 confirms that the synthesized sorbent is better than PLRP-S and Envi-chrom P but no better than LIChrolut EN for Ph recovery. A gradient separation was optimized to separate the ten compounds studied. The chromatogram obtained under the optimum conditions is shown in Fig. 2. A programme of wavelengths in the UV spectrophotometric detector enabled each compound to be detected at its maximum absorbance. Good linearity of response by direct injection of between 0.25–0.1 and 40 mg l21 was obtained for all compounds with good regression coefficients (R2) ranging from 0.9993 for atrazine to 0.9999 for simazine. Detection limits26 attained ranged from 10 mg l21 for oxamyl, methomyl, 2,4-DNP and simazine to 50 mg l21 for Ph, 2-CP and MCPA.In order to calculate the recoveries for each compound, different sample volumes (10, 25, 50 and 100 ml) of the analyte mixtures studied were preconcentrated on the synthesized sorbent. In Table 1 the recoveries of these compounds are shown. Good recoveries were obtained for all compounds when Fig. 1 Breakthrough curves for a standard solution of 10 mg l21 of Ph, adjusted to pH 2.5, obtained for the different sorbents used. 426 Analyst, May 1997, Vol. 122100 ml of a standard solution at the 20 mg l21 level were analysed with the synthesized sorbent.Only Ph has a recovery near 40%; the remainder of the compounds have values higher than 70% with RSDs lower than 3% for three replicate analyses. In this study, to obtain a good recovery of Ph, 25 ml of sample were selected for further studies. A comparative study of the different sorbents, preconcentrating 25 ml of a standard solution spiked at 8 mg l21 adjusted to pH 2.5 with HCl, was carried out.Table 2 shows the recoveries found. The values obtained for Ph with the synthesized sorbent are better than those obtained using commercial sorbents such as PLRP-S or Envi-chrom P, and similar to those obtained with LIChrolut EN. For the other compounds similar results were obtained for all sorbents when this volume of sample was analysed. When 100 ml of sample spiked at the 2 mg l21 level were preconcentrated on PLRP-S, the recoveries of oxamyl and methomyl were lower (23 and 15%, respectively) and Ph was eluted; however, when the same volume was preconcentrated on the synthesized sorbent, the recoveries of oxamyl and methomyl were 77 and 70%, respectively, and Ph had a recovery of nearly 40%.Table 3 shows that the synthesized sorbent is suitable for determining these compounds. When Ebro river water was analysed, similar recoveries were obtained. The results obtained for oxamyl and methomyl are higher than those obtained for carbon sorbents which are recommended for polar compounds.23 Linearity of the response for the total analytical system, including the preconcentration step, was checked for a sample volume of 25 ml of Ebro river water spiked at different concentrations. Good linearity was obtained from 0.8–4 to 50 mg l21.R2 values were between 0.9996 and 0.9999 and the detection limits26 were between 0.2 and 0.8 mg l21. The results obtained are shown in Table 4. In order to compare the selectivity of the synthesized sorbent with that of the other commercially available sorbents which are recommended for the determination of selected compounds, 25 ml of Ebro river water spiked with 8 mg l21 of each compound were analysed with the different sorbents chosen.With LIChrolut EN, additional peaks appeared and the peak at 14 min Fig. 2 Chromatogram corresponding to a standard solution of 10 mg l21 of each compound studied under optimum conditions. 1, Oxamyl; 2, methomyl; 3, Ph; 4, 4-NP; 5, 2,4-DNP; 6, 2-CP; 7, bentazone; 8, simazine; 9, MCPA; and 10, atrazine.Table 1 Recoveries obtained with the synthesized sorbent using different volumes of 20 mg l 21 standard solution. Values are the means of three determinations and are expressed as a percentage. For conditions, see text Recovery (%)* Compound 10 ml 25 ml 50 ml 100 ml Oxamyl 84 80 77 77 Methomyl 82 76 73 70 Ph 91 70 58 40 4-NP 81 78 78 78 2,4-DNP 81 81 80 85 2-CP 80 72 75 79 Bentazone 85 85 82 88 Simazine 80 80 79 83 MCPA 80 73 73 78 Atrazine 86 76 75 79 * % RSDs are lower than 9% in all instances.Table 2 Recoveries obtained with PLRP-S , Envi-chrom P, the synthesized sorbent and LIChrolut EN using 25 ml of standard solution spiked at the 8 mg l21 level. Values are the means of three determinations and are expressed as a percentage. For conditions, see text Recovery (%)* Envi-chrom Synthesized Lichrolut Compound PLRP-S P sorbent EN Oxamyl 75 65 74 70 Methomyl 68 64 72 74 Ph 33 47 71 78 4-NP 73 71 62 81 2,4-DNP 77 73 72 77 2-CP 75 71 80 83 Bentazone 86 78 79 83 Simazine 73 77 76 84 MCPA 81 79 74 83 Atrazine 77 75 77 83 * % RSDs are lower than 10% in all instances.Table 3 Mean recoveries and RSDs (n = 3) of SPE with the synthesized sorbent and PLRP-S for 100 ml of standard solution spiked with 2 mg l21 of each compound in Milli-Q and Ebro river water. All values in % Synthesized sorbent PLRP-S PLRP-S (Milli-Q water) (Milli-Q water) (River water) Re- Re- Re- Compound covery RSD covery RSD covery RSD Oxamyl 77 2 23 3 21 5 Methomyl 69 2 15 1 16 2 Ph 40 3 — — — — 4-NP 78 2 32 4 40 8 2,4-DNP 85 2 79 4 81 3 2-CP 79 3 71 1 69 1 Bentazone 88 1 86 6 80 1 Simazine 83 3 65 4 57 8 MCPA 78 1 75 1 89 2 Atrazine 78 3 75 3 80 1 Table 4 Study of the linearity range and detection limits of the method Linearity range/ Detection limit/ Compound mg l21 mg l21 Oxamyl 0.8–50 0.2 Methomyl 1–50 0.4 Ph 4–50 0.8 4-NP 1–50 0.4 2,4-DNP 0.8–50 0.2 2-CP 4–50 0.8 Bentazone 0.8–50 0.2 Simazine 0.8–50 0.2 MCPA 1–50 0.4 Atrazine 0.8–50 0.2 Analyst, May 1997, Vol. 122 427was larger with this sorbent than with the synthesized sorbent. Fig. 3 shows the chromatograms of 25 ml of sample and the same sample spiked with a standard solution of 4 mg l21 of each compound in tap and Ebro river water (3c and 3d, and 3a and 3b, respectively). In the river water chromatograms, four peaks with the same retention times as bentazone, oxamyl, methomyl and Ph appear; however, for the first three compounds, the signal was near the detection limits, whereas for bentazone, a concentration of 1.5 mg l21 was found.This pesticide has already been found in samples of the same origin in our laboratory using MS the detection technique in the chromatographic system. Conclusions It has been shown that the synthesized sorbent has higher recoveries for the determination of some phenolic compounds and pesticides in surface and tap water than other commercially available sorbents such as PLRP-S or Envi-chrom P.Compared with these commercial sorbents, the synthesized sorbent gives better results for oxamyl, methomyl and Ph, and similar results for the less polar compounds. The matrix effect was similar to that obtained when Envi-chrom P or LIChrolut EN was used. The authors thank the Direcci�o General de Recerca de la Generalitat de Catalunya for supporting this study. References 1 Elvira-Cozal, C., Cano-Faura, P., P�erez-Arribas, L.V., Le�on-Gonz`alez, M.E., and Polo-Diez, L. M., Chromatographia, 1995, 40, 91. 2 Ruana, J., Urbe, I., and Borrull, F., J. Chromatogr. A, 1993, 655, 217. 3 Puig, D., and Barcel�o, D., Chromatographia, 1995, 40, 435. 4 Fr�ebortov�a, J., and Tatarkovicov�a, V., Analyst, 1994, 119, 1519. 5 Galcer�an, M. T., and J�auregui, O., Anal. Chim. Acta, 1995, 304, 75. 6 Pocurull, E., S�anchez, G., Borrull, F., and Marc�e, R. M., J. Chromatogr. A, 1995, 696, 31. 7 Masqu�e, N., Gali`a, M., Marc�e, R. M., and Borrull, F., J. Chromatogr. A, in the press. 8 Lamprecht, G., and Huber, J. F. K., J. Chromatogr. A, 1994, 667, 47. 9 Marc�e, R. M., Prosen, H., Crespo, C., Calull, M., Borrull, F., and Brinkman, U. A. Th., J. Chromatogr. A, 1995, 696, 63. 10 Bagheri, H., Slobodnik, J., Marc�e, R. M., Ghijsen, R. T., and Brinkman, U. A. Th., Chromatographia, 1993, 37, 159. 11 Molina, C., Honning, M., and Barcel�o, D., Anal. Chem., 1994, 66, 4444. 12 Aguilar, C., Borrull, F., and Marc�e, R. M., J. Chromatogr. A, 1996, 754, 77. 13 Lis A. G. L., Lingeman, H., and Brinkman, U. A. Th., Int. J. Environ. Anal. Chem., 1992, 47, 267. 14 Hennion, M. C., Trends Anal. Chem., 1991, 10, 317. 15 Font, G., Ma�nes, J., Molt�o, J. C., and Pic�o, Y., J. Chromatogr., 1993, 642, 135. 16 Brinkman, U. A. Th., J. Chromatogr. A, 1994, 665, 217. 17 Pocurull, E., Marc�e, R. M., and Borrull, F., Chromatographia, 1995, 41, 521. 18 Pocurull, E., Marc�e, R. M., and Borrull, F., J. Chromatogr. A, 1996, 738, 1. 19 Liska, I., Kuthan, A., and Krupik, J., J. Chromatogr., 1990, 509, 123. 20 Pocurull, E., Calull, M., Marc�e, R. M., and Borrull, F., Chromatographia, 1994, 38, 579. 21 Pocurull, E., Calull, M., Marc�e, R. M., and Borrull, F., J. Chromatogr. A, 1996, 719, 105. 22 Puig, D., and Barcel�o, D., J. Chromatogr. A., 1996, 733, 371. 23 Slobodnik, J., � Oztezkizan, � O., Lingeman, H., and Brinkman, U. A. Th., J. Chromatogr., 1996, 750, 227. 24 Sun, J. J., and Fritz, J. S., J. Chromatogr., 1992, 590, 197. 25 Tsyurupa, M. P., Llyin, M. M., Andreeva, A. I., and Davankov, V. A., Fresenius’ J. Anal. Chem., 1995, 352, 672. 26 Boqu�e, R., and Rius, X., J. Chem. Educ., 1993, 70, 230. Paper 6/07504E Received November 4, 1996 Accepted February 6, 1997 Fig. 3 Chromatograms obtained by on-line trace enrichment with the synthesized sorbent of a 25 ml sample. (a) Ebro river water, (b) Ebro river water spiked with 4 mg l21 of each compound, (c) tap water and (d) tap water spiked with 4 mg l21 of each compound. For peak designation, see Fig. 2. 428 Analy

