摘要:

Critical Review Near-infrared spectroscopy in the pharmaceutical industry M. Blanco*, J. Coello, H. Iturriaga, S. Maspoch and C. de la Pezuela Departamento de Química, Unidad de Química Analítica, Universidad Autónoma de Barcelona, E-08193 Bellaterra, Barcelona, Spain Summary of contents Introduction Background and literature sources Specialized journals Internet addresses Previous reviews Fundamentals of the technique Principles of NIR spectroscopy NIR diffuse reflectance spectroscopy Operational procedures in the NIR Mathematical processing of signals Qualitative analysis Identification and qualification of raw materials and pharmaceutical preparations Determination of homogeneity Polymorphism and optical isomers Quantitative analysis Sample selection Multivariate calibration methods Determination of physical parameters Determination of moisture content Determination of active compounds and excipients Calibration transfer Miscellaneous applications Conclusions References Keywords: Near-infrared spectroscopy; pharmaceutical preparations; pharmaceutical analysis; quality control; qualitative analysis; multivariate calibration; spectral pre-treatment; review Introduction Although the near-infrared (NIR) was the first non-visible region discovered in the absorption spectrum (by Herschel in 1800)1–3 analytical chemists made little use of it until the 1950s.A review published in 19604 that reported a comprehensive compilation of band assignments to different functional groups included only about 40 references to analytical applications of the NIR region; this prompted Wetzel’s comment5 that NIR spectroscopy was ‘a sleeper among spectroscopic techniques’.However, analytical applications of the NIR technique have grown dramatically in number since the 1960s, so much so that a review of the topic was published in 1994 under the suggestive title ‘Near-Infrared Spectroscopy. The Giant is Running Strong’.6 Over the last 25 years, NIR spectroscopy has been increasingly used as an analytical tool, particularly by the food and agricultural industries, but also, to some extent, by the textile and polymer industries.Reported applications have been the subject of a number of reviews and books.7–12 Applications to process control have also been developed over this period.13 Although most of the ensuing analytical methods use some chemometric technique to correlate spectral data with physical or chemical properties of the samples, there are also recent uses of NIR for identifying impurities and elucidating structures from band assignments in addition to the earliest reported applications.14–16 The growing interest aroused by NIR spectroscopy in the industrial sector is probably a direct result of its two major advantages as an analytical tool for quality control.Thus, the low molar absorptivity of NIR bands permits operation in the reflectance mode and hence recording of spectra of solid samples with minimal or no pre-treatment, thereby substantially increasing the throughput.Also, the dual dependence of the analytical signal on the physical and chemical nature of the sample facilitates both its identification and the determination of physical and chemical parameters. Notwithstanding these advantages, the pharmaceutical industry has been slow to adopt the NIR technique as it lacks the ability of mid-infrared (MIR) spectroscopy to identify samples by the mere inspection of spectra. In addition, quantitative NIR analyses involve calibration by sophisticated mathematical techniques that have reached extensive use only recently with the advent of microcomputing and chemometrics.Despite the initial reluctance, NIR spectroscopy has aroused great interest in the last few years as a result of both instrumental breakthroughs (e.g., improved detectors, the development of fast-scan and Fourier transform instruments to replace filter instruments, the widespread use of fibre-optic probes and instruments for recording spectra of individual tablets, which minimize or avoid sample pre-treatment) and the incorporation into equipment-bundled software of mathematical procedures for processing NIR spectra—a review of chemometric methods for NIR spectroscopy has been published by Mark.17 In addition, the dependence of the NIR signal on Marcelo Blanco is Professor of Analytical Chemistry at Universitat Autonoma of Barcelona and Head of a group working on applied chemometrics.His major research interest is focused on molecular analytical spectrometry, including UV/VIS, FTIR, NIR, circular dichroism and molecular fluorescence. Topics under study in this group are the application of these techniques to the development of rapid methods for analytical control in several industrial fields (textiles, leather, electroplating, etc.) with special attention to pharmaceutical analysis by NIR applying multivariate calibration techniques.Other topics are the use of multivariate techniques to multicomponent resolution by kinetic measurements and chiral determination by capillary electrophoresis and circular dichroism. Analyst, August 1998, Vol. 123 (135R–150R) 135Rboth the chemical composition and some physical properties of the sample, which was formerly considered a hindrance, permits not only the identification of compounds but also the total characterization of samples and the determination of nonchemical parameters with precision comparable to that of conventional methodologies, all due to powerful mathematical treatments for complex signals.This paper is intended to provide readers with an overview of NIR uses by the pharmaceutical industry. To this end, the authors have divided references published up to 1996 into six different categories. The first lists specialized journals devoted to the NIR technique, some Internet addresses that can be used as literature sources and brief descriptions of other, previously published reviews.The second part is a brief introduction to the theoretical fundamentals of the technique and the third discusses the mathematical treatments used to process recorded signals prior to qualitative or quantitative analysis. The fourth and fifth parts include references to qualitative (identification of raw materials and end products, homogeneity studies, polymorphism and optical isomers) and quantitative applications (determination of physical parameters, water contents, active compounds and excipients); references are briefly commented on and practical aspects to be considered in addressing each type of analysis are discussed.The last part describes some applications that are not strictly pharmaceutical but might be of interest to the pharmaceutical industry. Background and literature sources Specialized journals Although virtually every analytical chemistry journal has published NIR applications, the growing interest they have aroused and the widespread use they have reached in recent years have promoted the appearance of two specialized journals, viz., NIR News and the Journal of Near Infrared Spectroscopy.NIR News is the official newsletter of the International Committee for Near Infrared Spectroscopy. It contains up-todate information about meetings, conferences, etc., on NIR spectroscopy and various regular features that contain short articles aimed at disseminating general aspects of the technique, a list of published NIR papers and an advice page that answers general and specific questions posed by users of the technique.The Journal of Near Infrared Spectroscopy, which started early in 1993, is a conventional-format publication consisting of papers, short communications and reviews of theoretical aspects of the technique and its uses in the industrial, agricultural, nutritional, polymer, textile and pharmaceutical fields, among others.Internet addresses The dramatic expansion of the World Wide Web (Internet) in the last few years has made a vast amount of information available in electronic format to the public. The acronym ‘NIR’ stands for ‘Networked Information Retrieval’ in Web jargon, so using it as a search string in any of the popular Web searching engines is bound to generate non-spectroscopic ‘hits’ (references to addresses containing information relevant to the target topic).Since late 1995, NIR News has included a special section devoted to Internet resources for NIR and related topics.18 In a popularizing spirit, its issues list newsgroup and mail-list addresses [also called ‘uniform resource locators’ (URLs)] for bodies and societies concerned with NIR spectroscopy, and describe the procedure for joining them. Tables 1 and 2 list newsgroups and mail-lists, respectively, with an interest in the use of NIR spectroscopy in various fields.19,20 Also worthy of mention here are electronic magazines (ezines in Web jargon), which are partly or fully published in electronic format.Some are electronic versions of conventional journals while others have no ‘paper parent.’. Thus, the magazine e-JNIRS is the electronic version of the Journal of Near Infrared Spectroscopy, in Adobe Acrobat format. Table 3 gives the URLs for some e-zines that publish papers on NIR spectroscopy and its uses.Some e-zines are published and managed by non-professional groups. Such is the case with Wave of the Future, which includes articles on the pharmaceutical uses of NIR. Comments, criticisms and suggestions are welcome, so readers act as true ‘on-line reviewers’; readers’ comments and authors’ replies are all recorded and accessible to all. The e-zine can be reached at http://kerouac.pharm.uky.edu/ ARSG/wave/wavehp.html. At http://kerouac.pharm.uky.edu/ARSG/wave/cnirs/Ir_spechtm, Kramer and Lodder have compiled URLs and resources related to MIR and NIR spectroscopy that include laboratories, departments, researchers’ personal pages, instrument manufacturers, courses, journals, societies, etc.The same authors used Wave of the Future to publish a paper on MIR and NIR resources on the Internet, the printed version of which can be found in refs. 21 and 22. Many NIR spectroscopic societies have their own addresses. Table 4 gives some of these.Finally, of special interest in relation to the pharmaceutical uses of NIR is Derksen’s home page (http://leden.tref.nl/mderksen). This is an excellent repository of news and pharmaceutical uses of NIR that contains many links to papers about NIR and pharmaceuticals, as well as a list of NIR instruments of service to the pharmaceutical industry. Table 2 Mail lists concerned with NIR spectroscopy List Administrator’s URL Instruction (Analysis Group) analysis-l maiser@fs4.in.umist.ac.uk analysis-l name (American Society for Testing and Materials) astmsrch listserv@uga.cc.uga.edu astmsrch (International Chemometrics Society) ics-l listserv@umdd.umd.edu ics-l name (Process Group) process-l maiser@fs4.in.umist.ac.uk process-l (Society for Applied Spectroscopy) applspec listserv@uga.cc.uga.edu applspec (Spectroscopy Group of the UK’s Institute of Physics) spectroscopy-group mailbase@mailbase.ac.uk spectroscopy-group (Statistics and Statistical Discussion) stat-l listserv@vm1.mcgill.ca stat-l Table 1 Internet newsgroups concerned with the uses of NIR spectroscopy in various industrial fields news://comp.ai.neural-nets news://sci.data.formats news://comp.soft-sys.matlab news://sci.environment news://comp.soft-sys.sas news://sci.optics.fiber news://comp.sys.mac.scitech news://sci.polymers news://misc.industry.quality news://sci.stat.math news://sci.agriculture news://sci.techniques.misc news://sci.bio.food-science news://sci.techniques.spectroscopy news://sci.chem.analytical 136R Analyst, August 1998, Vol. 123Previous reviews There are a multitude of reviews on NIR spectroscopy; few, however, deal exclusively with its pharmaceutical applications. The Handbook of Near-Infrared Analysis23 and the book Advances in Near-Infrared Measurements24 each include one chapter devoted exclusively to NIR analyses of pharmaceuticals. Below are briefly described literature reviews published in analytical chemistry or pharmaceutical journals containing more or less extensive sections on the uses of NIR by the pharmaceutical industry. In 1986, Stark et al.25 published a comprehensive review of the NIR technique as a tool for qualitative and quantitative analysis.The four parts of the work provide readers with information about the fundamentals of the technique, the equipment it uses, chemometric treatments for processing NIR signals and performing quantitative analyses, its advantages and a list of applications in a variety of industrial fields (pharmaceutical, textile, nutritional, biomedical, chemical).Pharmaceutical applications of NIR spectroscopy were first reviewed by Ciurczak in 1987.26 The review cited about 50 papers published up to 1987 but made no mention of the technique’s fundamentals or associated equipment. The 1991 review by Drennen et al.27 encompassed several previous reviews published up