ISSN:0003-2654    

DOI:10.1039/a607504e

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Evaluation of Hydromatrix and Magnesium Sulfate Drying Agents for Supercritical Fluid Extraction of Multiple Pesticides in Produce Konstantin I. Ellera and Steven J. Lehotay*b a Institute of Nutrition, Russian Academy of Medical Sciences, 2/14 Ustinsky Proezd, Moscow 109240, Russia b United States Department of Agriculture, Agricultural Research Service, Building 007, Room 224, 10300 Baltimore Avenue, Beltsville, MD 20705, USA.E-mail: slehotay@asrr.arsusda.gov The simultaneous extraction of relatively polar and nonpolar pesticides has been problematic in multiresidue analysis using supercritical fluid extraction (SFE) with carbon dioxide. In fruit and vegetable samples, which typically contain 80–95% water, moisture acts to increase SFE recoveries of many polar pesticides, but a drying agent should be used to control water in SFE.Hydromatrix, a prevalent drying agent, has many desirable characteristics, but it reduces recovery of certain important pesticides, such as methamidophos, acephate, and omethoate. MgSO4 has been shown previously to be applicable for the extraction of methamidophos and six other pesticides, but MgSO4 has practical disadvantages in its use.In this study, properties and SFE results with the individual drying agents and their combination were evaluated. Simultaneous recoveries for polar and nonpolar pesticides were achieved for 71 pesticides fortified in apple using a mixture of 2 + 1 + 2 MgSO4·H2O–Hydromatrix–sample for extraction. The advantages of each drying agent were maintained by their combination.The analysis of real samples, however, showed that more study was needed to improve recoveries of nonpolar pesticides. Keywords: Supercritical fluid extraction; drying agents; Hydromatrix; magnesium sulfate; multiple pesticide residue analysis; food Supercritical fluid extraction (SFE) is a promising method of extraction for pesticide residues in fruits and vegetables.1–15 The advantages of a lower solvent viscosity, higher rate of mass transfer, and a wide range of densities for supercritical fluids versus liquid solvents often lead to a reduced extraction time and increased selectivity in SFE.Other practical advantages of SFE versus organic-solvent-based methods of extraction include: no need for solvent evaporation, reduced solvent consumption, less hazardous waste generation, smaller space requirements, increased automation, and the use of less glassware.However, SFE methods have not yet been demonstrated to extract a comparable number of pesticides possible by traditional multiresidue methods.16–18 The polarity range of the pesticides that can be extracted by SFE is limited by the use of CO2, a relatively nonpolar medium, but other supercritical fluids are currently impractical in analytical commercial applications.Other current limitations of SFE include the matrix dependency of extraction, typically small sample size, and frequent need for solvent modifiers. A modifier solvent, such as methanol, can be added to the CO2 to increase the polarity of the extraction medium, but with fruit and vegetable samples, water already present in the sample serves to modify the fluid, and addition of other solvent modifiers has no effect.9 The use of modifiers also complicates the optimization of extraction and trapping conditions.6,19–21 Derivatization is another option to improve SFE recovery of polar pesticides, such as 2,4-D,22–25 but in multiresidue analysis, derivatization methods are not desirable.Fruit and vegetable samples typically consist of 80–95% water, which must be removed or controlled before SFE or instrument performance will be affected. Lyophilization of the sample is a poor option due to the time and equipment required, the concentration of matrix interferants, and the loss of volatile analytes.10 Furthermore, the presence of water in the sample is often beneficial to the extraction process.8,9,26 Therefore, a drying agent must be used to control sample moisture. Hopper and King were the first to apply Hydromatrix (HMX), a pelletized diatomaceous earth material that absorbs twice its mass in water, to SFE.1 HMX not only served to absorb moisture, but also increased the recovery of pesticides from food samples due to increased sample dispersion and reduced particle size. However, HMX can retain certain polar pesticides, such as methamidophos, omethoate, and sulfonyl urea herbicides. 1,11,21 Burford et al. evaluated several other drying agents, including inorganic salts, in the SFE of organic pollutants from moist samples,27 but pesticide analytes or food matrices were not included in the study.Valverde-Garc�ýa et al. demonstrated that methamidophos, chlorpyrifos, endosulfans, methiocarb, and chlorothalonil, were extracted from vegetables when using anhydrous MgSO4 as the drying agent.13,14 The objective of this study was to further describe the effect of drying agents in the SFE of pesticides from matrices with high water content, and to demonstrate multiresidue capabilities of using a MgSO4–HMX drying agent mixture for sample preparation in SFE.Experimental Apparatus A Model 7680T (Hewlett-Packard, Little Falls, DE, USA) and an Autoprep-44 (Suprex, Pittsburgh, PA, USA), both equipped with automated extraction capability, automated variable restrictors, and solid sorbent collection systems, were used in this study. Most data reported in this paper were obtained with the 7680T, except for a set of extractions of the carrot check sample which was performed using both instruments.The SFE parameters were: 320 atm extraction pressure (1 atm = 101.325 kPa), 60 °C temperature (CO2 density of 0.85 g ml21), 2 min static extraction followed by dynamic extraction of six vessel volumes of CO2 at 1.6 ml min21 flow rate, 50 °C restrictor temperature, extract collection on a sorbent trap at 9 °C, and elution with 1.5 ml of acetonitrile at 0.4 ml min21 and 30 °C.The vessel size was 7 ml for the 7680T and 10 ml for the Autoprep-44. For the 7680T, the 1 ml Hypersil octyldecylsilane (ODS) trap was rinsed to waste with 2 ml of ethyl acetate followed by 2 ml of acetonitrile at 1 ml min21 each to clean and regenerate the trap between extractions.For the Autoprep-44, the 1 ml trap consisted of a 1 + 1 ODS–Unibeads sorbent Analyst, May 1997, Vol. 122 (429–435) 429prepared by Suprex, and the trap was flushed with 4 ml of acetonitrile at 1.8 ml min21 between extractions (60 psi N2 gas pressure to blow the trap dry, 1 psi = 6894.76 Pa). After 8–16 extractions, the 7680T tubing (after the vessel) and trap were flushed with Å 10 ml of 1 + 1 methanol–water.A Model ITS40 gas chromatography–ion-trap mass spectrometric detector (GC–ITD) instrument (Finnigan MAT, San Jose, CA, USA), consisting of a Varian 3300/3400 GC and a CTC A200S autosampler was used for multiresidue analyses. Operating conditions for the GC–ITD were: DB-5ms (J&W Scientific, Folsom, CA, USA), 30 m, 0.25 mm id, 0.25 mm film thickness, capillary column, 5 m phenyl–methyl deactivated (Restek, Bellefonte, PA, USA) guard column (0.25 mm id), 1 ml injection volume, Model 1093 (Varian, Walnut Creek, CA, USA) septum programmable injector (SPI); 55 °C injection port for 30 s followed by ramping to 230 °C min at 250 °C min21; 11.5 psi He column head pressure; 55 °C initial oven temperature for 30 s, ramped to 130 °C at 50 °C min21, then to 165 °C at 1.5 °C min21 and to 250 °C at 4 °C min21, and held at 250 °C until 51 min total time elapsed; 240 °C transfer line temperature; and 205 °C ion-trap manifold temperature.Typical ITD operating conditions were: electron impact mode, 12 mA filament current, 1800 V electron multiplier tube, and automatic gain control (AGC) at 30 000.The GC–ITD utilized a Magnum version 2.4 software package loaded into a Gateway 2000 computer for data collection and analysis. The data collection range was 60–420 m/z from 5 to 51 min for the analysis of the 71 pesticides. A Model 5890 (Hewlett-Packard) GC, equipped with a Model 7673 autosampler and fla photometric detector (FPD), was used to analyze methamidophos, chlorpyrifos, and diazinon in the drying agent study.For the GC–FPD, a SPB-1 (Supelco, Bellefonte, PA, USA) 30 m, 0.25 mm id, 0.25 mm film thickness, capillary column was used, and operating conditions were: 1 ml splitless injection; 220 °C injection port; 110 °C initial oven temperature for 1 min, ramped to 226 °C at 15 °C min21, then to 250 °C at 3 °C min21; 250 °C FPD.Reagents SFC/SFE Grade CO2 (Air Products, Allentown, PA, USA), with 1800 psi He headspace (for the Autoprep-44) or without He headspace (for the 7680T), was used to perform SFE. Dry, commercial-grade CO2 was required for cryogenic cooling for the SFE instruments and the SPI on the GC–ITD. Anhydrous MgSO4 (99% purity) was obtained from Fisher (Fair Lawn, NJ, USA) and Aldrich (Milwaukee, WI, USA), and MgSO4·H2O was made by mixing 18 g of H2O with 120 g of MgSO4.HMX was obtained from Varian (Harbor City, CA, USA). All organic solvents were pesticide grade quality (Fisher), and all pesticides were obtained from the US Environmental Protection Agency (Research Park, NC, USA or Beltsville, MD, USA). Chrysened12 (Cambridge Isotope Laboratories, Woburn, MA, USA) was used as the internal standard for GC–ITD, and terbufos was used as the internal standard for GC–FPD.Individual stock solutions with typical concentrations of 2 mg ml21 were prepared in acetone, and these solutions were used to make three working standard mixtures of 100 mg ml21 for each pesticide in acetone consisting of ‘organochlorine,’ ‘organophosphorus,’ and ‘other’ pesticides.Sample Preparation Produce samples were purchased at a local supermarket to serve as blank or fortified samples. Apple, carrot, green bean, orange, and pea check samples were provided by the California Department of Food and Agriculture (CDFA) as part of a quality assurance protocol for the laboratories participating in the Pesticide Data Program,28 and green bean samples containing field residues of methamidophos and acephate were provided by the Michigan Department of Agriculture.The 100 g frozen samples were each mixed with 100 g of MgSO4·H2O and 50 g of HMX in a chopper (Black & Decker; Shelton, CT, USA). [Safety Note: A filter mask should be worn to prevent breathing in of small particles.] A small amount of dry ice was added to the samples during mixing to maintain frozen conditions.12 The 7 ml vessels were packed with 6 g of the cold homogenate (2.4 g of produce sample) and 10 ml vessels were packed with 8 g (3.2 g of produce sample).Filter paper, Whatman # 3 (Maidstone, Kent, UK) was cut into disks which were placed at both ends to keep particles from affecting sealing of the vessels. The vessels were kept cold before extraction in order to reduce losses of degradative pesticides.12 In the case of the fortified samples, a mortar and pestle were used to mix 10 g of sample (after fortification) with 10 g of MgSO4·H2O and 5 g of HMX.The homogenized mixtures, 6 g (2.4 g of sample), were loaded into the 7 ml vessels and extracted by SFE. Analysis The same solutions as the spiking solutions were diluted to make the calibration standards.After extraction, typically 15 ml of 100 mg ml21 chrysene-d12 solution were added to each 1.5 ml extract as an internal standard. The calibration standards were prepared in SFE extracts from sample blanks of the same matrix29 or the method of standard additions was used for check samples as described previously.11,12 Table 1 lists the pesticides, retention times, quantitation masses, and limits of detection (LODs) for the analysis of apple in this study.The pesticides consist of 28 organophosphorus, 19 organochlorine, 6 pyrethroid, 4 carbamate, and 13 other insecticides, herbicides, and fungicides. The LOD values, the concentration at which signal/noise is 3, were determined using calibration data from the pesticide recovery experiment in apple in the manner reported previously for GC–ITD.11 The LODs vary depending on matrix and the conditions of the GC–ITD; the reported values are typical of an apple matrix.Results and Discussion Drying Agents Before performing SFE of produce samples, experiments were carried out to compare drying agent properties of MgSO4 and HMX. To determine the amount of material transported by the supercritical CO2, the masses of the loaded vessels were measured before and after SFE.To determine the water retention capabilities of the drying agents, the vessel contents after SFE were transferred to a watch glass and were baked at 250 °C for 24 h. The amount of water extracted during SFE was determined by subtracting the post-SFE water content in the sample from the original water content. Table 2 contains the results of this experiment.The solubility of H2O in supercritical CO2 is Å 0.3%,30 which corresponds to Å 110 mg of H2O for the 38 g of CO2 used in our extractions. As shown in Table 2, up to 160 mg of water were lost from the drying agents during SFE. A conclusion from this experiment was that MgSO4 retained water more strongly than HMX.More importantly, these results documented how the drying agents do not retain all of the moisture in the system, and water dissolved in the supercritical CO2 is available as a ‘modifier’ in SFE. Not all of the mass loss could be accounted for by water alone; up to 90 mg of material was missing in some cases. After a series of extractions of samples containing MgSO4, a build-up of fine particles appeared at the frit between the vessel and restrictor.When using MgSO4 in routine applications, this frit should be changed frequently, and a filter paper with pores < 2 mm should be used to contain the sample in the SFE vessel. Plugging of the automated variable restrictor is a potential 430 Analyst, May 1997, Vol. 122problem with MgSO4, especially if there is no filter placed before the restrictor, but with linear restrictors, plugging due to MgSO4 is not a problem.13,14 Effect of Drying Agent :H2O Ratio Recovery studies of three representative pesticides with different degrees of polarity were performed in model matrices to determine the effect of water in SFE, and to optimize the drying agent :H2O (sample) ratio for produce samples.Fig. 1 gives the structures and the solubilities in water of the chosen organophosphorus insecticides, methamidophos, diazinon, and chlorpyrifos. Table 3 lists the results of matrices fortified at 1 mg g21 with the pesticides. On filter paper, water did not have a significant effect on the recoveries of diazinon and chlorpyrifos, but methamidophos recovery reached a maximum with 100–200 ml of H2O added to the vessel.Based on these results, methamidophos is slightly soluble in supercritical CO2 at density = 0.85 g ml21, but when the CO2 was ‘modified’ with H2O, methamidophos solubility increased. This amount of water, which ostensibly corresponded with the saturation of water in supercritical CO2, enabled the solubilization of methamidophos.However, the addition of more water reduced recovery of methamidophos, presumably due to the greater availability of water remaining in the vessel for the pesticide to partition into. Diazinon and chlorpyrifos are more soluble in supercritical CO2, and less soluble in water than methamidophos, and partitioned into the supercritical fluid phase. Note that the pesticides were spiked onto the filter paper while in the vessel (due to the inability to homogenize the sample).This spiking procedure often gives higher recoveries than if the spike solution is mixed into the sample before loading the vessel. In the case of HMX, methamidophos was not recovered, independent of the amount of water present. Nor was methamidophos recovered from anhydrous or monohydrated MgSO4, but recovery improved to 70–90% with a 2 + 1 or 1 + 1 MgSO4– H2O matrix. Based on results in Tables 2 and 3, Å 70 mg of water were required to dissolve methamidophos in CO2 under the SFE conditions.The differences between HMX and MgSO4 in the extraction of methamidophos were reported previously, 1,11,13,14 but no attempts of explanation were made.The most likely explanation is that HMX contains SiOH sites that can interact through hydrogen bonding with the amine group.31 MgSO4 contains no hydroxyl groups to interact with amines, and methamidophos is extracted from the matrix when enough water is present to modify the fluid. Furthermore, MgSO4 has a solubility in water of 34–90% by mass (depending on hydrated state and temperature);32 thus, available water is saturated with MgSO4 which acts to salt out methamidophos from the water phase into the supercritical CO2 phase.The same effect is observed in liquid–liquid partitioning when salt is added to an aqueous layer. In the case of HMX, absorbed water is more accessible than water hydrated with MgSO4, and due to its low solubility in water, HMX does not salt out methamidophos from the water. The pH of saturated aqueous solutions of both HMX and MgSO4 was neutral.Table 1 Pesticide retention times (tt), quantitation masses, and limits of detection (LOD) in SFE apple extracts using GC–ITD analysis LOD*/ LOD*/ No. Pesticide tt/s Masses (m/z) ng g21 No. Pesticide tt/s Masses (m/z) ng g21 1 Methamidophos 350 94† + 95 + 141 50 37 Malathion 1392 125 + 127 + 173† 20 2 Dichlorvos 358 109† + 185 2 38 Metolachlor 1398 162† + 238 7 3 Mevinphos 575 127† + 192 1 39 Chlorpyrifos 1404 197 + 199 + 314 6 4 Acephate 582 94 + 136† 200 40 Aldrin 1405 347–351 13 5 Tetrahydrophthalimide 683 79† + 151 8 41 Dacthal 1413 299–303 2 6 Pentachlorobenzene 725 248–252 0.3 42 Parathion 1421 97 + 109 + 291† 13 7 o-Phenylphenol 727 169 + 170† 2 43 Dicofol 1435 139† + 250 7 8 Omethoate 849 110 + 156† 100 44 Captan 1512 79† 7 9 Propoxur 883 110† + 152 4 45 Methidathion 1539 85 + 93 + 145† 0.9 10 Diphenylamine 916 167 + 168 + 169† 2 46 o,pA-DDE 1548 316–320 3 11 Ethoprop 934 97 + 158† + 243 3 47 Tetrachlorvinphos 1553 330–333 2 12 Chlorpropham 975 127† + 171 + 213 20 48 Disulfoton sulfone 1561 97 + 153 + 213† 16 13 Dicrotophos 988 67 + 127† + 237 200 49 Endosulfan I 1567 337–341 20 14 Trifluralin 993 264 + 306† 0.2 50 cis-Chlordane 1568 373–379 2 15 Monocrotophos 1018 67 + 127† + 223 200 51 Fenamiphos 1587 260 + 288 + 303† 50 16 Phorate 1032 75† + 121 + 260 3 52 p,pA-DDE 1618 316–320 2 17 Hexachlorobenzene 1046 282–290 0.9 53 Myclobutanil 1626 150 + 179† + 181 100 18 Dichloran 1078 176† + 206 7 54 Endosulfan II 1687 195† + 241 + 339 8 19 Dimethoate 1082 87† + 93 30 55 p,pA-DDD 1697 165 + 235† + 237 2 20 Carbofuran 1104 149 + 164† 9 56 cis-Nonachlor 1697 405–413 1 21 Quintozene (PCNB) 1121 293–301 5 57 Ethion 1698 97 + 153 + 231† 6 22 Atrazine 1122 200† 2 58 o,p-DDT 1701 165 + 235† + 237 2 23 Lindane (g-HCH) 1137 181 + 183† + 219 20 59 Carbophenothion 1748 157† + 199 + 342 11 24 Terbufos 1157 231† 2 60 Propargite 1819 135† + 335 + 350 30 25 Diazinon 1182 137 + 179† + 304 3 61 Iprodione 1871 314† + 316 10 26 Chorothalonil 1182 264–268 2 62 Phosmet 1887 160† 4 27 Disulfoton 1202 88† + 89 + 97 3 63 Methoxychlor 1913 227† 3 28 Phosphamidon 1277 127† + 264 12 64 Phosalone 1991 182† + 367 4 29 Propanil 1284 161† + 163 + 217 50 65 Azinphos-methyl 2003 77† + 132 + 160 50 30 Vinclozolin 1304 198 + 212† + 285 6 66 cis-Permethrin 2252 183† 10 31 Parathion-methyl 1307 109 + 125 + 263† 16 67 trans-Permethrin 2288 183† 10 32 Alachlor 1312 160 + 188† + 238 10 68 Cyfluthrin 2404 163† + 206 + 226 100 33 Carbaryl 1321 115 + 116 + 144† 10 69 Cypermethrin 2511 127 + 163† + 181 50 34 Heptachlor 1323 270 + 272† + 274 5 70 Fenvalerate 2878 125 + 225† + 419 40 35 Fenitrothion 1367 125 + 260 + 277† 3 71 Esfenvalerate 2986 125 + 225† + 419 40 36 Linuron 1379 61† + 248 5 IS‡ Chrysene-d12 1900 240† * As determined by SFE for a 2.4 g apple in 1.5 ml of final volume, 1 ml injection.† Base peak. ‡ IS, internal standard. Analyst, May 1997, Vol. 122 431As designated by bold script in Table 3, an interesting effect was also observed for diazinon and chlorpyrifos in dry HMX and MgSO4.Dry HMX retained chlorpyrifos to some extent, but MgSO4 did not, whereas anhydrous MgSO4 retained diazinon, but HMX did not. In both instances, the presence of water, even the small amount in the MgSO4·H2O matrix, served to enhance the recoveries of the pesticides. The pronounced effect of water on SFE recoveries has been observed previously, but not with relatively nonpolar pesticides such as diazinon and chlorpyrifos. Campbell et al.observed a decreased recovery of chlorpyrifosmethyl in dry wheat and resorted to the use of 2% methanol modifier to improve recovery.15 Our results emphasize how the modifying and dispersing effect of water can serve a similar purpose as the addition of an organic solvent.MgSO4–HMX Mixtures HMX possesses practical advantages of a much lower heat of hydration, slightly lower cost (4c per gram versus 6c per gram), higher water absorptivity per mass basis, and better sample consistency than MgSO4. Also, HMX does not contain as many fine particles. However, the hydration/salt effect described above and higher density (enabling more sample per fixed vessel volume) presents the advantages of MgSO4.A drawback of MgSO4 is the formation of agglomerates when mixed with water. The combination of HMX and MgSO4 drying agents greatly reduced the problems associated with the individual drying agents, and appeared to maintain their advantages. As shown in Table 3, the addition of HMX as a minor component with MgSO4 did not adversely affect pesticide recoveries, and served to improve the practical aspects of sample preparation for SFE.The ratio of 1 + 0.4 + 1 MgSO4–HMX–H2O was found to be most useful because higher HMX content affected methamidophos recovery, and higher water content made the matrix too moist, leading to reduced recoveries. More sample could be loaded in vessels than with HMX alone, and sample consistency remained powdery and dispersed. In the analysis of real samples, MgSO4·H2O was used rather than anhydrous MgSO4, and for convenience, the sampling ratio was changed to 1 + 0.5 + 1 MgSO4·H2O–HMX–sample. The use of MgSO4 reduced the heat generated when the sample was mixed with the drying agent.Analysis of Acephate and Methamidophos in Green Beans Extraction and analysis of acephate and methamidophos, insecticides that are commonly applied to green vegetables, has been troublesome by SFE and GC–ITD previously.11,12 In a test of the SFE method using MgSO4–HMX as the drying agent, analyses of green beans previously analyzed by the Michigan Department of Agriculture were performed.GC–ITD data for the green bean sample extracted by SFE is shown in Fig. 2, and quantitative results are listed in Table 4 (the LOD for acephate was lower in green bean than the value reported for apple in Table 1). Despite the relative difficulties involved in GC–ITD analysis of these pesticides, pesticide concentrations agreed within the error of analysis in each case. The large deviation (60% RSD) in the acephate concentration was mainly due to variations in GC–ITD analysis and not SFE.Acephate response fluctuated with the cleanliness of the GC system, and the variability in these results reflected the effect of matrix components in the injection liner and precolumn.29 The use of GC–FPD gives better peak shapes and a lower LOD than GC– ITD for these compounds, but GC–ITD makes mass spectral confirmation possible. Table 2 Amount of water transported from drying agents by SFE Water in Water in Water vessel vessel trans- Sample before after ported by Sample matrix size/g SFE/mg SFE/mg SFE/mg Hydromatrix (HMX), dry 2.0 0 0 0 HMX–H2O (2 + 1) 3.3 1100 980 120 HMX–H2O (1 + 1) 4.0 2000 1840 160 MgSO4, anhydrous 6.0 0 0 0 MgSO4–H2O (6.7 + 1) 6.3 820 810 10 MgSO4–H2O (2 + 1) 4.5 1500 1430 70 MgSO4–H2O (2 + 1) 6.0 2000 1920 80 MgSO4–H2O (1 + 1) 6.0 3000 2900 100 MgSO4–HMX–H2O (1 + 0.4 + 1) 5.0 2080 1940 140 MgSO4–HMX–H2O (1 + 0.4 + 1) 6.5 2710 2560 150 Fig. 1 Structures and solubilities in H2O of methamidophos, diazinon, and chlorpyrifos. Table 3 Effect of H2O on SFE % recoveries (± s) of organophosphorus pesticides on different matrices (1 mg g21 spikes); results from GC–FPD analysis; values in bold show the interesting results for diazinon and chlorpyrifos in dry HMX and MgSO4 Sample matrix n Methamidophos Diazinon Chlorpyrifos Filter paper*, dry 2 65.4 ± 3.2 102 ± 17 102 ± 17 Filter paper + 50 ml H2O 2 67 ± 16 91 ± 10 93.1 ± 7.7 Filter paper + 100 ml H2O 2 92 ± 16 87.5 ± 7.9 90.5 ± 2.8 Filter paper + 200 ml H2O 2 117 ± 13 82.2 ± 0.4 90.1 ± 0.3 Filter paper + 300 ml H2O 2 48 ± 31 84.5 ± 3.0 87.1 ± 2.5 Filter paper + 600 ml H2O 2 55.3 ± 3.6 81.3 ± 4.4 87.8 ± 3.9 HMX†, dry 3 ND‡ 87.6 ± 0.8 36.9 ± 0.8 HMX–H2O (2 + 1) 4 ND 98.0 ± 0.7 97.0 ± 1.3 HMX–H2O (1 + 1) 4 ND 86.0 ± 4.3 91.0 ± 2.7 MgSO4, anhydrous 2 ND 11.3 ± 3.0 86.8 ± 8.9 MgSO4–H2O (6.7 + 1) 2 ND 82.3 ± 0.4 84.1 ± 1.3 MgSO4–H2O (2 + 1) 3 86.6 ± 8.3 95 ± 11 88.1 ± 4.6 MgSO4–H2O (1 + 1) 3 72 ± 16 69 ± 16 72 ± 18 MgSO4–HMX (2 + 3), dry 2 ND 12.9 ± 1.4 54.5 ± 6.5 MgSO4–HMX–H2O (1 + 0.4 + 1) 2 90.5 ± 9.6 91.2 ± 7.8 89.3 ± 8.3 MgSO4–HMX–H2O (0.6 + 0.4 + 1) 2 85 ± 14 88 ± 16 88 ± 11 * Filter paper was spiked in the SFE vessel.† Drying agents were spiked outside the SFE vessel. ‡ ND, not detected. 432 Analyst, May 1997, Vol. 122Analysis of Fortified Samples Recovery studies of apple samples fortified at 1 mg g21 (high spike) and 0.05 mg g21 (low spike) were performed (triplicate extractions).As shown in Table 5, all 71 pesticides were determined at the high spiking level with an average recovery of 86.7% (5.9% average RSD), and only cis-nonachlor (60.8%) had a recovery < 70%. Fig. 3 is a GC–ITD total ion chromatogram of a 0.4 mg g21 apple standard for the pesticides, and it displays the low background associated with the SFE method. For the low spiking level, the fortification level was less than the LOD for several pesticides which were not detected, and confirmation criteria were not met for a few other pesticides as noted in Table 5.The 59 other pesticides were confirmed and quantified with an average recovery of 88.9% (6.9% average RSD). Six pesticides were recovered from the low spike at less than 70%, one of which was cis-nonachlor (57%) again.The likely source of the lower recovery of cis-nonachlor was incomplete elution of the trap with acetonitrile. The use of acetone was shown to elute nonpolar pesticides from the ODS Fig. 2 GC–ITD chromatogram peaks of methamidophos and acephate in green bean sample no. 2 after SFE. Mass spectral confirmation is presented for acephate. Table 5 Presticide recoveries from apple fortified at 1 mg g21 (high spike) and 0.05 mg g21 (low spike) using the SFE and GC–ITD method (n = 3) High spike Low spike High spike Low spike recovery recovery recovery recovery No. Pesticide (% RSD) (% RSD) No. Pesticide (% RSD) (% RSD) 1 Methamidophos 72.3 (7.5) NC* 37 Malathion 89.8 (4.3) 82 (8.8) 2 Dichlorvos 86.1 (9.5) 88 (6.0) 38 Metolachlor 80.5 (5.7) 94 (0) 3 Mevinphos 95.3 (1.2) 81 (5.1) 39 Chlorpyrifos 92.2 (2.8) 99 (7.1) 4 Acephate 86.5 (7.3) ND† 40 Aldrin 82.7 (0.9) 96 (2.1) 5 Tetrahydrophthalimide 92.10 92 (3.2) 41 Dacthal 81.3 (11) 101 (1.2) 6 Pentachlorobenzene 75.30 86 (4.6) 42 Parathion 92.1 (1.8) 79 (1.5) 7 o-Phenylphenol 89.7 (4.0) 103 (4.1) 43 Dicofol 85.7 (7.0) 95 (1.2) 8 Omethoate 102 (3.0) ND 44 Captan 91.9 (4.1) 60 (15) 9 Propoxur 89.1 (6.2) 95 (5.3) 45 Methidathion 88.9 (7.9) 83 (12) 10 Diphenylamine 80.7 (7.6) 87 (5.3) 46 o,pA-DDE 84.1 (2.9) 95 (2.4) 11 Ethoprop 86.5 (1.9) 93 (5.4) 47 Tetrachlorvinphos 93.3 (2.3) 95 (4.4) 12 Chlorpropham 80.9 (8.4) 114 (10) 48 Disulfoton sulfone 90.8 (7.0) 80 (22) 13 Dicrotophos 97.6 (4.1) ND 49 Endosulfan I 82.0 (3.2) 107 (7.5) 14 Trifluralin 93.2 (1.7) 78 (2.6) 50 cis-Chlordane 83.1 (1.5) 92 (6.5) 15 Monocrotophos 89.6 (4.7) ND 51 Fenamiphos 91.9 (15) ND 16 Phorate 87.0 (13) 83 (9.2) 52 p,pA-DDE 83.4 (2.2) 95 (4.4) 17 Hexachlorobenzene 82.7 (3.3) 95 (3.2) 53 Myclobutanil 87.8 (10) ND 18 Dicloran 91.5 (3.1) 89 (5.6) 54 Endosulfan II 84.8 (1.8) 132 (7.4) 19 Dimethoate 94.4 (5.8) 111 (7.3) 55 p,pA-DDD 86.9 (3.1) 99 (4.6) 20 Carbofuran 89.9 (5.9) 95 (3.2) 56 cis-Nonachlor 60.8 (9.2) 57 (16) 21 Quintozene 87.0 (3.0) 89 (1.3) 57 Ethion 92.9 (6.8) 71 (6.5) 22 Atrazine 86.6 (6.0) 98 (2.0) 58 o,pA-DDT 87.3 (14) 80 (6.6) 23 Lindane 85.1 (5.6) 123 (7.3) 59 Carbophenothion 92.1 (5.4) 71 (8.6) 24 Terbufos 89.8 (9.1) 65 (9.8) 60 Propargite 80.6 (8.9) 95 (2.4) 25 Diazinon 89.1 (3.3) 101 (5.0) 61 Iprodione 81.1 (12) 63 (51) 26 Chorothalonil 74.3 (3.2) 89 (1.3) 62 Phosmet 91.5 (4.6) 65 (12) 27 Disulfoton 87.9 (2.8) 82 (7.3) 63 Methoxychlor 85.0 (2.5) 73 (7.9) 28 Phosphamidon 97.5 (9.3) ND 64 Phosalone 92.2 (4.2) 75 (14) 29 Propanil 93.5 (2.9) NC 65 Azinphos-methyl 95.2 (3.6) NC 30 Vinclozolin 79.4 (7.6) 99 (8.1) 66 cis-Permethrin 80.1 (9.3) 80 (16) 31 Parathion-methyl 92.3 (2.8) 87 (3.5) 67 trans-Permethrin 82.3 (9.8) 79 (1.5) 32 Alachlor 79.8 (7.7) 92 (2.2) 68 Cyfluthrin 83.8 (16) ND 33 Carbaryl 92.0 (5.5) 89 (4.7) 69 Cypermethrin 84.0 (10) ND 34 Heptachlor 83.6 (1.6) 96 (4.2) 70 Fenvalerate 81.1 (15) 97 (8.4) 35 Fenitrothion 91.8 (1.7) 78 (2.6) 71 Esfenvalerate 85.4 (12) 115 (1.0) 36 Linuron 76.8 (5.0) 65 (7.1) average 86.7 (5.9) 88.9 (6.9) * NC, Detected, but confirmation criteria not met.† ND, Not detected (LOD > 0.05 mg g21). Table 4 Comparison of SFE method results versus traditional method results for acephate and methamidophos in green beans Sample Michigan lab. SFE results†/ pesticide results*/mg g21 mg g21 Sample— Acephate 0.65 0.98 ± 0.60 Methamidophos 0.12 0.115 ± 0.023 Sample 2— Acephate 0.18 0.23 ± 0.13 Methamidophos 0.074 0.087 ± 0.010 * Method of ref. 16, GC–FPD analysis. † GC–ITD analysis. Analyst, May 1997, Vol. 122 433trap in less volume than acetonitrile.33 Linuron, iprodione, and phosmet calibration data were not as linear as other pesticide calibration plots, and lower recoveries for these three pesticides at the 0.05 mg g21 level were believed to be a result of calibration bias.Captan readily degrades, and terbufos, as well as disulfoton, phorate, fenamiphos, and phosphamidon, may convert into sulfoxide and sulfone forms. Recoveries of the 52 other pesticides were independent of the concentration, and the slightly higher deviation in the values can be ascribed to the intricacies of analysis at lower concentrations.Analysis of Multiresidue Check Samples Based on the recovery data, the SFE and GC–ITD method using the HMX–MgSO4 mixture drying agent was ready to be tested in the extraction and analysis of actual samples. Table 6 presents the results of analyses of six check samples using the proposed method compared with results obtained by 7–9 state regulatory laboratories using traditional methods.16,17 In all, 10 check samples have been analyzed by SFE and GC–ITD,11,12 and there have been no instances of false negatives or false positives for the pesticides of interest (confirmation of pesticides by mass spectrometry is a prerequisite for reporting results). However, the quantitation of results between the traditional methods and the SFE and GC–ITD method agreed well in only six of the 20 cases.Of the 6 instances of agreement, only ethion (in orange) can be considered a nonpolar pesticide. Based on these results, the use of mixed HMX–MgSO4 drying agent gave low recoveries of nonpolar pesticides in real samples, and high recoveries of the relatively polar pesticides, omethoate, dimethoate, carbaryl, and diphenylamine. The cause of this difference between the fortified (Table 2) and check sample results (Table 6) was possibly related to the difference in sample preparation procedures of using a mortar and pestle for the small fortified samples versus a blender for the 100 g check samples.However, due to the agreeable results for polar pesticides, and good reproducibility of pesticide results for multiple subsamples, the blending procedure developed previously for HMX alone adequately homogenized the extracts.12 The more likely reason that nonpolar pesticides were not extracted as efficiently as the more polar pesticides in real samples is related to the higher degree of difficulty in extracting nonpolar analytes from aged samples than from fortified samples.34,35 Further investigations to address this problem will be conducted in the future.The USDA Agricultural Marketing Service (Pesticide Data Program) helped fund this research and provided the ITS40 instrument. The authors are also grateful to Hewlett-Packard and Suprex for providing use of the 7680T and Autoprep-44, respectively. Also, the authors thank the CDFA for providing the check samples, and the Department of Agriculture laboratories of California, Florida, Michigan, New York, North Carolina, Ohio, Texas, and Washington for performing traditional analyses, and the Michigan Department of Agriculture for providing and analyzing the green bean samples.The mention of specific items of equipment and chemicals by brand names do not constitute endorsement of a product by the USDA.Fig. 3 Total ion chromatogram of an SFE apple extract fortified at 0.4 mg g21 with 71 pesticides. Numbered peaks correspond to pesticides listed in Table 1, and peaks designated with * are matrix components. 434 Analyst, May 1997, Vol. 122References 1 Hopper, M. L., and King, J. W., J. Assoc. Off. Anal. Chem., 1991, 74, 661. 2 Nishikawa, Y., Anal. Sci., 1991, 7, 567. 3 Thomson, C. A., and Chesney, D. J., Anal. Chem., 1992, 64, 848. 4 King, J. W., Snyder, J. M., Taylor, S. L., Johnson, J. H., and Rowe, L. D., J. Chromatogr. Sci., 1993, 31, 1. 5 Wigfield, Y. Y., and Lanouette, M., J. Agric. Food Chem., 1993, 41, 84. 6 Howard, A. L., Braue, C., and Taylor, L. T., J. Chromatogr. Sci., 1993, 31, 323. 7 King, J.W., Hopper, M. L., Luchtenfeld, R. G., Taylor, S. L., and Orton, W. L., J. AOAC Int., 1993, 76, 857. 8 Skopec, Z. V., Clark, R., Harvey, P. M. A., and Wells, R. J., J. Chromatogr. Sci., 1993, 31, 445. 9 Aharonson, N., Lehotay, S. J., and Ibrahim, M. A., J. Agric. Food Chem., 1994, 42, 2817. 10 Lehotay, S. J., and Ibrahim, M. A., J. AOAC Int., 1995, 78, 445. 11 Lehotay, S.J., and Eller, K. I., J. AOAC Int., 1995, 78, 821. 12 Lehotay, S. J., Aharonson, N., Pfeil, E., and Ibrahim, M. A., J. AOAC Int., 1995, 78, 831. 13 Valverde-Garc�ýa, A., Fern�andez-Alba, A. R., Ag�uera A., and Contreras, M., J. AOAC Int., 1995, 78, 867. 14 Valverde-Garc�ýa, A., Fern�andez-Alba, A. R., Contreras, M., and Ag�uera A., J. Agric. Food Chem., 1996, 44, 1780. 15 Campbell, R.M., Meurnier, D. M., and Cortes, H. J., J. Microcolumn Sep., 1989, 1, 302. 16 Luke, M. A., and Masumoto, H. T., Anal. Methods Pestic. Plant Growth Regul., 1986, 15, 161. 17 Lee, S. M., Papathakis, M. L., Feng, H. M. C., Hunter, G. F., and Carr, J. E., Fresenius’ J. Anal. Chem., 1991, 339, 376. 18 Cairns, T., Luke, M. A., Chiu, K. S., Navarro, D., and Siegmund, E. G., Rapid Commun.Mass Spectrom., 1993, 7, 1070. 19 Langenfeld, J. L., Hawthorne, S. B., Miller, D. J., and Pawliszyn, J., Anal. Chem., 1994, 66, 909. 20 Mulcahey, L. J., and Taylor, L. T., Anal. Chem., 1992, 64, 2352. 21 Howard, A. L., and Taylor, L. T., J. High Res. Chromatogr. 1993, 16, 39. 22 Hawthorne, S. B., Miller, D. J., Nivens, D. E., and White, D. C., Anal. Chem., 1992, 64, 405. 23 Field, J. A., Miller, D. J., Field, T. M., Hawthorne, S. B., and Giger, W., Anal. Chem., 1992, 64, 3161. 24 Hillmann, R., and Bachman, K., J. High Res. Chromatogr., 1994, 17, 350. 25 Chatfield, S. N., Croft, M. Y., Dang, T., Murby, G. Y. F., and Wells, R. J., Anal. Chem., 1995, 67, 945. 26 Fahmy, T. M., Paulitis, M. E., Johnson, D. M., and McNally, M. E. P., Anal. Chem., 1993, 65, 1462. 27 Burford, M. D., Hawthorne, S. B., and Miller, D. J., J. Chromatogr. A, 1993, 657, 413. 28 Agricultural Marketing Service, Pesticide Data Program (PDP) Summary of 1994 Data, USDA, Washington, DC, 1996. 29 Erney, D. R., Gillespie, A. M., Gilvydis, D. M., and Poole, C. F., J. Chromatogr., 1993, 638, 57. 30 Kuk, M. S., and Montagna, J. C., Chemical Engineering at Supercritical Conditions, ed.Paulitis, M. E., Penninger, J. M., Gray, R. D., and Davison, K. P., Ann Arbor Science, Ann Arbor, MI, 1983, pp. 101–111. 31 Oostdyk, T. S., Grob, R. L., Snyder, J. L., and McNally, M. E. P., J. Chromatogr. Sci., 1993, 31, 177. 32 Lange’s Handbook of Chemistry, ed. Dean, J. A., McGraw-Hill, New York, 13th edn. 1985, pp. 4–74. 33 Lehotay, S. J., and Valverde-Garc�ýa, A., J.Chromatogr. A, in the press. 34 Burford, M. D., Hawthorne, S. B., and Miller, D. J., Anal. Chem., 1993, 65, 1497. 35 Camel, V., Tambut�e, M., and Caude, M., J. Chromatogr. A, 1995, 693, 101. Paper 6/07754A Received November 6, 1996 Accepted January 19, 1997 Table 6 Comparison of results (concentrations in mg g21) of SFE method versus traditional methods of analysis (refs. 16, 17) performed by state regulatory laboratories State lab.results Commodity Spiking SFE method pesticide conc. Conc. (n = 9) High Low conc. (n = 3) Apple— Azinphos-methyl incurred* 0.110 ± 0.045 0.20 0.033 0.064 ± 0.003 Chlorothalonil 0.16 0.180 ± 0.065 0.29 0.12 0.104 ± 0.025 Diphenylamine 0.82 0.64 ± 0.13 0.86 0.43 0.580 ± 0.088 Fenamiphos 0.088 0.060 ± 0.020 0.091 0.034 0.022 ± 0.006 o-Phenylphenol incurred NR† < 0.01 Phosmet incurred NR < 0.01 Apple— Carbaryl 0.482 0.451 ± 0.070 0.61 0.40 0.444 ± 0.036 Chlorpyrifos 0.114 0.122 ± 0.024 0.16 0.10 0.088 ± 0.007 Diphenylamine incurred 0.93 ± 0.23 1.2 0.55 0.97 ± 0.11 Propargite 0.814 0.686 ± 0.088 0.80 0.53 0.309 ± 0.034 Carrot— p,pA-DDE incurred 0.028 ± 0.022 0.050 0 0.020 ± 0.001‡ Parathion-methyl 0.117 0.110 ± 0.016 0.14 0.090 0.036 ± 0.010‡ Trifluralin incurred 0.111 ± 0.040 0.20 0.070 0.030 ± 0.007‡ Green bean— Chlorothalonil 0.16 0.120 ± 0.056 0.24 0.052 0.034 ± 0.010 p,pA-DDE incurred 0.023 ± 0.031 0.10 0.009 0.004 ± 0.003 Disulfoton 0.18 0.056 ± 0.020 0.079 0.027 0.022 ± 0.009 Disulfoton sulfone incurred NR < 0.01 Endosulfan I 0.052 0.053 ± 0.011 0.08 0.041 0.022 ± 0.004 Orange— Chlorothalonil 0.16 0.120 ± 0.060 0.22 0.04 0.104 ± 0.019 Ethion 0.075 0.071 ± 0.018 0.09 0.04 0.073 ± 0.007 Pea— Dimethoate 0.090 0.092 ± 0.018 0.12 0.060 0.081 ± 0.006 Methoxychlor 0.206 0.181 ± 0.048 0.25 0.080 0.058 ± 0.005 Omethoate 0.084 0.083 ± 0.019 0.10 0.050 0.098 ± 0.064 * Incurred, pesticide applied in field.