to 1988 and included an introduction to NIR equipment and its recent developments; it also described the theoretical foundation of qualitative and quantitative chemometric methods used in connection with NIR applications.In 1992, reviews by Martin28 and by Ciurczak and Drennen29 were published. Of special interest in Martin’s extensive work is the section dealing with quantitative analysis, with subsections devoted to sample selection, mathematical signal treatments, calibration techniques and their selection, and calibration transfer. Also worthy of note is the applications section, with references to NIR uses in food, cosmetic, polymer, textile and pharmaceutical analyses.In contrast, the review of Ciurczak and Drennen only covers, and rather briefly, 31 references to qualitative and quantitative applications of NIR spectroscopy in the pharmaceutical industry. In 1993, Corti et al.30 reviewed NIR diffuse reflectance spectroscopic uses in the pharmaceutical and biomedical fields reported up to 1992.In addition to about 70 references to specific applications, the review included a brief description of the more widely used chemometric methods for qualitative and quantitative analysis by NIR spectroscopy. The reviews by Workman13 and by MacDonald and Prebble31 were also published in 1993. The former included about 145 applications of NIR spectroscopy in various industrial fields, described in a chronological sequence and classified into different categories (in-line, in situ, on-line, noninvasive, remote and rapid NIR analyses). The latter review was defined by its authors as a ‘non-comprehensive overview of near-infrared reflectance analysis in the pharmaceutical industry’ and illustrates the technical potential of this technique in this industrial field. Recent advances in NIR spectroscopic equipment were compiled by McClure in 1994.6 He described commercially available NIR spectrophotometers from various manufacturers.Other reviews of NIR pharmaceutical uses were published by Morisseau and Rhodes,32 and by Kirsch and Drennen,33 both in 1995.The former authors reviewed far fewer references than did Corti et al.;30 however, they included a short, albeit interesting, introduction to NIR equipment and its manufacturers. Kirsch and Drennen placed special emphasis on direct analyses of solid pharmaceutical preparations (intact dosage forms) and included a brief introduction to chemometric methods used for qualitative NIR analysis and about 30 references to qualitative and quantitative studies in the pharmaceutical field. Fundamentals of the technique Properly using an instrumental technique for a specific analytical purpose entails knowing what is to be measured and in what way, since extracting the full informative potential from an instrument requires a sound knowledge of the physicochemical theories on which its measurements rely and of the instrumental principles involved.In order to introduce the readers to the technique, this section is intended to provide a basic knowledge of the theoretical foundation of NIR spectroscopy and diffuse reflectance measurements in the NIR region.Table 3 Uniform resource locators (URLs) for selected journals publishing articles about NIR spectroscopy Journal URL Analyst http://www.rsc.org/analyst Analytica Chimica Acta http://www.elsevier.com:80/locate/issn/03654877 Analytical Chemistry http://pubs.acs.org/journals/ancham/index.html Applied Spectroscopy http://esther.la.asu.edu/sas Applied Spectroscopy Newsletter http://esther.la.asu.edu/sas/epstein Chemometrics and Intelligent Laboratory Systems http://www.elsevier.com:80/locate/issn/01697439 e-JNIRS http://www.nirpublications.com/electron.html Food Testing and Analysis http://www.worldsys.com/labinfo/journal/fta/fta/htm Journal of Chemometrics http://www.wiley.com/journals/cem Journal of Near Infrared Spectroscopy http://www.nirpublications.com/jnirs.html Spectroscopy http://www.techexpo.com/toc/spectros.html Journal of Pharmaceutical Sciences http://pubs.acs.org/journals-sci/jsfa/index.html NIR News http://www.nirpublications.com/nirn.html Talanta http://www.elsevier.com:80/locate/issn/00399140 Trends in Analytical Chemistry http://www.elsevier.com:80/section/chemical/trac Table 4 Uniform resource locators (URLs) for selected societies concerned with NIR spectroscopy Society URL American Association of Cereal Chemists http://www.scisoc.org/aacc/info.html Council for Near Infrared Spectroscopy http://kerouac.pharm.uky.edu/ASRG/cnirs/cnirs.htm Society for Applied Spectroscopy http://esther.la.asu.edu/sas Analyst, August 1998, Vol. 123 137RInterested readers can find much more extensive descriptions of both NIR theory and equipment elsewhere.7,34,35 Principles of NIR spectroscopy The NIR lies between the visible and MIR regions of the electromagnetic spectrum and is defined by the American Society for Testing and Materials (ASTM) as the spectral region spanning the wavelength range 780–2526 nm (or the wavenumber range 12 820–3959 cm21).Light absorption in this region is primarily due to overtones and combinations of fundamental vibration bands occurring in the MIR region. For infrared light to be absorbed, its energy must be high enough to produce vibrational transitions in the molecules concerned, i.e., the light frequency should be exactly the same as a fundamental vibration frequency for a specific molecule and the molecule should undergo a change in its dipole moment by virtue of its fundamental vibration.The vibrational frequency f for a diatomic molecule can be determined on the assumption of the harmonic oscillator model, where an atom shifts from its equilibrium position with a strength proportional to the shift (Hooke’s law): f c k = 1 2p m where c is the speed of light, k the bonding force constant (a measure of the strength or rigidity of a chemical bond in its normal equilibrium position) and m the reduced mass.In this case, the variation of the potential energy with bond distance is a parabola centred about the equilibrium distance with evenly spaced vibrational energy levels. The energy Ev of each level will be given by Ev = f(v + 1 2) where f is the vibrational frequency and v the vibrational quantum number. Because the selection rule for harmonic oscillator transitions is Dv = ±1 and energy levels are evenly spaced, the energy difference between two consecutive levels will always be E(v + 1) – Ev = f, which is called the ‘fundamental frequency’ of the band.Vibrations in polyatomic molecules involve complex movements of their constituent atoms. The movements can be resolved into individual vibrations called ‘normal vibrations’. The motion of each atom is the result of its movements in the normal vibrations; the energy of each normal frequency is independent of the others, so the vibrational energy of the molecule is the sum of the individual energies: E f v v i i N = + = Â ( ) 1 2 0 In practice, molecular vibrations tend to be non-harmonic, i.e., vibrations about the equilibrium position are non-symmetric.The potential energy curve for real bonds is only roughly parabolic, with slight deviations at the lower energy levels that are more marked at the upper energy levels. Also, spacings between energy levels are not identical but rather decrease with increasing energy.One correction to the harmonic oscillator model that improves consistency between theoretical and experimental data involves additional terms of higher order than those used by Hooke’s law. Thus, the energy Ev for each level will be given by Ev = fe(v + 1 2) 2 fexe(v + 1 2) + higher-order terms where v is the vibrational quantum number, xe the nonharmonicity constant (which measures the deviation of the potential function from the parabola) and fe the uniform spacing between levels corresponding to a parabola with its centre at the equilibrium distance and the same curvature as the real potential energy function.If higher-order terms are neglected, then the frequency of a transition between adjacent energy levels (v ?n + 1) will depend on the vibrational quantum number: f = fe[1 2 2xe(v + 1)] One further consequence of introducing the quadratic term into Hooke’s law is that the selection rule becomes Dv = ±1, ±2, etc.; hence, in addition to the fundamental band (+1), other, higher frequencies called overtones or harmonics appear at frequencies two, three, etc., times higher than the fundamental frequency.The intensity of these bands decays abruptly since transition probability decreases markedly with increase in the vibrational quantum number and, in practice, each fundamental band only exhibits its first two or three overtones. Polyatomic molecules possess several fundamental frequencies so they may exhibit simultaneous changes in the energies of two or more vibrational modes; the frequency observed will be the sum of (f1 + f2, 2f1 + f2, etc.) or the difference between (f1 – f2, 2f1 – f2, etc.) the individual fundamental frequencies.This results in very weak bands that are called ‘combination’ and ‘subtraction’ bands—the latter are possible but rarely observed in room temperature NIR spectra. Non-harmonicity results in combination bands that are slightly smaller than the combined fundamental frequencies involved.Many NIR bands are overtones and combination bands for hydrogen bonds (C–H, N–H, O–H and S–H). The small mass and large force constants for hydrogen are the origin of the high fundamental frequencies in this atom; as a result, its first few overtones appear in the NIR region. CNO, C–C, C–F and C–Cl groups usually exhibit very weak or no bands in the NIR region; fundamental vibrations in these groups occur at low frequencies in the MIR region, where their first few overtones also appear as a result.NIR diffuse reflectance spectroscopy The low molar absorptivity of adsorption bands in the NIR region (typically between 0.01 and 0.1 l mol–1 cm–1) severely restricts sensitivity; however, it permits operation in the reflectance mode and hence the recording of spectra for solid samples. Reflectance spectroscopy measures the light reflected by the sample surface, which contains a specular component and a diffuse component.Specular reflectance, described by Fresnel’s law, contains little information about composition; consequently, its contribution to measurements is minimized by adjusting the detector’s position relative to the sample. On the other hand, diffuse reflectance, which is described by the Kubelka–Munk theory,36 is the basis for measurements by this technique. The Kubelka–Munk function, f(R°), is given by f R R R k s ( ) ( ) • • • = - = 1 2 2 where R° is the absolute reflectance of the sample (viz., the fraction of light impinging on it that is reflected), k its absorption coefficient and s its dispersion coefficient.In practice, relative reflectance (R), which is the ratio of the intensity of the light reflected by the sample to that by a standard, is preferred to absolute reflectance. The standard is usually a stable material with a high and fairly constant absolute 138R Analyst, August 1998, Vol. 123reflectance (e.g., Teflon, barium sulfate, magnesium oxide, high-purity alumina ceramics).The Kubelka–Munk equation can be rewritten in terms of the relative reflectance and the concentration of the absorbing analyte (c): f R R R k s c s c a ( ) ( ) = - = = = 1 2 2 e ln 10 where e is the molar absorptivity and a = s/2.303e. Thus, a plot of f(R) against c for samples conforming to this relationship will be a straight line of slope 1/a. However, if the matrix absorbs or the analyte exhibits strong absorption bands, the diffuse reflectance of the sample will not fit the Kubelka–Munk equation and the f(r) versus concentration plot will be nonlinear. As with Beer’s law, the Kubelka–Munk equation is acknowledged to be a boundary equation that is only applicable to weak absorption bands, or when the product of absorptivity times concentration is small.This is so in the NIR region; however, because the matrix frequently absorbs strongly at the same wavelength as the analyte, absorption by the latter cannot be resolved and deviations from the previous equation result.One widely used practical alternative is a relationship between concentration and relative reflectance similar to Beer’s law, namely: A R a c = = ¢ log 1 where A is apparent absorbance, R relative reflectance, c concentration and aA a proportionality constant. Although this relationship has no theoretical basis on the Kubelka–Munk equation, it provides highly satisfactory results under the typical conditions used in many diffuse reflectance spectroscopic applications.Operational procedures in the NIR The procedures used in the NIR region are much less labourintensive than those employed in the MIR and very similar to those used for liquid samples in the UV and visible regions. The absorbance of a liquid or solution can be readily measured by using quartz or sapphire cuvettes of variable pathlength or fibre-optic probes. No special precautions need be exercised since the absorption of NIR radiation obeys Beer’s law.The most suitable solvents in this context are those not containing O–H, N–H and C–H groups, which exhibit little or no absorption in this spectral region. The NIR spectrum of a solid sample can be obtained by using various types of device. The most frequent choices when the sample is in powder or grain form are reflectance cuvettes with a transparent window material (e.g., quartz) and fibre-optic probes.The latter considerably facilitate recording of spectra; however, light losses resulting from transport along the fibre result in increased signal noise. The spectra of samples in tablet form requiring no pre-treatment (e.g., powdering, sieving, homogenization) can be recorded by using three different types of equipment, namely: (a) specially designed reflectance cuvettes for tablets, which, however, provide spectra subject to marked scattering arising from dead spaces between tablets placed in the cuvette; (b) a commercially available instrument for recording reflectance spectra for individual tablets, inspired by the double-reflecting sample holder developed by Lodder and Hieftje in 1988;37 or (c) recently introduced instruments that allow the transmission spectra for individual tablets to be recorded. At this point, it is worth noting that the type of standard to be used for reflectance measurement remains a subject of strong debate.According to ASTM, the perfect standard for this purpose is a material that absorbs no light at any wavelength and reflects light at an angle identical with the incidence angle.Because no single material meets these requirements, the standards used in this context are stable, homogeneous, nontransparent, non-fluorescent materials of high, fairly constant relative reflectance. Springsteen and Ricker38,39 discussed the merits and pitfalls of materials such as barium sulfate, magnesium oxide, Teflon and ceramic plates as standards for reflectance measurements. Mathematical processing of signals The analytical signal obtained in NIR spectroscopy is a complex function that depends on both the physical and chemical properties of the sample; also, it is non-linear owing to scatter, stray light and inconsistency in the instrument response.This entails converting recorded data into apparent absorbance values [A = log (1/R)] or Kubelka–Munk (KM) units when measurements are made in the reflectance mode, and into absorbance units [A = log (1/T)] when made in the transmission mode.Osborne40 compared the ability to obtain linear calibrations from raw data, apparent absorbance values [log (1/R)] and KM units, and found that the last two provided calibrations that were not necessarily more linear than those obtained from raw data. He also found that the transformation choice was dictated by the particular data set.40 Griffiths41 claims that the choice of KM or log (1/R) depends on both the type of sample and the spectral region.For any type of diffuse reflectance measurement where the baseline is irreproducible, band intensities change with it when KM units are used but not when log (1/R) is employed. Dahm and Dahm42 re-assessed the pitfalls of KM units noted by Olinger and Griffiths.43 They explained why they did not share Griffiths’ view that, as a rule, log (1/R) versus analyte concentration plots are more linear in practice than KM graphs; however, they also claimed that the use of log (1/R) with powdered samples was effective.Converted values obtained from recorded data are markedly affected by scatter. This becomes apparent when NIR diffuse reflectance spectroscopy is used to analyse solid samples; in fact, the reflectance depends on the degree of scatter of incident light: the more marked the scatter is, the less deep light will penetrate into the sample and hence the smaller will be the absorption. Light scatter depends essentially on the physical properties of the sample (particle size, crystalline environment) and has a multiplicative effect on the amount of light that is absorbed by the sample, which combines with other additive effects such as baseline shifts or chemical absorption.The dependence of the signal on the physical properties of the sample, which is highly useful with a view to its characterization and makes the NIR technique a highly suitable tool for determining physical parameters, is a severe hindrance to qualitative analyses for identifying a product where physical appearance is not important, for detecting chemical deviations in the manufacturing process (e.g., heterogeneity, moisture, omission of some component of a preparation), or for the quantitative determination of chemical components—in fact, scatter may be largely responsible for variability between samples, which leads to high correlation among measurements at different wavelengths.These situations call for the prior mathematical processing of spectra in order to minimize the effects of those physical properties of the sample that influence an NIR spectrum and introduce variability that provides irrelevant chemical information.44–47 Some of the more widely used mathematical treatments for scatter in NIR spectra48 include normalization,49,50 derivation, 51–53 multiplicative scatter correction (MSC),54–56 piece- Analyst, August 1998, Vol. 123 139Rwise multiplicative scatter correction (PMSC),57 extended multiplicative signal correction (EMSC),58 optimized scaling (OS),59,60 standard normal variate (SNV),61 de-trending (DT),61 SNV followed by DT (SDT)62 and DT followed by SNV (DTS).62 Interested readers are referred to the cited references for information about each specific choice, a detailed description of which is obviously beyond the scope of this review. Papers concerned with available mathematical treatments for scatter in NIR can be classified into two broad categories, namely: (a) those that make empirical comparisons of various treatments or establish relationships among them; (b) those that examine the effects of different treatments on NIR spectral quantification.Prominent in the former group is the paper where Barnes et al.61 demonstrate the efficiency of SNV and DT treatments by application to sucrose samples of different particle sizes, and those where Dhanoa and co-workers demonstrate a linear relationship between SNV and MSC on the one hand,63 and DTS and SDT on the other.62 The latter group includes a large number of references, several of which are concerned with pharmaceuticals.Thus, Jacobsson et al.64 determined sulfasalazine by using a fibreoptic probe, partial least-squares regression and different mathematical treatments (derivative, MSC, PMSC). They found that MSC and PMSC provided the lowest errors of prediction; however, implementation of the latter required a preliminary study in order to optimize window size.Blanco et al.65 used the V/M ratio of Aucott et al.44 to assess and compare the efficiency of derivative, normalization, MSC, SNV, DT and DTS treatments with a view to reducing the effects of scatter on mixed-phase spectra for a pharmaceutical preparation. They found that the derivative and SNV treatments provided the best results, and normalization the worst, for the case considered.Qualitative analysis Identification and qualification of raw materials and pharmaceutical preparations Quality control involves implementing suitable procedures or measurements in order to ensure the identity of the materials at each stage of the manufacturing process, from the time the raw materials are received to that when the end products are released. NIR spectroscopy is an advantageous alternative to wet chemical methods and other instrumental techniques such as MIR spectrophotometry and NMR spectroscopy for this purpose.Similarly to MIR, the earliest NIR studies aimed at the identification of substances were concerned with structural elucidation. In 1954, Kaye66 assigned the bands in the spectra of bromoform, chloroform, benzene, methanol and m-toluidine, and studied the factors that influenced the position and intensity of the bands (inter- and intramolecular interactions, temperature, physical state of the sample).Several years later, Sinsheimer and Keuhnelian67 stated that NIR spectra were among the most effective means for distinguishing dissolved primary, secondary and tertiary amines; thus, primary amines differ from secondary and tertiary amines by the presence of a band at 2180 nm and the absence of another at 2050 nm. However, identifying a substance from the mere inspection of its NIR spectrum is usually difficult since this consists of very broad, usually overlapped bands that call for pattern recognition procedures (statistical treatments used to characterize spectra). 68 Essentially, the identification process involves two steps, viz., recording a series of analytical signals for the product and generating a so-called ‘spectral library’, and recording the sample signal and comparing it with those in the previously compiled spectral library on the basis of mathematical criteria for parametrizing spectral similarity. If the similarity level exceeds a pre-set threshold, then the spectra are considered identical and the sample is identified with the corresponding product in the library.Reliable identification of a product relies on correct choice of spectra for inclusion in the library. The spectra compiled for each product should contain every possible source of variability associated with spectral recording and the product’s manufacturing process. Spectral variability is considered by including spectra for the same sample recorded by different operators on different days; manufacturing variability is considered by including spectra for samples from different production batches.It is difficult to anticipate the exact number of spectra to be included in a ‘comprehensive’ library. For a product that is manufactured in a highly reproducible manner, manufacturing variability can be spanned by samples from 5 to 10 different batches and a total of 20–40 spectra. If manufacturing reproducibility is poor, the number of spectra required can easily double.One other important consideration in building a spectral library is checking that all the spectra included are correct. Uncontrolled factors (e.g., incompletely filled cuvettes, voltage drops at the time of recording) may result in spectral differences not ascribable to natural variability; any such spectra should be discarded. One widely used NIR mathematical treatment for expressing similarity is the correlation coefficient,69,70 which is defined as the cosine of the angle between the vectors for the sample spectrum and the average spectrum for each product included in the library: rjk ij j ik k i p ij j i p ik k i p x x x x x x x x = - - - - = = = Â Â Â ( )( ) ( ) ( ) 1 2 1 2 1 where p is the number of wavelengths; subscripts k and j denote the sample and reference product, respectively; xi is the measured value at wavelength i; �xj is the average spectrum of the reference product j; and �xk is the average spectrum of the sample.If the similarity coefficient exceeds a pre-set threshold, then the two spectra compared are considered identical and the sample is identified with the reference product. Theoretically, if the two spectra are coincident, the correlation coefficient should be unity; however, random noise associated with any type of spectral measurement precludes obtaining a coefficient of exactly 1. This parameter has the advantage that it is independent of library size and concentration changes, which permits correct identifications by use of libraries consisting of a small number of spectra.On the other hand, it is calculated from secondderivative spectra; hence, samples of the same product in different grain sizes will have the same correlation coefficient since particle size only affects band intensity in secondderivative spectra. Van der Vlies and co-workers used a correlation coefficient which they called the spectral match value (SMV) as a simple, expeditious and precise tool for identifying different types of cellulose71 and ampicillin trihydrate.72 Blanco et al.70 demonstrated the discriminating ability of a correlation coefficient that they called the match index (MI) in the identification of a pharmaceutical preparation by use of a library consisting of 163 substances including excipients, active compounds, amino acids and vitamins.One especially interesting application in this context is the non-invasive NIR method of Galante et al.73 for assessing microbiological contamination in injections.The method, based 140R Analyst, August 1998, Vol. 123on correlation measurements, is fast, detects contamination by various types of microbe (yeasts, mould and bacteria), avoids contamination by the analytical method itself and can be implemented