† NR, not reported; ‡ Results from two sets of analyses using different SFE instruments.Analyst, May 1997, Vol. 122 435 Evaluation of Hydromatrix and Magnesium Sulfate Drying Agents for Supercritical Fluid Extraction of Multiple Pesticides in Produce Konstantin I. Ellera and Steven J. Lehotay*b a Institute of Nutrition, Russian Academy of Medical Sciences, 2/14 Ustinsky Proezd, Moscow 109240, Russia b United States Department of Agriculture, Agricultural Research Service, Building 007, Room 224, 10300 Baltimore Avenue, Beltsville, MD 20705, USA.E-mail: slehotay@asrr.arsusda.gov The simultaneous extraction of relatively polar and nonpolar pesticides has been problematic in multiresidue analysis using sup extraction (SFE) with carbon dioxide. In fruit and vegetable samples, which typically contain 80–95% water, moisture acts to increase SFE recoveries of many polar pesticides, but a drying agent should be used to control water in SFE.Hydromatrix, a prevalent drying agent, has many desirable characteristics, but it reduces recovery of certain important pesticides, such as methamidophos, acephate, and omethoate. MgSO4 has been shown previously to be applicable for the extraction of methamidophos and six other pesticides, but MgSO4 has practical disadvantages in its use.In this study, properties and SFE results with the individual drying agents and their combination were evaluated. Simultaneous recoveries for polar and nonpolar pesticides were achieved for 71 pesticides fortified in apple using a mixture of 2 + 1 + 2 MgSO4·H2O–Hydromatrix–sample for extraction. The advantages of each drying agent were maintained by their combination. The analysis of real samples, however, showed that more study was needed to improve recoveries of nonpolar pesticides.Keywords: Supercritical fluid extraction; drying agents; Hydromatrix; magnesium sulfate; multiple pesticide residue analysis; food Supercritical fluid extraction (SFE) is a promising method of extraction for pesticide residues in fruits and vegetables.1–15 The advantages of a lower solvent viscosity, higher rate of mass transfer, and a wide range of densities for supercritical fluids versus liquid solvents often lead to a reduced extraction time and increased selectivity in SFE.Other practical advantages of SFE versus organic-solvent-based methods of extraction include: no need for solvent evaporation, reduced solvent consumption, less hazardous waste generation, smaller space requirements, increased automation, and the use of less glassware.However, SFE methods have not yet been demonstrated to extract a comparable number of pesticides possible by traditional multiresidue methods.16–18 The polarity range of the pesticides that can be extracted by SFE is limited by the use of CO2, a relatively nonpolar medium, but other supercritical fluids are currently impractical in analytical commercial applications. Other current limitations of SFE include the matrix dependency of extraction, typically small sample size, and frequent need for solvent modifiers.A modifier solvent, such as methanol, can be added to the CO2 to increase the polarity of the extraction medium, but with fruit and vegetable samples, water already present in the sample serves to modify the fluid, and addition of other solvent modifiers has no effect.9 The use of modifiers also complicates the optimization of extraction and trapping conditions.6,19–21 Derivatization is another option to improve SFE recovery of polar pesticides, such as 2,4-D,22–25 but in multiresidue analysis, derivatization methods are not desirable.Fruit and vegetable samples typically consist of 80–95% water, which must be removed or controlled before SFE or instrument performance will be affected. Lyophilization of the sample is a poor option due to the time and equipment required, the concentration of matrix interferants, and the loss of volatile analytes.10 Furthermore, the presence of water in the sample is often beneficial to the extraction process.8,9,26 Therefore, a drying agent must be used to control sample moisture. Hopper and King were the first to apply Hydromatrix (HMX), a pelletized diatomaceous earth material that absorbs twice its mass in water, to SFE.1 HMX not only served to absorb moisture, but also increased the recovery of pesticides from food samples due to increased sample dispersion and reduced particle size.However, HMX can retain certain polar pesticides, such as methamidophos, omethoate, and sulfonyl urea herbicides. 1,11,21 Burford et al. evaluated several other drying agents, including inorganic salts, in the SFE of organic pollutants from moist samples,27 but pesticide analytes or food matrices were not included in the study.Valverde-Garc�ýa et al. demonstrated that methamidophos, chlorpyrifos, endosulfans, methiocarb, and chlorothalonil, were extracted from vegetables when using anhydrous MgSO4 as the drying agent.13,14 The objective of this study was to further describe the effect of drying agents in the SFE of pesticides from matrices with high water content, and to demonstrate multiresidue capabilities of using a MgSO4–HMX drying agent mixture for sample preparation in SFE.Experimental Apparatus A Model 7680T (Hewlett-Packard, Little Falls, DE, USA) and an Autoprep-44 (Suprex, Pittsburgh, PA, USA), both equipped with automated extraction capability, automated variable restrictors, and solid sorbent collection systems, were used in this study.Most data reported in this paper were obtained with the 7680T, except for a set of extractions of the carrot check sample which was performed using both instruments. The SFE parameters were: 320 atm extraction pressure (1 atm = 101.325 kPa), 60 °C temperature (CO2 density of 0.85 g ml21), 2 min static extraction followed by dynamic extraction of six vessel volumes of CO2 at 1.6 ml min21 flow rate, 50 °C restrictor temperature, extract collection on a sorbent trap at 9 °C, and elution with 1.5 ml of acetonitrile at 0.4 ml min21 and 30 °C.The vessel size was 7 ml for the 7680T and 10 ml for the Autoprep-44. For the 7680T, the 1 ml Hypersil octyldecylsilane (ODS) trap was rinsed to waste with 2 ml of ethyl acetate followed by 2 ml of acetonitrile at 1 ml min21 each to clean and regenerate the trap between extractions.For the Autoprep-44, the 1 ml trap consisted of a 1 + 1 ODS–Unibeads sorbent Analyst, May 1997, Vol. 122 (429–435) 429prepared by Suprex, and the trap was flushed with 4 ml of acetonitrile at 1.8 ml min21 between extractions (60 psi N2 gas pressure to blow the trap dry, 1 psi = 6894.76 Pa).After 8–16 extractions, the 7680T tubing (after the vessel) and trap were flushed with Å 10 ml of 1 + 1 methanol–water. A Model ITS40 gas chromatography–ion-trap mass spectrometric detector (GC–ITD) instrument (Finnigan MAT, San Jose, CA, USA), consisting of a Varian 3300/3400 GC and a CTC A200S autosampler was used for multiresidue analyses.Operating conditions for the GC–ITD were: DB-5ms (J&W Scientific, Folsom, CA, USA), 30 m, 0.25 mm id, 0.25 mm film thickness, capillary column, 5 m phenyl–methyl deactivated (Restek, Bellefonte, PA, USA) guard column (0.25 mm id), 1 ml injection volume, Model 1093 (Varian, Walnut Creek, CA, USA) septum programmable injector (SPI); 55 °C injection port for 30 s followed by ramping to 230 °C min at 250 °C min21; 11.5 psi He column head pressure; 55 °C initial oven temperature for 30 s, ramped to 130 °C at 50 °C min21, then to 165 °C at 1.5 °C min21 and to 250 °C at 4 °C min21, and held at 250 °C until 51 min total time elapsed; 240 °C transfer line temperature; and 205 °C ion-trap manifold temperature.Typical ITD operating conditions were: electron impact mode, 12 mA filament current, 1800 V electron multiplier tube, and automatic gain control (AGC) at 30 000.The GC–ITD utilized a Magnum version 2.4 software package loaded into a Gateway 2000 computer for data collection and analysis. The data collection range was 60–420 m/z from 5 to 51 min for the analysis of the 71 pesticides. A Model 5890 (Hewlett-Packard) GC, equipped with a Model 7673 autosampler and flame photometric detector (FPD), was used to analyze methamidophos, chlorpyrifos, and diazinon in the drying agent study.For the GC–FPD, a SPB-1 (Supelco, Bellefonte, PA, USA) 30 m, 0.25 mm id, 0.25 mm film thickness, capillary column was used, and operating conditions were: 1 ml splitless injection; 220 °C injection port; 110 °C initial oven temperature for 1 min, ramped to 226 °C at 15 °C min21, then to 250 °C at 3 °C min21; 250 °C FPD.Reagents SFC/SFE Grade CO2 (Air Products, Allentown, PA, USA), with 1800 psi He headspace (for the Autoprep-44) or without He headspace or the 7680T), was used to perform SFE. Dry, commercial-grade CO2 was required for cryogenic cooling for the SFE instruments and the SPI on the GC–ITD.Anhydrous MgSO4 (99% purity) was obtained from Fisher (Fair Lawn, NJ, USA) and Aldrich (Milwaukee, WI, USA), and MgSO4·H2O was made by mixing 18 g of H2O with 120 g of MgSO4. HMX was obtained from Varian (Harbor City, CA, USA). All organic solvents were pesticide grade quality (Fisher), and all pesticides were obtained from the US Environmental Protection Agency (Research Park, NC, USA or Beltsville, MD, USA).Chrysened12 (Cambridge Isotope Laboratories, Woburn, MA, USA) was used as the internal standard for GC–ITD, and terbufos was used as the internal standard for GC–FPD. Individual stock solutions with typical concentrations of 2 mg ml21 were prepared in acetone, and these solutions were used to make three working standard mixtures of 100 mg ml21 for each pesticide in acetone consisting of ‘organochlorine,’ ‘organophosphorus,’ and ‘other’ pesticides.Sample Preparation Produce samples were purchased at a local supermarket to serve as blank or fortified samples. Apple, carrot, green bean, orange, and pea check samples were provided by the California Department of Food and Agriculture (CDFA) as part of a quality assurance protocol for the laboratories participating in the Pesticide Data Program,28 and green bean samples containing field residues of methamidophos and acephate were provided by the Michigan Department of Agriculture.The 100 g frozen samples were each mixed with 100 g of MgSO4·H2O and 50 g of HMX in a chopper (Black & Decker; Shelton, CT, USA). [Safety Note: A filter mask should be worn to prevent breathing in of small particles.] A small amount of dry ice was added to the samples during mixing to maintain frozen conditions.12 The 7 ml vessels were packed with 6 g of the cold homogenate (2.4 g of produce sample) and 10 ml vessels were packed with 8 g (3.2 g of produce sample).Filter paper, Whatman # 3 (Maidstone, Kent, UK) was cut into disks which were placed at both ends to keep particles from affecting sealing of the vessels.The vessels were kept cold before extraction in order to reduce losses of degradative pesticides.12 In the case of the fortified samples, a mortar and pestle were used to mix 10 g of sample (after fortification) with 10 g of MgSO4·H2O and 5 g of HMX. The homogenized mixtures, 6 g (2.4 g of sample), were loaded into the 7 ml vessels and extracted by SFE.Analysis The same solutions as the spiking solutions were diluted to make the calibration standards. After extraction, typically 15 ml of 100 mg ml21 chrysene-d12 solution were added to each 1.5 ml extract as an internal standard. The calibration standards were prepared in SFE extracts from sample blanks of the same matrix29 or the method of standard additions was used for check samples as described previously.11,12 Table 1 lists the pesticides, retention times, quantitation masses, and limits of detection (LODs) for the analysis of apple in this study.The pesticides consist of 28 organophosphorus, 19 organochlorine, 6 pyrethroid, 4 carbamate, and 13 other insecticides, herbicides, and fungicides. The LOD values, the concentration at which signal/noise is 3, were determined using calibration data from the pesticide recovery experiment in apple in the manner reported previously for GC–ITD.11 The LODs vary depending on matrix and the conditions of the GC–ITD; the reported values are typical of an apple matrix.Results and Discussion Drying Agents Before performing SFE of produce samples, experiments were carried out to compare drying agent properties of MgSO4 and HMX.To determine the amount of material transported by the supercritical CO2, the masses of the loaded vessels were measured before and after SFE. To determine the water retention capabilities of the drying agents, the vessel contents after SFE were transferred to a watch glass and were baked at 250 °C for 24 h.The amount of water extracted during SFE was determined by subtracting the post-SFE water content in the sample from the original water content. Table 2 contains the results of this experiment. The solubility of H2O in supercritical CO2 is Å 0.3%,30 which corresponds to Å 110 mg of H2O for the 38 g of CO2 used in our extractions. As shown in Table 2, up to 160 mg of water were lost from the drying agents during SFE.A conclusion from this experiment was that MgSO4 retained water more strongly than HMX. More importantly, these results documented how the drying agents do not retain all of the moisture in the system, and water dissolved in the supercritical CO2 is available as a ‘modifier’ in SFE. Not all of the mass loss could be accounted for by water alone; up to 90 mg of material was missing in some cases.After a series of extractions of samples containing MgSO4, a build-up of fine particles appeared at the frit between the vessel and restrictor. When using MgSO4 in routine applications, this frit should be changed frequently, and a filter paper with pores < 2 mm should be used to contain the sample in the SFE vessel.Plugging of the automated variable restrictor is a potential 430 Analyst, May 1997, Vol. 122problem with MgSO4, especially if there is no filter placed before the restrictor, but with linear restrictors, plugging due to MgSO4 is not a problem.13,14 Effect of Drying Agent :H2O Ratio Recovery studies of three representative pesticides with different degrees of polarity were performed in model matrices to determine the effect of water in SFE, and to optimize the drying agent :H2O (sample) ratio for produce samples.Fig. 1 gives the structures and the solubilities in water of the chosen organophosphorus insecticides, methamidophos, diazinon, and chlorpyrifos. Table 3 lists the results of matrices fortified at 1 mg g21 with the pesticides. On filter paper, water did not have a significant effect on the recoveries of diazinon and chlorpyrifos, but methamidophos recovery reached a maximum with 100–200 ml of H2O added to the vessel.Based on these results, methamidophos is slightly soluble in supercritical CO2 at density = 0.85 g ml21, but when the CO2 was ‘modified’ with H2O, methamidophos solubility increased. This amount of water, which ostensibly corresponded with the saturation of water in supercritical CO2, enabled the solubilization of methamidophos.However, the addition of more water reduced recovery of methamidophos, presumably due to the greater availability of water remaining in the vessel for the pesticide to partition into. Diazinon and chlorpyrifos are more soluble in supercritical CO2, and less soluble in water than methamidophos, and partitioned into the supercritical fluid phase.Note that the pesticides were spiked onto the filter paper while in the vessel (due to the inability to homogenize the sample). This spiking procedure often gives higher recoveries than if the spike solution is mixed into the sample before loading the vessel. In the case of HMX, methamidophos was not recovered, independent of the amount of water present.Nor was methamidophos recovered from anhydrous or monohydrated MgSO4, but recovery improved to 70–90% with a 2 + 1 or 1 + 1 MgSO4– H2O matrix. Based on results in Tables 2 and 3, Å 70 mg of water were required to dissolve methamidophos in CO2 under the SFE conditions. The differences between HMX and MgSO4 in the extraction of methamidophos were reported previously, 1,11,13,14 but no attempts of explanation were made.The most likely explanation is that HMX contains SiOH sites that can interact through hydrogen bonding with the amine group.31 MgSO4 contains no hydroxyl groups to interact with amines, and methamidophos is extracted from the matrix when enough water is present to modify the fluid.Furthermore, MgSO4 has a solubility in water of 34–90% by mass (depending on hydrated state and temperature);32 thus, available water is saturated with MgSO4 which acts to salt out methamidophos from the water phase into the supercritical CO2 phase. The same effect is observed in liquid–liquid partitioning when salt is added to an aqueous layer. In the case of HMX, absorbed water is more accessible than water hydrated with MgSO4, and due to its low solubility in water, HMX does not salt out methamidophos from the water.The pH of saturated aqueous solutions of both HMX and MgSO4 was neutral. Table 1 Pesticide retention times (tt), quantitation masses, and limits of detection (LOD) in SFE apple extracts using GC–ITD analysis LOD*/ LOD*/ No. Pesticide tt/s Masses (m/z) ng g21 No.Pesticide tt/s Masses (m/z) ng g21 1 Methamidophos 350 94† + 95 + 141 50 37 Malathion 1392 125 + 127 + 173† 20 2 Dichlorvos 358 109† + 185 2 38 Metolachlor 1398 162† + 238 7 3 Mevinphos 575 127† + 192 1 39 Chlorpyrifos 1404 197 + 199 + 314 6 4 Acephate 582 94 + 136† 200 40 Aldrin 1405 347–351 13 5 Tetrahydrophthalimide 683 79† + 151 8 41 Dacthal 1413 299–303 2 6 Pentachlorobenzene 725 248–252 0.3 42 Parathion 1421 97 + 109 + 291† 13 7 o-Phenylphenol 727 169 + 170† 2 43 Dicofol 1435 139† + 250 7 8 Omethoate 849 110 + 156† 100 44 Captan 1512 79† 7 9 Propoxur 883 110† + 152 4 45 Methidathion 1539 85 + 93 + 145† 0.9 10 Diphenylamine 916 167 + 168 + 169† 2 46 o,pA-DDE 1548 316–320 3 11 Ethoprop 934 97 + 158† + 243 3 47 Tetrachlorvinphos 1553 330–333 2 12 Chlorpropham 975 127† + 171 + 213 20 48 Disulfoton sulfone 1561 97 + 153 + 213† 16 13 Dicrotophos 988 67 + 127† + 237 200 49 Endosulfan I 1567 337–341 20 14 Trifluralin 993 264 + 306† 0.2 50 cis-Chlordane 1568 373–379 2 15 Monocrotophos 1018 67 + 127† + 223 200 51 Fenamiphos 1587 260 + 288 + 303† 50 16 Phorate 1032 75† + 121 + 260 3 52 p,pA-DDE 1618 316–320 2 17 Hexachlorobenzene 1046 282–290 0.9 53 Myclobutanil 1626 150 + 179† + 181 100 18 Dichloran 1078 176† + 206 7 54 Endosulfan II 1687 195† + 241 + 339 8 19 Dimethoate 1082 87† + 93 30 55 p,pA-DDD 1697 165 + 235† + 237 2 20 Carbofuran 1104 149 + 164† 9 56 cis-Nonachlor 1697 405–413 1 21 Quintozene (PCNB) 1121 293–301 5 57 Ethion 1698 97 + 153 + 231† 6 22 Atrazine 1122 200† 2 58 o,p-DDT 1701 165 + 235† + 237 2 23 Lindane (g-HCH) 1137 181 + 183† + 219 20 59 Carbophenothion 1748 157† + 199 + 342 11 24 Terbufos 1157 231† 2 60 Propargite 1819 135† + 335 + 350 30 25 Diazinon 1182 137 + 179† + 304 3 61 Iprodione 1871 314† + 316 10 26 Chorothalonil 1182 264–268 2 62 Phosmet 1887 160† 4 27 Disulfoton 1202 88† + 89 + 97 3 63 Methoxychlor 1913 227† 3 28 Phosphamidon 1277 127† + 264 12 64 Phosalone 1991 182† + 367 4 29 Propanil 1284 161† + 163 + 217 50 65 Azinphos-methyl 2003 77† + 132 + 160 50 30 Vinclozolin 1304 198 + 212† + 285 6 66 cis-Permethrin 2252 183† 10 31 Parathion-methyl 1307 109 + 125 + 263† 16 67 trans-Permethrin 2288 183† 10 32 Alachlor 1312 160 + 188† + 238 10 68 Cyfluthrin 2404 163† + 206 + 226 100 33 Carbaryl 1321 115 + 116 + 144† 10 69 Cypermethrin 2511 127 + 163† + 181 50 34 Heptachlor 1323 270 + 272† + 274 5 70 Fenvalerate 2878 125 + 225† + 419 40 35 Fenitrothion 1367 125 + 260 + 277† 3 71 Esfenvalerate 2986 125 + 225† + 419 40 36 Linuron 1379 61† + 248 5 IS‡ Chrysene-d12 1900 240† * As determined by SFE for a 2.4 g apple in 1.5 ml of final volume, 1 ml injection. † Base peak.‡ IS, internal standard.Analyst, May 1997, Vol. 122 431As designated by bold script in Table 3, an interesting effect was also observed for diazinon and chlorpyrifos in dry HMX and MgSO4. Dry HMX retained chlorpyrifos to some extent, but MgSO4 did not, whereas anhydrous MgSO4 retained diazinon, but HMX did not. In both instances, the presence of water, even the small amount in the MgSO4·H2O matrix, served to enhance the recoveries of the pesticides.The pronounced effect of water on SFE recoveries has been observed previously, but not with relatively nonpolar pesticides such as diazinon and chlorpyrifos. Campbell et al. observed a decreased recovery of chlorpyrifosmethyl in dry wheat and resorted to the use of 2% methanol modifier to improve recovery.15 Our results emphasize how the modifying and dispersing effect of water can serve a similar purpose as the addition of an organic solvent.MgSO4–HMX Mixtures HMX possesses practical advantages of a much lower heat of hydration, slightly lower cost (4c per gram versus 6c per gram), higher water absorptivity per mass basis, and better sample consistency than MgSO4. Also, HMX does not contain as many fine particles.However, the hydration/salt effect described above and higher density (enabling more sample per fixed vessel volume) presents the advantages of MgSO4. A drawback of MgSO4 is the formation of agglomerates when mixed with water. The combination of HMX and MgSO4 drying agents greatly reduced the problems associated with the individual drying agents, and appeared to maintain their advantages.As shown in Table 3, the addition of HMX as a minor component with MgSO4 did not adversely affect pesticide recoveries, and served to improve the practical aspects of sample preparation for SFE. The ratio of 1 + 0.4 + 1 MgSO4–HMX–H2O was found to be most useful because higher HMX content affected methamidophos recovery, and higher water content made the matrix too moist, leading to reduced recoveries.More sample could be loaded in vessels than with HMX alone, and sample consistency remained powdery and dispersed. In the analysis of real samples, MgSO4·H2O was used rather than anhydrous MgSO4, and for convenience, the sampling ratio was changed to 1 + 0.5 + 1 MgSO4·H2O–HMX–sample. The use of MgSO4 reduced the heat generated when the sample was mixed with the drying agent.Analysis of Acephate and Methamidophos in Green Beans Extraction and analysis of acephate and methamidophos, insecticides that are commonly applied to green vegetables, has been troublesome by SFE and GC–ITD previously.11,12 In a test of the SFE method using MgSO4–HMX as the drying agent, analyses of green beans previously analyzed by the Michigan Department of Agriculture were performed.GC–ITD data for the green bean sample extracted by SFE is shown in Fig. 2, and quantitative results are listed in Table 4 (the LOD for acephate was lower in green bean than the value reported for apple in Table 1). Despite the relative difficulties involved in GC–ITD analysis of these pesticides, pesticide concentrations agreed within the error of analysis in each case.The large deviation (60% RSD) in the acephate concentration was mainly due to variations in GC–ITD analysis and not SFE. Acephate response fluctuated with the cleanliness of the GC system, and the variability in these results reflected the effect of matrix components in the injection liner and precolumn.29 The use of GC–FPD gives better peak shapes and a lower LOD than GC– ITD for these compounds, but GC–ITD makes mass spectral confirmation possible.Table 2 Amount of water transported from drying agents by SFE Water in Water in Water vessel vessel trans- Sample before after ported by Sample matrix size/g SFE/mg SFE/mg SFE/mg Hydromatrix (HMX), dry 2.0 0 0 0 HMX–H2O (2 + 1) 3.3 1100 980 120 HMX–H2O (1 + 1) 4.0 2000 1840 160 MgSO4, anhydrous 6.0 0 0 0 MgSO4–H2O (6.7 + 1) 6.3 820 810 10 MgSO4–H2O (2 + 1) 4.5 1500 1430 70 MgSO4–H2O (2 + 1) 6.0 2000 1920 80 MgSO4–H2O (1 + 1) 6.0 3000 2900 100 MgSO4–HMX–H2O (1 + 0.4 + 1) 5.0 2080 1940 140 MgSO4–HMX–H2O (1 + 0.4 + 1) 6.5 2710 2560 150 Fig. 1 Structures and solubilities in H2O of methamidophos, diazinon, and chlorpyrifos. Table 3 Effect of H2O on SFE % recoveries (± s) of organophosphorus pesticides on different matrices (1 mg g21 spikes); results from GC–FPD analysis; values in bold show the interesting results for diazinon and chlorpyrifos in dry HMX and MgSO4 Sample matrix n Methamidophos Diazinon Chlorpyrifos Filter paper*, dry 2 65.4 ± 3.2 102 ± 17 102 ± 17 Filter paper + 50 ml H2O 2 67 ± 16 91 ± 10 93.1 ± 7.7 Filter paper + 100 ml H2O 2 92 ± 16 87.5 ± 7.9 90.5 ± 2.8 Filter paper + 200 ml H2O 2 117 ± 13 82.2 ± 0.4 90.1 ± 0.3 Filter paper + 300 ml H2O 2 48 ± 31 84.5 ± 3.0 87.1 ± 2.5 Filter paper + 600 ml H2O 2 55.3 ± 3.6 81.3 ± 4.4 87.8 ± 3.9 HMX†, dry 3 ND‡ 87.6 ± 0.8 36.9 ± 0.8 HMX–H2O (2 + 1) 4 ND 98.0 ± 0.7 97.0 ± 1.3 HMX–H2O (1 + 1) 4 ND 86.0 ± 4.3 91.0 ± 2.7 MgSO4, anhydrous 2 ND 11.3 ± 3.0 86.8 ± 8.9 MgSO4–H2O (6.7 + 1) 2 ND 82.3 ± 0.4 84.1 ± 1.3 MgSO4–H2O (2 + 1) 3 86.6 ± 8.3 95 ± 11 88.1 ± 4.6 MgSO4–H2O (1 + 1) 3 72 ± 16 69 ± 16 72 ± 18 MgSO4–HMX (2 + 3), dry 2 ND 12.9 ± 1.4 54.5 ± 6.5 MgSO4–HMX–H2O (1 + 0.4 + 1) 2 90.5 ± 9.6 91.2 ± 7.8 89.3 ± 8.3 MgSO4–HMX–H2O (0.6 + 0.4 + 1) 2 85 ± 14 88 ± 16 88 ± 11 * Filter paper was spiked in the SFE vessel. † Drying agents were spiked outside the SFE vessel.‡ ND, not detected. 432 Analyst, May 1997, Vol. 122Analysis of Fortified Samples Recovery studies of apple samples fortified at 1 mg g21 (high spike) and 0.05 mg g21 (low spike) were performed (triplicate extractions). As shown in Table 5, all 71 pesticides were determined at the high spiking level with an average recovery of 86.7% (5.9% average RSD), and only cis-nonachlor (60.8%) had a recovery < 70%.Fig. 3 is a GC–ITD total ion chromatogram of a 0.4 mg g21 apple standard for the pesticides, and it displays the low background associated with the SFE method. For the low spiking level, the fortification level was less than the LOD for several pesticides which were not detected, and confirmation criteria were not met for a few other pesticides as noted in Table 5.The 59 other pesticides were confirmed and quantified with an average recovery of 88.9% (6.9% average RSD). Six pesticides were recovered from the low spike at less than 70%, one of which was cis-nonachlor (57%) again. The likely source of the lower recovery of cis-nonachlor was incomplete elution of the trap with acetonitrile. The use of acetone was shown to elute nonpolar pesticides from the ODS Fig. 2 GC–ITD chromatogram peaks of methamidophos and acephate in green bean sample no. 2 after SFE. Mass spectral confirmation is presented for acephate. Table 5 Presticide recoveries from apple fortified at 1 mg g21 (high spike) and 0.05 mg g21 (low spike) using the SFE and GC–ITD method (n = 3) High spike Low spike High spike Low spike recovery recovery recovery recovery No.Pesticide (% RSD) (% RSD) No. Pesticide (% RSD) (% RSD) 1 Methamidophos 72.3 (7.5) NC* 37 Malathion 89.8 (4.3) 82 (8.8) 2 Dichlorvos 86.1 (9.5) 88 (6.0) 38 Metolachlor 80.5 (5.7) 94 (0) 3 Mevinphos 95.3 (1.2) 81 (5.1) 39 Chlorpyrifos 92.2 (2.8) 99 (7.1) 4 Acephate 86.5 (7.3) ND† 40 Aldrin 82.7 (0.9) 96 (2.1) 5 Tetrahydrophthalimide 92.10 92 (3.2) 41 Dacthal 81.3 (11) 101 (1.2) 6 Pentachlorobenzene 75.30 86 (4.6) 42 Parathion 92.1 (1.8) 79 (1.5) 7 o-Phenylphenol 89.7 (4.0) 103 (4.1) 43 Dicofol 85.7 (7.0) 95 (1.2) 8 Omethoate 102 (3.0) ND 44 Captan 91.9 (4.1) 60 (15) 9 Propoxur 89.1 (6.2) 95 (5.3) 45 Methidathion 88.9 (7.9) 83 (12) 10 Diphenylamine 80.7 (7.6) 87 (5.3) 46 o,pA-DDE 84.1 (2.9) 95 (2.4) 11 Ethoprop 86.5 (1.9) 93 (5.4) 47 Tetrachlorvinphos 93.3 (2.3) 95 (4.4) 12 Chlorpropham 80.9 (8.4) 114 (10) 48 Disulfoton sulfone 90.8 (7.0) 80 (22) 13 Dicrotophos 97.6 (4.1) ND 49 Endosulfan I 82.0 (3.2) 107 (7.5) 14 Trifluralin 93.2 (1.7) 78 (2.6) 50 cis-Chlordane 83.1 (1.5) 92 (6.5) 15 Monocrotophos 89.6 (4.7) ND 51 Fenamiphos 91.9 (15) ND 16 Phorate 87.0 (13) 83 (9.2) 52 p,pA-DDE 83.4 (2.2) 95 (4.4) 17 Hexachlorobenzene 82.7 (3.3) 95 (3.2) 53 Myclobutanil 87.8 (10) ND 18 Dicloran 91.5 (3.1) 89 (5.6) 54 Endosulfan II 84.8 (1.8) 132 (7.4) 19 Dimethoate 94.4 (5.8) 111 (7.3) 55 p,pA-DDD 86.9 (3.1) 99 (4.6) 20 Carbofuran 89.9 (5.9) 95 (3.2) 56 cis-Nonachlor 60.8 (9.2) 57 (16) 21 Quintozene 87.0 (3.0) 89 (1.3) 57 Ethion 92.9 (6.8) 71 (6.5) 22 Atrazine 86.6 (6.0) 98 (2.0) 58 o,pA-DDT 87.3 (14) 80 (6.6) 23 Lindane 85.1 (5.6) 123 (7.3) 59 Carbophenothion 92.1 (5.4) 71 (8.6) 24 Terbufos 89.8 (9.1) 65 (9.8) 60 Propargite 80.6 (8.9) 95 (2.4) 25 Diazinon 89.1 (3.3) 101 (5.0) 61 Iprodione 81.1 (12) 63 (51) 26 Chorothalonil 74.3 (3.2) 89 (1.3) 62 Phosmet 91.5 (4.6) 65 (12) 27 Disulfoton 87.9 (2.8) 82 (7.3) 63 Methoxychlor 85.0 (2.5) 73 (7.9) 28 Phosphamidon 97.5 (9.3) ND 64 Phosalone 92.2 (4.2) 75 (14) 29 Propanil 93.5 (2.9) NC 65 Azinphos-methyl 95.2 (3.6) NC 30 Vinclozolin 79.4 (7.6) 99 (8.1) 66 cis-Permethrin 80.1 (9.3) 80 (16) 31 Parathion-methyl 92.3 (2.8) 87 (3.5) 67 trans-Permethrin 82.3 (9.8) 79 (1.5) 32 Alachlor 79.8 (7.7) 92 (2.2) 68 Cyfluthrin 83.8 (16) ND 33 Carbaryl 92.0 (5.5) 89 (4.7) 69 Cypermethrin 84.0 (10) ND 34 Heptachlor 83.6 (1.6) 96 (4.2) 70 Fenvalerate 81.1 (15) 97 (8.4) 35 Fenitrothion 91.8 (1.7) 78 (2.6) 71 Esfenvalerate 85.4 (12) 115 (1.0) 36 Linuron 76.8 (5.0) 65 (7.1) average 86.7 (5.9) 88.9 (6.9) * NC, Detected, but confirmation criteria not met.† ND, Not detected (LOD > 0.05 mg g21). Table 4 Comparison of SFE method results versus traditional method results for acephate and methamidophos in green beans Sample Michigan lab.SFE results†/ pesticide results*/mg g21 mg g21 Sample— Acephate 0.65 0.98 ± 0.60 Methamidophos 0.12 0.115 ± 0.023 Sample 2— Acephate 0.18 0.23 ± 0.13 Methamidophos 0.074 0.087 ± 0.010 * Method of ref. 16, GC–FPD analysis. † GC–ITD analysis. Analyst, May 1997, Vol. 122 433trap in less volume than acetonitrile.33 Linuron, iprodione, and phosmet calibration data were not as linear as other pesticide calibration plots, and lower recoveries for these three pesticides at the 0.05 mg g21 level were believed to be a result of calibration bias.Captan readily degrades, and terbufos, as well as disulfoton, phorate, fenamiphos, and phosphamidon, may convert into sulfoxide and sulfone forms. Recoveries of the 52 other pesticides were independent of the concentration, and the slightly higher deviation in the values can be ascribed to the intricacies of analysis at lower concentrations.Analysis of Multiresidue Check Samples Based on the recovery data, the SFE and GC–ITD method using the HMX–MgSO4 mixture drying agent was ready to be tested in the extraction and analysis of actual samples.Table 6 presents the results of analyses of six check samples using the proposed method compared with results obtained by 7–9 state regulatory laboratories using traditional methods.16,17 In all, 10 check samples have been analyzed by SFE and GC–ITD,11,12 and there have been no instances of false negatives or false positives for the pesticides of interest (confirmation of pesticides by mass spectrometry is a prerequisite for reporting results).However, the quantitation of results between the traditional methods and the SFE and GC–ITD method agreed well in only six of the 20 cases. Of the 6 instances of agreement, only ethion (in orange) can be considered a nonpolar pesticide. Based on these results, the use of mixed HMX–MgSO4 drying agent gave low recoveries of nonpolar pesticides in real samples, and high recoveries of the relatively polar pesticides, omethoate, dimethoate, carbaryl, and diphenylamine. The cause of this difference between the fortified (Table 2) and check sample results (Table 6) was possibly related to the difference in sample preparation procedures of using a mortar and pestle for the small fortified samples versus a blender for the 100 g check samples.However, due to the agreeable results for polar pesticides, and good reproducibility of pesticide results for multiple subsamples, the blending procedure developed previously for HMX alone adequately homogenized the extracts.12 The more likely reason that nonpolar pesticides were not extracted as efficiently as the more polar pesticides in real samples is related to the higher degree of difficulty in extracting nonpolar analytes from aged samples than from fortified samples.34,35 Further investigations to address this problem will be conducted in the future.The USDA Agricultural Marketing Service (Pesticide Data Program) helped fund this research and provided the ITS40 instrument. The authors are also grateful to Hewlett-Packard and Suprex for providing use of the 7680T and Autoprep-44, respectively. Also, the authors thank the CDFA for providing the check samples, and the Department of Agriculture laboratories of California, Florida, Michigan, New York, North Carolina, Ohio, Texas, and Washington for performing traditional analyses, and the Michigan Department of Agriculture for providing and analyzing the green bean samples.The mention of specific items of equipment and chemicals by brand names do not constitute endorsement of a product by the USDA. Fig. 3 Total ion chromatogram of an SFE apple extract fortified at 0.4 mg g21 with 71 pesticides. Numbered peaks correspond to pesticides listed in Table 1, and peaks designated with * are matrix components. 434 Analyst, May 1997, Vol. 122References 1 Hopper, M. L., and King, J. W., J. Assoc. Off. Anal. Chem., 1991, 74, 661. 2 Nishikawa, Y., Anal. Sci., 1991, 7, 567. 3 Thomson, C. A., and Chesney, D. J., Anal. Chem., 1992, 64, 848. 4 King, J. W., Snyder, J. M., Taylor, S. L., Johnson, J. H., and Rowe, L. D., J. Chromatogr. Sci., 1993, 31, 1. 5 Wigfield, Y. Y., and Lanouette, M., J.Agric. Food Chem., 1993, 41, 84. 6 Howard, A. L., Braue, C., and Taylor, L. T., J. Chromatogr. Sci., 1993, 31, 323. 7 King, J. W., Hopper, M. L., Luchtenfeld, R. G., Taylor, S. L., and Orton, W. L., J. AOAC Int., 1993, 76, 857. 8 Skopec, Z. V., Clark, R., Harvey, P. M. A., and Wells, R. J., J. Chromatogr. Sci., 1993, 31, 445. 9 Aharonson, N., Lehotay, S. J., and Ibrahim, M. A., J.Agric. Food Chem., 1994, 42, 2817. 10 Lehotay, S. J., and Ibrahim, M. A., J. AOAC Int., 1995, 78, 445. 11 Lehotay, S. J., and Eller, K. I., J. AOAC Int., 1995, 78, 821. 12 Lehotay, S. J., Aharonson, N., Pfeil, E., and Ibrahim, M. A., J. AOAC Int., 1995, 78, 831. 13 Valverde-Garc�ýa, A., Fern�andez-Alba, A. R., Ag�uera A., and Contreras, M., J. AOAC Int., 1995, 78, 867. 14 Valverde-Garc�ýa, A., Fern�andez-Alba, A.R., Contreras, M., and Ag�uera A., J. Agric. Food Chem., 1996, 44, 1780. 15 Campbell, R. M., Meurnier, D. M., and Cortes, H. J., J. Microcolumn Sep., 1989, 1, 302. 16 Luke, M. A., and Masumoto, H. T., Anal. Methods Pestic. Plant Growth Regul., 1986, 15, 161. 17 Lee, S. M., Papathakis, M. L., Feng, H. M. C., Hunter, G. F., and Carr, J. E., Fresenius’ J. Anal. Chem., 1991, 339, 376. 18 Cairns, T., Luke, M. A., Chiu, K. S., Navarro, D., and Siegmund, E. G., Rapid Commun. Mass Spectrom., 1993, 7, 1070. 19 Langenfeld, J. L., Hawthorne, S. B., Miller, D. J., and Pawliszyn, J., Anal. Chem., 1994, 66, 909. 20 Mulcahey, L. J., and Taylor, L. T., Anal. Chem., 1992, 64, 2352. 21 Howard, A. L., and Taylor, L. T., J. High Res. Chromatogr. 1993, 16, 39. 22 Hawthorne, S. B., Miller, D. J., Nivens, D. E., and White, D. C., Anal. Chem., 1992, 64, 405. 23 Field, J. A., Miller, D. J., Field, T. M., Hawthorne, S. B., and Giger, W., Anal. Chem., 1992, 64, 3161. 24 Hillmann, R., and Bachman, K., J. High Res. Chromatogr., 1994, 17, 350. 25 Chatfield, S. N., Croft, M. Y., Dang, T., Murby, G. Y. F., and Wells, R. J., Anal. Chem., 1995, 67, 945. 26 Fahmy, T. M., Paulitis, M. E., Johnson, D. M., and McNally, M. E. P., Anal. Chem., 1993, 65, 1462. 27 Burford, M. D., Hawthorne, S. B., and Miller, D. J., J. Chromatogr. A, 1993, 657, 413. 28 Agricultural Marketing Service, Pesticide Data Program (PDP) Summary of 1994 Data, USDA, Washington, DC, 1996. 29 Erney, D. R., Gillespie, A. M., Gilvydis, D. M., and Poole, C. F., J. Chromatogr., 1993, 638, 57. 30 Kuk, M. S., and Montagna, J. C., Chemical Engineering at Supercritical Conditions, ed. Paulitis, M. E., Penninger, J. M., Gray, R. D., and Davison, K. P., Ann Arbor Science, Ann Arbor, MI, 1983, pp. 101–111. 31 Oostdyk, T. S., Grob, R. L., Snyder, J. L., and McNally, M. E. P., J. Chromatogr. Sci., 1993, 31, 177. 32 Lange’s Handbook of Chemistry, ed. Dean, J. A., McGraw-Hill, New York, 13th edn. 1985, pp. 4–74. 33 Lehotay, S. J., and Valverde-Garc�ýa, A., J. Chromatogr. A, in the press. 34 Burford, M. D., Hawthorne, S. B., and Miller, D. J., Anal. Chem., 1993, 65, 1497. 35 Camel, V., Tambut�e, M., and Caude, M., J. Chromatogr. A, 1995, 693, 101. Paper 6/07754A Received November 6, 1996 Accepted January 19, 1997 Table 6 Comparison of results (concentrations in mg g21) of SFE method versus traditional methods of analysis (refs. 16, 17) performed by state regulatory laboratories State lab. results Commodity Spiking SFE method pesticide conc. Conc. (n = 9) High Low conc. (n = 3) Apple— Azinphos-methyl incurred* 0.110 ± 0.045 0.20 0.033 0.064 ± 0.003 Chlorothalonil 0.16 0.180 ± 0.065 0.29 0.12 0.104 ± 0.025 Diphenylamine 0.82 0.64 ± 0.13 0.86 0.43 0.580 ± 0.088 Fenamiphos 0.088 0.060 ± 0.020 0.091 0.034 0.022 ± 0.006 o-Phenylphenol incurred NR† < 0.01 Phosmet incurred NR < 0.01 Apple— Carbaryl 0.482 0.451 ± 0.070 0.61 0.40 0.444 ± 0.036 Chlorpyrifos 0.114 0.122 ± 0.024 0.16 0.10 0.088 ± 0.007 Diphenylamine incurred 0.93 ± 0.23 1.2 0.55 0.97 ± 0.11 Propargite 0.814 0.686 ± 0.088 0.80 0.53 0.309 ± 0.034 Carrot— p,pA-DDE incurred 0.028 ± 0.022 0.050 0 0.020 ± 0.001‡ Parathion-methyl 0.117 0.110 ± 0.016 0.14 0.090 0.036 ± 0.010‡ Trifluralin incurred 0.111 ± 0.040 0.20 0.070 0.030 ± 0.007‡ Green bean— Chlorothalonil 0.16 0.120 ± 0.056 0.24 0.052 0.034 ± 0.010 p,pA-DDE incurred 0.023 ± 0.031 0.10 0.009 0.004 ± 0.003 Disulfoton 0.18 0.056 ± 0.020 0.079 0.027 0.022 ± 0.009 Disulfoton sulfone incurred NR < 0.01 Endosulfan I 0.052 0.053 ± 0.011 0.08 0.041 0.022 ± 0.004 Orange— Chlorothalonil 0.16 0.120 ± 0.060 0.22 0.04 0.104 ± 0.019 Ethion 0.075 0.071 ± 0.018 0.09 0.04 0.073 ± 0.007 Pea— Dimethoate 0.090 0.092 ± 0.018 0.12 0.060 0.081 ± 0.006 Methoxychlor 0.206 0.181 ± 0.048 0.25 0.080 0.058 ± 0.005 Omethoate 0.084 0.083 ± 0.019 0.10 0.050 0.098 ± 0.064 * Incurred, pesticide applied in field. † NR, not reported; ‡ Results from two sets of analyses using different SFE instruments. Analyst, May 19

ISSN:0003-2654    

DOI:10.1039/a607554a

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Selective Precipitation Separation and Inductively Coupled Plasma Mass Spectrometric Determination of Trace Metal Impurities in High Purity Silver Yuh-Chang Suna, Jerzy Mierzwaa, Chien-Feng Lina, T. I. Yehb, and Mo-Hsiung Yang*a a Department of Nuclear Science, National Tsing Hua University, 30043 Hsinchu, Taiwan b Center for Measurement Standards, Industrial Technology Research Institute, 30043 Hsinchu, Taiwan A simple and rapid method for the determination of some trace element impurities in high purity silver, combining the isolation of analytes from the silver matrix with selective precipitation followed by ICP-MS determination was developed.On the basis of an extreme difference in the solubilities of the chlorides of silver and the other accompanying trace elements, silver can be separated completely through the addition of hydrochloric acid. The sample of silver was at first dissolved in 7 M nitric acid followed by addition of hydrochloric acid to remove the silver matrix by formation of a silver chloride precipitate, while leaving the trace element impurities in the solution, which was subsequently analysed by ICP-MS.Eleven elements (Al, Au, Cd, Co, Cu, Fe, Mg, Mn, Ni, Pb and Sn) were determined with good accuracy and precision. The limits of detection (based on the 3s criterion) of these elements were 1021–1023 ng g21. The proposed method was successfully applied to the determination of metal impurities in high-purity silver samples (EM9465 and EM 9343) and validated by the analysis of NIST SRM 8171 (Fine Silver FS 14).Keywords: Selective precipitation; inductively coupled plasma mass spectrometry; silver To control and to improve the manufacturing technology of high purity materials (e.g., high purity metals), it is necessary to develop methods for the determination of trace element impurities at levels as low as ng g21. One of the additional reasons for the determination of trace element impurities in high purity silver is the effect of impurities on the freezing and melting properties when silver is used for primary and secondary temperature calibration.1 Spark source mass spectrometry and instrumental neutron activation analysis are widely used for the routine determination of trace impurities in high purity metals.2 However, owing to the unavailability of suitable standards and the difficulties connected with matrix interference, the quantitative applications of these physical methods are severely restricted,3 whereas the combination of chemical pre-treatment processes with instrumental analysis always gives the possibility of achieving better reliability in the trace analysis of some special materials.By using a suitable chemical treatment, the sample matrix or coexisting elements can be separated effectively and the aqueous calibration standards can easily be applied owing to the prior simplification of the sample solutions.Atomic spectrometric methods, e.g., ETAAS4–10 and ICP-OES,10,11 and electrochemical methods,12–14 sometimes combined with preconcentration and/or separation procedures, are also employed for the precise determination of trace impurities. With the advent of ICP-MS, many procedures11,15 for the determination of trace element impurities in high purity materials have appeared; however, the interferences resulting from the parent matrix and the limits of detection obtained usually raise the difficulty of the direct determination of trace impurities without a preliminary treatment.Methods of analysis of high purity materials (high purity metals) involve the separation of trace impurities from the dissolved parent matrix by, e.g., solid–liquid extraction,9 ion exchange,13 coprecipitation4,6 and electrodeposition.5,16–20 However, the determination of trace impurities in high purity materials involving multi-stage combined procedures increases the risk of contamination and worsening of the detection limits.To attain high sensitivity and reliability, the analytical blank and systematic error inherent in extreme trace analysis should be very critically controlled.21 An alternative way is to use spectroanalytical techniques with direct solid (slurry) sampling; e.g., Hinds22 used slurry and solid ETAAS for the determination of gold, palladium and platinum in high purity silver. Kogan et al.23 studied the determination of some trace metals in gold and silver samples by laser ablation ICP-MS without matrix matched standards.In order to increase the detection power of the proposed analytical method and to avoid the possibility of contamination, our efforts were directed to developing a relatively simple chemical separation procedure for the elimination of the parent matrix elements. A simple precipitation procedure based on the formation of silver chloride precipitate and remaining trace element impurities in the solution can be performed by the addition of high purity hydrochloric acid to precipitate silver ions in nitrate solution.The elements Al, Au, Cu, Cd, Co, Fe, Mg, Mn, Ni, Pb and Sn can be quantitatively separated from the precipitate of silver chloride and very low blank values can be achieved. The ICP-MS measurements were optimized and potential interferences were also investigated. Experimental Apparatus An Elan 5000 inductively coupled plasma mass spectrometer (Perkin-Elmer SCIEX, Thornhill, ON, Canada) equipped with a conventional pneumatic nebulization system was used.The main operating parameters for ICP-MS measurements are given in Table 1. Reagents, Containers and Samples The high purity water used in this study was obtained by purification through de-ionization, double distillation and subboiling distillation. The purification of nitric acid and hydrochloric acid was carried out by recycled sub-boiling distillation of the analytical-reagent grade acids three times.PTFE and glass containers were used throughout the work and were cleaned by immersion in concentrated HNO3 overnight and in concentrated HCl overnight and then steaming successively with HNO3 for 8 h and water vapour for 8 h. Samples of high purity silver EM9343 (6A9 grade silver shot) and EM9465 (5A9 grade silver shot) were obtained from Johnson Analyst, May 1997, Vol. 122 (437–440) 437Matthey Electronics (UK). Silver SRM 8171 (Fine Silver FS 14 Block) was obtained from NIST (Gaithersburg, MD, USA).This material was analysed and prepared by the Royal Canadian Mint (Ottawa, Canada). Sample Cleaning A 20.0 g sample of high-purity silver was weighed into a 30 ml Pyrex beaker and 25 ml of cold 0.1 m nitric acid were added with approximately 10 min of agitation, followed by thorough rinsing in doubly distilled, de-ionized water. After drying by purging with nitrogen in a Class-100 clean bench, the sample was collected in a desiccator.A similar procedure was also applied to silver SRM 8171. Sample Pre-treatment A 1.00 g sample of high purity silver (0.1–0.3 g of silver SRM) was weighed and 3 ml of water and 3 ml of concentrated nitric acid (1 ml of water + 1 ml of concentrated nitric acid) were added to it in a 20 ml PTFE beaker. The sample was heated below the boiling-point of nitric acid until complete dissolution of the silver sample was achieved, then 4 ml of 2.3 m hydrochloric acid were added progressively to form a fine precipitate of silver chloride.After addition of 2 ml of concentrated hydrochloric acid and about 20 h of agitation, the solution was filtered with a 0.45 mm membrane filter. The filtrate was made up to 25 ml with high purity water in a calibrated flask and then trace metals in the filtrate were determined by ICP-MS. Results and Discussion Precipitation Separation Procedure To achieve high sensitivity and accuracy of the determination of trace impurities in silver, an analytical method based on the separation of the matrix element prior to the determination of isolated trace elements was developed. It is noteworthy that coprecipitation should be avoided during the establishment of a precipitation process which permits the effective separation of impurities from the sample matrix.It was found that 40–60% of spiked elements of interest were lost during the formation of the precipitate of silver chloride under neutral conditions.The presence of acid was found to be an effective way to prevent the occurrence of coprecipitation and to release the adsorbed trace impurities ions from the precipitate of silver chloride. The effects of nitric acid and hydrochloric acid concentration on the recovery of trace elements are shown in Figs. 1 and 2, respectively. The experiments were carried out in doubly distilled water spiked with 10 ppb of the elements of interest. Quantitative recoveries of most of the spiked elements from the silver chloride precipitate were observed for concentrations of nitric acid up to 3 m and concentrations of hydrochloric acid from 2 to 3 m.It can be seen from Fig. 1 that gold remains in the precipitate in spite of the addition of nitric acid. The possible reason is that gold cannot be dissolved by nitric acid, whereas it can be easily dissolved by hydrochloric acid owing to the formation of AuCl24 ,24 as shown in Fig. 2. Partial adsorption of cadmium and lead on the precipitate was observed when hydrochloric acid was used (see Fig. 2). Fortunately, the adsorbed cadmium and lead could be readily released from the precipitate by addition of nitric acid. With a view to achieving the complete separation of trace element impurities from the silver matrix and minimizing the amounts of reagents, a compromise composition of the precipitant for precipitation separation and ICP-MS measurements was chosen as a mixture of 2.8 m nitric acid and 2.4 m hydrochloric acid.Table 1 Main operating parameters of ICP-MS instrument ICP radiofrequency power 1.0 kW Nebulizer type Cross-flow Plasma gas flow rate 15 l min21 Auxiliary gas flow rate 0.8 l min21 Nebulizer gas flow rate 0.9 l min21 Sampler/skimmer Nickel Sampling depth 18 mm Data acquisition— Dwell time 200 ms Scan mode Peak hopping Points per peak 3 Signal measurement Integrated counts Fig. 1 Effect of nitric acid concentration on the recovery of metal impurities. Fig. 2 Effect of hydrochloric acid concentration on the recovery of metal impurities. 438 Analyst, May 1997, Vol. 122Determination by ICP-MS The excellent detection power afforded by ICP-MS has been used to great advantage in many analytical applications. The determination of trace element impurities in acids of high concentration is particularly demanding, requiring rigorous optimization of the analytical instrument prior to the measurement.Therefore, the effects of nitric acid and hydrochloric acid concentrations on the ICP-MS measurements were investigated. With a view to releasing the coprecipitated trace elements from the silver chloride precipitate, a mixture of concentrated nitric and hydrochloric acids was used. However, some of the elements analysed might suffer from spectral interference from polyatomic species, the disturbance of plasma equilibrium and the variation of transport efficiency due to the introduction of more concentrated acids.25 Fortunately, isobaric interferences were not found in the ICP-MS measurement.Figs. 3 and 4 show the effect of nitric acid and hydrochloric acid, respectively, on the relative intensity of Mn, Ni, Cu, and Au isotopes. It was found that the introduction of nitric acid and hydrochloric acid at concentrations in the range 0–3 m into the ICP-MS system does not cause a very significant suppression of enhancement of the signals of the elements tested.The moderate decrease in the analytical signals of these elements can be ascribed to the alteration of the transport efficiency resulting from the increase in acid concentration. Analytical Figures of Merit The proposed method involves considerable pre-treatment of the analyte, hence it is necessary to assess the blank value in order to evaluate the limits of detection. The mean blank values for the elements of interest in purified nitric and hydrochloric acid (when using an evaporation method to elevate the concentrations 10-fold for ICP-MS determination) are shown in Table 2.The results indicate that correction for the elemental concentration in the parent material is advisable, although not always necessary because the concentrations of the elements studied were 10–100-fold lower in the reagents than those in the high purity silver analysed. Table 3 gives the limits of detection of the investigated elements calculated using the 3s criterion.The results for the determination of Al, Au, Co, Cu, Cd, Fe, Ni, Mg, Mn, Pb and Sn at ng g21 levels in SRM 8171 (Fine Silver FS 14) and samples of high purity silver (Johnson Matthey, EM 9343 and EM 9465) are given in Table 4. The analytical data obtained for Fine Silver FS 14 are generally in good agreement with the certified values (only moderate agreement for lead). Additionally, the accuracy of the proposed procedure was evaluated by the analysis of spiked samples to which elements of interest were added at concentrations of 10 ppb each.The recoveries of the elements were established by measurements of the concentration of the impurities in unspiked and spiked samples pre-treated by the proposed procedure. Table 5 gives the results of the determination of some impurities in two high purity silver samples and the recoveries for the spiked samples. It was calculated that the errors of determinations are < 10% and the spike recoveries are within the range 95–110%.Fig. 3 Effect of nitric acid concentration on the ICP-MS signal intensity of Mn-55, Ni-60 and Cu-63 isotopes. Fig. 4 Effect of hydrochloric acid concentration on ICP-MS signal intensity of Mn-55, Ni-60, Cu-63 and Au-197 isotopes. Table 2 Results of the determination of trace impurities in hydrochloric and nitric acid Concentration/ng ml21 Nitric Hydrochloric Element acid acid Mg 0.013 0.035 Al 0.021 0.015 Fe 0.025 0.029 Mn 0.076 0.023 Co 0.010 0.004 Ni 0.041 0.038 Cu 0.045 0.085 Cd 0.013 0.004 Pb 0.037 0.068 Sn 0.054 0.086 Au 0.009 0.026 Table 3 Detection limits of the elements determined in high purity silver Detection Detection Element limit/ng g21 Element limit/ng g21 Mg 0.02 Cu 0.03 Al 0.006 Cd 0.05 Mn 0.01 Pb 0.009 Fe 0.73 Sn 0.09 Co 0.004 Au 0.05 Ni 0.02 Analyst, May 1997, Vol. 122 439The proposed procedure appears to be very useful for the determination of sub-ppb levels of trace element impurities in high purity silver samples.Conclusion A relatively simple method for the determination of trace metal impurities in high purity silver involving a combination of precipitation separation and ICP-MS measurement was established. Silver in the sample solution is separated from the trace impurities by the formation of a silver chloride precipitate under the appropriate experimental conditions and the trace elements are subsequently determined by ICP-MS. Using the procedure developed, ng g21 levels of trace impurities in high purity silver can be determined with very good accuracy and precision.This procedure has been successfully applied to the routine determination of trace element impurities in high purity silver samples. The authors are grateful to the National Science Council (Taipei, Taiwan) for partial financial support. References 1 MacLaren, E. H., Can. J. Phys., 1957, 35, 1086. 2 Kudermenn, G., Blanfuss, K. H., L�uhrs, C., Vielhaber, W., and Collisi, V., Fresenius’ J.Anal. Chem., 1992, 343, 734. 3 T�olg, G., Pure Appl. Chem., 1978, 50, 1075. 4 Mullen, J. D., Talanta, 1976, 23, 846. 5 Tanaka, T., Maki, Y., Kobayashi, Y., and Mizuike, A., Anal. Chim. Acta, 1991, 252, 211. 6 Reichel, W., and Bleakley, B. G., Anal. Chem., 1974, 46, 59. 7 Lund, W., Larsen, B. V., and Gundersen, N., Anal. Chim. Acta, 1976, 81, 319. 8 Hinds, M. W., J. Anal. At. Spectrom., 1992, 7, 685. 9 Hiraide, M., Mikuni, Y., and Kawaguchi, H., Fresenius’ J.Anal. Chem., 1996, 354, 212. 10 Yudelevich, I. G., Zakda, B. I., Shabarova, V. P., and Chereko, A. S., At. Spectrosc., 1992, 13, 108. 11 Kudermann, G., Fresenius’ Z. Anal. Chem., 1988, 331, 697. 12 Naumann, R., Schmidt, W., and H�ohl, G., Fresenius’ J. Anal. Chem., 1990, 347, 133. 13 Kayasth, S. R., Basu, A. K., Chattopadhyay, N., and Desai, H. B., Anal. Chim. Acta, 1990, 231, 133. 14 Matsuda, T., and Nagai, T., Anal. Sci., 1990,. 15 Stummeyer, J., and W�unsch, G., Fresenius’ J.Anal. Chem., 1991, 340, 269. 16 Mizuike, A., Enrichment Techniques for Inorganic Trace Analysis, Springer, Berlin, 1983. 17 Chirnside, R. C., Cluley, H. J., and Proffitt, P. M. C., Analyst, 1957, 82, 18. 18 Mizuike, A., Mirsuya, N., and Yammagai, K., Bull. Chem. Soc. Jpn., 1969, 42, 253. 19 Vassos, B. H., Hirsch, R. F., and Letlterman, H., Anal. Chem., 1973, 45, 792. 20 Malissa, H., and Man, I. L., Mikrochim. Acta, 1971, 2, 241. 21 Tsch�opel, P., and T�olg, G., J.Trace Microprobe Tech., 1982, 1, 1. 22 Hinds, M. W., Spectrochim. Acta, Part B, 1993, 48, 435. 23 Kogan, V. V., Hinds, W. M., and Ramendik, G. I., Spectrochim. Acta, Part B, 1994, 49, 333. 24 Burns, D. T., Townshend, A., and Carter, A. H., Inorganic Reaction Chemistry, Wiley, New York, 1981. 25 Holland, G., and Eaton, A. N., Application of Plasma Source Mass Spectrometry, Royal Society of Chemistry, Cambridge, 1991. Paper 6/00139H Received January 6, 1997 Accepted January 21, 1997 Table 4 Results (mg g21) for NIST SRM 8171 (Fine Silver FS 14) and Johnson Matthey high purity silver samples (n = 3) EM9465 EM9343 Silver FS 14 Producer Producer Reference Element This work value* This work value† This work value Fe 1.85 ± 0.08 2.0 0.26 ± 0.02 0.3 47.6 ± 2.2 48.9 ± 2.6 Cu 0.132 ± 0.004 0.1 0.078 ± 0.002 0.1 61.8 ± 3.1 65.2 ± 3.7 Mg 0.087 ± 0.005 0.1 0.064 ± 0.006 0.07 Au 1.15 ± 0.03 1.0 0.044 ± 0.006 20.8 ± 1.6 26.7 ± 6.4 Pb 0.597 ± 0.008 0.5 0.011 ± 0.002 33.8 ± 2.9 38.8 ± 2.2 Al 0.082 ± 0.007 0.024 ± 0.001 Mn 0.011 ± 0.001 0.0070 ± 0.0003 Co 0.0024 ± 0.0003 N.D.‡ Ni 0.011 ± 0.001 0.007 ± 0.001 53.9 ± 3.0 57.0 ± 3.5 Sn 0.006 ± 0.001 0.039 ± 0.004 44.0 ± 2.1 46.1 ± 6.8 Cd 0.018 ± 0.006 0.0120 ± 0.0009 * EM9465 sample certificate of compliance and analysis by Johnson Matthey Electronics (June 24, 1991). † EM9343 sample certificate of compliance and analysis by Johnson Matthey Electronics (January 4, 1991).‡ Not detected.Table 5 Analytical results and recovery of trace levels of metals in two high purity silver samples (n = 3) Sample 1 Sample 2 Concentra- Spike Concentra- Spike Element tion/mg g21 recovery (%) tion/mg g21 recovery (%) Fe 0.25 ± 0.01 100.5 ± 8.1 1.41 ± 0.08 103 ± 5 Cu 0.029 ± 0.002 105.5 ± 3.2 0.131 ± 0.004 106.1 ± 8.5 Mg 0.064 ± 0.002 95.9 ± 5.5 0.067 ± 0.005 95.1 ± 5.8 Au 0.144 ± 0.006 97.4 ± 3.2 1.145 ± 0.034 101.9 ± 3.8 Pb 0.011 ± 0.002 101.8 ± 2.1 0.099 ± 0.008 105.6 ± 3.6 Al 0.023 ± 0.001 101.0 ± 7.3 0.092 ± 0.007 95.7 ± 6.5 Mn 0.007 ± 0.001 106.5 ± 2.8 0.011 ± 0.002 101.3 ± 1.9 Co 0.23 ± 0.05 108.4 ± 2.4 0.37 ± 0.06 109.1 ± 2.4 Ni 0.007 ± 0.001 102.7 ± 4.7 0.011 ± 0.001 100.0 ± 5.6 Cd 0.011 ± 0.001 98.9 ± 4.2 0.018 ± 0.006 96.9 ± 3.5 Sn 0.039 ± 0.004 107.1 ± 3.4 0.006 ± 0.001 103.0 ± 6.0 440 Analyst, May 1997, Vol. 122 Selective Precipitation Separation and Inductively Coupled Plasma Mass Spectrometric Determination of Trace Metal Impurities in High Purity Silver Yuh-Chang Suna, Jerzy Mierzwaa, Chien-Feng Lina, T.I. Yehb, and Mo-Hsiung Yang*a a Department of Nuclear Science, National Tsing Hua University, 30043 Hsinchu, Taiwan b Center for Measurement Standards, Industrial Technology Research Institute, 30043 Hsinchu, Taiwan A simple and rapid method for the determination of some trace element impurities in high purity silver, combining the isolation of analytes from the silver matrix with selective precipitation followed by ICP-MS determination was developed.On the basis of an extreme difference in the solubilities of the chlorides of silver and the other accompanying trace elements, silver can be separated completely through the addition of hydrochloric acid. The sample of silver was at first dissolved in 7 M nitric acid followed by addition of hydrochloric acid to remove the silver matrix by formation of a silver chloride precipitate, while leaving the trace element impurities in the solution, which was subsequently analysed by ICP-MS. Eleven elements (Al, Au, Cd, Co, Cu, Fe, Mg, Mn, Ni, Pb and Sn) were determined with good accuracy and precision.The limits of detection (based on the 3s criterion) of these elements were 1021–1023 ng g21. The proposed method was successfully applied to the determination of metal impurities in high-purity silver samples (EM9465 and EM 9343) and validated by the analysis of NIST SRM 8171 (Fine Silver FS 14). Keywords: Selective precipitation; inductively coupled plasma mass spectrometry; silver To control and to improve the manufacturing technology of high purity materials (e.g., high purity metals), it is necessary to develop methods for the determination of trace element impurities at levels as low as ng g21.One of the additional reasons for the determination of trace element impurities in high purity silver is the effect of impurities on the freezing and melting properties when silver is used for primary and secondary temperature calibration.1 Spark source mass spectrometry and instrumental neutron activation analysis are widely used for the routine determination of trace impurities in high purity metals.2 However, owing to the unavailability of suitable standards and the difficulties connected with matrix interference, the quantitative applications of these physical methods are severely restricted,3 whereas the combination of chemical pre-treatment processes with instrumental analysis always gives the possibility of achieving better reliability in the trace analysis of some special materials.By using a suitable chemical treatment, the sample matrix or coexisting elements can be separated effectively and the aqueous calibration standards can easily be applied owing to the prior simplification of the sample solutions. Atomic spectrometric methods, e.g., ETAAS4–10 and ICP-OES,10,11 and electrochemical methods,12–14 sometimes combined with preconcentration and/or separation procedures, are also employed for the precise determination of trace impurities.With the advent of ICP-MS, many procedures11,15 for the determination of trace element impurities in high purity materials have appeared; however, the interferences resulting from the parent matrix and the limits of detection obtained usually raise the difficulty of the direct determination of trace impurities without a preliminary treatment. Methods of analysis of high purity materials (high purity metals) involve the separation of trace impurities from the dissolved parent matrix by, e.g., solid–liquid extraction,9 ion exchange,13 coprecipitation4,6 and electrodeposition.5,16–20 However, the determination of trace impurities in high purity materials involving multi-stage combined procedures increases the risk of contamination and worsening of the detection limits.To attain high sensitivity and reliability, the analytical blank and systematic error inherent in extreme trace analysis should be very critically controlled.21 An alternative way is to use spectroanalytical techniques with direct solid (slurry) sampling; e.g., Hinds22 used slurry and solid ETAAS for the determination of gold, palladium and platinum in high purity silver.Kogan et al.23 studied the determination of some trace metals in gold and silver samples by laser ablation ICP-MS without matrix matched standards. In order to increase the detection power of the proposed analytical method and to avoid the possibility of contamination, our efforts were directed to developing a relatively simple chemical separation procedure for the elimination of the parent matrix elements.A simple precipitation procedure based on the formation of silver chloride precipitate and remaining trement impurities in the solution can be performed by the addition of high purity hydrochloric acid to precipitate silver ions in nitrate solution.The elements Al, Au, Cu, Cd, Co, Fe, Mg, Mn, Ni, Pb and Sn can be quantitatively separated from the precipitate of silver chloride and very low blank values can be achieved. The ICP-MS measurements were optimized and potential interferences were also investigated. Experimental Apparatus An Elan 5000 inductively coupled plasma mass spectrometer (Perkin-Elmer SCIEX, Thornhill, ON, Canada) equipped with a conventional pneumatic nebulization system was used.The main operating parameters for ICP-MS measurements are given in Table 1. Reagents, Containers and Samples The high purity water used in this study was obtained by purification through de-ionization, double distillation and subboiling distillation. The purification of nitric acid and hydrochloric acid was carried out by recycled sub-boiling distillation of the analytical-reagent grade acids three times. PTFE and glass containers were used throughout the work and were cleaned by immersion in concentrated HNO3 overnight and in concentrated HCl overnight and then steaming successively with HNO3 for 8 h and water vapour for 8 h.Samples of high purity silver EM9343 (6A9 grade silver shot) and EM9465 (5A9 grade silver shot) were obtained from Johnson Analyst, May 1997, Vol. 122 (437–440) 437Matthey Electronics (UK). Silver SRM 8171 (Fine Silver FS 14 Block) was obtained from NIST (Gaithersburg, MD, USA). This material was analysed and prepared by the Royal Canadian Mint (Ottawa, Canada).Sample Cleaning A 20.0 g sample of high-purity silver was weighed into a 30 ml Pyrex beaker and 25 ml of cold 0.1 m nitric acid were added with approximately 10 min of agitation, followed by thorough rinsing in doubly distilled, de-ionized water. After drying by purging with nitrogen in a Class-100 clean bench, the sample was collected in a desiccator. A similar procedure was also applied to silver SRM 8171. Sample Pre-treatment A 1.00 g sample of high purity silver (0.1–0.3 g of silver SRM) was weighed and 3 ml of water and 3 ml of concentrated nitric acid (1 ml of water + 1 ml of concentrated nitric acid) were added to it in a 20 ml PTFE beaker. The sample was heated below the boiling-point of nitric acid until complete dissolution of the silver sample was achieved, then 4 ml of 2.3 m hydrochloric acid were added progressively to form a fine precipitate of silver chloride. After addition of 2 ml of concentrated hydrochloric acid and about 20 h of agitation, the solution was filtered with a 0.45 mm membrane filter.The filtrate was made up to 25 ml with high purity water in a calibrated flask and then trace metals in the filtrate were determined by ICP-MS. Results and Discussion Precipitation Separation Procedure To achieve high sensitivity and accuracy of the determination of trace impurities in silver, an analytical method based on the separation of the matrix element prior to the determination of isolated trace elements was developed.It is noteworthy that coprecipitation should be avoided during the establishment of a precipitation process which permits the effective separation of impurities from the sample matrix. It was found that 40–60% of spiked elements of interest were lost during the formation of the precipitate of silver chloride under neutral conditions. The presence of acid was found to be an effective way to prevent the occurrence of coprecipitation and to release the adsorbed trace impurities ions from the precipitate of silver chloride.The effects of nitric acid and hydrochloric acid concentration on the recovery of trace elements are shown in Figs. 1 and 2, respectively. The experiments were carried out in doubly distilled water spiked with 10 ppb of the elements of interest. Quantitative recoveries of most of the spiked elements from the silver chloride precipitate were observed for concentrations of nitric acid up to 3 m and concentrations of hydrochloric acid from 2 to 3 m.It can be seen from Fig. 1 that gold remains in the precipitate in spite of the addition of nitric acid. The possible reason is that gold cannot be dissolved by nitric acid, whereas it can be easily dissolved by hydrochloric acid owing to the formation of AuCl24 ,24 as shown in Fig. 2. Partial adsorption of cadmium and lead on the precipitate was observed when hydrochloric acid was used (see Fig. 2). Fortunately, the adsorbed cadmium and lead could be readily released from the precipitate by addition of nitric acid. With a view to achieving the complete separation of trace element impurities from the silver matrix and minimizing the amounts of reagents, a compromise composition of the precipitant for precipitation separation and ICP-MS measurements was chosen as a mixture of 2.8 m nitric acid and 2.4 m hydrochloric acid. Table 1 Main operating parameters of ICP-MS instrument ICP radiofrequency power 1.0 kW Nebulizer type Cross-flow Plasma gas flow rate 15 l min21 Auxiliary gas flow rate 0.8 l min21 Nebulizer gas flow rate 0.9 l min21 Sampler/skimmer Nickel Sampling depth 18 mm Data acquisition— Dwell time 200 ms Scan mode Peak hopping Points per peak 3 Signal measurement Integrated counts Fig. 1 Effect of nitric acid concentration on the recovery of metal impurities. Fig. 2 Effect of hydrochloric acid concentration on the recovery of metal impurities. 438 Analyst, May 1997, Vol. 122Determination by ICP-MS The excellent detection power afforded by ICP-MS has been used to great advantage in many analytical applications. The determination of trace element impurities in acids of high concentration is particularly demanding, requiring rigorous optimization of the analytical instrument prior to the measurement. Therefore, the effects of nitric acid and hydrochloric acid concentrations on the ICP-MS measurements were investigated.With a view to releasing the coprecipitated trace elements from the silver chloride precipitate, a mixture of concentrated nitric and hydrochloric acids was used. However, some of the elements analysed might suffer from spectral interference from polyatomic species, the disturbance of plasma equilibrium and the variation of transport efficiency due to the introduction of more concentrated acids.25 Fortunately, isobaric interferences were not found in the ICP-MS measurement.Figs. 3 and 4 show the effect of nitric acid and hydrochloric acid, respectively, on the relative intensity of Mn, Ni, Cu, and Au isotopes. It was found that the introduction of nitric acid and hydrochloric acid at concentrations in the range 0–3 m into the ICP-MS system does not cause a very significant suppression of enhancement of the signals of the elements tested. The moderate decrease in the analytical signals of these elements can be ascribed to the alteration of the transport efficiency resulting from the increase in acid concentration. Analytical Figures of Merit The proposed method involves considerable pre-treatment of the analyte, hence it is necessary to assess the blank value in order to evaluate the limits of detection.The mean blank values for the elements of interest in purified nitric and hydrochloric acid (when using an evaporation method to elevate the concentrations 10-fold for ICP-MS determination) are shown in Table 2. The results indicate that correction for the elemental concentration in the parent material is advisable, although not always necessary because the concentrations of the elements studied were 10–100-fold lower in the reagents than those in the high purity silver analysed.Table 3 gives the limits of detection of the investigated elements calculated using the 3s criterion. The results for the determination of Al, Au, Co, Cu, Cd, Fe, Ni, Mg, Mn, Pb and Sn at ng g21 levels in SRM 8171 (Fine Silver FS 14) and samples of high purity silver (Johnson Matthey, EM 9343 and EM 9465) are given in Table 4.The analytical data obtained for Fine Silver FS 14 are generally in good agreement with the certified values (only moderate agreement for lead). Additionally, the accuracy of the proposed procedure was evaluated by the analysis of spiked samples to which elements of interest were added at concentrations of 10 ppb each. The recoveries of the elements were established by measurements of the concentration of the impurities in unspiked and spiked samples pre-treated by the proposed procedure.Table 5 gives the results of the determination of some impurities in two high purity silver samples and the recoveries for the spiked samples. It was calculated that the errors of determinations are < 10% and the spike recoveries are within the range 95–110%. Fig. 3 Effect of nitric acid concentration on the ICP-MS signal intensity of Mn-55, Ni-60 and Cu-63 isotopes. Fig. 4 Effect of hydrochloric acid concentration on ICP-MS signal intensity of Mn-55, Ni-60, Cu-63 and Au-197 isotopes. Table 2 Results of the determination of trace impurities in hydrochloric and nitric acid Concentration/ng ml21 Nitric Hydrochloric Element acid acid Mg 0.013 0.035 Al 0.021 0.015 Fe 0.025 0.029 Mn 0.076 0.023 Co 0.010 0.004 Ni 0.041 0.038 Cu 0.045 0.085 Cd 0.013 0.004 Pb 0.037 0.068 Sn 0.054 0.086 Au 0.009 0.026 Table 3 Detection limits of the elements determined in high purity silver Detection Detection Element limit/ng g21 Element limit/ng g21 Mg 0.02 Cu 0.03 Al 0.006 Cd 0.05 Mn 0.01 Pb 0.009 Fe 0.73 Sn 0.09 Co 0.004 Au 0.05 Ni 0.02 Analyst, May 1997, Vol. 122 439The proposed procedure appears to be very useful for the determination of sub-ppb levels of trace element impurities in high purity silver samples. Conclusion A relatively simple method for the determination of trace metal impurities in high purity silver involving a combination of precipitation separation and ICP-MS measurement was established.Silver in the sample solution is separated from the trace impurities by the formation of a silver chloride precipitate under the appropriate experimental conditions and the trace elements are subsequently determined by ICP-MS. Using the procedure developed, ng g21 levels of trace impurities in high purity silver can be determined with very good accuracy and precision. This procedure has been successfully applied to the routine determination of trace element impurities in high purity silver samples.The authors are grateful to the National Science Council (Taipei, Taiwan) for partial financial support. References 1 MacLaren, E. H., Can. J. Phys., 1957, 35, 1086. 2 Kudermenn, G., Blanfuss, K. H., L�uhrs, C., Vielhaber, W., and Collisi, V., Fresenius’ J. Anal. Chem., 1992, 343, 734. 3 T�olg, G., Pure Appl. Chem., 1978, 50, 1075. 4 Mullen, J. D., Talanta, 1976, 23, 846. 5 Tanaka, T., Maki, Y., Kobayashi, Y., and Mizuike, A., Anal. Chim. Acta, 1991, 252, 211. 6 Reichel, W., and Bleakley, B. G., Anal. Chem., 1974, 46, 59. 7 Lund, W., Larsen, B. V., and Gundersen, N., Anal. Chim. Acta, 1976, 81, 319. 8 Hinds, M. W., J. Anal. At. Spectrom., 1992, 7, 685. 9 Hiraide, M., Mikuni, Y., and Kawaguchi, H., Fresenius’ J. Anal. Chem., 1996, 354, 212. 10 Yudelevich, I. G., Zakda, B. I., Shabarova, V. P., and Chereko, A. S., At. Spectrosc., 1992, 13, 108. 11 Kudermann, G., Fresenius’ Z. Anal. Chem., 1988, 331, 697. 12 Naumann, R., Schmidt, W., and H�ohl, G., Fresenius’ J. Anal. Chem., 1990, 347, 133. 13 Kayasth, S. R., Basu, A. K., Chattopadhyay, N., and Desai, H. B., Anal. Chim. Acta, 1990, 231, 133. 14 Matsuda, T., and Nagai, T., Anal. Sci., 1990, 7, 75. 15 Stummeyer, J., and W�unsch, G., Fresenius’ J. Anal. Chem., 1991, 340, 269. 16 Mizuike, A., Enrichment Techniques for Inorganic Trace Analysis, Springer, Berlin, 1983. 17 Chirnside, R.C., Cluley, H. J., and Proffitt, P. M. C., Analyst, 1957, 82, 18. 18 Mizuike, A., Mirsuya, N., and Yammagai, K., Bull. Chem. Soc. Jpn., 1969, 42, 253. 19 Vassos, B. H., Hirsch, R. F., and Letlterman, H., Anal. Chem., 1973, 45, 792. 20 Malissa, H., and Man, I. L., Mikrochim. Acta, 1971, 2, 241. 21 Tsch�opel, P., and T�olg, G., J. Trace Microprobe Tech., 1982, 1, 1. 22 Hinds, M. W., Spectrochim. Acta, Part B, 1993, 48, 435. 23 Kogan, V. V., Hinds, W. M., and Ramendik, G. I., Spectrochim. Acta, Part B, 1994, 49, 333. 24 Burns, D. T., Townshend, A., and Carter, A. H., Inorganic Reaction Chemistry, Wiley, New York, 1981. 25 Holland, G., and Eaton, A. N., Application of Plasma Source Mass Spectrometry, Royal Society of Chemistry, Cambridge, 1991. Paper 6/00139H Received January 6, 1997 Accepted January 21, 1997 Table 4 Results (mg g21) for NIST SRM 8171 (Fine Silver FS 14) and Johnson Matthey high purity silver samples (n = 3) EM9465 EM9343 Silver FS 14 Producer Producer Reference Element This work value* This work value† This work value Fe 1.85 ± 0.08 2.0 0.26 ± 0.02 0.3 47.6 ± 2.2 48.9 ± 2.6 Cu 0.132 ± 0.004 0.1 0.078 ± 0.002 0.1 61.8 ± 3.1 65.2 ± 3.7 Mg 0.087 ± 0.005 0.1 0.064 ± 0.006 0.07 Au 1.15 ± 0.03 1.0 0.044 ± 0.006 20.8 ± 1.6 26.7 ± 6.4 Pb 0.597 ± 0.008 0.5 0.011 ± 0.002 33.8 ± 2.9 38.8 ± 2.2 Al 0.082 ± 0.007 0.024 ± 0.001 Mn 0.011 ± 0.001 0.0070 ± 0.0003 Co 0.0024 ± 0.0003 N.D.‡ Ni 0.011 ± 0.001 0.007 ± 0.001 53.9 ± 3.0 57.0 ± 3.5 Sn 0.006 ± 0.001 0.039 ± 0.004 44.0 ± 2.1 46.1 ± 6.8 Cd 0.018 ± 0.006 0.0120 ± 0.0009 * EM9465 sample certificate of compliance and analysis by Johnson Matthey Electronics (June 24, 1991). † EM9343 sample certificate of compliance and analysis by Johnson Matthey Electronics (January 4, 1991). ‡ Not detected. Table 5 Analytical results and recovery of trace levels of metals in two high purity silver samples (n = 3) Sample 1 Sample 2 Concentra- Spike Concentra- Spike Element tion/mg g21 recovery (%) tion/mg g21 recovery (%) Fe 0.25 ± 0.01 100.5 ± 8.1 1.41 ± 0.08 103 ± 5 Cu 0.029 ± 0.002 105.5 ± 3.2 0.131 ± 0.004 106.1 ± 8.5 Mg 0.064 ± 0.002 95.9 ± 5.5 0.067 ± 0.005 95.1 ± 5.8 Au 0.144 ± 0.006 97.4 ± 3.2 1.145 ± 0.034 101.9 ± 3.8 Pb 0.011 ± 0.002 101.8 ± 2.1 0.099 ± 0.008 105.6 ± 3.6 Al 0.023 ± 0.001 101.0 ± 7.3 0.092 ± 0.007 95.7 ± 6.5 Mn 0.007 ± 0.001 106.5 ± 2.8 0.011 ± 0.002 101.3 ± 1.9 Co 0.23 ± 0.05 108.4 ± 2.4 0.37 ± 0.06 109.1 ± 2.4 Ni 0.007 ± 0.001 102.7 ± 4.7 0.011 ± 0.001 100.0 ± 5.6 Cd 0.011 ± 0.001 98.9 ± 4.2 0.018 ± 0.006 96.9 ± 3.5 Sn 0.039 ± 0.004 107.1 ± 3.4 0.006 ± 0.001 103.0 ± 6.0 440 Analyst, May 19