on-line with the manufacturing process. Replacing conventional identification techniques, while important, is not the sole advantage of NIR spectroscopy for qualitative analytical purposes.In fact, this technique also affords qualification. The pharmaceutical industry must guarantee the correct dosage, manufacture and stability of each product, so it must carefully control raw materials and each step of the manufacturing process for factors such as potency, moisture, density, viscosity and particle size in order to detect potential deviations and correct them in a timely manner. Controls can rely on numerical determinations of the target parameters by using qualitative methods of analysis or comparing the NIR spectrum for the sample with the body of spectra for samples complying with the specifications and encompassing every possible source of natural and manufacturing variability. This latter choice is known as ‘qualification’ and involves expressing similarity in distance terms in order to determine whether a sample falls within the normal variability range or is subject to manufacturing deviations that call for comprehensive analysis.Distance-based methods rely on a compromise between the maximum number of wavelengths that can be used—in fact, if the wavelengths are correlated, increasing their number will increase the distance without providing additional information—and the minimum number required to encompass all possible sources of manufacturing variability in the product. One of the most widely used qualificatiothods is probably the wavelength distance method,70 which assumes that measurements at each wavelength are distributed according to the normal law.It generally uses the second-derivative of spectra from a library that defines the accepted variability for the product to obtain an average spectrum and the standard deviation at each wavelength. The distance between the unknown sample and the average spectrum for the reference product at each wavelength is calculated and the most unfavourable situation (viz., the wavelength that results in the maximum distance) is determined from the following equation: d x x s kj kp jp ij = - max where subscripts k and j denote sample and reference product, respectively; xkp is the measured sample value at wavelength p; �xjp is the average spectrum of reference product j at wavelength p; and sjp is the standard deviation of the measured values for reference product j at wavelength p.If the sample belongs to the same population as the reference product, then there will be a probability of 99.7% that the distance will be less than three times the standard deviation.If the maximum distance does not meet this criterion, then the sample must belong to a different population (i.e., it will not meet the qualification criterion). The qualification criterion based on the expression Dmax @3s is usually too conservative; it is often more practical to have users decide on the most suitable limit for their own problems and working methods.Correct usage of this method entails exhaustive control of the instrument in order to ensure that noise remains roughly constant, since measurements at individual wavelengths and derivative spectra tend to introduce noise and wavelength shifts. One shortcoming of this method is the risk of false-negatives at wavelengths coinciding with x-intercepts in second-derivative spectra (zero cross-over). If the standard deviation for the average spectrum at a given wavelength is very small, then the distance at that wavelength will be very large and a negative qualification will result.This may be the case when secondderivative values are very close to zero. This problem can be circumvented by using the wavelength library stabilization method,74 where the average spectrum and its standard deviation behave as though each second-derivative spectrum had been shifted by a fraction of a nanometre to the left and right along the wavelength axis (stabilization constant) in such a way that the standard deviation at zero cross-over points will be increased and false-negative qualifications avoided.Plugge and van der Vlies72,75 used the wavelength distance method to determine what they called the conformity index (CI), which is seemingly highly sensitive to sample impurities and occasionally allows one to pinpoint the sources of the inability to qualify a raw material or product by using a C-PLOT (viz., a plot of the absolute distance at each wavelength as a function of the wavelength itself).One constraint of the CPLOT is that it does not provide a sign of manufacturing deviations. Based on the C-PLOT, González and Pous76 developed DISPLOT (a plot of the distance, sign included, as a function of wavelength), which identifies the sign of small chemical and/or physical deviations introduced during the manufacture of the mixed phase of a product. One alternative to the wavelength distance method is to use the whole information contained in the spectrum by calculating the Mahalanobis distance.77–79 This parameter is useful for cluster analysis and can be calculated for multi-dimensional spaces.The distance between the sample and the centre of the cluster formed by the spectra of the reference product is defined as D2 = (Xj 2 �X k)TC(Xj 2 �X k) where Xj is the vector describing the spectrum of sample j, Xk is the vector for the average spectrum of reference k, C is the matrix that describes distance measurements in the multidimensional space studied and superscript T denotes transpose matrix.Usually, if the distance thus obtained is less than three times the standard deviation, the sample meets the qualification criterion (i.e., the manufacturer’s specifications). In the early 1990s, Corti and co-workers published several papers reporting on the use of the Mahalanobis distance for quality control in various pharmaceutical preparations.They qualified chloroform extracts of samples containing 0.05% estrogen and 0.25% progesterone or only one of them80 and discriminated among creams containing the same active principle but different excipient proportions.81 They showed that the Mahalanobis distance was a highly effective choice for qualifying antibiotics82,83 as it allows one to discriminate between their crystalline and amorphous forms, as well as among mixtures containing variable concentrations of the same antibiotic.Finally, they obtained a high reproducibility in the classification of organic and inorganic raw materials for which spectra had been recorded by different operators under different conditions.84 Dreassi and co-workers used the Mahalanobis distance to discriminate among samples of the same active principle differing in some physical and/or chemical property,85 as well as for quality control in the production of an antibiotic, where their method allows samples to be characterized at different stages of the process86 and distinguishes them from other products manufactured in the same production area.86,87 Recently, van der Vlies et al.88 developed a procedure for the qualification of pharmaceuticals based on the conversion of NIR spectra to polar coordinates and the subsequent calculation of the corresponding Mahalanobis distance (the polar qualification system).The method uses spectra, which facilitates relating the distribution of the products to their spectral features (e.g., Analyst, August 1998, Vol. 123 141Rthe presence of an impurity absorbing in a specific region); also, graphs are two-dimensional and hence easy to interpret. Notwithstanding these advantages, it remains to be proved whether this method surpasses existing alternatives in practice and is applicable to extensive spectral libraries—in fact, its use is seemingly restricted to the discrimination of similar products with known spectral features. Plugge and van der Vlies89 showed that the method allows one to discriminate chemically identical substances from different suppliers and also to detect differences in physical properties among samples from the same supplier.One alternative to direct spectral computations, where correlations and distances are calculated in the wavelength space, is the use of principal component analysis (PCA)90 to reduce variables (in PCA, correlations and distances are calculated from scores in the space bound by the principal components).Because the number of data involved is smaller— a large number of wavelengths is replaced with a few principal components (PCs)—library searches are much faster; however, all spectra in a library influence PC calculations, so every time a new spectrum is included in or an existing one is excluded from the library, PCs must be recalculated and the modified library validated. In addition, equipment-bundled software usually restricts the maximum number of PCs that can be used in the calculations, which in turn limits the number of different products that a spectral library can contain.Lo and Brown91 used the correlation coefficient in the PC space to identify components in mixtures of organic solvents. The ensuing method is selective, requires no prior knowledge of the mixture composition and avoids variability due to spectral noise. Wu et al.46 identified tablet blisters containing different amounts of an active principle by using PCA and different mathematical treatments of the spectra.Second-derivative calculations proved to be the most effective transformation as regards discriminating power. Shah and Gemperline92 qualified different batches of Avicel PH101 microcrystalline cellulose by using the Mahalanobis distance in the PC space. Their criterion was to assume that a sample was qualified when the probability level for a c2ibution fell in the range 1.0–0.05.One other classification procedure used in some reported applications is the soft independent modelling of class analogy (SIMCA).93 Each of the products in the sample is subjected to PCA and Fisher’s test is subsequently applied in order to estimate the likelihood of a sample belonging to the class defined by the spectra of the reference product. The residual variance for a spectrum k to be identified (S2k ) that is assumed to belong to class j (defined by the spectra of the reference product j) is divided into the total variance for the samples belonging to class j (S02 ) in order to obtain the following variance relationship: F S S n n a k = - - 2 0 2 1 where n is the number of spectra for the reference product and a the number of PCs used to construct the class model.Gemperline et al.94 used SIMCA to qualify 400 firstderivative spectra for six raw materials and found it to be sensitive to the presence of impurities such as production intermediates and degradation products, as well as to particle size.In subsequent work, Gemperline and Boyer95 identified and qualified samples adulterated with small amounts of impurities by using libraries of variable size and the Mahalanobis distance, maximum distance and SIMCA method. The maximum distance proved to be the most suitable choice for identifying samples with small spectral libraries, but performed worse for qualification purposes as it was insensitive to impurities at levels below 2%.Shah and Gemperline96 used the Mahalanobis distance and the SIMCA method to classify different types of cellulose and detect impurities at levels of 0.1–2%. They found that temporal changes in instrument response influenced the limit of detection; consequently, reliable detection of impurities required considering such changes in the spectral libraries. Dempster et al.97 used the maximum distance to confirm the identity of tablet blisters containing different concentrations of the active principle (5, 10 and 20% m/m).They compared three different spectral recording procedures, namely: (a) extracting tablets from their blisters prior to measurement; (b) making measurements through blisters, using the horizontal set-up presentation module; and (c) using a fibre-optic probe for measurements. The first procedure proved to be the most sensitive as it distinguished among the three concentration levels tested and the placebo; on the other hand, the other two failed to discriminate the 5% sample and the placebo but had the advantage that they were non-invasive.Subsequently, they used the fibre-optic probe in conjunction with the maximum distance, Mahalanobis distance and SIMCA to confirm the identity of coated and uncoated tablet blisters.98 They used only those spectral zones where differences among products were maximum and the results were optimum, provided that the tablets and blisters to be qualified had been manufactured under the same conditions as those included in the library.Ciurczak and Maldacker99 compared the ability of crosscorrelation spectral reconstruction methods and that of discriminant analysis based on the Mahalanobis distance to classify tablets in terms of their active principle. The spectral