ISSN:0003-2654    

DOI:10.1039/a700139h

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Clean Method for the Simultaneous Determination of Propyphenazone and Caffeine in Pharmaceuticals by Flow Injection Fourier Transform Infrared Spectrometry Zouhair Bouhsain†, Salvador Garrigues and Miguel de la Guardia* Department of Analytical Chemistry, University of Valencia, 50 Dr. Moliner St, 46100 Burjassot, Valencia, Spain A procedure is proposed for the simultaneous FTIR determination of propyphenazone (PFZ) and caffeine (CAF) in pharmaceuticals. The method involves the dissolution of the active principles in CHCl3, followed by filtration of sample solutions to remove the excipients.PFZ is then determined by absorbance measurements at 1595 cm21, using a baseline established between 2000 and 890 cm21, and CAF by using the first-derivative values at 1712 cm21, using solutions of PFZ and CAF for external calibration. The method was applied in both the stopped-flow and flow-injection modes, providing precise and accurate results for the analysis of real samples.The incorporation of a distillation unit for the on-line recycling of the CHCl3 used as carrier and solvent provides an environmentally friendly analytical methodology which makes possible an injection frequency of 120 samples h21 and reduces the cost and side-effects of the production of laboratory waste. Keywords: Propyphenazone; caffeine; flow analysis; Fourier transform infrared spectrometry; clean analytical methodology Propyphenazone (4-isopropyl-1,5-dimethyl-2-phenyl-4-pyrazolin- 3-one) and caffeine (7-methyltheophylline) are important analgesic and stimulant drugs, respectively, which are commonly used and combined in pharmaceutical formulations.1,2 Many methods have been developed for the simultaneous determination of propyphenazone (PFZ) and caffeine (CAF) in pharmaceuticals together and in the presence of other compounds involving in general the use of chromatographic techniques, such as TLC,3 GC4 and HPLC.5,6 Additionally, multivariate calibration, particularly the partial least-squares methodology, has been employed for the treatment of UV data in order to determine PFZ, CAF and phenacetin in the presence of papaverine and phenobarbitone.7 However, there are no reports on the simultaneous determination of PFZ and CAF by FTIR spectrometry.FTIR has a better selectivity than UV/VIS spectrometry and provides an excellent tool for the determination of mixtures of compounds. On the other hand, in the last few years flow injection (FI)-FTIR has been demonstrated to be a synergetic combination highly appropriate for the quality control of drugs;8–13 however, only a few papers have been published on the simultaneous FI-FTIR determination of mixtures of compounds in pharmaceutical formulations, both directly using specific bands14,15 or using a partial least-squares regression data treatment.16 In general, the analytical procedures developed for IR determinations require the use of organic solvents, and frequently the chlorinated solvents, such as CCl4, CHCl3 and CH2Cl2, are those that provide the best transparency in the IR range and the lowest limits of detection.However, these reagents are highly toxic and hence it is essential to control the waste from this type of measurement; this increases the cost of the analysis, and also involves environmental and operator risks. The FI-FTIR methodology not only offers a fast and simple way for the automation of the sample treatment and measurement operations, but also possibilities for the development of a sustainable analytical chemistry, based on the on-line decontamination of laboratory waste.17,18 Recently, our group has developed an on-line distillation manifold, which permits the recovery of carrier solvents employed in pharmaceutical analysis, providing a reduction in the consumption of reagents and the elimination of wastes.19,20 The main objective of this work was to develop a clean analytical procedure for the simultaneous determination of PFZ and CAF in commercial pharmaceutical formulations based on the use of FI and FTIR.Experimental Apparatus and Reagents A Mattson (Madison, WI, USA) Model Research 1 FTIR spectrophotometer was employed to obtain the FTIR spectra in the range 4000–400 cm21. The manifold employed (see Fig. 1) is a monochannel assembly with a Gilson (Worthington, OH, USA) Minipuls peristaltic pump with solvent-resistant Viton (iso-versinic) pump tubes of 0.1 cm internal diameter and 0.3 cm external diameter.A Rheodyne (Cotati, CA, USA) Type 50 injection valve with various fixed volume loops was employed for the introduction of samples and standards; the length of the tube from the valve to the cell was 19 cm. The manifold includes a semi-micro-distillation unit which permits the on-line recycling of the carrier organic solvent and thus avoids the generation and accumulation of toxic wastes.† Permanent address: Tetouan University, Morocco. Fig. 1 Manifold employed for the clean simultaneous FI-FTIR determination of PFZ and CAF. Analyst, May 1997, Vol. 122 (441–445) 441The distillation unit includes a 100 ml balloon, a Liebig condenser of 14 cm length and a double flask arrangement to recover the distillate, the second part of which is cooled in order to ensure that the carrier solution is at room temperature. The total volume of the organic solvent used in the distillation unit is 70 ml.The distillation temperature employed was 62 °C. All the measurements were carried out using a micro-flow cell from Graseby Specac (Orlando, FL, USA) with KBr windows and the connection tubes were of PTFE with 0.8 mm internal diameter. PFZ was supplied by Guinama (Valencia, Spain) and CAF by Fluka (Buchs, Switzerland). Both products were of USP-grade and were employed without further purification. Analytical-reagent grade CHCl3, stabilized with b-amylene, and CH2Cl2, stabilized with ethanol, were obtained from Scharlau (Barcelona, Spain).The following commercial samples were acquired from Spanish pharmacies: Megral tablets, with a nominal content of 300 mg of PFZ, 50 mg of CAF and excipient, and Sedalmerck tablets, containing 300 mg of PFZ, 50 mg of CAF, 5 mg of ephedrine, 5 mg of ethylmorphine and excipient. In both instances, the nature of the excipient was not indicated. General Procedure Ten tablets were weighed and the average mass of one tablet was determined. The tablets were powdered and an exact appropriate amount was taken for analysis by both FTIR and a comparative procedure based on the multivariate calibration of UV absorbance data of untreated samples by partial leastsquares (PLS-UV).7 FI-FTIR Determination An appropriate amount of sample (between 100 and 150 mg) was weighed, placed in a beaker with 7 ml of CHCl3 and shaken manually for a few seconds.The solution was filtered through a 0.45 mm nylon membrane and made up to a final volume of 10 ml.For flow measurements, 300 ml of the sample solution were injected into a carrier stream (1.23 ml min21) of CHCl3 and the FTIR spectrum was continuously recorded, accumulating two scans. For the stopped-flow mode, the flow cell was filled with the CHCl3 solution of the sample and 16 scans were accumulated. In both instances, a 0.5 mm bandpass and a nominal resolution of 4 cm21 were employed, and pure CHCl3 was used to establish the background.For PFZ determination, peak height absorbance measurements at 1595 cm21, corrected with a baseline established between 2000 and 890 cm21, were employed, whereas for CAF determination, first derivative peak height measurements at 1712 cm21 were used. Standards of PFZ and CAF, dissolved in CHCl3 and treated in the same way as the samples, were used in order to establish the appropriate calibration graph. Reference Procedure An appropriate amount of sample containing about 60 mg of PFZ and 10 mg of CAF was weighed accurately and dissolved in 250 ml of doubly distilled water.A 2.5–3.0 ml volume of the filtered solution was diluted to 25 ml and absorbance measurements were made in the UV range. For calibration, four synthetic mixtures of PFZ and CAF at two concentration levels dissolved in doubly distilled water were employed. The upper and lower concentrations used for calibration were selected in order to include the maximum levels (±10%) of the active principles in the pharmaceuticals analysed.Results and Discussion FTIR Spectra of PFZ and CAF The FTIR absorbance spectra of PFZ and CAF, obtained in a CHCl3 solution, present well defined and intense bands between 1800 and 1000 cm21 (see Fig. 2). The specific bands at 1595, 1497 and 1136 cm21 for PFZ, and at 1705 and 1555 cm21 for CAF offer considerable possibilities for the direct simultaneous FTIR determination of the two analytes in the same sample.The appropriate selection of the solvent is important in IR determinations in order to obtain a total dissolution of the compounds to be determined and a good signal-to-noise ratio. Data in Table 1 summarize the analytical features corresponding to the determination of PFZ and CAF using CHCl3 and CH2Cl2 as solvent. Carbon tetrachloride was rejected because CAF is not soluble in this solvent. In order to establish the best conditions for the two analytes, the previously indicated characteristic wavenumber bands were tested, using, for CHCl3, two different calibrations lines, one obtained with pure PFZ and CAF solutions and the other with binary standards containing PFZ and CAF in a ratio of 5 : 1 tested.As can be seen, the two solvents tested provide a comparable sensitivity, dynamic range and linearity for PFZ and CAF. The limit of detection and the relative standard deviation for PFZ are better in CHCl3 than in CH2Cl2, while the results for CAF are similar in both solvents.Because of this, and taking into consideration the low boiling temperature of CH2Cl2, this solvent was discarded, and additional experiments were carried out in CHCl3. A comparison was made of the analytical curves obtained for pure solutions of PFZ and CAF and those obtained for a 5 : 1 proportion of these compounds (which corresponds to the common composition of pharmaceuticals). It was found that PFZ can be directly determined in the presence of CAF (using the band at 1595 cm21).PFZ interferes with the determination of CAF but the band at 1705 cm21 suffers less interference than that at 1555 cm21. The upper limits of the dynamic ranges for PFZ and CAF are 50 and 10 mg ml21, respectively, with a bandpass of 0.1 mm; in order to improve the limit of detection and the sensitivity, a bandpass of 0.5 mm was used. Study of Interferences A systematic study of the interference of PFZ on the determination of CAF was carried out by evaluating the effect of increasing concentrations of PFZ on the absorbance measure- Fig. 2 FTIR spectra of CHCl3 solutions of PFZ, CAF and a real sample containing PFZ and CAF. Spectra are the average of 16 accumulated scans obtained with a nominal resolution of 4 cm21. 442 Analyst, May 1997, Vol. 122ments of a fixed concentration of 1 mg ml21 of CAF. As can be seen in Fig. 3, concentrations higher than 2 mg ml21 cause a positive interference on the absorbance of CAF, providing excess errors of the order of 7%.However, this interference can be corrected by using the measurements in the first-derivative mode for CAF at 1712 cm21, particularly for the proportion at which these active compounds are usually combined (see Fig. 3). Therefore, for the determination of CAF in the presence of PFZ, the use of the first-derivative spectra is recommended. An additional study carried out on the interference of pharmaceutical excipients, such as glucose, sucrose, starch, lactose, fructose, maltose and talc, demonstrated that none of these compounds interfered with the FTIR determination of PFZ or CAF.In fact, CHCl3 solutions of mixtures of 50 mg ml21 of each excipient, 5 mg ml21 of PFZ and 1 mg ml21 of CAF, after filtration, provided the same absorbance values as the corresponding solutions without excipient, demonstrating the low solubility in CHCl3 of typical excipients employed in pharmaceutical preparations.Effect of the FI and Instrumental Parameters on the FTIR Determination of PFZ and CAF The effect of FI parameters, such as the carrier flow rate and sample volume injected, and of instrumental parameters, such as resolution and number of accumulated spectra, on the FTIR measurements was studied in the monoparametric mode by modifying each parameter in turn while keeping the other parameters constant, using a bandpass of 0.5 mm. Fig. 4 shows, as an example, the effect of FI parameters on the first-derivative measurements of CAF.As can be seen, the use of carrier flow rates higher than 1.36 ml min21 leads to a decrease in sensitivity and also a loss of precision in the firstderivative measurements. On the other hand, an increase in the sample volume injected increases the sensitivity, but also increases the time required to carry out the analysis by increasing the peak width. Hence, in order to obtain a compromise between sensitivity, repeatability and sample throughput, a carrier flow rate of 1.23 ml min21 and a sample volume of 300 ml were selected.The effect of resolution on the FI absorbance measurements is shown in Fig. 5. For both PFZ and CAF, an increase in the nominal resolution leads to a decrease in the absorbance and first-derivative values, respectively, and, additionally, for a resolution of 2 cm21, a poor repeatability was obtained. For these reasons a nominal resolution of 4 cm21 was selected for analytical purposes.A study of the effect of the number of spectra accumulated demonstrated that this had only a slight influence on the sensitivity and precision of the absorbance measurements for both analytes assayed, providing for 1–8 accumulated spectra relative errors lower than 1.8%. Hence, the number of spectra accumulated was two so as to reduce the analysis time without sacrificing the analytical precision. Figures of Merit of the Simultaneous FTIR Determination of PFZ and CAF Table 2 shows the comparison between the analytical features found for the simultaneous determination of PFZ and CAF in both the stopped-flow and FI modes, using a bandpass of 0.5 mm.As can be seen, the sensitivity obtained using the stoppedflow mode is slightly higher than that obtained by FI. The limit of detection of CAF is better in the stopped-flow than in the FI mode as it corresponds to a higher sensitivity, but the opposite is the case with PFZ determination because of the higher stability of the blank measurements in the FI mode.The relative standard deviation of five independent analyses of the two analytes is lower in the stopped-flow than in the FI mode. However, the fact that by using the latter approach a sampling frequency of 120 h21 can be obtained for the simultaneous determination of both analytes is significant for a routine methodology in quality control analysis in the pharmaceutical industry.The inter-day repeatability of the FI procedure was checked using a solution containing 1.5 mg ml21 of CAF and 5.5 mg ml21 of PFZ. A relative standard deviation of 1.2% was Table 1 Analytical features of the stopped-flow FTIR determination of PFZ and CAF dissolved in CHCl3 and CH2Cl2 * Parameter Wavenumber/ A = a + bC LOD‡/ RSD§ Dynamic range¶/ Solvent Analyte cm21 (C in mg ml21)† r mg ml21 (%) mg ml21 CHCl3 PFZ 1595 Aa = 0.005 + 0.0107 C 0.9998 0.09 0.18 2.5–50 Ab = 0.004 + 0.0111 C 0.9998 — — — 1497 Aa = 0.010 + 0.0104 C 0.9992 0.09 0.15 2.5–50 Ab = 0.0116 + 0.0120 C 0.9993 — — — 1136 Aa = 0.0011 + 0.0061 C 0.9999 0.3 0.3 2.5–50 Ab = 0.0013 + 0.0062 C 0.9999 — — — CAF 1705 Ac = 0.0047 + 0.0524 C 0.9998 0.02 0.2 0.25–10 Ab = 0.0056 + 0.0557 C 0.9998 — — — 1555 Ac = 0.0020 + 0.0163 C 0.9999 0.2 0.2 0.25–10 Ab = 0.0025 + 0.0199 C 0.9998 — — — CH2Cl2 PFZ 1595 Aa = 0.0047 + 0.0097 C 0.9997 0.2 0.5 2.5–50 1497 Aa = 0.0109 + 0.0105 C 0.9991 0.4 1.0 2.5–50 1136 Aa = 0.0011 + 0.0061 C 0.9999 0.3 1.9 2.5–50 CAF 1705 Ac = 0.004 + 0.0561 C 0.9998 0.1 0.11 0.25–10 1555 Ac = 0.0008 + 0.0166 C 0.9998 0.3 0.60 0.25–10 * All the measurements were carried out using a lead spacer with a bandpass of 0.1 mm and a general baseline was established between 2000 and 890 cm21.† Calibration lines were obtained for pure PFZ solution (a), solutions containing PFZ and CAF in a 5 : 1 concentration proportion (b) and pure CAF solution (c). ‡ LOD: Limit of detection established for K = 3 (a probability level of 99.6%).§ RSD: Relative standard deviation in per cent. for five independent measurements carried out at a concentration level of 30 mg ml21 for PFZ and 6 mg ml21 for CAF. ¶ The upper limit of the dynamic range was established from adjustment of the data in order to obtain a linear adjustment of the analytical data. Analyst, May 1997, Vol. 122 443obtained for both compounds for four independent day sessions performing five measurements each day.On-line Treatment of Wastes The method developed for the FTIR determination of PFZ and CAF in pharmaceuticals involves a pre-treatment step, consisting in their dissolution in CHCl3 followed by filtration, which is very simple and does not provide toxic residues. However, after the FTIR determination, the samples analysed and the standard solutions, as well as the CHCl3 carrier employed, constitute a toxic and dangerous residue which must be stored and treated to avoid the release into the environment of this halogenated hydrocarbon which is a carcinogen and can contaminate water effluents and the laboratory atmosphere.Hence, the use of a closed flow system, such as that indicated in the manifold of Fig. 1, could be helpful in order to reduce the amount of the solvent required and to eliminate vapour leaks or the storage of large amounts of CHCl3. The basic idea of coupling on-line a small distillation unit to the determination procedure is to avoid the storage of used solvents by providing a continuous recycling of the carrier stream.Additionally, after measurements, the remaining volumes of sample solutions are employed to feed the distillation balloon in order to ensure the continuous use of the distillation unit throughout the analysis. Different distillation unit volumes were tested in previous studies of the clean analytical determination of ketoprofen by FTIR, performing the on-line distillation of CCl4,19 and by UV spectrophotometry, using methanol as a solvent,20 from which a total volume of 70 ml was established as the most convenient for the on-line continuous treatment of carrier flows from 0.5 to 5 ml min21, without the need to replenish the CHCl3 inside the distillation unit.Throughout this study, the aforementioned manifold was employed and no analytical problems were found on using the on-line distilled solvent several times, thus helping to reduce the cost of the analysis and also providing an environmentally friendly procedure without pollution side-effects.Analysis of Real Samples The samples described under Experimental were analysed by FTIR using the stopped-flow and FI modes and also by a reference PLS-UV spectrophotometric procedure;7 the results obtained for the simultaneous determination of PFZ and CAF are summarized in Table 3. A statistical treatment of the data reported in Table 3 provided the following regression equations between data found by FTIR and those obtained by PLS-UV: y = 25.29 + 1.044x for the stopped-flow procedure and y = 25.89 + 1.068x for FI, thus indicating that the FTIR procedures do not require any blank correction nor present constant relative errors as compared with the reference procedure.On the other hand, a comparison of the experimental data found for each of the samples considered showed (from the Snedecor treatment) that in most cases the data obtained by FTIR and those obtained by UV are comparable for a probability level of 95%, the exceptions being CAF in Megral by FI-FTIR or PFZ in Megral and CAF in Sedalmerk by Fig. 3 Effect of PFZ on the absorbance of CAF at 1705 cm21 corrected with a baseline established between 2000 and 890 cm21 (5), and on the first-derivative measurements at 1712 cm21 (/). The dotted lines indicate the ±3 times the standard deviation of the absorbance or first-derivative values found in the absence of PFZ.Fig. 4 Effect of the FI parameters on the first-derivative values of CAF. (a) Effect of the carrier flow for an injection volume of 300 ml. (b) Effect of the injection volume for a carrier flow of 1.23 ml min21. In all instances, two accumulated spectra, a nominal resolution of 4 cm21 and a bandpass of 0.5 mm were employed, for a CAF concentration of 1 mg ml21. Fig. 5 Effect of the spectral resolution of the FTIR measurements of PFZ (dotted line) and CAF (solid line) using a carrier flow of 1.23 ml min21, a 300 ml injection volume, two accumulated spectra and a bandpass of 0.5 mm.The concentrations of PFZ and CAF were 5 and 1 mg ml21, respectively. 444 Analyst, May 1997, Vol. 122stopped-flow FTIR, where the comparability is for a 99% level. However, all the values obtained are comparable within the range tolerated for these types of drugs by the pharmacopoeia, thus demonstrating the applicability of the methodology developed.Conclusion PFZ and CAF can be rapidly and simultaneously determined by FTIR using univariate calibration. The determination can be carried out simply by leaching the two active principles in CHCl3 and measuring the absorbance at 1595 cm21, corrected with a baseline established between 2000 and 890 cm21, and the first derivative at 1712 cm21 for PFZ and CAF, respectively. The proposed method is clean (no toxic waste is produced), rapid and reproducible and provides values comparable to those obtained with a PLS-UV methodology.In the FI mode, throughput of 120 samples h21 can be achieved. Thus, the developed methodology offers a good alternative for the simultaneous determination of PFZ and CAF in the quality control analysis of pharmaceuticals. The authors gratefully acknowledge the financial support of the Spanish DGICYT Project 92-0870, the Generalitat Valenciana Project 1021/93 and the technical collaboration of ATI-Unicam (Spain). References 1 Clark’s Isolation and Identification of Drugs, ed.Moffat, A. C., The Pharmaceutical Society of Great Britain, London, 1986. 2 British Pharmacopoeia, HM Stationery Office, London, 1993. 3 Tomankova, H., and Vasatova, M., Pharmazie, 1989, 44, 197. 4 Markovic, S., and Kusec, Z., Pharmazie, 1990, 45, 935. 5 Mamolo, M. G., Vio, L., and Maurich, V., J. Pharm. Biomed. Anal., 1985, 3, 157. 6 Voroshilova, O, I., and Kaiser, R. E., Chromatographia, 1993, 36, 47. 7 Otto, M., and Sporreiter, K., Mikrochim.Acta, 1985, III, 167. 8 Garrigues, S., Gallignani, M., and de la Guardia, M., Talanta, 1993, 40, 89. 9 Bouhsain, Z., Hassan, B. A., Garrigues, S., and de la Guardia, M., Quim. Anal., 1995, 14, 96. 10 Bouhsain, Z., Garrigues, S., and de la Guardia, M., Analyst, 1996, 121, 635. 11 McKittrick, P. T., Danielson, N. D., and Katon, J. E., Microchem. J., 1991, 44, 105. 12 Morgan, D. K., Danielson, N. D., and Katon, J. E., Anal. Lett., 1985, 18(A16), 1979. 13 Ramos, M. L., Tyson, J. F., and Curran, D. J., Anal. Proc., 1995, 32, 175. 14 Garrigues, S., Gallignani, M., and de la Guardia, M., Talanta, 1993, 40, 1799. 15 Miller, B. E, Danielson, N. D., and Katon, J. E., Appl. Spectrosc., 1988, 42(3), 401. 16 Bouhsain, Z., Garrigues, S., and de la Guardia, M., Analyst, 1996, 121, 1935. 17 de la Guardia, M., and R°u�zi�cka, J., Analyst, 1995, 120, 17N. 18 de la Guardia, M., Khalaf, K. D., Carbonell, V., and Morales- Rubio, A., Anal. Chim.Acta., 1995, 308, 462. 19 S�anchez-Das�ý, J., Cervera, M. L., Garrigues, S., and de la Guardia, M., paper presented at the First Mediterranean Basin Conference on Analytical Chemistry, C�ordoba, November, 1995. 20 S�anchez-Das�ý, J., Cervera, M. L., Garrigues, S., and de la Guardia, M., paper presented at the Clean Technology Symposium, London, June, 1996. Paper 6/07109K Received October 18, 1996 Accepted February 3, 1997 Table 2 Analytical features of the FTIR determination of PFZ and CAF in CHCl3 in both stopped-flow and FI modes using a bandpass of 0.5 mm Wavenumber (Baseline)/ Mode Analyte cm21 Parameter Stopped-flow FI A = a + bC A = 0.0114 + 0.0530 C A=20.0001 + 0.0484 C (C in mg ml21) r 0.9997 0.99990 1595 Absorbance PFZ LOD/mg ml21 0.06 0.03 (2000D 0.5% A = 0.223 ± 0.001 2.1% A = 0.388 ± 0.008 C = 4 mg ml21 C = 8 mg ml21 dA/dn = a + bC dA/dn = 20.000320.0209 C dA/dn = 20.000120.0205 C (C in mg ml21) r 20.9998 20.9998 1712 Derivative CAF LOD/mg ml21 0.45 1.5 — RSD 0.5% dA/dn = 20.0341 ± 0.0001 1.0% dA/dn = 20.0327 ± 0.0003 C = 1.6 mg ml21 C = 1.6 mg ml21 Table 3 Simultaneous determination of PFZ and CAF in pharmaceuticals Found*/mg Declared/ Stopped- Analyte Sample mg flow FI PLS-UV Megral 300 297 ± 3 300 ± 7 283 ± 1 PFZ Sedalmerck 300 301 ± 2 307 ± 7 296 ± 1 Megral 50 50.7 ± 0.7 49.3 ± 0.8 51.2 ± 0.2 CAF Sedalmerck 50 49.0 ± 0.2 47.8 ± 1.6 52.2 ± 0.2 * Values indicated are the average of three independent analyses ± the corresponding standard deviation.Analyst, May 1997, Vol. 122 445 Clean Method for the Simultaneous Determination of Propyphenazone and Caffeine in Pharmaceuticals by Flow Injection Fourier Transform Infrared Spectrometry Zouhair Bouhsain†, Salvador Garrigues and Miguel de la Guardia* Department of Analytical Chemistry, University of Valencia, 50 Dr. Moliner St, 46100 Burjassot, Valencia, Spain A procedure is proposed for the simultaneous FTIR determination of propyphenazone (PFZ) and caffeine (CAF) in pharmaceuticals.The method involves the dissolution of the active principles in CHCl3, followed by filtration of sample solutions to remove the excipients. PFZ is then determined by absorbance measurements at 1595 cm21, using a baseline established between 2000 and 890 cm21, and CAF by using the first-derivative values at 1712 cm21, using solutions of PFZ and CAF for external calibration. The method was applied in both the stopped-flow and flow-injection modes, providing precise and accurate results for the analysis of real samples.The incorporation of a distillation unit for the on-line recycling of the CHCl3 used as carrier and solvent provides an environmentally friendly analytical methodology which makes possible an injection frequency of 120 samples h21 and reduces the cost and side-effects of the production of laboratory waste. Keywords: Propyphenazone; caffeine; flow analysis; Fourier transform infrared spectrometry; clean analytical methodology Propyphenazone (4-isopropyl-1,5-dimethyl-2-phenyl-4-pyrazolin- 3-one) and caffeine (7-methyltheophylline) are important analgesic and stimulant drugs, respectively, which are commonly used and combined in pharmaceutical formulations.1,2 Many methods have been developed for the simultaneous determination of propyphenazone (PFZ) and caffeine (CAF) in pharmaceuticals together and in the presence of other compounds involving in general the use of chromatographic techniques, such as TLC,3 GC4 and HPLC.5,6 Additionally, multivariate calibration, particularly the partial least-squares methodology, has been employed for the treatment of UV data in order to determine PFZ, CAF and phenacetin in the presence of papaverine and phenobarbitone.7 However, there are no reports on the simultaneous determination of PFZ and CAF by FTIR spectrometry.FTIR has a better selectivity than UV/VIS spectrometry and provides an excellent tool for the determination of mixtures of compounds.On the other hand, in the last few years flow injection (FI)-FTIR has been demonstrated to be a synergetic combination highly appropriate for the quality control of drugs;8–13 however, only a few papers have been published on the simultaneous FI-FTIR determination of mixtures of compounds in pharmaceutical formulations, both directly using specific bands14,15 or using a partial least-squares regression data treatment.16 In general, the analytical procedures developed for IR determinations require the use of organic solvents, and frequently the chlorinated solvents, such as CCl4, CHCl3 and CH2Cl2, are those that provide the best transparency in the IR range and the lowest limits of detection.However, these reagents are highly toxic and hence it is essential to control the waste from this type of measurement; this increases the cost of the analysis, and also involves environmental and operator risks.The FI-FTIR methodology not only offers a fast and simple way for the automation of the sample treatment and measurement operations, but also possibilities for the development of a sustainable analytical chemistry, based on the on-line decontamination of laboratory waste.17,18 Recently, our group has developed an on-line distillation manifold, which permits the recovery of carrier solvents employed in pharmaceutical analysis, providing a reduction in the consumption of reagents and the elimination of wastes.19,20 The main objective of this work was to develop a clean analytical procedure for the simultaneous determination of PFZ and CAF in commercial pharmaceutical formulations based on the use of FI and FTIR.Experimental Apparatus and Reagents A Mattson (Madison, WI, USA) Model Research 1 FTIR spectrophotometer was employed to obtain the FTIR spectra in the range 4000–400 cm21. The manifold employed (see Fig. 1) is a monochannel assembly with a Gilson (Worthington, OH, USA) Minipuls peristaltic pump with solvent-resistant Viton (iso-versinic) pump tubes of 0.1 cm internal diameter and 0.3 cm external diameter.A Rheodyne (Cotati, CA, USA) Type 50 injection valve with various fixed volume loops was employed for the introduction of samples and standards; the length of the tube from the valve to the cell was 19 cm. The manifold includes a semi-micro-distillation unit which permits the on-line recycling of the carrier organic solvent and thus avoids the generation and accumulation of toxic wastes.† Permanent address: Tetouan University, Morocco. Fig. 1 Manifold employed for the clean simultaneous FI-FTIR determination of PFZ and CAF. Analyst, May 1997, Vol. 122 (441–445) 441The distillation unit includes a 100 ml balloon, a Liebig condenser of 14 cm length and a double flask arrangement to recover the distillate, the second part of which is cooled in order to ensure that the carrier solution is at room temperature.The total volume of the organic solvent used in the distillation unit is 70 ml. The distillation temperature employed was 62 °C. All the measurements were carried out using a micro-flow cell from Graseby Specac (Orlando, FL, USA) with KBr windows and the connection tubes were of PTFE with 0.8 mm internal diameter. PFZ was supplied by Guinama (Valencia, Spain) and CAF by Fluka (Buchs, Switzerland). Both products were of USP-grade and were employed without further purification.Analytical-reagent grade CHCl3, stabilized with b-amylene, and CH2Cl2, stabilized with ethanol, were obtained from Scharlau (Barcelona, Spain). The following commercial samples were acquired from Spanish pharmacies: Megral tablets, with a nominal content of 300 mg of PFZ, 50 mg of CAF and excipient, and Sedalmerck tablets, containing 300 mg of PFZ, 50 mg of CAF, 5 mg of ephedrine, 5 mg of ethylmorphine and excipient. In both instances, the nature of the excipient was not indicated.General Procedure Ten tablets were weighed and the average mass of one tablet was determined. The tablets were powdered and an exact appropriate amount was taken for analysis by both FTIR and a comparative procedure based on the multivariate calibration of UV absorbance data of untreated samples by partial leastsquares (PLS-UV).7 FI-FTIR Determination An appropriate amount of sample (between 100 and 150 mg) was weighed, placed in a beaker with 7 ml of CHCl3 and shaken manually for a few seconds.The solution was filtered through a 0.45 mm nylon membrane and made up to a final volume of 10 ml. For flow measurements, 300 ml of the sample solution were injected into a carrier stream (1.23 ml min21) of CHCl3 and the FTIR spectrum was continuously recorded, accumulating two scans. For the stopped-flow mode, the flow cell was filled with the CHCl3 solution of the sample and 16 scans were accumulated.In both instances, a 0.5 mm bandpass and a nominal resolution of 4 cm21 were employed, and pure CHCl3 was used to establish the background. For PFZ determination, peak height absorbance measurements at 1595 cm21, corrected with a baseline established between 2000 and 890 cm21, were employed, whereas for CAF determination, first derivative peak height measurements at 1712 cm21 were used. Standards of PFZ and CAF, dissolved in CHCl3 and treated in the same way as the samples, were used in order to establish the appropriate calibration graph.Reference Procedure An appropriate amount of sample containing about 60 mg of PFZ and 10 mg of CAF was weighed accurately and dissolved in 250 ml of doubly distilled water. A 2.5–3.0 ml volume of the filtered solution was diluted to 25 ml and absorbance measurements were made in the UV range. For calibration, four synthetic mixtures of PFZ and CAF at two concentration levels dissolved in doubly distilled water were employed.The upper and lower concentrations used for calibration were selected in order to include the maximum levels (±10%) of the active principles in the pharmaceuticals analysed. Results and Discussion FTIR Spectra of PFZ and CAF The FTIR absorbance spectra of PFZ and CAF, obtained in a CHCl3 solution, present well defined and intense bands between 1800 and 1000 cm21 (see Fig. 2). The specific bands at 1595, 1497 and 1136 cm21 for PFZ, and at 1705 and 1555 cm21 for CAF offer considerable possibilities for the direct simultaneous FTIR determination of the two analytes in the same sample.The appropriate selection of the solvent is important in IR determinations in order to obtain a total dissolution of the compounds to be determined and a good signal-to-noise ratio. Data in Table 1 summarize the analytical features corresponding to the determination of PFZ and CAF using CHCl3 and CH2Cl2 as solvent. Carbon tetrachloride was rejected because CAF is not soluble in this solvent.In order to establish the best conditions for the two analytes, the previously indicated characteristic wavenumber bands were tested, using, for CHCl3, two different calibrations lines, one obtained with pure PFZ and CAF solutions and the other with binary standards containing PFZ and CAF in a ratio of 5 : 1 tested. As can be seen, the two solvents tested provide a comparable sensitivity, dynamic range and linearity for PFZ and CAF.The limit of detection and the relative standard deviation for PFZ are better in CHCl3 than in CH2Cl2, while the results for CAF are similar in both solvents. Because of this, and taking into consideration the low boiling temperature of CH2Cl2, this solvent was discarded, and additional experiments were carried out in CHCl3. A comparison was made of the analytical curves obtained for pure solutions of PFZ and CAF and those obtained for a 5 : 1 proportion of these compounds (which corresponds to the common composition of pharmaceuticals). It was found that PFZ can be directly determined in the presence of CAF (using the band at 1595 cm21).PFZ interferes with the determination of CAF but the band at 1705 cm21 suffers less interference than that at 1555 cm21. The upper limits of the dynamic ranges for PFZ and CAF are 50 and 10 mg ml21, respectively, with a bandpass of 0.1 mm; in order to improve the limit of detection and the sensitivity, a bandpass of 0.5 mm was used.Study of Interferences A systematic study of the interference of PFZ on the determination of CAF was carried out by evaluating the effect of increasing concentrations of PFZ on the absorbance measure- Fig. 2 FTIR spectra of CHCl3 solutions of PFZ, CAF and a real sample containing PFZ and CAF. Spectra are the average of 16 accumulated scans obtained with a nominal resolution of 4 cm21. 442 Analyst, May 1997, Vol. 122ments of a fixed concentration of 1 mg ml21 of CAF.As can be seen in Fig. 3, concentrations higher than 2 mg ml21 cause a positive interference on the absorbance of CAF, providing excess errors of the order of 7%. However, this interference can be corrected by using the measurements in the first-derivative mode for CAF at 1712 cm21, particularly for the proportion at which these active compounds are usually combined (see Fig. 3). Therefore, for the determination of CAF in the presence of PFZ, the use of the first-derivative spectra is recommended.An additional study carried out on the interference of pharmaceutical excipients, such as glucose, sucrose, starch, lactose, fructose, maltose and talc, demonstrated that none of these compounds interfered with the FTIR determination of PFZ or CAF. In fact, CHCl3 solutions of mixtures of 50 mg ml21 of each excipient, 5 mg ml21 of PFZ and 1 mg ml21 of CAF, after filtration, provided the same absorbance values as the corresponding solutions without excipient, demonstrating the low solubility in CHCl3 of typical excipients employed in pharmaceutical preparations.Effect of the FI and Instrumental Parameters on the FTIR Determination of PFZ and CAF The effect of FI parameters, such as the carrier flow rate and sample volume injected, and of instrumental parameters, such as resolution and number of accumulated spectra, on the FTIR measurements was studied in the monoparametric mode by modifying each parameter in turn while keeping the other parameters constant, using a bandpass of 0.5 mm.Fig. 4 shows, as an example, the effect of FI parameters on the first-derivative measurements of CAF. As can be seen, the use of carrier flow rates higher than 1.36 ml min21 leads to a decrease in sensitivity and also a loss of precision in the firstderivative measurements. On the other hand, an increase in the sample volume injected increases the sensitivity, but also increases the time required to carry out the analysis by increasing the peak width.Hence, in order to obtain a compromise between sensitivity, repeatability and sample throughput, a carrier flow rate of 1.23 ml min21 and a sample volume of 300 ml were selected. The effect of resolution on the FI absorbance measurements is shown in Fig. 5. For both PFZ and CAF, an increase in the nominal resolution leads to a decrease in the absorbance and first-derivative values, respectively, and, additionally, for a resolution of 2 cm21, a poor repeatability was obtained. For these reasons a nominal resolution of 4 cm21 was selected for analytical purposes.A study of the effect of the number of spectra accumulated demonstrated that this had only a slight influence on the sensitivity and precision of the absorbance measurements for both analytes assayed, providing for 1–8 accumulated spectra relative errors lower than 1.8%. Hence, the number of spectra accumulated was two so as to reduce the analysis time without sacrificing the analytical precision.Figures of Merit of the Simultaneous FTIR Determination of PFZ and CAF Table 2 shows the comparison between the analytical features found for the simultaneous determination of PFZ and CAF in both the stopped-flow and FI modes, using a bandpass of 0.5 mm. As can be seen, the sensitivity obtained using the stoppedflow mode is slightly higher than that obtained by FI. The limit of detection of CAF is better in the stopped-flow than in the FI mode as it corresponds to a higher sensitivity, but the opposite is the case with PFZ determination because of the higher stability of the blank measurements in the FI mode.The relative standard deviation of five independent analyses of the two analytes is lower in the stopped-flow than in the FI mode. However, the fact that by using the latter approach a sampling frequency of 120 h21 can be obtained for the simultaneous determination of both analytes is significant for a routine methodology in quality control analysis in the pharmaceutical industry.The inter-day repeatability of the FI procedure was checked using a solution containing 1.5 mg ml21 of CAF and 5.5 mg ml21 of PFZ. A relative standard deviation of 1.2% was Table 1 Analytical features of the stopped-flow FTIR determination of PFZ and CAF dissolved in CHCl3 and CH2Cl2 * Parameter Wavenumber/ A = a + bC LOD‡/ RSD§ Dynamic range¶/ Solvent Analyte cm21 (C in mg ml21)† r mg ml21 (%) mg ml21 CHCl3 PFZ 1595 Aa = 0.005 + 0.0107 C 0.9998 0.09 0.18 2.5–50 Ab = 0.004 + 0.0111 C 0.9998 — — — 1497 Aa = 0.010 + 0.0104 C 0.9992 0.09 0.15 2.5–50 Ab = 0.0116 + 0.0120 C 0.9993 — — — 1136 Aa = 0.0011 + 0.0061 C 0.9999 0.3 0.3 2.5–50 Ab = 0.0013 + 0.0062 C 0.9999 — — — CAF 1705 Ac = 0.0047 + 0.0524 C 0.9998 0.02 0.2 0.25–10 Ab = 0.0056 + 0.0557 C 0.9998 — — — 1555 Ac = 0.0020 + 0.0163 C 0.9999 0.2 0.2 0.25–10 Ab = 0.0025 + 0.0199 C 0.9998 — — — CH2Cl2 PFZ 1595 Aa = 0.0047 + 0.0097 C 0.9997 0.2 0.5 2.5–50 1497 Aa = 0.0109 + 0.0105 C 0.9991 0.4 1.0 2.5–50 1136 Aa = 0.0011 + 0.0061 C 0.9999 0.3 1.9 2.5–50 CAF 1705 Ac = 0.004 + 0.0561 C 0.9998 0.1 0.11 0.25–10 1555 Ac = 0.0008 + 0.0166 C 0.9998 0.3 0.60 0.25–10 * All the measurements were carried out using a lead spacer with a bandpass of 0.1 mm and a general baseline was established between 2000 and 890 cm21.† Calibration lines were obtained for pure PFZ solution (a), solutions containing PFZ and CAF in a 5 : 1 concentration proportion (b) and pure CAF solution (c).‡ LOD: Limit of detection established for K = 3 (a probability level of 99.6%). § RSD: Relative standard deviation in per cent. for five independent measurements carried out at a concentration level of 30 mg ml21 for PFZ and 6 mg ml21 for CAF. ¶ The upper limit of the dynamic range was established from adjustment of the data in order to obtain a linear adjustment of the analytical data.Analyst, May 1997, Vol. 122 443obtained for both compounds for four independent day sessions performing five measurements each day. On-line Treatment of Wastes The method developed for the FTIR determination of PFZ and CAF in pharmaceuticals involves a pre-treatment step, consisting in their dissolution in CHCl3 followed by filtration, which is very simple and does not provide toxic residues. However, after the FTIR determination, the samples analysed and the standard solutions, as well as the CHCl3 carrier employed, constitute a toxic and dangerous residue which must be stored and treated to avoid the release into the environment of this halogenated hydrocarbon which is a carcinogen and can contaminate water effluents and the laboratory atmosphere.Hence, the use of a closed flow system, such as that indicated in the manifold of Fig. 1, could be helpful in order to reduce the amount of the solvent required and to eliminate vapour leaks or the storage of large amounts of CHCl3.The basic idea of coupling on-line a small distillation unit to the determination procedure is to avoid the storage of used solvents by providing a continuous recycling of the carrier stream. Additionally, after measurements, the remaining volumes of sample solutions are employed to feed the distillation balloon in order to ensure the continuous use of the distillation unit throughout the analysis. Different distillation unit volumes were tested in previous studies of the clean analytical determination of ketoprofen by FTIR, performing the on-line distillation of CCl4,19 and by UV spectrophotometry, using methanol as a solvent,20 from which a total volume of 70 ml was established as the most convenient for the on-line continuous treatment of carrier flows from 0.5 to 5 ml min21, without the need to replenish the CHCl3 inside the distillation unit.Throughout this study, the aforementioned manifold was employed and no analytical problems were found on using the on-line distilled solvent several times, thus helping to reduce the cost of the analysis and also providing an environmentally friendly procedure without pollution side-effects. Analysis of Real Samples The samples described under Experimental were analysed by FTIR using the stopped-flow and FI modes and also by a reference PLS-UV spectrophotometric procedure;7 the results obtained for the simultaneous determination of PFZ and CAF are summarized in Table 3.A statistical treatment of the data reported in Table 3 provided the following regression equations between data found by FTIR and those obtained by PLS-UV: y = 25.29 + 1.044x for the stopped-flow procedure and y = 25.89 + 1.068x for FI, thus indicating that the FTIR procedures do not require any blank correction nor present constant relative errors as compared with the reference procedure. On the other hand, a comparison of the experimental data found for each of the samples considered showed (from the Snedecor treatment) that in most cases the data obtained by FTIR and those obtained by UV are comparable for a probability level of 95%, the exceptions being CAF in Megral by FI-FTIR or PFZ in Megral and CAF in Sedalmerk by Fig. 3 Effect of PFZ on the absorbance of CAF at 1705 cm21 corrected with a baseline established between 2000 and 890 cm21 (5), and on the first-derivative measurements at 1712 cm21 (/). The dotted lines indicate the ±3 times the standard deviation of the absorbance or first-derivative values found in the absence of PFZ.Fig. 4 Effect of the FI parameters on the first-derivative values of CAF. (a) Effect of the carrier flow for an injection volume of 300 ml. (b) Effect of the injection volume for a carrier flow of 1.23 ml min21. In all instances, two accumulated spectra, a nominal resolution of 4 cm21 and a bandpass of 0.5 mm were employed, for a CAF concentration of 1 mg ml21. Fig. 5 Effect of the spectral resolution of the FTIR measurements of PFZ (dotted line) and CAF (solid line) using a carrier flow of 1.23 ml min21, a 300 ml injection volume, two accumulated spectra and a bandpass of 0.5 mm. The concentrations of PFZ and CAF were 5 and 1 mg ml21, respectively. 444 Analyst, May 1997, Vol. 122stopped-flow FTIR, where the comparability is for a 99% level. However, all the values obtained are comparable within the range tolerated for these types of drugs by the pharmacopoeia, thus demonstrating the applicability of the methodology developed.Conclusion PFZ and CAF can be rapidly and simultaneously determined by FTIR using univariate calibration. The determination can be carried out simply by leaching the two active principles in CHCl3 and measuring the absorbance at 1595 cm21, corrected with a baseline established between 2000 and 890 cm21, and the first derivative at 1712 cm21 for PFZ and CAF, respectively.The proposed method is clean (no toxic waste is produced), rapid and reproducible and provides values comparable to those obtained with a PLS-UV methodology. In the FI mode, throughput of 120 samples h21 can be achieved. Thus, the developed methodology offers a good alternative for the simultaneous determination of PFZ and CAF in the quality control analysis of pharmaceuticals. The authors gratefully acknowledge the financial support of the Spanish DGICYT Project 92-0870, the Generalitat Valenciana Project 1021/93 and the technical collaboration of ATI-Unicam (Spain).References 1 Clark’s Isolation and Identification of Drugs, ed. Moffat, A. C., The Pharmaceutical Society of Great Britain, London, 1986. 2 British Pharmacopoeia, HM Stationery Office, London, 1993. 3 Tomankova, H., and Vasatova, M., Pharmazie, 1989, 44, 197. 4 Markovic, S., and Kusec, Z., Pharmazie, 1990, 45, 935. 5 Mamolo, M. G., Vio, L., and Maurich, V., J. Pharm. Biomed. Anal., 1985, 3, 157. 6 Voroshilova, O, I., and Kaiser, R. E., Chromatographia, 1993, 36, 47. 7 Otto, M., and Sporreiter, K., Mikrochim. Acta, 1985, III, 167. 8 Garrigues, S., Gallignani, M., and de la Guardia, M., Talanta, 1993, 40, 89. 9 Bouhsain, Z., Hassan, B. A., Garrigues, S., and de la Guardia, M., Quim. Anal., 1995, 14, 96. 10 Bouhsain, Z., Garrigues, S., and de la Guardia, M., Analyst, 1996, 121, 635. 11 McKittrick, P. T., Danielson, N. D., and Katon, J. E., Microchem. J., 1991, 44, 105. 12 Morgan, D. K., Danielson, N. D., and Katon, J. E., Anal. Lett., 1985, 18(A16), 1979. 13 Ramos, M. L., Tyson, J. F., and Curran, D. J., Anal. Proc., 1995, 32, 175. 14 Garrigues, S., Gallignani, M., and de la Guardia, M., Talanta, 1993, 40, 1799. 15 Miller, B. E, Danielson, N. D., and Katon, J. E., Appl. Spectrosc., 1988, 42(3), 401. 16 Bouhsain, Z., Garrigues, S., and de la Guardia, M., Analyst, 1996, 121, 1935. 17 de la Guardia, M., and R°u�zi�cka, J., Analyst, 1995, 120, 17N.de la Guardia, M., Khalaf, K. D., Carbonell, V., and Morales- Rubio, A., Anal. Chim. Acta., 1995, 308, 462. 19 S�anchez-Das�ý, J., Cervera, M. L., Garrigues, S., and de la Guardia, M., paper presented at the First Mediterranean Basin Conference on Analytical Chemistry, C�ordoba, November, 1995. 20 S�anchez-Das�ý, J., Cervera, M. L., Garrigues, S., and de la Guardia, M., paper presented at the Clean Technology Symposium, London, June, 1996. Paper 6/07109K Received October 18, 1996 Accepted February 3, 1997 Table 2 Analytical features of the FTIR determination of PFZ and CAF in CHCl3 in both stopped-flow and FI modes using a bandpass of 0.5 mm Wavenumber (Baseline)/ Mode Analyte cm21 Parameter Stopped-flow FI A = a + bC A = 0.0114 + 0.0530 C A=20.0001 + 0.0484 C (C in mg ml21) r 0.9997 0.99990 1595 Absorbance PFZ LOD/mg ml21 0.06 0.03 (2000–890) RSD 0.5% A = 0.223 ± 0.001 2.1% A = 0.388 ± 0.008 C = 4 mg ml21 C = 8 mg ml21 dA/dn = a + bC dA/dn = 20.000320.0209 C dA/dn = 20.000120.0205 C (C in mg ml21) r 20.9998 20.9998 1712 Derivative CAF LOD/mg ml21 0.45 1.5 — RSD 0.5% dA/dn = 20.0341 ± 0.0001 1.0% dA/dn = 20.0327 ± 0.0003 C = 1.6 mg ml21 C = 1.6 mg ml21 Table 3 Simultaneous determination of PFZ and CAF in pharmaceuticals Found*/mg Declared/ Stopped- Analyte Sample mg flow FI PLS-UV Megral 300 297 ± 3 300 ± 7 283 ± 1 PFZ Sedalmerck 300 301 ± 2 307 ± 7 296 ± 1 Megral 50 50.7 ± 0.7 49.3 ± 0.8 51.2 ± 0.2 CAF Sedalmerck 50 49.0 ± 0.2 47.8 ± 1.6 52.2 ± 0.2 * Values indicated are the average of three independent analyses ± the corresponding standard deviation. Analyst, May 1997, Vo

ISSN:0003-2654    

DOI:10.1039/a607109k

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Modified Fluorimetric Assay for Estimating Ampicilloate Concentrations and its Use for Detecting b-Lactamase and Penicillin Acylase Activity in Bacteria W. L. Baker School of Chemical Sciences, Swinburne University of Technology, John Street, Hawthorn, Melbourne, Victoria 3122, Australia Sodium ampicilloate concentrations were estimated fluorimetrically by heating solutions with ascorbic acid, EDTA and a modified Lowry A reagent which was prepared by including copper sulfate and potassium sodium tartrate in 0.5 mol dm23 acetate buffer at pH 4.A concentration range of 0.5–50 mmol dm23 was used for the estimations. The reaction was used to estimate b-lactamase activity on ampicillin but the substrate also showed some fluorescence and a calculation was required to determine the amount of ampicilloate formed when both substances were present in the one reaction mixture. The b-lactamase was inhibited by treatment with trichloroacetic acid so the procedure could be used to assay the enzyme activity after a fixed time. 6-Aminopenicillanic acid did not fluoresce on treatment with the modified reagent and organisms which contained penicillin acylase lowered the amount of ampicillin which could be converted to ampicilloate. When penicillin acylase and b-lactamase co-existed in the one organism, the respective activities were determined by use of the copper–ascorbate–EDTA fluorescence assay for ampicilloate coupled with a fluorescamine assay for 6-aminopenicillanic acid determinations.On prolonged incubation, some organisms containing penicillin acylases lowered the amount of ampicilloate which formed a fluorescent product. This effect was attributed to deacylation of ampicilloate by the penicillin acylases. Keywords: b-Lactamase; penicillin acylase; ampicilloate; fluorescence assay; Lowry A reagent; ascorbate b-Lactamase activity may be measured by a macro-iodometric titrimetric procedure,1 a micro-iodometric procedure,2,3 a hydroxylamine reaction4 or colorimetric cephalosporins.5 Activity may also be measured by the rate of reduction of CuII using neocuproine to estimate the CuI formed6 or a fixed time assay using bicinchoninic acid (BCA) to estimate the CuI formed.7 In this latter technique the added CuII hydrolysed a small amount of the penicillin and a calculation was required to obtain accurate results. a-Aminopenicillins were found to react with the biuret reagent to give a green colour.8 Addition of ascorbate, and rigidly controlled conditions, gave linear standard curves in the concentration range of 50–600 mg cm23.Later observations have revealed that a fluorescent product can be detected in solutions containing ampicillin which had been incubated for prolonged periods. At the pH of the biuret reaction some penicilloate of ampicillin (ampicilloate) would be formed so the reaction of this substance with the biuret, and other copper reagents was therefore examined.A fluorimetric method was developed for determination of ampicilloate concentrations based on observations arising from this investigation. The procedure has been used as the basis of a fluorescent assay for estimating b-lactamase activity and also for penicillin acylase activity including activity of the latter enzyme in deacylating ampicilloate. Experimental Chemicals Penicillins and derivatives came from the sources previously described.8 Copper sulfate was obtained from Ajax, Homebush, Sydney, Australia, while potassium sodium tartrate and disodium ethylenediaminetetraacetic acid (EDTA) were obtained from May and Baker, Footscray, Melbourne, Australia.Reduced glutathione (GSH), sodium bicinchoninate (BCA) and a blactamase from Bacillus cereus were obtained from Sigma, St. Louis, MO, USA, while 6-aminopenicillanic acid (6-APA) was obtained from Aldrich, Milwaukee, WI, USA. Ampicilloate and the open b-lactam ring form of other penicillin and cephalosporin derivatives were formed immediately before use by treatment with NaOH followed by neutralisation. Instruments A Shimadzu (Tokyo, Japan) SP500 spectrophotofluorimeter was used for fluorescence readings.A Metrohm AG620 pH meter (Metrohm, Herisau, Switzerland) was used for pH measurements and a Cary 3E spectrophotometer (Varian Instruments, Melbourne, Australia) was used for absorbance readings. Analytical Procedures The reaction of ampicillin with the biuret reagent was effected, in the presence or absence of ascorbate, as previously described.8 A modification of the Lowry reagent9 was used for the fluorescence assay.Two volumes of 2% potassium sodium tartrate and two volumes of 1% CuSO4·5H2O were made to 100 volumes using 0.5 mol dm23 acetate buffer of pH 4 (or buffers of other pH values and strengths as required). b-Lactamase assays were performed by mixing 0.4 cm3 of cell suspension or enzyme with 1.6 cm3 of 6.25 mmol dm23 ampicillin in 0.05 mmol dm23 phosphate buffer of pH 7 (final concentration of ampicillin 5 mmol dm23).At specific time intervals 0.1 cm3 of reaction mixture was added to 0.1 cm3 of cold 8% trichloroacetic acid (TCA) and allowed to stand for 1 min to inactivate the enzyme. The solution was then diluted to 10 cm3 with 0.5 mol dm23 acetate buffer of pH 4 and the concentration of ampicilloate was estimated by the fluorimetric procedure described below. Phosphate buffer (0.05 mol dm23, pH 7) was used as the diluent when ampicilloate concentrations were estimated by either the BCA or micro-iodometric procedures.The fluorimetric procedure was effected by adding 5 cm3 of the modified Lowry A reagent at pH 4 to 1 cm3 of the diluted reaction mixture (1 : 100) prepared as described above. One cm3 Analyst, May 1997, Vol. 122 (447–453) 447of 50 mmol dm23 EDTA containing ascorbate (0.5 mg) was then added and the solution mixed. The tubes were heated in a boiling water bath for exactly 30 min, removed, cooled and the fluorescence read at an excitation wavelength of 344 nm and an emission wavelength of 450 nm.A calculation, which accounted for the fluorescence of both ampicillin and ampicilloate, was required to determine b-lactamase activity. The expression X = (Ftest – Famp )50 (100 – Famp ) was used where 100 is the relative fluorescent intensity given by 50 nmol of ampicilloate (in 1 cm3), Famp is the fluorescence given by 50 nmol of ampicillin (in 1 cm3) obtained from standard curves, Ftest is the the fluorescence of the test solution and x is the number of nanomoles of ampicilloate formed by the enzyme and has to be multiplied by the dilution factor (usually 100) to obtain the amount of ampicilloate formed in the original test solution (mmol dm23).Colorimetric estimations were performed by treating 1 cm3 of diluted solution with 2 cm3 of BCA reagent at 37 °C, incubating for 10 min, adding 0.1 cm3 of 0.2 mol dm23 EDTA and reading the absorbance at 562 nm.7 Estimations of the ampicilloate formed were made against a standard curve of ampicilloate in the concentration range of 2–40 mg cm23.Ampicilloate gave a less intense reaction with the BCA reagent than benzylpenicilloate. A modification of the micro-iodometric procedure was also used. Diluted reaction mixture (0.6 cm3) containing both ampicilloate and ampicillin, up to the total equivalent of 30 nmol, was added to 2.4 cm3 of micro-iodometric reagent,3 from which the b-lactamase enzyme and benzylpenicillin substrate were omitted and 0.4 cm3 of additional buffer was added, then the solution was allowed to decolorise to equilibrium.Standard curves were constructed under the same conditions using varying concentrations of ampicilloate between 5 and 30 nmol cm23. Growth of Organisms For b-lactamase activity all organisms were grown on nutrient agar after flood inoculation. After 16 h at 37 °C the cells were washed from the surface using 2.5 cm3 of 0.05 mol dm23 phosphate buffer of pH 7 and collected.Penicillin acylase activity in cells was induced using a solid medium containing phenylacetic acid (PAA).7 Visual Detection of b-Lactamase b-Lactamase activity in organisms was detected visually after mixing 1 cm3 of solution of b-lactamase or bacterial cell suspension with 1 cm3 of ampicillin solution (3 mg cm23) in 0.05 mol dm23 phosphate buffer of pH 7.This solution was inoculated into holes bored in 20 cm3 of 2% agar containing 3 cm3 of 50 mmol dm23 EDTA, 0.3 cm3 of 1% copper sulfate solution, 0.3 cm3 of 2% potassium sodium tartrate, 2 cm3 of ascorbic acid solution (added at 50 °C) and 14.4 cm3 of 0.5 mmol dm23 acetate buffer of pH 4. The plate was incubated at 37 °C for 16 h and the presence of zones of fluorescence was recorded and compared with controls of cells alone and ampicillin alone. Photographs of plates (or the agar gels alone) were prepared on Kodacolor 100 (ASA100) using a distance of 38 cm, an f stop of 5.6, dual UV side lighting, light of wavelength 366 nm and a UV filter on the camera.An exposure time of 1 s was used. Penicillin Acylase Activity Penicillin acylase activity against benzylpenicillin was measured by the fluorescamine procedure.10 Activity of this enzyme against ampicillin was also measured by mixing 0.4 cm3 of cells of Escherichia coli ATCC 9637, Escherichia coli ATCC 11105 or Alcaligenes faecalis ATCC 15246, which had been grown in the presence of PAA, with 1.6 cm3 of 6.25 mmol dm23 ampicillin at pH 7.8 and shaking in a reciprocating water bath at 37 °C.Samples (0.1 cm3) were taken, the enzyme inactivated using TCA and diluted and analysed for ampicilloate concentrations as described above. Since the pH of the dilute solution was 4 the presence of 6-APA was also conveniently examined using the modification of the fluorescamine assay for ampicillin as a substrate.11,12 When both penicillin acylase and b-lactamase co-existed in the one organism, and showed significant activity, the following procedure was adopted. 6-APA concentrations, formed by the activity of penicillin acylase, were inferred by the loss of fluorescence in solutions which had been treated with NaOH and diluted with 0.5 mol dm23 of acetate buffer of pH 4 and analysed for ampicilloate. They were accurately estimated using fluorescamine analysis at pH 4.A second sample (0.1 cm3) was treated directly with 0.1 cm3 of cold 8% TCA and diluted to 10 cm3 with the same buffer but not treated with NaOH. Duplicate 1 cm3 volumes of each of these solutions were also analysed for ampicilloate and contained a total concentration of ampicilloate and ampicillin which amounted to 50 mmol dm23 less the concentration of 6-APA formed. The amount of ampicillin and ampicilloate remaining in the solution was then calculated from the observed fluorescence of the test solution (Ftest) using the above expression and obtaining Famp and FampOA, from 50 mmol dm23 standards of ampicillin and ampicilloate, respectively.Penicillin Acylase Activity Against Ampicilloate The activity of penicillin acylase activity against ampicilloate was determined simply by substituting 1.6 cm3 of 6.25 mmol dm23 ampicilloate concentration for the ampicillin solutions used in the b-lactamase determinations and estimating the loss of fluorescence in dilutions after incubation.Results Reaction of Ampicillin with Copper Reagents Ampicillin forms a fluorescent product when heated with formaldehyde.13 When treated with the biuret reagent and ascorbate8 solutions of ampicillin (600 mg cm23) developed intense fluorescence. Results were inconsistent and curves of fluorescence versus concentration were exponential in the absence of ascorbate.14 As solutions containing biuret reagent were intensely coloured, the Lowry reagent A was also examined using ascorbate.Fluorescence also developed with this solution but consistent standard curves were not obtained. Nevertheless, this reagent was examined further because it contained less copper and was only faintly coloured. Parameters of the Reaction Between Ampicilloate and Lowry A Reagent Penicilloate would be formed at the pH of the Lowry A reagent (12.2) and the biuret reagent (11.4) and it seemed that the observed fluorescence may have developed from ampicilloate.Reformulation of Lowry A reagent by adding copper and potassium sodium tartrate to buffers of varying pH, and examining the fluorescence after heating, indicated that maximum ampicilloate fluorescence developed at pH 4 and pH 5 [Fig. 1(A)]. Ampicillin showed only a small amount of fluorescence within the pH range 3–11 [Fig. 1(A)]. At pH 12.2 the fluorescence of ampicillin and ampicilloate were identical, confirming that the ampicilloate was the analyte which developed a fluorescent product. 448 Analyst, May 1997, Vol. 122At pH 4, the fluorescence developed slowly in this reaction at room temperature and was maximum after about 2.5 h. It then faded relatively slowly. Variation in the acetate concentration of the buffer showed that the fluorescence increased as the concentration range was raised to 0.5 mol dm23 at pH 4 [Fig. 1(B)]. Variation of the copper ion concentration showed that the maximum fluorescence was obtained at the same concentration of copper ion as the Lowry A reagent [Fig. 1(C)]. On excitation a maximum was observed at 344 nm (Fig. 2) and this wavelength was used for the analyses. A second peak was also observed at a maximum of 379 nm. A peak at this wavelength was not observed in the previous analytical technique.13 The maximum wavelength of emission was 450 nm which is greater than that observed in the method using formaldehyde.13 The Effect of Heating the Reaction Mixture With Other Reducing Agents Heating was not essential for the development of fluorescence but other workers prepared fluorescent products from ampicillin by heating with citrate buffer and formaldehyde at acidic pH values.13,15 Formaldehyde is volatile and toxic and its use is undesirable if heating is required.Therefore, heating the copper–acetate–ascorbate system was examined as a method of developing the fluorescence in the presence of reducing agents such as ascorbic acid and reduced glutathione.Since free CuII rapidly oxidises both ascorbate16 and thiols,17 and copper ions may quench fluorescence,18 the effect of EDTA was also examined. The most intense fluorescence was obtained when the EDTA was added to the ascorbate reagent to prevent its oxidation by copper (Table 1). Intact ampicillin showed a smaller amount of fluorescence. After cooling the fluorescence of the solution was stable. Standard Curves A standard curve, constructed within the concentration range of 5–50 mmol dm23 of ampicilloate (Fig. 3), was linear.This range was used routinely in enzyme experiments. The small fluorescence of ampicillin, treated in the same manner and within the same concentration range as the ampicilloate, was also recorded and used for calculations. The equation for the line of Fig. 3 was y = 2.6 x 2 2. The mean of 40 estimations at a concentration of 25 mmol dm23 was 25.25 mmol dm23 with a standard deviation of 0.76 mmol dm23.A concentration range of 0.5–5 mmol dm23 ampicilloate (and ampicillin) could also be used and gave reproducible results. Results below this concentration range were inconsistent. Reaction With Other Penicilloates The reaction with the intact and the open b-lactam ring forms of amoxycillin, 6-APA and cephalothin in addition to ampicillin are shown in Table 2. At the wavelengths used the fluorescence of the penicilloate of amoxycillin was small compared with ampicillin while the fluorescence given by 6-APA and cephalothin, and their open lactam ring forms [penicic acid Fig. 1 Factors affecting the fluorescence of ampicillin and ampicilloate with the modified copper reagent. A, The effect of pH on (2) 100 nmol of ampicilloate and (5) 100 nmol of ampicillin. Copper reagent contained 0.1 cm3 of 2% potassium sodium tartrate, 0.1 cm3 of 1% copper sulfate, 0.5 mg of ascorbic acid, 50 mmol of EDTA and 240 mmol of buffer of the respective pH in a total volume of 6 cm3.B, The effect of acetate concentration at pH 4 on the fluorescence of the reaction. The reaction mixtures contained the same concentration of potassium sodium tartrate and copper sulfate as shown in the solutions used for A. C, The effect of copper ion concentration on fluorescence using 0.5 mol dm23 acetate buffer of pH 4 and the same concentration of potassium sodium tartrate and copper sulfate as shown in the legend to A. Fig. 2 Excitation and emission fluorescence spectra of the compound(s) formed by heating 50 nmol cm23 of ampicilloate (———) or ampicillin (------) with a modified Lowry A reagent at pH 4 containing ascorbate and EDTA.Table 1 Effect of heating in the presence of reducing agents* Relative fluorescence Chemical treatment Ampicillin Ampicilloate No other additions 3.3 1.2 Ascorbate 8 18.4 GSH 2.8 8 EDTA 5.5 62 Ascorbate + EDTA 12 100 GSH + EDTA 3 65 * All solutions contained 50 nmol cm23 of ampicillin or ampicilloate in a final volume of 7 cm3 as described under Experimental.EDTA was added at a concentration of 50 mmol per 7 cm3 whilst ascorbate and other thiol additives were added at a concentration of 0.5 mg cm23. Fig. 3 Relationship between relative fluorescence intensity and either ampicilloate concentration (5) and ampicillin concentration (2) within the concentration range of 5–50 mmol dm23. Duplicate readings were used for the standard curve. The most concentrated solution was used to set the 100% relative fluorescence intensity reference point.Analyst, May 1997, Vol. 122 449(6-APOA), and cephalothinoate, respectively] was negligible. The open b-lactam ring form of amoxycillin (amoxycilloate) fluoresced strongly at longer excitation and emission wavelengths. Ampicilloate and amoxycilloate have a free a-amino group. Ampicillin is metabolised to a piperazine derivative19 which does not contain a free amino group. Benzylpenicillin and phenoxymethylpenicillin do not contain a free a-amino group and did not develop fluorescence under the conditions of this work.Ampicillin was a suitable substrate for the fluorimetric assay of b-lactamase activity. Penicillin acylase activity on ampicillin could be estimated from a loss of fluorescence upon formation of 6-APA, and penicillin acylase activity on ampicilloate as a substrate could also be estimated by the loss of fluorescence of solutions. Detection of b-Lactamase Activity in Micro-organisms b-Lactamase activity in solutions and culture broths of organisms was detected by an intense blue fluorescent circle surrounding the bore-holes in agar plates filled with ampicillin (1.5 mg cm23) and b-lactamase solution or culture broth of the organism (Fig. 4). Enzyme solution and bacterial suspensions alone did not cause fluorescence. Estimation of b-Lactamase Activity Table 3 indicates the results of b-lactamase activity of the commercial concentrated enzyme from Bacillus cereus using ampicillin as a substrate and three methods of determination of enzyme activity.The measured activities were similar using each method. b-Lactamase activity in Aerobacter faecalis ATCC 15246 and Klebsiella pneumoniae against both ampicillin and 6-APA is shown in Fig. 5. Both organisms hydrolysed the b-lactam ring of ampicillin (measured by the formation of ampicilloate) but only K. pneumoniae hydrolysed the b-lactam ring of 6-APA (measured by loss of fluorescence on treatment with fluorescamine at pH 4).b-Lactamase activity of several other organisms (not subjected to induction with methicillin) was estimated by the formation of fluorescent reacting products from ampicillin (Table 4). In most cases the enzyme activity of the intact organisms was low and long incubation times were required contrasting with the enzymes found in Alc. faecalis and K. pneumoniae. The elevated activity of the culture of B. cereus suggested a ready access of the b-lactamase to the substrate, possibly due to an extracellular secretion.The nature of the enzyme was not investigated at this stage but could have been a mixture of type I and type II metallo-b-lactamases.20 Determination of Penicillin Acylase Activity Penicillin acylase activity against benzylpenicillin was estimated by the fluorescamine procedure.10 Results are shown in Fig. 6(A) for uninduced and induced cells of E. coli ATCC 9637 and A. faecalis ATCC 15246. The results obtained after incubation of the penicillin acylase producing E.coli ATCC 9637 with ampicillin, and conversion of residual penicillin to ampicilloate (NaOH), confirm that Table 2 Reaction of penicilloates of penicillins and derivatives* Compound used Relative fluorescence intensity 6-APOA 0.3 6-APA 0 Amoxycilloate 14.5 Amoxycillin 1 Cephalothin 0 Cephalothinoate 0 Ampicillin 11.2 Ampicilloate 100 * Penicilloates were used at a concentration of 100 mg cm23 and were treated with 5 cm3 of a mixture containing 0.1 cm3 of 2% potassium sodium tartrate and 0.1 cm3 of 1% CuSO4·5H2O in 0.5 mol dm23 acetate buffer of pH 4.One cm3 of a mixture of ascorbic acid (0.5 mg cm23) in 50 mmol dm23 EDTA was added, the solution heated for 30 min and then cooled and the the fluorescence was read using an excitation wavelength of 344 nm and an emission wavelength of 450 nm. Fig. 4 Detection of b-lactamase activity from B. subtilis on ampicillin. Procedures are described under materials and methods.T indicates a solution containing 289 mg cm23 of b-lactamase and 1.5 mg cm23 of ampicillin which was placed in the bore-hole. E indicates an enzyme control while A indicates an ampicillin control. Table 3 Comparison of results of b-lactamase activity* Enzyme activity/ Time of sample/ mmol dm23 Method used min ampicilloate formed Micro-iodometric 5 1.98 10 2.98 15 3.63 BCA 5 2.06 10 2.93 15 3.67 Fluorimetric 5 2.21 10 2.44 15 3.79 * Samples of b-lactamase [0.4 cm3 of b-lactamase from B.cereus (Sigma) 14.9 mg cm23 which showed two bands on SDS-PAGE at 22.5 kD and 28 kD indicative of a mixture of type I and type II enzymes20] were mixed with 1.6 cm3 of of 6.25 mmol dm23 ampicillin in 0.05 mol dm23 phosphate buffer of pH 7 and allowed to react at 37 °C. At the times indicated two 0.1 cm3 samples were taken and added to 0.1 cm3 of 8% TCA to inactivate the enzyme. One sample was then diluted to 10 cm3 with 0.5 mol dm23 acetate buffer of pH 4 and treated as described for the fluorimetric assay.The second sample was diluted with 0.05 mol dm23 phosphate buffer of pH 7 and duplicate 1 cm3 samples were added to the BCA reagent at pH 7. This reaction mixture was raised to 37 °C, and 0.1 cm3 of 20 mmol dm23 copper sulfate was added and mixed and after 10 min reaction 0.1 cm3 of 0.2 mol dm23 EDTA was then added. Absorbance was read at 562 nm. Further duplicate 0.6 cm3 samples of this reaction mixture were added to 2.4 cm3 of microiodometric reagent prepared by mixing 1 cm3 of starch–iodine and 1.4 cm3 of 0.1 mol dm23 phosphate buffer of pH 7 and the colour allowed to fade for 5 min prior to reading the absorbance at 620 nm. 450 Analyst, May 1997, Vol. 1226-APA production is accompanied by loss of fluorescence [Fig. 6(B)]. Using this organism the same downward trend was observed in solutions where sodium hydroxide treatment was not used to convert residual ampicillin to ampicilloate.The blactamase activity of this organism is very low7,21 and was barely detectable when ampicilloate estimations were made. The loss of ampicilloate and the formation of 6-APA (measured by using fluorescamine) were quantitatively related [Fig. 6(B)]. Alc. faecalis ATCC 15246, on the other hand, contains an active penicillin acylase as well as an active b-lactamase. This organism showed a similar downward trend in the loss of ampicillin and ampicilloate reacting activity to that observed using E.coli ATCC9637 after solutions had been treated with NaOH. The active b-lactamase resulted in an increase in fluorescence in solutions not treated with NaOH as a significant amount of ampicilloate was formed. In this case only smaller amounts of 6-APA were formed by the action of the penicillin acylase activity on ampicillin when compared with the cells of E. coli. This was thought to reflect the differences in rates of activities of the respective enzymes and the inability of the blactamase in Alc.faecalis to hydrolyse 6-APA (Fig. 5). Penicillin Acylase Activity and Ampicilloate In prolonged incubations using Acl. faecalis ATCC 15246 the calculated concentrations of ampicilloate fell. This effect was attributed to penicillin acylase activity on the ampicilloate formed. Therefore, ampicilloate was incubated with cells of penicillin acylase producing organisms and the activity of the enzyme against this substrate was measured by the loss of fluorescence.The results (Table 5) certainly indicate that E. coli ATCC 9637 and ATCC 11105 and A. faecalis ATCC 15246 did hydrolyse the amide linkage between the 6-APOA and phenylglycine moieties resulting in two products which could not give the fluorescence on heating solutions. In all cases penicillin acylase activity against ampicilloate occurred at a much slower rate than against benzylpenicillin and was still incomplete after 24 h reaction time.Discussion Spectrofluorimetry has many analytical advantages over conventional spectrophotometry. Sensitivity depends on the fluorophore and may be varied by attenuation of the detection system of the instrumentation. Selective wavelengths of excitation and emission allow specificity and there may not be a requirement for sample clean up when assays are used with biological fluids. b-Lactamases of many organisms use ampicillin as a substrate and hydrolysis frequently occurs at similar rates to benzylpenicillin.22 Ampicilloate, the end product of enzyme activity, forms a fluorescent product under specific conditions and in the present work the toxic formaldehyde used by the earlier workers13,15 was omitted from the heating step.Mercuric ion, which was used to catalyse formation of a fluorescent compound from ampicilloate,23 was replaced by copper ion as well as EDTA and fluorescence was generated in the presence of the non-toxic biological reducing agent, ascorbate.The blactamases in Gram negative organisms are considered to be more active against ampicillin than against benzylpenicillin, so ampicillin seems to be an excellent substrate to measure this enzyme’s activity. Ampicillin is acid stable22 so it was possible to inhibit b-lactamase activity by TCA prior to development of fluorescence. This permitted measurement of b-lactamase activity after a fixed time. Acid inactivation overcomes some of the objections for estimation of b-lactamase activity using Fig. 5 Hydrolysis with time of ampicillin by the b-lactamase in K. pneumoniae (2) and Alc. faecalis (5) and 6-APA by the enzyme in K. pneumoniae (8) and Alc. faecalis (-). Ampicillin or 6-APA (6.25 mmol dm23, 1.6 cm3) and cells (0.4 cm3) were mixed and shaken at 37 °C. Samples (0.1 cm3) were taken and added to 0.1 cm3 of cold 8% TCA and diluted to 10 cm3 with 0.5 mmol dm23 acetate buffer of pH 4. Further procedures were as described under Experimental using fluorescamine to estimate the disappearance of 6-APA or heating in the presence of copper– ascorbate and EDTA to estimate the appearance of ampicilloate.Table 4 Effect of b-lactamase in bacteria on ampicillin* Activity/mmol of ampicilloate Bacterium formed per cm3 per 16 h Proteus vulgaris 0.58 Bacillus subtilis 0.62 Bacillus cereus 4.89 Escherichia coli K12 0.7 Erwinia carotovora 0.7 Serratia marscescens 0.32 Pseudomonas aeruginosa 1.42 Pseudomonas fluorescens 0.35 * Cell suspensions of organisms, grown on nutrient agar plates, were prepared as described under Experimental and 0.4 cm3 mixed with 1.6 cm3 of 6.25 mmol dm23 ampicillin in 0.05 mol dm23 phosphate buffer at pH 7. The mixtures then shaken in a reciprocating water bath at 37 °C and 0.1 cm3 samples taken at the time indicated, added to 0.1 cm3 of cold 8% TCA and diluted to 10 cm3 with 0.5 mmol dm23 acetate buffer of pH 4.Fig. 6 A, Effect of penicillin acylase activity in E.coli ATCC 9637 and Alc. faecalis ATCC 15246 against benzylpenicillin. Symbol (2) is E. coli grown on phenylacetic acid medium7 and (5) is the organism grown on nutrient agar. Symbol (8) is Alc. faecalis grown on phenylacetic acid medium and symbol (-) is the organism grown on nutrient agar. Activities were measured by the fluorescamine procedure. B, Indicates penicillin acylase activity against ampicillin. The open symbols refer to E. coli and A. faecalis as described for A and indicate the estimated ampicillin remaining in solution after treatment with NaOH and heating dilute solutions with the copper reagent.The closed symbols refer to the amount of 6-APA formed by E. coli (5) and Alc. faecalis (-) measured by use of fluorescamine while the symbol (D) indicates an estimate of the amount of ampicilloate formed in the solution containing Alc. faecalis by the b-lactamase present in this organism. Results of ampicilloate formation by the activity of the blactamase of E. coli ATCC 9637 on ampicillin were small and are not shown.Analyst, May 1997, Vol. 122 451iodine because some enzymes (or some forms of the enzyme) are not inhibited by the addition of iodine and sometimes the iodine reacts with the tyrosine residues of the enzyme.24,25 The present technique is suitable for use with concentrated enzyme solutions or with those organisms in which b-lactamase is very active (e.g., K. pneumoniae and Alc.faecalis). It can be used in organisms which contain low b-lactamase activity but in these cases long incubation times are required. The major drawback with the analytical technique is that the lesser amount of fluorescence given by the intact ampicillin in the solutions must be accounted for by a calculation in a similar manner to the absorbance given by intact penicillin in the BCA assay.7 Intact ampicillin is converted to a penicillenate derivative after heating at a pH near to 526 (Scheme 1).Penicilloates do not form penicillenates.26 2-Hydroxy-3-phenyl-6-methylpyrazine (Scheme 1) has been identified as the fluorescent compound formed by heating ampicillin with formaldehyde.15 The spectral properties (Fig. 2) of the compound(s) formed in the present work are different from those of other fluorescent derivatives of ampicillin.13 The fluorescent compound(s) and its mechanism of formation is not known but it seems possible that a substituted piperazine ring is formed and is subsequently oxidised to a pyrazine.The ability to form a fluorescent compound is lost when 6-APA is formed from ampicillin by the action of penicillin acylase. Furthermore, the action of penicillin acylase on ampicilloate results in loss of ability to form a fluorescent compound. Hence, by examining the activity of penicillin acylase producing organisms against ampicilloate the production of 6-APOA may be be inferred. The production of 6-APOA from benzylpenicilloate occurs more slowly than the production of 6-APA from benzylpenicillin.21,27 The type 2 penicillin acylases in the organisms used in this work also acted on ampicilloate at a slower rate than benzylpenicillin but it was possible to make an accurate quantitative estimate of the deacylation of this penicilloate after the incubating reaction mixtures for prolonged periods.References 1 Perrett, C. J., Nature (London), 1954, 174, 1012. 2 Novick, R. P., Biochem.J., 1962, 83, 236. 3 Sykes, R. B., and Nordstrom, K., Antimicrob. Agents Chemother., 1972, 1, 94. 4 Batchelor, F. R., Cameron-Wood, J., Chain, E. B., and Rolinson, G. N., Proc. R. Soc., London, B, 1961, 154, 514. 5 O’Callaghan, C. H., Morris, A., Kirby, S. M., and Shingler, A., Antimicrob. Agents Chemother., 1972, 1, 283. 6 Cohenford, M. A., Abraham, J., and Medeiros, A. E., Anal. Biochem., 1988, 168, 252. 7 Baker, W. L., J. Appl. Bacteriol., 1992, 73, 14. 8 Baker, W. L., Antonie van Leeuwenhoek., 1983, 49, 551. Table 5 Effect of penicillin producing organisms on ampicilloate* Penicillin acylase activity against substrate/ mmol dm23 per cm3 cell suspension per time Growth Benzylpenicillin/ Ampicillin/ Ampicilloate/ Source of enzyme medium 4 h 4 h 24 h E.coli ATCC 9637 A 1.27 0.27 0 B 4.6 0.45 0.83 E. coli ATCC 11105 A 1.31 0.31 0.97 B 4.31 1.18 1.25 Alc. faecalis ATCC15246 A 1.69 0.16 2.30 B 1.63 0.26 3.41 * A cell suspension of organisms (0.4 cm3) was obtained after growth on A, nutrient agar or B, solidified medium containing phenylacetic acid and mixed with 1.6 cm3 of 6.25 mmol dm23 benzylpenicillin, ampicillin or sodium ampicilloate, respectively, (prepared by NaOH treatment) in 0.05 mol dm23 of phosphate buffer of pH 7.8 and incubated on a reciprocating water bath at 37 °C.Samples (0.1 cm3) were taken at the times indicated and mixed with 0.1 cm3 of cold 8% TCA and then diluted to 10 cm3 with 0.5 mol dm23 acetate buffer of pH 4.The diluted solution was analysed for 6-APA concentrations using the fluorescamine assay (benzylpenicillin and ampicillin) or disappearance of ampicilloate concentrations according to the procedure described in Experimental. Scheme 1 Proposed reactions of ampicillin and ampicilloate thought to occur under the conditions used in this work. Reaction A is a typical hydrolysis of the b-lactam ring caused by the enzyme b-lactamase. Reaction B occurs only with intact penicillins at a pH of about 5 and especially in the presence of copper.26 Reaction C seems to involve a ring closure.A fluorescent compound was formed when ampicillin and its ampicilloate was heated with formaldehyde13 or ampicilloate was treated with mercuric ion.23 A fluorescent compound with the structural formula shown above was isolated.15 452 Analyst, May 1997, Vol. 1229 Lowry, O. H., Rosebrough, N. J., Farr, A. L., and Randall, R. J., J. Biol. Chem., 1951, 193, 265. 10 Baker, W. L., Antimicrob. Agents Chemother., 1983, 23, 26. 11 Baker, W. L., Aust. J. Biol. Sci., 1984, 37, 257. 12 Baker, W. L., and Havlicek, P. J., J. Gen. Appl. Microbiol., 1985, 31, 107. 13 Jusko, W. J., J. Pharm. Sci., 1971, 60, 728. 14 Baker, W. L., J. Appl. Bacteriol., 1980, 49, 225. 15 Barbhaiya, R. H., Brown, R. C., Payling, D. W., and Turner, P., J. Pharm. Pharmacol., 1978, 30, 224. 16 Taqui Khan, M. M., and Martell, A. E., J. Am. Chem. Soc., 1967, 89, 4176. 17 Jocelyn, P. C., Biochemistry of the SH Group: The Occurrence, Chemical Properties, Metabolism and Biological Function of Thiols and Disulphides, Academic Press, New York, 1972, p. 95. 18 Lakowicz, J., Principles of Fluorescence Spectroscopy, Plenum Press, New York, 1983, p. 259. 19 Haginaka, J., and Wakai, J., J. Pharm. Pharmacol., 1986, 38, 225. 20 Thatcher, D. R., in Methods in Enzymology, ed. Hash, J. H., Academic Press, New York, 1975, vol. 43, pp. 640–664. 21 Pruess, D. L., and Johnson, M.J., J. Bacteriol., 1965, 90, 380. 22 Rolinson, G., and Stevens, S., Br. Med. J., 1961, ii, 191. 23 Miyazaki, K., Ogino, O., and Arita, T., 1974, Chem. Pharm. Bull., 1974, 22, 1910. 24 Citri, N., Biochim. Biophys. Acta, 1958, 27, 277. 25 Csanyi, V., Acta Physiol. Acad. Sci. Hungary, 1961, 18, 261. 26 Smith, J. W. G., de Grey, G. E., and Patel, V. J., Analyst, 1967, 92, 247. 