reconstruction method, developed by Honigs et al.,100 allows one to obtain the individual spectrum for each mixture component and determine the nature of interactions among analytes; however, it is less suitable for classification purposes.One qualification alternative similar to that using the Mahalanobis distance is the boostrap error-adjusted singlesample technique (BEAST),101,102 which uses reflectance values obtained at pre-set wavelengths to obtain a multidimensional data distribution. The chief difference between the Mahalanobis distance and BEAST is that, in the latter, the confidence limits used to define the clustering limits consider asymmetry in the sample distribution rather than the symmetric distribution assumed in the Mahalanobis distance.The most severe shortcoming of BEAST is that it requires extensive data storage resources. Lodder and co-workers showed that the use of NIR spectroscopy in combination with BEAST provides a rapid method for detecting adulterants (Fe2O3, NaF, NaCN, KCN and As2O3) in capsules103 and allows discrimination among aspirin tablets from different manufacturers37 with no need for sample pre-treatment—and hence with minimal manipulation errors.This overview of the qualitative applications of NIR spectroscopy in the pharmaceutical industry would be incomplete if no mention were made of the fact that this technique has been endorsed by several agencies in their official methods of analysis (Table 5). Determination of homogeneity One important operation in manufacturing solid pharmaceuticals is monitoring of the homogenization process, which determines the encapsulation or compression quality of the end product.Almost every application of NIR spectroscopy in this field has been reported recently and uses one of the qualification methods described in the previous section. Ciurczak104 developed three different approaches to the monitoring of the homogenization of aspirin–vitamin B12 mixtures by use of fibre optics, namely: (a) visual comparison of second-derivative spectra recorded at different homogenization times; (b) calcula- 142R Analyst, August 1998, Vol. 123tion of the correlation coefficient; and (c) calculation of the maximum distance. The last proved to be the most reliable method for determining the end-point of the homogenization process as it discriminated between the penultimate and last mixtures. The qualification concept was used by Wargo and Drennen105 to verify the homogeneity of solid mixtures and determine the optimum homogenization time for a preparation containing hydrochlorothiazide as the active principle.Qualitative analytical algorithms based on BEAST proved to be more sensitive to variations in sample homogeneity than did a c2 test as the former uses the entire NIR spectrum. van der Vlies and co-workers converted NIR spectra into polar coordinates and used these to calculate the Mahalanobis distance88 in order to identify non-homogeneous samples.89 They found that analysis of variance (ANOVA) was an effective choice for validating homogenization processes.Hailey et al.47 and Sekulic et al.106 developed systems for monitoring the homogenization of solid mixtures based on measurements via a fibre-optic probe fitted to the mixer. The most salient advantage of these systems is that they permit the determination of the end-point of the homogenization process in real time and in a non-invasive manner. In both cases, mixture homogeneity is determined by plotting the standard deviation for several replicates against the homogenization time.Polymorphism and optical isomers NIR spectroscopy was used by Gimet and Luong107 for the qualitative control of a dimorphic analgesic. The pure forms exhibit NIR spectra that are sufficiently different in their maximum wavelengths and absorbances to allow quantification. Mixed spectra confirm that quantitative analyses are possible even in the absence of qualitative differences between the spectra. The ensuing method is applicable to substantial amounts of product, which avoids errors arising from sampling and sample heterogeneity.The ability of the Mahalanobis distance and SIMCA to identify and assess the polymorphic quality of a drug was compared by Aldridge et al.108 The former proved to be the more effective choice since, in addition to discriminating between the polymorph of interest and other substances with highly similar spectra, it is more sensitive to the presence of low levels of impurities in the polymorph.Norris et al.109 used NIR spectroscopy to monitor polymorphic conversion. Their method subjects spectra recorded over the course of the reaction to PCA and allows the end-point of the process to be determined in real time. Buchanan et al.110 determined the enantiomeric purity of the optically active forms of valine. Mixtures containing d- and lvaline in different proportions exhibited identical spectra except for baseline shifts resulting from differences in particle size.However, when the mixtures were dissolved and recrystallized, the resulting spectra exhibited qualitative and quantitative differences that permitted the determination of enantiomeric purity in the starting product. b-Cyclodextrin and silica gel were used as chiral selectors for distinguishing the (1R)-(+) and (1S)-(–) enantiomers of apinene by NIR transmission spectroscopy.111 The bond between the (+) enantiomer and a reagent is different from that between the (2) enantiomer and the same reagent, which facilitates discrimination of the enantiomers by PCA.Quantitative analysis NIR spectra typically contain broad, overlapping bands that cannot always be ascribed to an individual sample component. As a result, whenever the NIR technique is used for quantitative purposes—whatever the physical or chemical property of the sample to be determined—a calibration must be performed by using an existing multivariate procedure.69,90 Essentially, the procedure for quantification using multivariate calibration involves the following steps: (a) selecting a representative sample set; (b) acquiring the analytical signals and obtaining the reference values; (c) mathematical processing of the signals; (d) selecting the model that relates the property to be determined and the signals; and (e) validating the model.Each step is described in detail below, with special emphasis on the problems arising from NIR analyses of pharmaceuticals. Halsey112 devised a protocol for developing quantitative NIR methods for the pharmaceutical industry.Although the protocol is based on the NSAS software package, bundled with NIRSystems instruments, it can provide users of equipment from other manufacturers with basic concepts to be considered in developing an NIR analytical method. Sample selection The starting point for every calibration technique is a set of samples which have previously been analysed by a reference method, span the working concentration range and are representative of the manufacturing variability sources that are bound to influence the NIR spectra.One of the problems encountered in using NIR spectroscopy for the quantitative analysis of pharmaceuticals is the need to obtain a sample set that can be used to establish a calibration model. As a rule, all available production samples contain the active principle and excipient in amounts very close to the nominal values; this precludes spanning a wide enough concentration range for calibration.One way of circumventing this shortcoming is by using a set consisting of production and laboratory-made samples; the former will introduce the variability sources typical of the production process while the latter will expand the narrow range spanned by the former. Therefore, the two essential questions that arise when developing an NIR quantification method are as follows: what concentration range is the sample set to span? and how can preparation of the laboratory samples be approached? Regarding the former question, Corti and co-workers83,84,113 claim that a sample set spanning a concentration range about ±5% of the nominal value affords precisely and reproducibly sufficient calibration for quality control purposes.However, such a range may be too narrow if the manufacturer’s tolerated limits are greater than ± 5% of the nominal value. In order to expand the concentration range without altering any physical properties potentially affecting NIR spectra, one can make the samples at a pilot plant,114,115 prepare laboratory samples containing each compo- Table 5 NIR methods endorsed by various official agencies Agency Method Food and Drug Administration (FDA) Identification, quantification and determination of moisture content in ampicillin trihydrate (Gist Brocades) Health Protection Branch (HPB) Identification of raw materials and packaging components (Merck Frosst Canada) Norwegian Medicines Control Authority (SLK) Identification and quantification of paracetamol, and determination of moisture content, in Paracet 500 mg (Wieders Farmasoytiske) Medicinal Controls Agency (MCA) in UK Identification of Zovirax 200 mg (Glaxo Wellcome) Analyst, August 1998, Vol. 123 143Rnent of the pharmaceutical at concentrations over the manufacturer’s specified ranges,116 or over- and underdose samples from different production batches with small amounts of the active principle or excipient, respectively, to obtain the desired concentration range.70,86,87 The first approach is probably that providing the samples that are closest in composition to production samples; however, it is also the least feasible in practice as it is rarely possible to manufacture production batches suited to particular needs, nor is it possible to ensure that a small-scale process will be comparable to the actual production process.The principal advantage of the second approach is the ease with which the different concentrations needed to span the required range can be obtained; however, the grinding, mixing and other miscellaneous processes used in the laboratory may differ substantially from those used in the manufacturing process and hence lead to samples differing markedly—NIR spectra included—from production samples. Over- and underdosing production sample make expanding the concentration range a labour-intensive, care-demanding task; provided that strict control is exercised, however, variability in the physical features of the samples can be much smaller than in the previous case.Blanco et al.117 found that the use of over- and underdosed samples did not alter the quality of the results for production samples; they quantified production samples by using calibration sets consisting of an increasing number of laboratory-made samples. The same authors118 compared two different approaches, viz., preparing samples by (a) weighing of all components and (b) over- and underdosing samples from different production batches, to quantify the active principle in the mixed phase of a commercially available preparation.Although the results obtained with both approaches were similar, the over- and underdosing approach resulted in simpler calibration models and in slightly smaller errors of prediction. Once the calibration set has been established, it is split into two sub-sets, viz., a calibration set consisting of a small number of samples that are representative of the entire set and allows the determinand to be related to the analytical measurement, and a prediction set composed of the remainder of samples that is used to assess the predictive ability of the model.In splitting the original set, the questions arise as to what the optimum number of samples to be included in the calibration set is and how such samples should be chosen.The use of a small number of samples in the calibration set may result in some source of variability in the product being excluded and hence in spurious results in analysing new samples. Several workers claim that the optimum number of samples depends on their complexity, on the concentration range to be spanned and on the particular calibration method used.113,119 Thus, when the aim is to quantify 1–4 components and the samples exhibit no large differences in their physical and chemical properties, a calibration set consisting of a minimum of 15–20 samples will be more than adequate.As regards the second question, there are two types of approach, viz., those focusing on general aspects of sample selection for NIR calibration69,120 and those based on comparisons among available choices. Although, in general, the latter have been developed for and applied to food samples, they are worth mentioning here because they are also applicable to pharmaceuticals.Honigs et al.121 used a sample selection method similar to Gaussian elimination, They selected ‘unique’ samples in a sequential manner in order to identify that exhibiting the highest NIR absorbance. The selected sample was removed from the remaining set and the process