27 Hamilton-Miller, J. M. T., Bacteriol. Rev., 1966, 30, 761. Paper 6/08053G Received November 28, 1996 Accepted February 17, 1997 Analyst, May 1997, Vol. 122 453 Modified Fluorimetric Assay for Estimating Ampicilloate Concentrations and its Use for Detecting b-Lactamase and Penicillin Acylase Activity in Bacteria W. L. Baker School of Chemical Sciences, Swinburne University of Technology, John Street, Hawthorn, Melbourne, Victoria 3122, Australia Sodium ampicilloate concentrations were estimated fluorimetrically by heating solutions with ascorbic acid, EDTA and a modified Lowry A reagent which was prepared by including copper sulfate and potassium sodium tartrate in 0.5 mol dm23 acetate buffer at pH 4.A concentration range of 0.5–50 mmol dm23 was used for the estimations. The reaction was used to estimate b-lactamase activity on ampicillin but the substrate also showed some fluorescence and a calculation was required to determine the amount of ampicilloate formed when both substances were present in the one reaction mixture. The b-lactamase was inhibited by treatment with trichloroacetic acid so the procedure could be used to assay the enzyme activity after a fixed time. 6-Aminopenicillanic acid did not fluoresce on treatment with the modified reagent and organisms which contained penicillin acylase lowered the amount of ampicillin which could be converted to ampicilloate. When penicillin acylase and b-lactamase co-existed in the one organism, the respective activities were determined by use of the copper–ascorbate–EDTA fluorescence assay for ampicilloate coupled with a fluorescamine assay for 6-aminopenicillanic acid determinations. On prolonged incubation, some organisms containing penicillin acylases lowered the amount of ampicilloate which formed a fluorescent product.This effect was attributed to deacylation of ampicilloate by the penicillin acylases. Keywords: b-Lactamase; penicillin acylase; ampicilloate; fluorescence assay; Lowry A reagent; ascorbate b-Lactamase activity may be measured by a macro-iodometric titrimetric procedure,1 a micro-iodometric procedure,2,3 a hydroxylamine reaction4 or colorimetric cephalosporins.5 Activity may also be measured by the rate of reduction of CuII using neocuproine to estimate the CuI formed6 or a fixed time assay using bicinchoninic acid (BCA) to estimate the CuI formed.7 In this latter technique the added CuII hydrolysed a small amount of the penicillin and a calculation was required to obtain accurate results.a-Aminopenicillins were found to react with the biuret reagent to give a green colour.8 Addition of ascorbate, and rigidly controlled conditions, gave linear standard curves in the concentration range of 50–600 mg cm23. Later observations have revealed that a fluorescent product can be detected in solutions containing ampicillin which had been incubated for prolonged periods. At the pH of the biuret reaction some penicilloate of ampicillin (ampicilloate) would be formed so the reaction of this substance with the biuret, and other copper reagents was therefore examined.A fluorimetric method was developed for determination of ampicilloate concentrations based on observations arising from this investigation. The procedure has been used as the basis of a fluorescent assay for estimating b-lactamase activity and also for penicillin acylase activity including activity of the latter enzyme in deacylating ampicilloate. Experimental Chemicals Penicillins and derivatives came from the sources previously described.8 Copper sulfate was obtained from Ajax, Homebush, Sydney, Australia, while potassium sodium tartrate and disodium ethylenediaminetetraacetic acid (EDTA) were obtained from May and Baker, Footscray, Melbourne, Australia.Reduced glutathione (GSH), sodium bicinchoninate (BCA) and a blactamase from Bacillus cereus were obtained from Sigma, St. Louis, MO, USA, while 6-aminopenicillanic acid (6-APA) was obtained from Aldrich, Milwaukee, WI, USA.Ampicilloate and the open b-lactam ring form of other penicillin and cephalosporin derivatives were formed immediately before use by treatment with NaOH followed by neutralisation. Instruments A Shimadzu (Tokyo, Japan) SP500 spectrophotofluorimeter was used for fluorescence readings. A Metrohm AG620 pH meter (Metrohm, Herisau, Switzerland) was used for pH measurements and a Cary 3E spectrophotometer (Varian Instruments, Melbourne, Australia) was used for absorbance readings.Analytical Procedures The reaction of ampicillin with the biuret reagent was effected, in the presence or absence of ascorbate, as previously described.8 A modification of the Lowry reagent9 was used for the fluorescence assay. Two volumes of 2% potassium sodium tartrate and two volumes of 1% CuSO4·5H2O were made to 100 volumes using 0.5 mol dm23 acetate buffer of pH 4 (or buffers of other pH values and strengths as required). b-Lactamase assays were performed by mixing 0.4 cm3 of cell suspension or enzyme with 1.6 cm3 of 6.25 mmol dm23 ampicillin in 0.05 mmol dm23 phosphate buffer of pH 7 (final concentration of ampicillin 5 mmol dm23).At specific time intervals 0.1 cm3 of reaction mixture was added to 0.1 cm3 of cold 8% trichloroacetic acid (TCA) and allowed to stand for 1 min to inactivate the enzyme. The solution was then diluted to 10 cm3 with 0.5 mol dm23 acetate buffer of pH 4 and the concentration of ampicilloate was estimated by the fluorimetric procedure described below.Phosphate buffer (0.05 mol dm23, pH 7) was used as the diluent when ampicilloate concentrations were estimated by either the BCA or micro-iodometric procedures. The fluorimetric procedure was effected by adding 5 cm3 of the modified Lowry A reagent at pH 4 to 1 cm3 of the diluted reaction mixture (1 : 100) prepared as described above. One cm3 Analyst, May 1997, Vol. 122 (447–453) 447of 50 mmol dm23 EDTA containing ascorbate (0.5 mg) was then added and the solution mixed.The tubes were heated in a boiling water bath for exactly 30 min, removed, cooled and the fluorescence read at an excitation wavelength of 344 nm and an emission wavelength of 450 nm. A calculation, which accounted for the fluorescence of both ampicillin and ampicilloate, was required to determine b-lactamase activity. The expression X = (Ftest – Famp )50 (100 – Famp ) was used where 100 is the relative fluorescent intensity given by 50 nmol of ampicilloate (in 1 cm3), Famp is the fluorescence given by 50 nmol of ampicillin (in 1 cm3) obtained from standard curves, Ftest is the the fluorescence of the test solution and x is the number of nanomoles of ampicilloate formed by the enzyme and has to be multiplied by the dilution factor (usually 100) to obtain the amount of ampicilloate formed in the original test solution (mmol dm23).Colorimetric estimations were performed by treating 1 cm3 of diluted solution with 2 cm3 of BCA reagent at 37 °C, incubating for 10 min, adding 0.1 cm3 of 0.2 mol dm23 EDTA and reading the absorbance at 562 nm.7 Estimations of the ampicilloate formed were made against a standard curve of ampicilloate in the concentration range of 2–40 mg cm23. Ampicilloate gave a less intense reaction with the BCA reagent than benzylpenicilloate. A modification of the micro-iodometric procedure was also used. Diluted reaction mixture (0.6 cm3) containing both ampicilloate and ampicillin, up to the total equivalent of 30 nmol, was added to 2.4 cm3 of micro-iodometric reagent,3 from which the b-lactamase enzyme and benzylpenicillin substrate were omitted and 0.4 cm3 of additional buffer was added, then the solution was allowed to decolorise to equilibrium.Standard curves were constructed under the same conditions using varying concentrations of ampicilloate between 5 and 30 nmol cm23. Growth of Organisms For b-lactamase activity all organisms were grown on nutrient agar after flood inoculation.After 16 h at 37 °C the cells were washed from the surface using 2.5 cm3 of 0.05 mol dm23 phosphate buffer of pH 7 and collected. Penicillin acylase activity in cells was induced using a solid medium containing phenylacetic acid (PAA).7 Visual Detection of b-Lactamase b-Lactamase activity in organisms was detected visually after mixing 1 cm3 of solution of b-lactamase or bacterial cell suspension with 1 cm3 of ampicillin solution (3 mg cm23) in 0.05 mol dm23 phosphate buffer of pH 7.This solution was inoculated into holes bored in 20 cm3 of 2% agar containing 3 cm3 of 50 mmol dm23 EDTA, 0.3 cm3 of 1% copper sulfate solution, 0.3 cm3 of 2% potassium sodium tartrate, 2 cm3 of ascorbic acid solution (added at 50 °C) and 14.4 cm3 of 0.5 mmol dm23 acetate buffer of pH 4. The plate was incubated at 37 °C for 16 h and the presence of zones of fluorescence was recorded and compared with controls of cells alone and ampicillin alone.Photographs of plates (or the agar gels alone) were prepared on Kodacolor 100 (ASA100) using a distance of 38 cm, an f stop of 5.6, dual UV side lighting, light of wavelength 366 nm and a UV filter on the camera. An exposure time of 1 s was used. Penicillin Acylase Activity Penicillin acylase activity against benzylpenicillin was measured by the fluorescamine procedure.10 Activity of this enzyme against ampicillin was also measured by mixing 0.4 cm3 of cells of Escherichia coli ATCC 9637, Escherichia coli ATCC 11105 or Alcaligenes faecalis ATCC 15246, which had been grown in the presence of PAA, with 1.6 cm3 of 6.25 mmol dm23 ampicillin at pH 7.8 and shaking in a reciprocating water bath at 37 °C.Samples (0.1 cm3) were taken, the enzyme inactivated using TCA and diluted and analysed for ampicilloate concentrations as described above. Since the pH of the dilute solution was 4 the presence of 6-APA was also conveniently examined using the modification of the fluorescamine assay for ampicillin as a substrate.11,12 When both penicillin acylase and b-lactamase co-existed in the one organism, and showed significant activity, the following procedure was adopted. 6-APA concentrations, formed by the activity of penicillin acylase, were inferred by the loss of fluorescence in solutions which had been treated with NaOH and diluted with 0.5 mol dm23 of acetate buffer of pH 4 and analysed for ampicilloate.They were accurately estimated using fluorescamine analysis at pH 4. A second sample (0.1 cm3) was treated directly with 0.1 cm3 of cold 8% TCA and diluted to 10 cm3 with the same buffer but not treated with NaOH. Duplicate 1 cm3 volumes of each of these solutions were also analysed for ampicilloate and contained a total concentration of ampicilloate and ampicillin which amounted to 50 mmol dm23 less the concentration of 6-APA formed. The amount of ampicillin and ampicilloate remaining in the solution was then calculated from the observed fluorescence of the test solution (Ftest) using the above expression and obtaining Famp and FampOA, from 50 mmol dm23 standards of ampicillin and ampicilloate, respectively. Penicillin Acylase Activity Against Ampicilloate The activity of penicillin acylase activity against ampicilloate was determined simply by substituting 1.6 cm3 of 6.25 mmol dm23 ampicilloate concentration for the ampicillin solutions used in the b-lactamase determinations and estimating the loss of fluorescence in dilutions after incubation.Results Reaction of Ampicillin with Copper Reagents Ampicillin forms a fluorescent product when heated with formaldehyde.13 When treated with the biuret reagent and ascorbate8 solutions of ampicillin (600 mg cm23) developed intense fluorescence. Results were inconsistent and curves of fluorescence versus concentration were exponential in the absence of ascorbate.14 As solutions containing biuret reagent were intensely coloured, the Lowry reagent A was also examined using ascorbate.Fluorescence also developed with this solution but consistent standard curves were not obtained. Nevertheless, this reagent was examined further because it contained less copper and was only faintly coloured. Parameters of the Reaction Between Ampicilloate and Lowry A Reagent Penicilloate would be formed at the pH of the Lowry A reagent (12.2) and the biuret reagent (11.4) and it seemed that the observed fluorescence may have developed from ampicilloate.Reformulation of Lowry A reagent by adding copper and potassium sodium tartrate to buffers of varying pH, and examining the fluorescence after heating, indicated that maximum ampicilloate fluorescence developed at pH 4 and pH 5 [Fig. 1(A)]. Ampicillin showed only a small amount of fluorescence within the pH range 3–11 [Fig. 1(A)]. At pH 12.2 the fluorescence of ampicillin and ampicilloate were identical, confirming that the ampicilloate was the analyte which developed a fluorescent product. 448 Analyst, May 1997, Vol. 122At pH 4, the fluorescence developed slowly in this reaction at room temperature and was maximum after about 2.5 h. It then faded relatively slowly. Variation in the acetate concentration of the buffer showed that the fluorescence increased as the concentration range was raised to 0.5 mol dm23 at pH 4 [Fig. 1(B)]. Variation of the copper ion concentration showed that the maximum fluorescence was obtained at the same concentration of copper ion as the Lowry A reagent [Fig. 1(C)]. On excitation a maximum was observed at 344 nm (Fig. 2) and this wavelength was used for the analyses. A second peak was also observed at a maximum of 379 nm. A peak at this wavelength was not observed in the previous analytical technique.13 The maximum wavelength of emission was 450 nm which is greater than that observed in the method using formaldehyde.13 The Effect of Heating the Reaction Mixture With Other Reducing Agents Heating was not essential for the development of fluorescence but other workers prepared fluorescent products from ampicillin by heating with citrate buffer and formaldehyde at acidic pH values.13,15 Formaldehyde is volatile and toxic and its use is undesirable if heating is required.Therefore, heating the copper–acetate–ascorbate system was examined as a method of developing the fluorescence in the presence of reducing agents such as ascorbic acid and reduced glutathione.Since free CuII rapidly oxidises both ascorbate16 and thiols,17 and copper ions may quench fluorescence,18 the effect of EDTA was also examined. The most intense fluorescence was obtained when the EDTA was added to the ascorbate reagent to prevent its oxidation by copper (Table 1). Intact ampicillin showed a smaller amount of fluorescence. After cooling the fluorescence of the solution was stable.Standard Curves A standard curve, constructed within the concentration range of 5–50 mmol dm23 of ampicilloate (Fig. 3), was linear. This range was used routinely in enzyme experiments. The small fluorescence of ampicillin, treated in the same manner and within the same concentration range as the ampicilloate, was also recorded and used for calculations. The equation for the line of Fig. 3 was y = 2.6 x 2 2. The mean of 40 estimations at a concentration of 25 mmol dm23 was 25.25 mmol dm23 with a standard deviation of 0.76 mmol dm23.A concentration range of 0.5–5 mmol dm23 ampicilloate (and ampicillin) could also be used and gave reproducible results. Results below this concentration range were inconsistent. Reaction With Other Penicilloates The reaction with the intact and the open b-lactam ring forms of amoxycillin, 6-APA and cephalothin in addition to ampicillin are shown in Table 2. At the wavelengths used the fluorescence of the penicilloate of amoxycillin was small compared with ampicillin while the fluorescence given by 6-APA and cephalothin, and their open lactam ring forms [penicic acid Fig. 1 Factors affecting the fluorescence of ampicillin and ampicilloate with the modified copper reagent. A, The effect of pH on (2) 100 nmol of ampicilloate and (5) 100 nmol of ampicillin. Copper reagent contained 0.1 cm3 of 2% potassium sodium tartrate, 0.1 cm3 of 1% copper sulfate, 0.5 mg of ascorbic acid, 50 mmol of EDTA and 240 mmol of buffer of the respective pH in a total volume of 6 cm3.B, The effect of acetate concentration at pH 4 on the fluorescence of the reaction. The reaction mixtures contained the same concentration of potassium sodium tartrate and copper sulfate as shown in the solutions used for A. C, The effect of copper ion concentration on fluorescence using 0.5 mol dm23 acetate buffer of pH 4 and the same concentration of potassium sodium tartrate and copper sulfate as shown in the legend to A.Fig. 2 Excitation and emission fluorescence spectra of the compound(s) formed by heating 50 nmol cm23 of ampicilloate (———) or ampicillin (------) with a modified Lowry A reagent at pH 4 containing ascorbate and EDTA. Table 1 Effect of heating in the presence of reducing agents* Relative fluorescence Chemical treatment Ampicillin Ampicilloate No other additions 3.3 1.2 Ascorbate 8 18.4 GSH 2.8 8 EDTA 5.5 62 Ascorbate + EDTA 12 100 GSH + EDTA 3 65 * All solutions contained 50 nmol cm23 of ampicillin or ampicilloate in a final volume of 7 cm3 as described under Experimental.EDTA was added at a concentration of 50 mmol per 7 cm3 whilst ascorbate and other thiol additives were added at a concentration of 0.5 mg cm23. Fig. 3 Relationship between relative fluorescence intensity and either ampicilloate concentration (5) and ampicillin concentration (2) within the concentration range of 5–50 mmol dm23. Duplicate readings were used for the standard curve.The most concentrated solution was used to set the 100% relative fluorescence intensity reference point. Analyst, May 1997, Vol. 122 449(6-APOA), and cephalothinoate, respectively] was negligible. The open b-lactam ring form of amoxycillin (amoxycilloate) fluoresced strongly at longer excitation and emission wavelengths. Ampicilloate and amoxycilloate have a free a-amino group. Ampicillin is metabolised to a piperazine derivative19 which does not contain a free amino group.Benzylpenicillin and phenoxymethylpenicillin do not contain a free a-amino group and did not develop fluorescence under the conditions of this work. Ampicillin was a suitable substrate for the fluorimetric assay of b-lactamase activity. Penicillin acylase activity on ampicillin could be estimated from a loss of fluorescence upon formation of 6-APA, and penicillin acylase activity on ampicilloate as a substrate could also be estimated by the loss of fluorescence of solutions. Detection of b-Lactamase Activity in Micro-organisms b-Lactamase activity in solutions and culture broths of organisms was detected by an intense blue fluorescent circle surrounding the bore-holes in agar plates filled with ampicillin (1.5 mg cm23) and b-lactamase solution or culture broth of the organism (Fig. 4).Enzyme solution and bacterial suspensions alone did not cause fluorescence. Estimation of b-Lactamase Activity Table 3 indicates the results of b-lactamase activity of the commercial concentrated enzyme from Bacillus cereus using ampicillin as a substrate and three methods of determination of enzyme activity. The measured activities were similar using each method.b-Lactamase activity in Aerobacter faecalis ATCC 15246 and Klebsiella pneumoniae against both ampicillin and 6-APA is shown in Fig. 5. Both organisms hydrolysed the b-lactam ring of ampicillin (measured by the formation of ampicilloate) but only K. pneumoniae hydrolysed the b-lactam ring of 6-APA (measured by loss of fluorescence on treatment with fluorescamine at pH 4).b-Lactamase activity of several other organisms (not subjected to induction with methicillin) was estimated by the formation of fluorescent reacting products from ampicillin (Table 4). In most cases the enzyme activity of the intact organisms was low and long incubation times were required contrasting with the enzymes found in Alc. faecalis and K.pneumoniae. The elevated activity of the culture of B. cereus suggested a ready access of the b-lactamase to the substrate, possibly due to an extracellular secretion. The nature of the enzyme was not investigated at this stage but could have been a mixture of type I and type II metallo-b-lactamases.20 Determination of Penicillin Acylase Activity Penicillin acylase activity against benzylpenicillin was estimated by the fluorescamine procedure.10 Results are shown in Fig. 6(A) for uninduced and induced cells of E. coli ATCC 9637 and A. faecalis ATCC 15246. The results obtained after incubation of the penicillin acylase producing E. coli ATCC 9637 with ampicillin, and conversion of residual penicillin to ampicilloate (NaOH), confirm that Table 2 Reaction of penicilloates of penicillins and derivatives* Compound used Relative fluorescence intensity 6-APOA 0.3 6-APA 0 Amoxycilloate 14.5 Amoxycillin 1 Cephalothin 0 Cephalothinoate 0 Ampicillin 11.2 Ampicilloate 100 * Penicilloates were used at a concentration of 100 mg cm23 and were treated with 5 cm3 of a mixture containing 0.1 cm3 of 2% potassium sodium tartrate and 0.1 cm3 of 1% CuSO4·5H2O in 0.5 mol dm23 acetate buffer of pH 4.One cm3 of a mixture of ascorbic acid (0.5 mg cm23) in 50 mmol dm23 EDTA was added, the solution heated for 30 min and then cooled and the the fluorescence was read using an excitation wavelength of 344 nm and an emission wavelength of 450 nm. Fig. 4 Detection of b-lactamase activity from B. subtilis on ampicillin. Procedures are described under materials and methods. T indicates a solution containing 289 mg cm23 of b-lactamase and 1.5 mg cm23 of ampicillin which was placed in the bore-hole. E indicates an enzyme control while A indicates an ampicillin control. Table 3 Comparison of results of b-lactamase activity* Enzyme activity/ Time of sample/ mmol dm23 Method used min ampicilloate formed Micro-iodometric 5 1.98 10 2.98 15 3.63 BCA 5 2.06 10 2.93 15 3.67 Fluorimetric 5 2.21 10 2.44 15 3.79 * Samples of b-lactamase [0.4 cm3 of b-lactamase from B.cereus (Sigma) 14.9 mg cm23 which showed two bands on SDS-PAGE at 22.5 kD and 28 kD indicative of a mixture of type I and type II enzymes20] were mixed with 1.6 cm3 of of 6.25 mmol dm23 ampicillin in 0.05 mol dm23 phosphate buffer of pH 7 and allowed to react at 37 °C. At the times indicated two 0.1 cm3 samples were taken and added to 0.1 cm3 of 8% TCA to inactivate the enzyme.One sample was then diluted to 10 cm3 with 0.5 mol dm23 acetate buffer of pH 4 and treated as described for the fluorimetric assay. The second sample was diluted with 0.05 mol dm23 phosphate buffer of pH 7 and duplicate 1 cm3 samples were added to the BCA reagent at pH 7. This reaction mixture was raised to 37 °C, and 0.1 cm3 of 20 mmol dm23 copper sulfate was added and mixed and after 10 min reaction 0.1 cm3 of 0.2 mol dm23 EDTA was then added. Absorbance was read at 562 nm.Further duplicate 0.6 cm3 samples of this reaction mixture were added to 2.4 cm3 of microiodometric reagent prepared by mixing 1 cm3 of starch–iodine and 1.4 cm3 of 0.1 mol dm23 phosphate buffer of pH 7 and the colour allowed to fade for 5 min prior to reading the absorbance at 620 nm. 450 Analyst, May 1997, Vol. 1226-APA production is accompanied by loss of fluorescence [Fig. 6(B)]. Using this organism the same downward trend was observed in solutions where sodium hydroxide treatment was not used to convert residual ampicillin to ampicilloate.The blactamase activity of this organism is very low7,21 and was barely detectable when ampicilloate estimations were made. The loss of ampicilloate and the formation of 6-APA (measured by using fluorescamine) were quantitatively related [Fig. 6(B)]. Alc. faecalis ATCC 15246, on the other hand, contains an active penicillin acylase as well as an active b-lactamase. This organism showed a similar downward trend in the loss of ampicillin and ampicilloate reacting activity to that observed using E.coli ATCC9637 after solutions had been treated with NaOH. The active b-lactamase resulted in an increase in fluorescence in solutions not treated with NaOH as a significant amount of ampicilloate was formed. In this case only smaller amounts of 6-APA were formed by the action of the penicillin acylase activity on ampicillin when compared with the cells of E.coli. This was thought to reflect the differences in rates of activities of the respective enzymes and the inability of the blactamase in Alc. faecalis to hydrolyse 6-APA (Fig. 5). Penicillin Acylase Activity and Ampicilloate In prolonged incubations using Acl. faecalis ATCC 15246 the calculated concentrations of ampicilloate fell. This effect was attributed to penicillin acylase activity on the ampicilloate formed. Therefore, ampicilloate was incubated with cells of penicillin acylase producing organisms and the activity of the enzyme against this substrate was measured by the loss of fluorescence.The results (Table 5) certainly indicate that E. coli ATCC 9637 and ATCC 11105 and A. faecalis ATCC 15246 did hydrolyse the amide linkage between the 6-APOA and phenylglycine moieties resulting in two products which could not give the fluorescence on heating solutions. In all cases penicillin acylase activity against ampicilloate occurred at a much slower rate than against benzylpenicillin and was still incomplete after 24 h reaction time.Discussion Spectrofluorimetry has many analytical advantages over conventional spectrophotometry. Sensitivity depends on the fluorophore and may be varied by attenuation of the detection system of the instrumentation. Selective wavelengths of excitation and emission allow specificity and there may not be a requirement for sample clean up when assays are used with biological fluids.b-Lactamases of many organisms use ampicillin as a substrate and hydrolysis frequently occurs at similar rates to benzylpenicillin.22 Ampicilloate, the end product of enzyme activity, forms a fluorescent product under specific conditions and in the present work the toxic formaldehyde used by the earlier workers13,15 was omitted from the heating step. Mercuric ion, which was used to catalyse formation of a fluorescent compound from ampicilloate,23 was replaced by copper ion as well as EDTA and fluorescence was generated in the presence of the non-toxic biological reducing agent, ascorbate.The blactamases in Gram negative organisms are considered to be more active against ampicillin than against benzylpenicillin, so ampicillin seems to be an excellent substrate to measure this enzyme’s activity. Ampicillin is acid stable22 so it was possible to inhibit b-lactamase activity by TCA prior to development of fluorescence. This permitted measurement of b-lactamase activity after a fixed time.Acid inactivation overcomes some of the objections for estimation of b-lactamase activity using Fig. 5 Hydrolysis with time of ampicillin by the b-lactamase in K. pneumoniae (2) and Alc. faecalis (5) and 6-APA by the enzyme in K. pneumoniae (8) and Alc. faecalis (-). Ampicillin or 6-APA (6.25 mmol dm23, 1.6 cm3) and cells (0.4 cm3) were mixed and shaken at 37 °C. Samples (0.1 cm3) were taken and added to 0.1 cm3 of cold 8% TCA and diluted to 10 cm3 with 0.5 mmol dm23 acetate buffer of pH 4.Further procedures were as described under Experimental using fluorescamine to estimate the disappearance of 6-APA or heating in the presence of copper– ascorbate and EDTA to estimate the appearance of ampicilloate. Table 4 Effect of b-lactamase in bacteria on ampicillin* Activity/mmol of ampicilloate Bacterium formed per cm3 per 16 h Proteus vulgaris 0.58 Bacillus subtilis 0.62 Bacillus cereus 4.89 Escherichia coli K12 0.7 Erwinia carotovora 0.7 Serratia marscescens 0.32 Pseudomonas aeruginosa 1.42 Pseudomonas fluorescens 0.35 * Cell suspensions of organisms, grown on nutrient agar plates, were prepared as described under Experimental and 0.4 cm3 mixed with 1.6 cm3 of 6.25 mmol dm23 ampicillin in 0.05 mol dm23 phosphate buffer at pH 7.The mixtures then shaken in a reciprocating water bath at 37 °C and 0.1 cm3 samples taken at the time indicated, added to 0.1 cm3 of cold 8% TCA and diluted to 10 cm3 with 0.5 mmol dm23 acetate buffer of pH 4.Fig. 6 A, Effect of penicillin acylase activity in E. coli ATCC 9637 and Alc. faecalis ATCC 15246 against benzylpenicillin. Symbol (2) is E. coli grown on phenylacetic acid medium7 and (5) is the organism grown on nutrient agar. Symbol (8) is Alc. faecalis grown on phenylacetic acid medium and symbol (-) is the organism grown on nutrient agar. Activities were measured by the fluorescamine procedure. B, Indicates penicillin acylase activity against ampicillin.The open symbols refer to E. coli and A. faecalis as described for A and indicate the estimated ampicillin remaining in solution after treatment with NaOH and heating dilute solutions with the copper reagent. The closed symbols refer to the amount of 6-APA formed by E. coli (5) and Alc. faecalis (-) measured by use of fluorescamine while the symbol (D) indicates an estimate of the amount of ampicilloate formed in the solution containing Alc. faecalis by the b-lactamase present in this organism.Results of ampicilloate formation by the activity of the blactamase of E. coli ATCC 9637 on ampicillin were small and are not shown. Analyst, May 1997, Vol. 122 451iodine because some enzymes (or some forms of the enzyme) are not inhibited by the addition of iodine and sometimes the iodine reacts with the tyrosine residues of the enzyme.24,25 The present technique is suitable for use with concentrated enzyme solutions or with those organisms in which b-lactamase is very active (e.g., K.pneumoniae and Alc. faecalis). It can be used in organisms which contain low b-lactamase activity but in these cases long incubation times are required. The major drawback with the analytical technique is that the lesser amount of fluorescence given by the intact ampicillin in the solutions must be accounted for by a calculation in a similar manner to the absorbance given by intact penicillin in the BCA assay.7 Intact ampicillin is converted to a penicillenate derivative after heating at a pH near to 526 (Scheme 1). Penicilloates do not form penicillenates.26 2-Hydroxy-3-phenyl-6-methylpyrazine (Scheme 1) has been identified as the fluorescent compound formed by heating ampicillin with formaldehyde.15 The spectral properties (Fig. 2) of the compound(s) formed in the present work are different from those of other fluorescent derivatives of ampicillin.13 The fluorescent compound(s) and its mechanism of formation is not known but it seems possible that a substituted piperazine ring is formed and is subsequently oxidised to a pyrazine.The ability to form a fluorescent compound is lost when 6-APA is formed from ampicillin by the action of penicillin acylase. Furthermore, the action of penicillin acylase on ampicilloate results in loss of ability to form a fluorescent compound. Hence, by examining the activity of penicillin acylase producing organisms against ampicilloate the production of 6-APOA may be be inferred.The production of 6-APOA from benzylpenicilloate occurs more slowly than the production of 6-APA from benzylpenicillin.21,27 The type 2 penicillin acylases in the organisms used in this work also acted on ampicilloate at a slower rate than benzylpenicillin but it was possible to make an accurate quantitative estimate of the deacylation of this penicilloate after the incubating reaction mixtures for prolonged periods. References 1 Perrett, C.J., Nature (London), 1954, 174, 1012. 2 Novick, R. P., Biochem. J., 1962, 83, 236. 3 Sykes, R. B., and Nordstrom, K., Antimicrob. Agents Chemother., 1972, 1, 94. 4 Batchelor, F. R., Cameron-Wood, J., Chain, E. B., and Rolinson, G. N., Proc. R. Soc., London, B, 1961, 154, 514. 5 O’Callaghan, C. H., Morris, A., Kirby, S. M., and Shingler, A., Antimicrob. Agents Chemother., 1972, 1, 283. 6 Cohenford, M. A., Abraham, J., and Medeiros, A. E., Anal. Biochem., 1988, 168, 252. 7 Baker, W. L., J. Appl. Bacteriol., 1992, 73, 14. 8 Baker, W. L., Antonie van Leeuwenhoek., 1983, 49, 551. Table 5 Effect of penicillin producing organisms on ampicilloate* Penicillin acylase activity against substrate/ mmol dm23 per cm3 cell suspension per time Growth Benzylpenicillin/ Ampicillin/ Ampicilloate/ Source of enzyme medium 4 h 4 h 24 h E. coli ATCC 9637 A 1.27 0.27 0 B 4.6 0.45 0.83 E. coli ATCC 11105 A 1.31 0.31 0.97 B 4.31 1.18 1.25 Alc. faecalis ATCC15246 A 1.69 0.16 2.30 B 1.63 0.26 3.41 * A cell suspension of organisms (0.4 cm3) was obtained after growth on A, nutrient agar or B, solidified medium containing phenylacetic acid and mixed with 1.6 cm3 of 6.25 mmol dm23 benzylpenicillin, ampicillin or sodium ampicilloate, respectively, (prepared by NaOH treatment) in 0.05 mol dm23 of phosphate buffer of pH 7.8 and incubated on a reciprocating water bath at 37 °C. Samples (0.1 cm3) were taken at the times indicated and mixed with 0.1 cm3 of cold 8% TCA and then diluted to 10 cm3 with 0.5 mol dm23 acetate buffer of pH 4. The diluted solution was analysed for 6-APA concentrations using the fluorescamine assay (benzylpenicillin and ampicillin) or disappearance of ampicilloate concentrations according to the procedure described in Experimental. Scheme 1 Proposed reactions of ampicillin and ampicilloate thought to occur under the conditions used in this work. Reaction A is a typical hydrolysis of the b-lactam ring caused by the enzyme b-lactamase. Reaction B occurs only with intact penicillins at a pH of about 5 and especially in the presence of copper.26 Reaction C seems to involve a ring closure. A fluorescent compound was formed when ampicillin and its ampicilloate was heated with formaldehyde13 or ampicilloate was treated with mercuric ion.23 A fluorescent compound with the structural formula shown above was isolated.15 452 Analyst, May 1997, Vol. 1229 Lowry, O. H., Rosebrough, N. J., Farr, A. L., and Randall, R. J., J. Biol. Chem., 1951, 193, 265. 10 Baker, W. L., Antimicrob. Agents Chemother., 1983, 23, 26. 11 Baker, W. L., Aust. J. Biol. Sci., 1984, 37, 257. 12 Baker, W. L., and Havlicek, P. J., J. Gen. Appl. Microbiol., 1985, 31, 107. 13 Jusko, W. J., J. Pharm. Sci., 1971, 60, 728. 14 Baker, W. L., J. Appl. Bacteriol., 1980, 49, 225. 15 Barbhaiya, R. H., Brown, R. C., Payling, D. W., and Turner, P., J. Pharm. Pharmacol., 1978, 30, 224. 16 Taqui Khan, M. M., and Martell, A. E., J. Am. Chem. Soc., 1967, 89, 4176. 17 Jocelyn, P. C., Biochemistry of the SH Group: The Occurrence, Chemical Properties, Metabolism and Biological Function of Thiols and Disulphides, Academic Press, New York, 1972, p. 95. 18 Lakowicz, J., Principles of Fluorescence Spectroscopy, Plenum Press, New York, 1983, p. 259. 19 Haginaka, J., and Wakai, J., J. Pharm. Pharmacol., 1986, 38, 225. 20 Thatcher, D. R., in Methods in Enzymology, ed. Hash, J. H., Academic Press, New York, 1975, vol. 43, pp. 640–664. 21 Pruess, D. L., and Johnson, M. J., J. Bacteriol., 1965, 90, 380. 22 Rolinson, G., and Stevens, S., Br. Med. J., 1961, ii, 191. 23 Miyazaki, K., Ogino, O., and Arita, T., 1974, Chem. Pharm. Bull., 1974, 22, 1910. 24 Citri, N., Biochim. Biophys. Acta, 1958, 27, 277. 25 Csanyi, V., Acta Physiol. Acad. Sci. Hungary, 1961, 18, 261. 26 Smith, J. W. G., de Grey, G. E., and Patel, V. J., Analyst, 1967, 92, 247. 27 Hamilton-Miller, J. M. T., Bacteriol. Rev., 1966, 30, 761. Paper 6/08053G Received November 28, 1996 Accepted February 17, 1997 Analyst, May 1997, Vol. 122 453