was repeated until the desired number of samples was chosen or the absorbance values of the remaining spectra were below a pre-set limit. Næs122 and Isaksson and Næs123 developed a method for selecting samples based on an unsupervised pattern recognition algorithm that is applicable to highly collinear data. They identified clusters of closely related samples by constructing a dendrogram based on the PCA scores for the sample spectra.From each cluster, the sample falling at the greatest distance from the cluster centre was chosen. A similar sample selection system was reported by Puchwein.124 The sample with the greatest Mahalanobis distance from the cluster centre was selected first and those samples with a Mahalanobis distance similar to the selected sample were left out.Subsequent samples were chosen similarly from among the remainder. The normalized Mahalanobis distance was used by Mark125 to select samples on the basis of discrete wavelengths. One disadvantage of this method is the difficulty involved in selecting a suitable wavelength. Ferré and Rius126 reported a procedure for selecting the best calibration set for principal component regression (PCR) based on a D-optimum design and on instrumental responses alone.Calibration investment and effort are reduced if the reference method is applied to the selected samples only. This approach was criticized by Davies,127 who stated that the prediction set provided by the method was strongly dependent on the calibration set and thus a poor choice for assessing the predictive ability of the model. Blanco et al.70 compared flat calibration,128 which involves spanning the working concentration range with a large number of samples, and the sample selection subroutine included in the NSAS software package,129 with a view to selecting calibration samples for the quantification of the active principle in a commercially available preparation.The flat calibration approach provided more robust calibration models. The same authors proposed using PCA to select the production batches best representing variability in the manufacturing process, which must be included in the calibration set in addition to laboratory-made samples.116,118 Multivariate calibration methods The calibration methods most frequently used in NIR spectroscopy in order to relate the property to be measured to the analytical signals acquired are multiple linear regression (MLR),130,131 PCR90 and partial least-squares regression (PLSR).90 Most of the earliest quantitative applications of NIR spectroscopy rely on MLR because spectra were then recorded on filter instruments, which afforded measurements at a relatively small number of wavelengths only.Applications involving PCR and PLSR have proliferated after the introduction of commercially available instruments that allow the whole NIR region to be scanned. The choice of the calibration method is dictated by the nature of the sample, the number of components to be simultaneously determined, the a priori knowledge of the system studied and available data on it.Below are briefly described the features of the different calibration options. The MLR technique is the usual choice with filter instruments and is also occasionally used with instruments that record whole spectra. It is an effective calibration approach when the analytical signal is linearly related to the concentration, spectral noise is low and the analyte does not interact with other sample components. The MLR technique also affords modelling some non-linear relationships as it assumes that modelling errors arise from concentrations.However, it can only be used at a small number of wavelengths, which, if incorrectly selected, may result in overfitting (i.e., in modelling of noise or random errors). Also, if spectral data are highly collinear, then the precision of the results suffers appreciably. A detailed description of available procedures for determining how many and which wavelengths should be used can be found elsewhere. 90,132,133 144R Analyst, August 1998, Vol. 123Whole-spectrum methodologies (viz., PCR and PLSR) have the advantage that they use every single wavelength in a recorded spectrum with no prior selection. Also, they allow the simultaneous determination of several components in the same sample and avoid the problems associated with collinearity among spectral data and with noise-related variability. As noted earlier, non-linearity in NIR signals is ascribed to non-linear detector responses that result in curved signal–concentration plots, as well as to physical and/or chemical factors giving rise to shifts and width changes in spectral bands.134,135 In some cases, non-linearity is so marked that a non-linear calibration methodology such as neural networks,136–140 locally weighted regression,141,142 projection pursuit regression143,144 or quadratic versions of the PCR or PLSR algorithms145,146 must be used.Determination of physical parameters Particle size determinations are of paramount importance to the pharmaceutical industry as incorrect grain size analyses can lead to altered properties such as coating power and colour, hinder subsequent mixing of powders (for tablet formulations) or powders and liquids (suspensions), and result in defective pressing of solid mixtures for making tablets.Because particle size is one of the physical parameters most markedly influencing NIR spectra, the NIR technique is an effective alternative to the traditional methods involving sieving, light scattering by suspensions, gas adsorption on solid surfaces or direct inspection under a microscope.Ciurczak et al.147 used the linear dependence of band intensity at a constant concentration on the average particle size at a pre-set wavelength to determine pure substances and granules. Absorbance versus particle size plots at different wavelengths exhibited two linear segments. These authors postulated that the sample’s absorption coefficient in the Kubelka–Munk function was large below 85 mm and ascribed the abrupt decrease in absorbance from 250 to 85 mm to the influence of such a coefficient.Consequently, the effect of particle size on reflectance measurements was significantly reduced below 80 mm. Blanco et al.148 determined the average particle size of Piracetam over the wavelength range 175–325 mm with an error of 15 mm, based on the assumption that an increase in particle size would produce an increase in absorbance that could be measured and used to quantify the former by MLR and PLSR calibration. They found that spectral reproducibility varied in an exponential manner with particle size and that sample compactness was the most influential factor on particle size.Ilari et al.149 investigated the feasibility of improving the determination of the average particle size of two highly reflecting inorganic compounds (viz., crystalline and amorphous NaCl) and an NIR-absorbing species (amorphous sorbitol), using the intercept and slope obtained by subjecting spectra to MSC treatment as input parameters for PLSR.While particle size continues to be the physical property of samples most frequently determined by NIR spectroscopy, several other parameters such as the dissolution rate and the thickness and hardness of the ethylcellulose coating on theophylline tablets have also been determined, all with good errors of prediction. 150 Determination of moisture content The presence of crystallization or adsorbed water in pure substances and pharmaceutical preparations, whether during treatment of the sample or its storage, causes significant changes in those properties that influence chemical decay rates, crystal dimensions, solubility and compaction power, among others. NIR spectroscopy is an effective alternative to traditional analytical methods such as thermogravimetry and Karl– Fischer (KF) titration as water gives a characteristic absorption spectrum the mere visual inspection of which allows one to determine, for example, if different batches of a given substance contain also different amounts of moisture.151 The NIR spectrum of water exhibits five absorption maxima at 760, 970, 1190, 1450 and 1940 nm; the positions of these bands can be slightly shifted by temperature changes152–154 or hydrogen bonding between the analyte and the matrix.155,156 The bands at 760, 970 and 1450 nm correspond to the first three overtones of O–H stretching bands, whereas the other two arise from combinations of O–H oscillations and stretching. The specific band to be used for determining water depends on the desired sensitivity and selectivity levels.157 As a rule, the overtone bands are appropriate for this purpose when using solutions in solvents containing no O–H groups; on the other hand, the band at 1940 nm provides increased sensitivity. In Tables 6 and 7, available NIR methods for determining moisture are classified according to whether they rely on transmittance or reflectance measurements, respectively.NIR transmittance methods are mainly used to determine water in solvents. Their earliest applications to solid preparations entailed dissolving the sample in a solvent with little or no absorption in this spectral region. All methods of this type use least-squares calibration to construct a straight line from absorbance values at the absorption maximum at about 1900 nm for solutions containing variable concentrations of the target species.Table 6 Applications of NIR transmittance spectroscopy to the determination of moisture content Sample type Remarks Ref. Solid The most suitable solvents for determining water in solid samples are pyridine and methanol, which exhibit no absorption band at 1900 nm; also, their mixtures with water obey Beer’s law over a wide composition range 155 Solid Determination of trace amounts of water (0.05%) in mono-, di- and triglycerides using chloroform as solvent.Free from interferences from triglyceride OH groups 158 Solid Use of the NIR technique in conjunction with dimethyl sulfoxide as solvent provides an expeditious, accurate and precise alternative to the traditional method for the determination of water in starch, which is time-consuming and involves cumbersome manipulations of the viscous, sticky samples involved 159 Solid Methanol is used to determine water in organic compounds and pharmaceutical preparations.Results are consistent with those provided by the conventional dehydration method and KF titration. The spectrophotometric method is more reproducible, simple and expeditious 156 Liquid Determination of small amounts of water in solvents (acetonitrile, propionitrile, tetrahydrofuran and dimethylformamide). The NIR method is less sensitive than KF titration and its LOD is about 20 ppm. The former is more rapid and flexible, and involves less extensive sample manipulation 160 Liquid Flow injection analysis method for the determination of water in dichloromethane and isobutyl methyl ketone, the LODs for which are 0.01 and 0.005% v/v, respectively.Free from sample contamination by environmental moisture 161 Analyst, August 1998, Vol. 123 145RDetermination of active compounds and excipients The number of determinations of active compounds and excipients has grown enormously in recent years.Table 8 summarizes reported uses of the NIR technique for this purpose. Calibration transfer In previous sections, we discussed the most important considerations in developing a quantitative NIR method. However, this review would be incomplete if one of the major hindrances to the application of developed methods, viz., transferability of calibration models among instruments,177–179 were not mentioned. Because detector responses are not uniform, the signals recorded by two different instruments may differ owing to wavelength shifts and/or changes in measured intensities.This precludes the use of one instrument’s calibration model by another. The problem can be overcome in three different ways, namely: (a) by recording spectra for calibration samples on each instrument and constructing a calibration model for each, which is not feasible when a large number of samples are to be processed, the instruments are very distant from each other or the samples are unstable; (b) by applying mathematical corrections to the spectra recorded by one instrument so that they can be used in the calibrations obtained with the another; (c) by transferring the calibration model from one instrument to another.The principal calibration transfer methods.180 are briefly commented on below. Some come from fields other than the pharmaceutical field but are indeed applicable in many sectors. The method of Shenk and Westerhaus181,182 uses a large number of stable samples for transfer and corrects the response of an instrument at each wavelength with reference to that of a primary instrument of identical resolution; after wavelengths have been corrected, the spectral intensity is adjusted.The applicability of this method has been assessed by several workers. Dardenne et al.183 focused on the problems encountered in transferring calibrations among different types of instrument and on the sample set required for this purpose.Bouveresse and co-workers found that the method provided satisfactory results as long as the transfer samples span the same spectral intensity range and are of the same nature as those to be subsequently analysed;184 they investigated whether altering the spectral intensity correction algorithm improved the quality of the results when some of the previous conditions were not fulfilled.185 Mark and Workman186 developed a method suitable for MLR calibration that uses no transfer sample set; rather, calibration transfer relies on a model that is constructed from those wavelengths that remain unchanged relative to spectral shift as independent variables.The previous two methods are applicable at a relatively small number of wavelengths and are usually incompatible with multivariate calibration based on whole spectra. Although the method of Shenk and Westerhaus181 affords whole-spectrum correction, its applicability to multivariate calibration approaches is restricted by the fact that corrections rely on a univariate scheme.Wang et al.187 developed four calibration transfer methods based on a multivariate scheme; all four use whole spectra and an unrestricted number of wavelengths in the calibration model. Two of them correct the calibration model (one with respect to a classical calibration model and the other relative to an inverse calibration model); the other two, based on direct standardization (DS) and piecewise direct standardization (PDS), correct the response of an instrument so that its spectra can be used by another. The DS method uses a PCA to obtain the transformation matrix that relates the responses of both instruments, but has the disadvantage that it requires a large number of samples.The PDS method188,189 relies on the fact that spectral variations are usually restricted to a small region; hence it reconstructs each point in the spectrum from one instrument by using several measurements through a small window in the spectrum from another.As a result, this method requires a fairly small number of transfer samples that need not span the whole concentration range spanned by the calibration samples. Bouveresse et al.190 compared the PDS method with the slope/bias correction method used by Jones et al.;191 when differences between instrument responses are small, the latter method provides good results, the quality of which can be assessed by Fisher’s test if the number of transfer samples used exceeds five.The method of Forina and co-workers192,193 uses PLSR to establish the relationship between the transfer samples processed with both instruments and then the regression equation for the first instrument. Blank et al.194 used a calibration transfer method based on a finite response filter to relate the response of a spectrophotometer to that from a second instrument without the need to record spectra for transfer samples on the latter.Miscellaneous applications This section comments on some uses of NIR spectroscopy with their roots outside the pharmaceutical field but of potential interest to the pharmaceutical industry. Table 7 Applications of NIR reflectance spectroscopy to the determination of moisture content Sample type Calibration Remarks Ref. Injection MLR PLSR Comparison of calibration methods by using products containing different amounts of active compound (0.5 and 1.5 mg per vial).Spectra are recorded through vial bottoms, using the horizontal set-up sample presentation module 162 Solid MLR Determination of moisture content (11.5–15%) in ceftazidime, with errors of 0–5%. The calibration set encompasses ±10% of the nominal value and consists of samples from different production batches and laboratory-made samples in 1:1 ratios 163 Solid MLR NIR method for determining moisture in the antibiotic ampicillin trihydrate; approved by FDA in 1992 72 Solid MLR PLSR Determination of moisture in the active compound ferrous lactate dihydrate (11.1–14.6%), using a fibre-optic probe.The two calibration methods used provide similar results, with errors less than 1.5% 164 Solid Tablets MLR Determination of moisture at different production stages (mixed phase, cores and tablets), using a single calibration equation that provides prediction errors less than 4%. Moisture contents below 1% can seemingly not be detected by reflectance measurements 87 Tablets MLR Determination of moisture in tablets with a maximum certified content of 2%.The results obtained over a one year period reveal that the method is suitable for quality control analyses 165 146R Analyst, August 1998, Vol. 123Reeve studied changes in NIR spectra due to various factors (moisture, pH and state of aggregation of the sample, among others).195,196 The presence of moisture in the samples was found to shift absorption bands to an extent dependent on the type of compound concerned–shifts are especially prominent with alcohols and ketones, and less marked with acids; also, the spectral features of solid samples vanish on dissolution, which destroys the crystal structure.197 Reeve also investigated the interactions between monomers and polymers of carbohydrates (glucose and sucrose with amylose, and amylopectin and cellulose with starch) that affect NIR spectra and assessed their effects on calibration methods.198 In several papers, Lin and Brown152,199–201 showed that MLR and PCR allow one to construct calibration models for determining NaCl in aqueous solutions as they afford measurements of the small intensity changes undergone by bands in the presence of the electrolyte.PCR was found to provide the smaller errors of prediction and models less markedly influenced by the temperature at which spectra were recorded. NIR spectroscopy with fibre-optic probes and PLSR or PCR is an alternative to gas chromatography and mass spectrometry for the in situ analysis of solvent mixtures on account of its high responsiveness, low maintenance costs and the need for no sample treatment.There are references to the analysis of mixtures of ethanol, acetone, acetic acid and water in ethyl acetate;202 methanol and water in hexane;202 methanol, ethanol and propanol;203 and ethanol, propan-1-ol and propan-2-ol in methanol.204 Martens and Stark58 demonstrated the significance of the prior mathematical treatment of signals as a means of suppressing multiplicative (pathlength variations) and additive changes (baseline shifts, spectral overlap) encountered in the determination of toluene in mixtures of benzene and xylene; use of a mathematical treatment led to simpler models of increased predictive capacity.Table 8 Applications of NIR spectroscopy to the determination of active compounds and excipients Analyte Sample type Calibration Remarks Ref.Glycerol, ethanol, phenazone, sodium thiosulfate Liquid MLR Errors less than 3% in major components (glycerol, ethanol and phenazone) and of 5–10% in minor components (lidocaine and sodium thiosulfate) 166 Glucose, fructose, maltose Syrup Comparison of NIR, FTIR and HPLC techniques. The accuracy of the spectrophotometric techniques is lower 167 Acetaminophen, codeine phosphate Syrup MLR Comparison of NIR spectroscopy and HPLC in terms of accuracy and throughput.The NIR technique is recommended for components at contents of at least 1% 168 Meprobamate (200 and 400 mg) Suspension Injection MLR Absorbance measurements at 1960 nm of the drug extract in chloroform 169 Cloxacillin benzathine (12.7%) Cream MLR Quantification in creams containing variable proportions of the same excipients, using a single calibration. Errors less than 3.5% 81 Nicotinamide Solid MLR Reproducible results (comparable to those of HPLC) obtained by using two different wavelengths 170 Ceftazidime (77%) Solid MLR Determination of a major active compound with errors of 0–3%. Variability in the raw materials used over a one year period has no effect on the quality of the results 163 Streptomycin sulfate; Cloxacillin benzathine Solid MLR Errors less than 4% in both components that change little on expanding the concentration range used for calibration 84 Ketoprofen (33%) Solid MLR Prior extraction of the drug into chloroform.Errors less than 3.5% 113 Ranitidine hydrochloride Solid MLR Errors less than ±2% and calibration transferability 87 Erythromycin ethylsuccinate (12.9, 19.9 and 34.3%) Granules MLR Errors less than 2.5% with a single calibration for the three presentations 83 Cimetidine (71.8%) Granules MLR The reproducibility of the NIR method (RSD = 0.16%) is comparable to that of a UV method (RSD = 0.15%) and is not influenced by particle size or grain colour 171 Vitamin C Granules MLR Errors of 1–2%.The PLSR method provides slightly smaller errors 117 (16.7, 22.9 and 40%) Tablets PLSR Ranitidine hydrochloride (53.5%) Tablets MLR Errors less than 5%. The precision is not operator-dependent 165 Pirisudanol dimaleate (88%) Tablets PLSR Comparison of recording systems (spinning cuvette and fibre-optic probe). Errors less than 1% and similar in both cases 70 Metoprolol succinate (47.5%) Tablets PLSR Comparison of NIR transmission and diffuse reflectance measurements.The former uses more favourable sample volumes but suffers from spectral noise above 1350 nm 172 Acetylsalicylic acid Tablets PCR Correlation of NIR spectra with the amount of salicylic acid formed by hydrolysis of acetylsalicylic acid. Prediction errors of ±0.04% of the tablet mass 173 Active principle (0.6, 1.2 and 2.4%) Tablets PLSR Automation of an NIR transmission spectroscopic method for determining content uniformity. The reproducibility of the autosampler used was studied 174 Aminodarone hydrochloride ketone (52%) Tablets MLR Errors less than 0.5%.The reproducibility was studied at different temperatures 175 SB 216469-S (1.5, 3 and 6%) Tablets PLSR Quantitative control of the active compound at the different production stages, with no sample pre-treatment 176 Cefuroxime acetyl Tablets MLR Quantitative control of the active compound at the different production stages. The PLSR 86 (66.8%) PLSR method provided lower errors in all cases Analyst, August 1998, Vol. 123 147RConclusions Recent breakthroughs in analytical instrumentation and available techniques for processing complex signals have fostered the development of new uses of NIR spectroscopy in various industrial fields, prominent among which is the pharmaceutical industry. The large number of references cited in this review testifies to the potential of NIR spectroscopy for qualitative and quantitative analysis of pharmaceutical preparations.References 1 Herschel, W., Philos. Trans. R. Soc. London, 1800, 90, 255. 2 Herschel, W., Philos. Trans. R. Soc. London, 1800, 90, 284. 3 Herschel, W., Philos. Trans. R. Soc. London, 1800, 90, 293. 4 Goddu, R. F., in Advances in Analytical Chemistry and Instrumentation, ed. Reilly, Ch. N., Interscience, New York, 1960, pp. 347–424. 5 Wetzel, D. L., Anal. Chem., 1983, 55, 1165A. 6 McClure, W. F., Anal. Chem.,1994, 66, 43A. 7 Osborne, B. G., and Fearn, T., Near Infrared Spectroscopy in Food Analysis, Longman Scientific & Technical, Harlow, Essex, 1986. 8 Williams, P. C., and Norris, K. H., in Near-Infrared Technology in the Agricultural and Food Industries, ed. Williams, P. C., and Norris, K. H., American Association of Cereal Chemists, St. Paul, MN, 1987, ch. 15, pp. 241–246. 9 Ghosh, S., and Rodgers, J., in Handbook of Near-Infrared Analysis, Practical Spectroscopy Series, vol.13, ed. Burns, D. A., and Ciurczak, E. W., Marcel Dekker, New York, 1992, vol. 1, ch. 18, pp. 495–526. 10 Hammersley, M. J., in Handbook of Near-infrared Analysis, Practical Spectroscopy Series, ed. Burns, D. A., and Ciurczak, E. W., Marcel Dekker Inc., New York, 1992, vol. 13, ch. 17, pp. 475–494. 11 Kradjel, C., and McDermott, L., in Handbook of Near-Infrared Analysis, Practical Spectroscopy Series, ed. Burns, D. A., and Ciurczak, E. W., Marcel Dekker, New York, 1992, vol. 13, ch. 21, pp. 565–608. 12 Bunding Lee, K. A., Appl. Spectrosc. Rev., 1993, 28, 231. 13 Workman, J., Jr., J. Near Infrared Spectrosc., 1993, 1, 221. 14 Workman J., Jr., Appl. Spectrosc. Rev., 1996, 31, 251. 15 Weyer, L. G., Appl. Spectrosc. Rev., 1985, 21, 1. 16 Lipp, E. D., Appl. Spectrosc. Rev., 1992, 27, 385. 17 Mark, H., Anal. Chim. Acta, 1989, 223, 75. 18 Michel, I., NIR News, 1995, 6(6), 4. 19 Hermiller, J., NIR News, 1996, 7(2), 4. 20 Hermiller, J., NIR News, 1996, 7(1), 4. 21 Kraemer, E. G., and Lodder, R. A., Spectroscopy, 1996, 11(7), 24. 22 Kraemer, E.G. and Lodder, R. A., Spectroscopy, 1996, 11(8), 17. 23 Ciurczak, E. W., in Handbook of Near-Infrared Analysis, Practical Spectroscopy Series, ed. Burns, D. A., and Ciurczak, E. W., Marcel Dekker, New York, 1992, vol. 13, ch. 20, pp. 549–564. 