ISSN:0003-2654    

DOI:10.1039/a608053g

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Study of a Lanthanide Fluorescence System With a Coupled Reaction Based on Hemin Catalysis Xue-Ying Zhenga, Jian-Zhong Lua, Qing-Zhi Zhua, Jin-Gou Xu*a and Qing-Ge Lib a Research Laboratory of SEDC of Analytical Science for Material and Life Chemistry, Department of Chemistry, Xiamen University, Xiamen 361005, China b Department of Biology, Xiamen University, Xiamen 361005, China The oxidation reactions of three substrates, p-hydroxybenzoic acid (HBA), p-hydroxyphenylacetic acid, and p-hydroxyphenylpropionic acid, with H2O2 catalysed by hemin were studied.It was significant that only the oxidation product of HBA, di-p,pA-hydroxybenzoic acid, forms a long-lived fluorescent complex with Tb3+. The three substrates themselves do not form fluorescent complexes with Tb3+ although they have chemical structural similarities. Based on this, a novel lanthanide fluorescence system with a coupled reaction was developed. The possible reactive mechanism is discussed.HBA was concluded to be well suited to several types of application from hemin determination to bovine serum albumin labeling. Keywords: p-Hydroxybenzoic acid; terbium fluorescence; hemin catalysis The use of chelates of rare earth elements as a label in fluorescence immunoassay offers a great improvement in signal-to-noise ratio over previously used fluorophores and has received much attention in clinical applications.1–4 In timeresolved fluorimetric immunoassays, it is desirable to use a chelate label that can be detected down to the sub-picomolar level.5 Alternatively, multiple labeling strategies can be used in order to achieve sub-picomolar analyte sensitivity.1,6 Direct labeling of a smaller protein with multiple fluorescent chelates usually does not provide sufficient sensitivity (with detection limits in the femtomole range for individual chelate molecules7) and the strategy of heavy multiple labeling of a larger protein with a fluorescent europium chelate which further improves the sensitivity1 can cause unfavorable increases in non-specific binding when the detected molecules are applied to membrane formats.8 Combining the background-rejecting capabilities in time-resolved fluorescence detection of lanthanide chelates with the amplification provided by the use of an enzyme as a label has potential as an alternative non-isotopic detection method for enzymes and enzyme-labeled biomolecules and is appropriate for a variety of assay formats and instrumentation. 8–10 The use of enzymes has some advantages, in particular with regard to specificity, but some enzymes in solution are not stable enough and fresh reagent needs to be prepared daily even when refrigerated; especially the high molecular mass of an enzyme would lead to steric hindrance to antigen–antibody reactions when an enzyme is used as a labeling reagent. Therefore, an alternative is to seek an understanding of enzyme behavior and thus provide a substitute for enzymes.Metalloporphyrins and metal–oxime complexes have been used in such determinations.11,12 The best among them, the MnIII complex, provides 84% of the activity of horseradish peroxidase (HRP). Hemin is a naturally occuring iron–porphine complex, and is an inexpensive product from bovine blood, and can catalyze the luminol chemiluminescence reaction as a substitute for HRP.13,14 In this work, hemin was used as a substitute for HRP and a method involving lanthanide fluorescence immunoassay with a coupled reaction was developed for the determination of hemin and other applications. It was found that the substrate, phydroxybenzoic acid (HBA), which does not form a long-lived fluorescent complex with Tb3+, could be oxidized by H2O2 in the presence of hemin to give a chelator, di-p,pA-hydroxybenzoic acid, which forms a fluorescent complex with Tb3+.This method has a limit of detection of 8.0 3 10210 mol dm23 of hemin and was used in the determination of bovine serum albumin–hemin conjugate.Experimental Instrumentation A Hitachi (Tokyo, Japan) Model 650-10S spectrofluorimeter equipped with a plotter unit and a 1 cm quartz cell was used for recording spectra and making fluorescence measurements. A digital pH meter (Extech, Boston, MA, USA) was also used in the experiment. Reagents All chemicals were of analytical-reagent grade and all aqueous solutions were prepared in distilled, de-ionized water. A hemin stock standard solution was prepared by dissolving 25 mg of hemin (Shanghai Dongfeng Biochemical Reagent Co., Shanghai, China) in 25 cm3 of 0.1 mol dm23 NaOH.This solution is stable for at least 6 months when refrigerated. Working standard solutions were prepared by serially diluting the stock standard solution. HBA solution (4.0 3 1023 mol dm23) was prepared by dissolving an appropriate amount of HBA in water and diluting to the required volume. This solution was stable for at least 1 week if kept in the dark.A Tb3+ stock standard solution (1.0 31022 mol dm23) was prepared as follows: 93 mg of Tb4O7 (99.9%) was dissolved in a minimum amount of 6 m HCl solution, then the mixture was heated with stirring and evaporated to dryness, and finally the white residue was dissolved in 50.0 cm3 of 0.1 mol dm23 HCl. The concentration of Tb3+–EDTA (molar ratio 1 : 1) working standard solution was 1.0 3 1023 mol dm23. A hydrogen peroxide working standard solution was prepared by appropriately diluting a 30% stock standard solution (Shanghai Chemical Reagents Factory, Shanghai, China) with water and standardized by titration with secondary standard KMnO4.Water-soluble carbodiimide was obtained from Sigma (St. Louis, MO, USA). Polyethylene glycol 20 000 was purchased from Shanghai Reagent Co. (Shanghai, China). Sephdex G-50 was supplied by SABC (China). Dimethyl sulfoxide (DMSO) was a product of Shengyang Xinxi Reagent Factory (China).Bovine serum albumin (BSA) was obtained from Shanghai Biochemicals (Shanghai, China). A pH 7.2 TRIS–HCl buffer solution was prepared by mixing 25.0 cm3 of 0.2 mol dm23 TRIS and 45.0 cm3 0.1 mol dm23 HCl. Analyst, May 1997, Vol. 122 (455–458) 455Preparation of Hemin-labeled BSA Hemin was bound to BSA with carbodiimide (EDC) as a coupling agent. A 5 mg amount of hemin was dissolved in 1 cm3 of DMSO and 13 mg of EDC were added. The mixture was incubated for 10 min at 25 °C with magnetic stirring, then 4 cm3 of 6.0 mg cm23 BSA solution (BSA dissolved in 1% m/v sodium chloride solution) was added and the mixture was stirred for 3 h at 25 °C.The pH of the solution was maintained at 4.5 ± 0.5 during the coupling reaction by addition of 0.1 mol dm23 HCl and NaOH. The mixture was centrifuged at 4 °C, then 10 cm3 of 0.1 mol dm23 phosphate buffer solution (pH 7.0) were added to the precipitate to retrieve the soluble product, followed by centrifugation at 4 °C.The whole retrieved product was saturated by adding an appropriate amount of ammonium sulfate and left at 4 °C overnight. The precipitate was dialyzed six times against 0.1 mol dm23 phosphate buffer solution (pH 7.0) at 4 °C (each time for 1 h against 1.0 dm3 of phosphate buffer solution). The end product, a pale brown solution, was concentrated to 2 cm3 with polyethylene glycol 20 000. The final solution was passed through a 50 3 1.5 cm id column packed with Sephdex G-50. The eluent was 0.1 mol dm23 phosphate buffer solution (pH 7.0).The conjugate gave a broad peak at 390 nm and a shoulder at 280 nm. The concentrations of BSA and hemin were determined by measurement of the corresponding absorbances at 280 and 390 nm on the assumption that the value of A1% 280nm for BSA was 6.6, and that the molar absorptivities of hemin at 280 and 390 nm were 3 3 104 and 9.3 3 104 dm3 mol21 cm21, respectively. The approximate molar ratio for the conjugate was 2.Procedure A 0.5 cm3 amount of 4.0 3 1023 mol dm23 p-hydroxybenzoic acid solution, 1.0 cm3 of 0.01 mol dm23 hydrogen peroxide solution, 2.0 cm3 of buffer solution (pH 7.2) and a fixed volume of hemin standard solution were added sequentially to a 10 cm3 calibrated flask. The mixture was allowed to stand for 45 min in the dark. A 1.5 cm3 amount of 1.0 3 1023 mol dm23 Tb3+– EDTA solution and 0.5 ml of 1.0 mol dm23 NaOH were added to the flask and the solution was diluted to the mark with water.The relative fluorescence intensity at 549 nm with excitation at 320 nm was measured. Results and Discussion Spectral Characteristics As shown in Fig. 1, the excitation and emission spectra of the HBA–H2O2 and HBA–H2O2–hemin systems were similar, with the maximum excitation and emission at 285 and 340 nm, respectively. Furthermore, the fluorescence intensities for these two systems were not obviously different. It is significant that the oxidation product of HBA in the presence of hemin can form a highly fluorescent chelate with terbium, which may be detected using the time-resolved or normal fluorescence method, and under the conditions of the fluorescence measurement the substrate, HBA, does not form such a chelate.In addition, the fluorescence signal of the system was also very small in the absence of hemin. In the presence of hemin, the emission spectrum was characteristic of the Tb3+ ion, with the maximum intensity occurring at 549 nm15 and the excitation wavelength red shifted to 320 nm, as shown in Fig. 2. Optimization of Reaction Conditions The concentrations of the various reagents used and the pH range were optimized through investigation of their effects on the coupled reaction. Two buffer solutions, aqueous NaH2PO4– Na2HPO4 and TRIS–HCl, were tested as the reaction medium and the results indicated that the fluorescence intensity in TRIS–HCl buffer solution was higher than that in NaH2PO4– Na2HPO4 buffer solution (see Fig. 3), probably owing to the amine-containing buffer which promotes the hemin catalytic activity,16 and the maximum fluorescence intensity occurred over the pH range 6.5–7.8. A 2.0 cm3 volume of TRIS–HCl Fig. 1 Excitation and emission spectra of the HBA–H2O2–hemin system (A, AA) and the HBA–H2O2 system (B, BA). Hemin, 2 3 1028 mol dm23; H2O2, 1.0 3 1023 mol dm23; HBA, 2.0 3 1024 mol dm23. Fig. 2 Excitation (dashed curve) and emission (solid curve) spectra of the HBA–H2O2–hemin–Tb3+–EDTA system (A, AA) and the HBA–H2O2– Tb3+–EDTA system (B, BA).Hemin, 2 31028 mol dm23; H2O2, 1.0 31023 mol dm23; HBA, 2.0 3 1024 mol dm23; Tb3+–EDTA, 1.5 3 1023 mol dm23. Fig. 3 Effect of buffer solutions on fluorescence intensity: A, TRIS–HCl; and B, NaH2PO4–Na2HPO4. 456 Analyst, May 1997, Vol. 122buffer solution (pH 7.2) was adopted for use in subsequent experiments. The optimum concentrations of H2O2 and HBA were also tested separately and the results showed that the maximum fluorescence intensity was reached when their concentrations were in the ranges 7.5 3 1024–2.0 3 1023 and 1.5 3 1024–3.0 3 1024 mol dm23, respectively, and the fluorescence intensity decreased outside this range.Hence 1.0 31023 mol dm23 H2O2 and 2.0 3 1024 mol dm23 HBA were adopted. The influence of the oxidation reaction time on the fluorescence intensity was investigated. The intensity approximated to the highest value after reaction for 40 min and remained constant for about 10 min.In addition, the reaction times were not obviously different at various hemin concentrations. Therefore, a 45 min reaction time was chosen. The reaction temperature did not affect the reaction, the fluorescence intensity was constant over the temperature range 20–35 °C; room temperature was therefore used. The pH range for terbium–product–EDTA system was optimized. The maximum fluorescence intensity was reached in 0.03–0.08 mol dm23 NaOH solution, so 0.05 mol dm23 NaOH was used in subsequent work.The optimum concentration of Tb3+–EDTA was also tested and the maximum emission occurred when its concentration was in the range 5.0 3 1025–2.0 3 1024 mol dm23. Therefore, 1.5 3 1024 mol dm23 Tb3+–EDTA was adopted. Suggested Reaction Mechanism In order to explain the possible mechanism, two traditionally similar substrates, p-hydroxyphenylacetic acid (HPAA and phydroxyphenylpropionic acid (HPPA), were compared.Surprisingly, no Tb3+ emission occurred in the systems containing HPAA and HPPA, even though they only involved changes in the substituents from –COOH to –CH2COOH and –CH2CH2COOH and had similar oxidation products to HBA.17 This implies that the complexing site with Tb3+ was the –COOH group, not the –OH group in the oxidation product, di-p,pAhydroybenzoic acid. When one or two methylene groups existed between the hydroxyphenyl and carboxyl groups as in HPAA and HPPA, it was not possible for the oxidation product to form a fluorescent complex with Tb3+ because of ring instability.The proposed reaction scheme is outlined in Fig. 4. Calibration Graph for Hemin The calibration graph obtained under the optimum conditions had a linear range of 0–5.0 3 1028 mol dm23 hemin. The linear regression equation was DF = 0.35 + 11.07[hemin], where the concentration unit of hemin is 1028 mol dm23. The detection limit (3s) for hemin was calculated from the SD of the blank (n = 6) as 8.0 3 10210 mol dm23.The correlation coefficient was 0.998 and the relative RSD was 2.9% (n = 6) for the determination of 2.0 3 1028 mol dm23 hemin. Two calibration curves for the determination of hemin, which were obtained by measuring the fluorescence intensity of the substrate at 340 nm and that of the product with Tb3+–EDTA at 549 nm, respectively, are shown in Fig. 5. It can be seen that the fluorescence intensity of the substrate at 340 nm cannot be used for the determination.Determination of BSA–Hemin Conjugate A simple label assay was performed to demonstrate the feasibility of utilizing the hemin-catalyzed oxidation reaction coupled with the terbium fluorescence reaction system as a means of detecting the hemin–BSA conjugate. The linear range for the determination of the hemin–BSA conjugate was 0–1.0 3 1027 mol dm23 and the detection limit was 2.0 3 10210 mol dm23 (calculated as above, n = 6).The regression equation was DF = 20.38 + 33.76 [hemin–BSA], where the concentration unit of hemin–BSA is in 1028 mol dm23, and the correlation coefficient was 0.9995. The RSD was 2.8% (n = 6) for the determination of 1.0 3 1028 mol dm23 hemin–BSA complex. The authors thank the Postdoctoral Science Fund of China for its support of this research. References 1 Bailey, M. P., Rocks, B. F., and Riley, C., Analyst, 1984, 109, 1449. 2 Saavedra, S. S., and Picozza, E. G., Analyst, 1989, 114, 835. 3 Morton, R. C., and Diamandis, E. P., Anal. Chem., 1990, 62, 1841. 4 Diamandis, E. P., and Christopoulos, T. K., Anal. Chem., 1990, 62, 1149A. 5 Hemmila, J., Dakuku, S., Mukkala, V. M., Siitari, H., and Lovgren, T., Anal. Biochem., 1984, 137, 335. 6 Diamandis, E. P., Clin. Chem., 1991, 37, 1486. 7 Evangelista, R. A., Pollak, A., Allore, B., Templeton, E. F., Morton, R. C., and Diamandis, E. P., Clin. Biochem., 1988, 21, 173. 8 Evangelista, R. A., Pollak, A., and Templeton, E.F. G., Anal. Biochem., 1991, 197, 213. 9 Christopoulos, T. K., and Diamandis, E. P., Anal. Chem., 1992, 64, 342. Fig. 4 Proposed mechanism of terbium fluorescence with coupled reaction. Fig. 5 Dependence of fluorescence intensity on the hemin concentration: A, for the substrate (at 340 nm); and B, for the oxidation product with Tb3+– EDTA (at 549 nm). H2O2, 1.0 3 1023 mol dm23; HBA, 2.0 3 1024 mol dm23; Tb3+–EDTA, 1.5 3 1023 mol dm23. Analyst, May 1997, Vol. 122 45710 Diamandis, E.P., Analyst, 1992, 117, 1879. 12 Ikariyama, Y., Suzuki, S., and Alzawa, M., Anal. Chem., 1982, 54, 1126. 12 Ci, Y.-X., Tie, J.-K., Yao, F.-J., Liu, Z.-L., Lin, S., and Zheng, W.-Q., Anal. Chim. Acta., 1993, 227, 67. 13 Yoshizumi, K., Aoki, K., Nouchi, I., Okita, T., Kobayashi, T., Kamakura, S., and Tajima, M., Atmos. Environ., 1984, 18, 395. 14 Ikariyama, Y., Suzuki, S., and Aizawa, M., Anal. Chim. Acta, 1984, 156, 245. 15 Sinha, A. P. B., Spectrosc. Inorg. Chem., 1971, 2, 255. 16 Zhang, G., and Dasgupta, P. K., Anal. Chem., 1992, 64, 517. 17 Genfa, Z., Dasgupta, P. K., Edgemond, W. S., and Marx, J. N., Anal. Chim. Acta, 1991, 243, 207. Paper 6/06560K Received Sepetember 24, 1996 Accepted January 6, 1997 458 Analyst, May 1997, Vol. 122 Study of a Lanthanide Fluorescence System With a Coupled Reaction Based on Hemin Catalysis Xue-Ying Zhenga, Jian-Zhong Lua, Qing-Zhi Zhua, Jin-Gou Xu*a and Qing-Ge Lib a Research Laboratory of SEDC of Analytical Science for Material and Life Chemistry, Department of Chemistry, Xiamen University, Xiamen 361005, China b Department of Biology, Xiamen University, Xiamen 361005, China The oxidation reactions of three substrates, p-hydroxybenzoic acid (HBA), p-hydroxyphenylacetic acid, and p-hydroxyphenylpropionic acid, with H2O2 catalysed by hemin were studied.It was significant that only the oxidation product of HBA, di-p,pA-hydroxybenzoic acid, forms a long-lived fluorescent complex with Tb3+.The three substrates themselves do not form fluorescent complexes with Tb3+ although they have chemical structural similarities. Based on this, a novel lanthanide fluorescence system with a coupled reaction was developed. The possible reactive mechanism is discussed. HBA was concluded to be well suited to several types of application from hemin determination to bovine serum albumin labeling. Keywords: p-Hydroxybenzoic acid; terbium fluorescence; hemin catalysis The use of chelates of rare earth elements as a label in fluorescence immunoassay offers a great improvement in signal-to-noise ratio over previously used fluorophores and has received much attention in clinical applications.1–4 In timeresolved fluorimetric immunoassays, it is desirable to use a chelate label that can be detected down to the sub-picomolar level.5 Alternatively, multiple labeling strategies can be used in order to achieve sub-picomolar analyte sensitivity.1,6 Direct labeling of a smaller protein with multiple fluorescent chelates usually does not provide sufficient sensitivity (with detection limits in the femtomole range for individual chelate molecules7) and the strategy of heavy multiple labeling of a larger protein with a fluorescent europium chelate which further improves the sensitivity1 can cause unfavorable increases in non-specific binding when the detected molecules are applied to membrane formats.8 Combining the background-rejecting capabilities in time-resolved fluorescence detection of lanthanide chelates with the amplification provided by the use of an enzyme as a label has potential as an alternative non-isotopic detection method for enzymes and enzyme-labeled biomolecules and is appropriate for a variety of assay formats and instrumentation. 8–10 The use of enzymes has some advantages, in particular with regard to specificity, but some enzymes in solution are not stable enough and fresh reagent needs to be prepared daily even when refrigerated; especially the high molecular mass of an enzyme would lead to steric hindrance to antigen–antibody reactions when an enzyme is used as a labeling reagent.Therefore, an alternative is to seek an understanding of enzyme behavior and thus provide a substitute for enzymes. Metalloporphyrins and metal–oxime complexes have been used in such determinations.11,12 The best among them, the MnIII complex, provides 84% of the activity of horseradish peroxidase (HRP).Hemin is a naturally occuring iron–porphine complex, and is an inexpensive product from bovine blood, and can catalyze the luminol chemiluminescence reaction as a substitute for HRP.13,14 In this work, hemin was used as a substitute for HRP and a method involving lanthanide fluorescence immunoassay with a coupled reaction was developed for the determination of hemin and other applications. It was found that the substrate, phydroxybenzoic acid (HBA), which does not form a long-lived fluorescent complex with Tb3+, could be oxidized by H2O2 in the presence of hemin to give a chelator, di-p,pA-hydroxybenzoic acid, which forms a fluorescent complex with Tb3+.This method has a limit of detection of 8.0 3 10210 mol dm23 of hemin and was used in the determination of bovine serum albumin–hemin conjugate. Experimental Instrumentation A Hitachi (Tokyo, Japan) Model 650-10S spectrofluorimeter equipped with a plotter unit and a 1 cm quartz cell was used for recording spectra and making fluorescence measurements.A digital pH meter (Extech, Boston, MA, USA) was also used in the experiment. Reagents All chemicals were of analytical-reagent grade and all aqueous solutions were prepared in distilled, de-ionized water. A hemin stock standard solution was prepared by dissolving 25 mg of hemin (Shanghai Dongfeng Biochemical Reagent Co., Shanghai, China) in 25 cm3 of 0.1 mol dm23 NaOH. This solution is stable for at least 6 months when refrigerated.Working standard solutions were prepared by serially diluting the stock standard solution. HBA solution (4.0 3 1023 mol dm23) was prepared by dissolving an appropriate amount of HBA in water and diluting to the required volume. This solution was stable for at least 1 week if kept in the dark. A Tb3+ stock standard solution (1.0 31022 mol dm23) was prepared as follows: 93 mg of Tb4O7 (99.9%) was dissolved in a minimum amount of 6 m HCl solution, then the mixture was heated with stirring and evaporated to dryness, and finally the white residue was dissolved in 50.0 cm3 of 0.1 mol dm23 HCl.The concentration of Tb3+–EDTA (molar ratio 1 : 1) working standard solution was 1.0 3 1023 mol dm23. A hydrogen peroxide working standard solution was prepared by appropriately diluting a 30% stock standard solution (Shanghai Chemical Reagents Factory, Shanghai, China) with water and standardized by titration with secondary standard KMnO4.Water-soluble carbodiimide was obtained from Sigma (St. Louis, MO, USA). Polyethylene glycol 20 000 was purchased from Shanghai Reagent Co. (Shanghai, China). Sephdex G-50 was supplied by SABC (China). Dimethyl sulfoxide (DMSO) was a product of Shengyang Xinxi Reagent Factory (China). Bovine serum albumin (BSA) was obtained from Shanghai Biochemicals (Shanghai, China). A pH 7.2 TRIS–HCl buffer solution was prepared by mixing 25.0 cm3 of 0.2 mol dm23 TRIS and 45.0 cm3 0.1 mol dm23 HCl.Analyst, May 1997, Vol. 122 (455–458) 455Preparation of Hemin-labeled BSA Hemin was bound to BSA with carbodiimide (EDC) as a coupling agent. A 5 mg amount of hemin was dissolved in 1 cm3 of DMSO and 13 mg of EDC were added. The mixture was incubated for 10 min at 25 °C with magnetic stirring, then 4 cm3 of 6.0 mg cm23 BSA solution (BSA dissolved in 1% m/v sodium chloride solution) was added and the mixture was stirred for 3 h at 25 °C. The pH of the solution was maintained at 4.5 ± 0.5 during the coupling reaction by addition of 0.1 mol dm23 HCl and NaOH.The mixture was centrifuged at 4 °C, then 10 cm3 of 0.1 mol dm23 phosphate buffer solution (pH 7.0) were added to the precipitate to retrieve the soluble product, followed by centrifugation at 4 °C. The whole retrieved product was saturated by adding an appropriate amount of ammonium sulfate and left at 4 °C overnight. The precipitate was dialyzed six times against 0.1 mol dm23 phosphate buffer solution (pH 7.0) at 4 °C (each time for 1 h against 1.0 dm3 of phosphate buffer solution).The end product, a pale brown solution, was concentrated to 2 cm3 with polyethylene glycol 20 000. The final solution was passed through a 50 3 1.5 cm id column packed with Sephdex G-50. The eluent was 0.1 mol dm23 phosphate buffer solution (pH 7.0). The conjugate gave a broad peak at 390 nm and a shoulder at 280 nm. The concentrations of BSA and hemin were determined by measurement of the corresponding absorbances at 280 and 390 nm on the assumption that the value of A1% 280nm for BSA was 6.6, and that the molar absorptivities of hemin at 280 and 390 nm were 3 3 104 and 9.3 3 104 dm3 mol21 cm21, respectively.The approximate molar ratio for the conjugate was 2. Procedure A 0.5 cm3 amount of 4.0 3 1023 mol dm23 p-hydroxybenzoic acid solution, 1.0 cm3 of 0.01 mol dm23 hydrogen peroxide solution, 2.0 cm3 of buffer solution (pH 7.2) and a fixed volume of hemin standard solution were added sequentially to a 10 cm3 calibrated flask.The mixture was allowed to stand for 45 min in the dark. A 1.5 cm3 amount of 1.0 3 1023 mol dm23 Tb3+– EDTA solution and 0.5 ml of 1.0 mol dm23 NaOH were added to the flask and the solution was diluted to the mark with water. The relative fluorescence intensity at 549 nm with excitation at 320 nm was measured. Results and Discussion Spectral Characteristics As shown in Fig. 1, the excitation and emission spectra of the HBA–H2O2 and HBA–H2O2–hemin systems were similar, with the maximum excitation and emission at 285 and 340 nm, respectively. Furthermore, the fluorescence intensities for these two systems were not obviously different. It is significant that the oxidation product of HBA in the presence of hemin can form a highly fluorescent chelate with terbium, which may be detected using the time-resolved or normal fluorescence method, and under the conditions of the fluorescence measurement the substrate, HBA, does not form such a chelate.In addition, the fluorescence signal of the system was also very small in the absence of hemin. In the presence of hemin, the emission spectrum was characteristic of the Tb3+ ion, with the maximum intensity occurring at 549 nm15 and the excitation wavelength red shifted to 320 nm, as shown in Fig. 2. Optimization of Reaction Conditions The concentrations of the various reagents used and the pH range were optimized through investigation of their effects on the coupled reaction.Two buffer solutions, aqueous NaH2PO4– Na2HPO4 and TRIS–HCl, were tested as the reaction medium and the results indicated that the fluorescence intensity in TRIS–HCl buffer solution was higher than that in NaH2PO4– Na2HPO4 buffer solution (see Fig. 3), probably owing to the amine-containing buffer which promotes the hemin catalytic activity,16 and the maximum fluorescence intensity occurred over the pH range 6.5–7.8.A 2.0 cm3 volume of TRIS–HCl Fig. 1 Excitation and emission spectra of the HBA–H2O2–hemin system (A, AA) and the HBA–H2O2 system (B, BA). Hemin, 2 3 1028 mol dm23; H2O2, 1.0 3 1023 mol dm23; HBA, 2.0 3 1024 mol dm23. Fig. 2 Excitation (dashed curve) and emission (solid curve) spectra of the HBA–H2O2–hemin–Tb3+–EDTA system (A, AA) and the HBA–H2O2– Tb3+–EDTA system (B, BA). Hemin, 2 31028 mol dm23; H2O2, 1.0 31023 mol dm23; HBA, 2.0 3 1024 mol dm23; Tb3+–EDTA, 1.5 3 1023 mol dm23.Fig. 3 Effect of buffer solutions on fluorescence intensity: A, TRIS–HCl; and B, NaH2PO4–Na2HPO4. 456 Analyst, May 1997, Vol. 122buffer solution (pH 7.2) was adopted for use in subsequent experiments. The optimum concentrations of H2O2 and HBA were also tested separately and the results showed that the maximum fluorescence intensity was reached when their concentrations were in the ranges 7.5 3 1024–2.0 3 1023 and 1.5 3 1024–3.0 3 1024 mol dm23, respectively, and the fluorescence intensity decreased outside this range.Hence 1.0 31023 mol dm23 H2O2 and 2.0 3 1024 mol dm23 HBA were adopted. The influence of the oxidation reaction time on the fluorescence intensity was investigated. The intensity approximated to the highest value after reaction for 40 min and remained constant for about 10 min. In addition, the reaction times were not obviously different at various hemin concentrations. Therefore, a 45 min reaction time was chosen.The reaction temperature did not affect the reaction, the fluorescence intensity was constant over the temperature range 20–35 °C; room temperature was therefore used. The pH range for terbium–product–EDTA system was optimized. The maximum fluorescence intensity was reached in 0.03–0.08 mol dm23 NaOH solution, so 0.05 mol dm23 NaOH was used in subsequent work. The optimum concentration of Tb3+–EDTA was also tested and the maximum emission occurred when its concentration was in the range 5.0 3 1025–2.0 3 1024 mol dm23.Therefore, 1.5 3 1024 mol dm23 Tb3+–EDTA was adopted. Suggested Reaction Mechanism In order to explain the possible mechanism, two traditionally similar substrates, p-hydroxyphenylacetic acid (HPAA and phydroxyphenylpropionic acid (HPPA), were compared. Surprisingly, no Tb3+ emission occurred in the systems containing HPAA and HPPA, even though they only involved changes in the substituents from –COOH to –CH2COOH and –CH2CH2COOH and had similar oxidation products to HBA.17 This implies that the complexing site with Tb3+ was the –COOH group, not the –OH group in the oxidation product, di-p,pAhydroybenzoic acid.When one or two methylene groups existed between the hydroxyphenyl and carboxyl groups as in HPAA and HPPA, it was not possible for the oxidation product to form a fluorescent complex with Tb3+ because of ring instability. The proposed reaction scheme is outlined in Fig. 4. Calibration Graph for Hemin The calibration graph obtained under the optimum conditions had a linear range of 0–5.0 3 1028 mol dm23 hemin. The linear regression equation was DF = 0.35 + 11.07[hemin], where the concentration unit of hemin is 1028 mol dm23. The detection limit (3s) for hemin was calculated from the SD of the blank (n = 6) as 8.0 3 10210 mol dm23. The correlation coefficient was 0.998 and the relative RSD was 2.9% (n = 6) for the determination of 2.0 3 1028 mol dm23 hemin.Two calibration curves for the determination of hemin, which were obtained by measuring the fluorescence intensity of the substrate at 340 nm and that of the product with Tb3+–EDTA at 549 nm, respectively, are shown in Fig. 5. It can be seen that the fluorescence intensity of the substrate at 340 nm cannot be used for the determination. Determination of BSA–Hemin Conjugate A simple label assay was performed to demonstrate the feasibility of utilizing the hemin-catalyzed oxidation reaction coupled with the terbium fluorescence reaction system as a means of detecting the hemin–BSA conjugate.The linear range for the determination of the hemin–BSA conjugate was 0–1.0 3 1027 mol dm23 and the detection limit was 2.0 3 10210 mol dm23 (calculated as above, n = 6). The regression equation was DF = 20.38 + 33.76 [hemin–BSA], where the concentration unit of hemin–BSA is in 1028 mol dm23, and the correlation coefficient was 0.9995. The RSD was 2.8% (n = 6) for the determination of 1.0 3 1028 mol dm23 hemin–BSA complex.The authors thank the Postdoctoral Science Fund of China for its support of this research. References 1 Bailey, M. P., Rocks, B. F., and Riley, C., Analyst, 1984, 109, 1449. 2 Saavedra, S. S., and Picozza, E. G., Analyst, 1989, 114, 835. 3 Morton, R. C., and Diamandis, E. P., Anal. Chem., 1990, 62, 1841. 4 Diamandis, E. P., and Christopoulos, T. K., Anal. Chem., 1990, 62, 1149A. 5 Hemmila, J., Dakuku, S., Mukkala, V. M., Siitari, H., and Lovgren, T., Anal. Biochem., 1984, 137, 335. 6 Diamandis, E. P., Clin. Chem., 1991, 37, 1486. 7 Evangelista, R. A., Pollak, A., Allore, B., Templeton, E. F., Morton, R. C., and Diamandis, E. P., Clin. Biochem., 1988, 21, 173. 8 Evangelista, R. A., Pollak, A., and Templeton, E. F. G., Anal. Biochem., 1991, 197, 213. 9 Christopoulos, T. K., and Diamandis, E. P., Anal. Chem., 1992, 64, 342. Fig. 4 Proposed mechanism of terbium fluorescence with coupled reaction. Fig. 5 Dependence of fluorescence intensity on the hemin concentration: A, for the substrate (at 340 nm); and B, for the oxidation product with Tb3+– EDTA (at 549 nm). H2O2, 1.0 3 1023 mol dm23; HBA, 2.0 3 1024 mol dm23; Tb3+–EDTA, 1.5 3 1023 mol dm23. Analyst, May 1997, Vol. 122 45710 Diamandis, E. P., Analyst, 1992, 117, 1879. 12 Ikariyama, Y., Suzuki, S., and Alzawa, M., Anal. Chem., 1982, 54, 1126. 12 Ci, Y.-X., Tie, J.-K., Yao, F.-J., Liu, Z.-L., Lin, S., and Zheng, W.-Q., Anal. Chim. Acta., 1993, 227, 67. 13 Yoshizumi, K., Aoki, K., Nouchi, I., Okita, T., Kobayashi, T., Kamakura, S., and Tajima, M., Atmos. Environ., 1984, 18, 395. 14 Ikariyama, Y., Suzuki, S., and Aizawa, M., Anal. Chim. Acta, 1984, 156, 245. 15 Sinha, A. P. B., Spectrosc. Inorg. Chem., 1971, 2, 255. 16 Zhang, G., and Dasgupta, P. K., Anal. Chem., 1992, 64, 517. 17 Genfa, Z., Dasgupta, P. K., Edgemond, W. S., and Marx, J. N., Anal. Chim. Acta, 1991, 243, 207. Paper 6/06560K Received Sepetember 24, 1996 Accepted January 6, 1997 458 Analyst, May 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a606560k

出版商:RSC    

年代:1997

数据来源: RSC