24 Drennen, J. K., and Lodder, R. A., in Advances in Near-Infrared Measurements, ed. Patonay, G., Jai Press, Greenwich, CT, 1993, vol. 1, pp. 93–112. 25 Stark, E., Luchter, K., and Margoshes, M., Appl. Spectrosc. Rev., 1986, 22, 335. 26 Ciurczak, E. W., Appl. Spectrosc. Rev., 1987, 23, 147. 27 Drennen, J. K., Kraemer, E. G., and Lodder, R. A., Crit. Rev. Anal. Chem., 1991, 22, 443. 28 Martin, K. A., Appl. Spectrosc. Rev., 1992, 27, 325. 29 Ciurczak, E. W., and Drennen, J. K., Spectroscopy, 1992, 7(6), 12. 30 Corti, P., Dreassi, E., and Lonardi, S., Il Farmaco, 1993, 48, 3. 31 MacDonald, B. F., and Prebble, K. A., J. Pharm. Biomed. Anal., 1993, 11, 1077. 32 Morisseau, K. M., and Rhodes, C. T., Drug Dev. Ind. Pharm., 1995, 21, 1071. 33 Kirsch, J. D., and Drennen, J. K., Appl. Spectrosc. Rev., 1995, 30, 139. 34 McClure, W. F., in Near-Infrared Technology in the Agricultural and Food Industries, ed. Williams, P., and Norris, K. H., American Association of Cereal Chemists, St. Paul, MN, 1987, ch. 5, pp. 89–105. 35 Williams, P. C., in Near-infrared Technology in the Agricultural and Food Industries, ed. Williams, P., and Norris, K. H., American Association of Cereal Chemists, St. Paul, MN, 1987, ch. 6, pp. 107–142. 36 Kubelka, P., and Munk, F., Z. Tech. Phys., 1931, 12, 593. 37 Lodder, R. A., and Hieftje, G. M., Appl. Spectrosc., 1988, 42, 556. 38 Springsteen, A., and Ricker, T., NIR News, 1996, 7(4), 6. 39 Springsteen, A., and Ricker, T., NIR News, 1996, 7(5), 12. 40 Osborne, B. G., Analyst, 1988, 113, 263. 41 Griffiths, P. J., J. Near Infrared Spectrosc., 1995, 3, 60. 42 Dahm, D. J., and Dahm, K. D., J. Near Infrared Spectrosc., 1995, 3, 53. 43 Olinger, J. M., and Griffiths, P.R., in Handbook of Near-Infrared Analysis, ed. Burns, D. A., and Ciurczak, E. W., Marcel Dekker, New York, 1992, ch. 3, pp. 13–35. 44 Aucott, L. S., Garthwaite, P. H., and Buckland, S. T., Analyst, 1988, 113, 1849. 45 Salamin, P. A., Cornelis, Y., and Bartels, H., Chemom. Intell. Lab. Syst., 1988, 3, 329. 46 Wu, W., Walczak, B., Massart, D. L., Prebble, K. A., and Last, I. R., Anal. Chim. Acta, 1995, 315, 243. 47 Hailey, P. A., Doherty, P., Tapsell, P., Oliver, T., and Aldridge, P.K., J. Pharm. Biomed. Anal., 1996, 14, 551. 48 Mobley, P. R., Kowalski, B. R., Workman, J. J., Jr., and Bro, R., Appl. Spectrosc. Rev., 1996, 31, 347. 49 The Unscrambler User’s Guide, Computer Aided Modelling A/S, Trondheim, Norway, 1993. 50 Downey, G., Robert, P., and Bertrand, D., Anal. Proc., 1992, 29, 8. 51 McClure, W. F., NIR News, 1993, 4(6), 12. 52 McClure, W. F., NIR News, 1994, 5(1), 12. 53 McClure, W. F., NIR News, 1994, 5(4), 14. 54 Isaksson, T., and Næs, T., Appl.Spectrosc., 1988, 42, 1273. 55 Geladi, P., MacDougall, D., and Martens, H., Appl. Spectrosc., 1985, 39, 491. 56 Næs, T., Isaksson, T., and Kowalski, B., Anal. Chem., 1990, 62, 664. 57 Isaksson, T., and Kowalski, B., Appl. Spectrosc., 1993, 47, 702. 58 Martens, H., and Stark, E., J. Pharm. Biomed. Anal., 1991, 9, 625. 59 Karstang, T. V., and Manne, R., Chemom. Intell. Lab. Syst., 1992, 14, 165. 60 Isaksson, T., Wang, Z., and Kowalski, B., J. Near Infrared Spectrosc., 1993, 1, 85. 61 Barnes, R. J., Dhanoa, M. S., and Lister, S. J., Appl. Spectrosc., 1989, 43, 772. 62 Dhanoa, M. S., Lister, S. J., and Barnes, R. J., Appl. Spectrosc., 1995, 49, 765. 63 Dhanoa, M. S., Lister, S. J., Sanderson, R., and Barnes, R. J., J. Near Infrared Spectrosc., 1994, 2, 43. 64 Jacobsson, S. P., Carlsson, M., Jönsson, U., and Nilsson, G., J. Pharm. Biomed. Anal., 1995, 13, 415. 65 Blanco, M., Coello, J., Iturriaga, H., Maspoch, S., and de la Pezuela, C., Appl.Spectrosc., 1997, 51, 240. 66 Kaye, W., Spectrochim. Acta, 1954, 6, 257. 67 Sinsheimer, J. E., and Keuhnelian, A. M., J. Pharm. Sci., 1966, 55, 1240. 68 Massart, D. L., Vandegiste, B. G. M., Deming, S. N., Michotte, Y., and Kaufman, L., Chemometrics: A Textbook, Elsevier, Amsterdam, 1988. 69 Workman, J. J., Jr., Mobley, P. R., Kowalski, B. R., and Bro, R., Appl. Spectrosc. Rev., 1996, 31, 73. 70 Blanco, M., Coello, J., Iturriaga, H., Maspoch, S., de la Pezuela, C., and Russo, E., Anal.Chim. Acta, 1994, 298, 183. 71 Brik, H., and van der Vlies, C., Pharmaeuropa, 1991, 3(1), 3. 72 Plugge, W., and van der Vlies, C., J. Pharm. Biomed. Anal., 1993, 11, 435. 73 Galante, L. J., Brinkley, M. A., Drennen, J. K., and Lodder, R. A., Anal. Chem., 1990, 62, 2514. 74 Perstorp Analytical, Technical Note 0554. 75 Plugge, W., and van der Vlies, C., J. Pharm. Biomed. Anal., 1992, 10, 797. 76 González, F., and Pous, R., J. Pharm. Biomed. Anal., 1995, 13, 419. 77 Mark, H.L., and Tunnell, D., Anal. Chem., 1985, 57, 1449. 78 Mark, H. L., Anal. Chem., 1987, 59, 790. 79 Tunnell, D. A., Anal. Proc., 1990, 27, 59. 80 Corti, P., Dreassi, E., Corbini, G., Lonardi, S., and Gravina, S., Analusis, 1990, 18, 112. 81 Corti, P., Dreassi, E., Corbini, G., Montecchi, L., and Paggi, J., Analusis, 1990, 18, 117. 148R Analyst, August 1998, Vol. 12382 Corti, P., Dreassi, E., Ceramelli, G., Lonardi, S., Viviani, R., and Gravina, S., Analusis, 1991, 19, 198. 83 Corti, P., Savini, L., Dreassi, E., Petriconi, S., Genga, R., Montecchi, L., and Lonardi S., Proc. Cont. Qual., 1992, 2, 131. 84 Corti, P., Savini, L., Dreassi, E., Ceramelli, G., Montecchi, L., and Lonardi, S., Pharm. Acta Helv., 1992, 67, 57. 85 Dreassi, E., Ceramelli, G., Corti, P., Lonardi, S., and Perruccio, P. L., Analyst, 1995, 120, 1005. 86 Dreassi, E., Ceramelli, G., Savini, L., Corti, P., Perruccio, P. L., and Lonardi, S., Analyst, 1995, 120, 319. 87 Dreassi, E., Ceramelli, G., Corti, P., Perruccio, P.L., and Lonardi, S., Analyst, 1996, 121, 219. 88 van der Vlies, C., Kaffka, K. J., and Plugge, W., Pharm. Technol. Eur., 1995, 7(4), 43. 89 Plugge, W., and van der Vlies, C., J. Pharm. Biomed. Anal., 1996, 14, 891. 90 Martens, H., and Næs, T., Multivariate Calibration, Wiley, New York, 1991. 91 Lo, S., and Brown, C. W., Appl. Spectrosc., 1992, 46, 790. 92 Shah, N. K., and Gemperline, P. J., Trends Anal. Chem., 1989, 8, 357. 93 Wold, S., and Sjostrom, M., in Chemometrics: Theory and Applications, ed.Kowalski, B. R., American Chemical Society, Washington DC, 1977. 94 Gemperline, P. J., Webber, L. D., and Lox, I. O., Anal. Chem., 1989, 61, 138. 95 Gemperline, P. J., and Boyer, N. R., Anal. Chem., 1995, 67, 160. 96 Shah, N. K., and Gemperline, P. J., Anal. Chem., 1990, 62, 465. 97 Dempster, M. A., Jones, J. A., Last, I. R., MacDonald, B. F., and Prebble, K. A., J. Pharm. Biomed. Anal., 1993, 11, 1087. 98 Dempster, M. A., MacDonald, B.F., Gemperline, P. J., and Boyer, N. R., Anal. Chim. Acta, 1995, 310, 43. 99 Ciurczak, E. W., and Maldacker, T. A., Spectroscopy, 1986, 1(1), 36. 100 Honigs, D. E., Hieftje, G. M., and Hirschfeld, T., Appl. Spectrosc., 1984, 38, 317. 101 Lodder, R. A., and Hieftje, G. M., Appl. Spectrosc., 1988, 42, 1351. 102 Lodder, R. A., and Hieftje, G. M., Appl. Spectrosc., 1988, 42, 1500. 103 Lodder, R. A., Selby, M., and Hieftje, G. M., Anal. Chem., 1987, 59, 1921. 104 Ciurczak, E. W., Pharm. Technol., 1991, 15, 140. 105 Wargo, D. J., and Drennen, J. K., J. Pharm. Biomed. Anal., 1996, 14, 1415. 106 Sekulic, S. S., Ward, H. W., Brannegan, D. R., Stanley, E. D., Evans, C. L., Sciavolino, S. T., Hailey, P. A., and Aldridge, P. K., Anal. Chem., 1996, 68, 509. 107 Gimet, R., and Luong, A. T., J. Pharm. Biomed. Anal., 1987, 5, 205. 108 Aldridge, P. K., Evans, C. L., Ward, H. W., Colgan, S. T., Boyer, N., and Gemperline, P. J., Anal. Chem., 1996, 68, 997. 109 Norris, T., Aldridge, P. K., and Sekulic, S. S., Analyst, 1997, 122, 549. 110 Buchanan, B. R., Ciurczak, E. W., Grunke, A. Q., and Honigs, D. E., Spectroscopy, 1988, 3, 54. 111 MacDonald, S. A., and Hieftje, G. M., Appl. Spectrosc., 1996, 50, 1161. 112 Halsey, S. A., Technical Note, NIRSystems, Silver Spring, MD, 1994. 113 Corti, P., Dreassi, E., Murratzu, C., Corbini, G., Ballerini, L., and Gravina, S., Pharm. Acta Helv., 1989, 64, 140. 114 Jouan-Rimbaud, D., Khots, M.S., Massart, D. L., Last, I. R., and Prebble, K. A., Anal. Chim. Acta, 1995, 315, 257. 115 Jouan-Rimbaud, D., Walczak, B., Massart, D. L., Last, I. R., and Prebble, K. A., Anal. Chim. Acta, 1995, 304, 285. 116 Blanco, M., Coello, J., Iturriaga, H., Maspoch, S., and de la Pezuela, C., Anal. Chim. Acta, 1996, 333, 147. 117 Blanco, M., Coello, J., Iturriaga, H., Maspoch, S., and de la Pezuela, C., Talanta, 1993, 40, 1671. 118 Blanco, M., Coello, J., Iturriaga, H., Maspoch, S., and de la Pezuela, C., Analyst, 1997, 122, 761. 119 Buchanan, B., and Honigs, D., Trends Anal. Chem., 1986, 5, 154. 120 Næs, T., and Isaksson, T., NIR News, 1994, 5(6), 16. 121 Honigs, D. E., Hieftje, G. M., Mark, H. L., and Hirschfeld, T. B., Anal. Chem., 1985, 57, 2299. 122 Næs, T., J. Chemom., 1987, 1, 121. 123 Isaksson, T., and Næs, T., NIR News, 1995, 6(1), 13. 124 Puchwein, G., Anal. Chem., 1988, 60, 569. 125 Mark, H., Anal. Chem., 1987, 59, 790. 126 Ferré, J., and Rius, F. X., Anal.Chem., 1996, 68, 1565. 127 Davies, T., Spectroscopy, 1996, 8(4), 27. 128 Fearn, T., Research report No. 105, University College, London, 1992. 129 Near Infrared Spectral Analysis Software. User’s Guide, NIRSystems, Silver Spring, MD, 1990. 130 Thomas, E. V., and Haaland, D. M., Anal. Chem., 1990, 62, 1091. 131 Draper, N., and Smith, H., Applied Regression Analysis, Wiley, New York, 2nd edn., 1981. 132 Honigs, D. E., Hieftje, G. M., and Hirschfeld, T. B., Appl. Spectrosc., 1983, 38, 317. 133 Cowe, I. A., McNicol, J. W., and Cuthbertson, D. C., Analyst, 1985, 110, 1233. 134 Næs, T., and Isaksson, T., NIR News, 1993,4(3), 14. 135 Miller, C. E., NIR News, 1993, 4(6), 3. 136 Long, R. L., Gregoriov, V. G., and Gemperline, P. J., Anal. Chem., 1990, 62, 1791. 137 Gemperline, P. J., Long, J. R., and Gregoriov, V. G., Anal. Chem., 1991, 63, 2313. 138 Næs, T., Kvaal, K., Isaksson, T., and Miller, C., J. Near Infrared Spectrosc., 1993, 1, 1. 139 Walczak, B., and Wegscheimer, W., Anal.Chim. Acta, 1993, 283, 508. 140 Liu, Y., Upadhyaya, B. R., and Naghedolfeizi, M., Appl. Spectrosc., 1993, 47, 12. 141 Næs, T., Isaksson, T., and Kowalski, B. R., Anal. Chem., 1990, 62, 664. 142 Næs, T., and Isaksson, T., Appl. Spectrosc., 1992, 46, 34. 143 Friedman, J. H., and Stuetzle, W., J. Am. Stat. Assoc., 1981, 76, 817. 144 Beebe, K. R., and Kowalski, B. R., Anal. Chem., 1988, 60, 2273. 145 Vogt, N. B., Chemom. Intell. Lab. Syst., 1989, 7, 119. 146 Wold, S., Wold, N. K., and Skagerberg, B., Chemom. Int. Lab. Syst., 1989, 7, 53. 147 Ciurczak, E. W., Torlini, R. P., and Demkowicz, M. P., Spectroscopy, 1986, 1(7), 36. 148 Blanco, M., Coello, J., Iturriaga, H., Maspoch, S., González, F., and Pous, R., in Near Infra-red Spectroscopy (Bridging Gap between Data Analysis and NIR Applications), ed. Hildrum, K. I., Isaksson, T., Næs, T., and Tandberg, A., Ellis Horwood, Chichester, 1992, ch. 64, pp. 401–406. 149 Ilari, J. L., Martens, H.and Isaksson, T., Appl. Spectrosc., 1988, 42, 722. 150 Kirsch, J. D., and Drennen, J. K., J. Pharm. Biomed. Anal., 1995, 13, 1273. 151 Langkilde, F. W., and Svantesson, A., J. Pharm. Biomed. Anal., 1995, 13, 409. 152 Lin, L., and Brown, C. W., Appl. Spectrosc., 1992, 46, 1809. 153 Delwiche, S. R., Norris, K. H., and Pitt, R. E., Appl. Spectrosc., 1992, 46, 782. 154 Lin, J., and Brown, C. W., Appl. Spectrosc., 1993, 47, 62. 155 Sinsheimer, J. E., and Poswalk, N. M., J. Pharm. Sci., 1968, 57, 2007. 156 Issa, R. M., El-Marsafy, K. M., and Gohar, M. M., Ann. Quím., 1988, 84, 312. 157 Böhme, W., Liekmeier, W., Horn, K., and Wilhelm, C., Labor Praxis, 1989, January/February, 44. 158 Warren, R. J., Zarembo, J. E., Chong, C. W., and Robinson, M. J., J. Pharm. Sci., 1970, 59, 109. 159 Vomhof, D. W., and Thomas, J. H., Anal. Chem., 1970, 42, 1230. 160 Ludvik, J., Hilgard, S., and Volke, J., Analyst, 1988, 113, 1729. 161 Garrigues, S., Gallignani, M., and de la Guardia, M., Anal. Chim. Acta, 1993, 281, 259. 162 Last, I. R., and Prebble, K. A., J. Pharm. Biomed. Anal., 1993, 11, 1071. 163 Lonardi, S., Viviani, R., Mosconi, L., Bernuzzi, M., Corti, P., Dreassi, E., Murratzu, C., and Corbini, G., J. Pharm. Biomed. Anal., 1989, 7, 303. 164 Blanco, M., Coello, J., Iturriaga, H., Maspoch, S., and Rovira, E., J. Pharm. Biomed. Anal., 1997, 16, 255. 165 Corti, P., Dreassi, E., Corbini, G., Lonardi, S., Viviani, R., Mosconi, L., and Bernuzzi, M., Pharm. Acta Helv., 1990, 65, 28. Analyst, August 1998, Vol. 123 149R166 Dubois, P., Martínez, J. R., and Levillain, P., Analyst, 1987, 112, 1675. 167 Meurens, M., Dupuy, N., Mirouze, F. D., Legrand, P., and Huvenne, P. J., in Near Infra-red Spectroscopy (Bridging Gap between Data Analysis and NIR Applications), ed. Hildrum, K. I., Isaksson, T., Næs, T., and Tandberg, A., Ellis Horwood, Chichester, 1992, ch. 7, pp. 47–49. 168 Ciurczak, E. W., and Torlini, R. P., Spectroscopy, 1987, 2(3), 41. 169 Zappala, A. F., and Post, A., J. Pharm. Sci., 1977, 66, 292. 170 Osborne, B. G., Analyst, 1987, 112, 313. 171 Chasseur, J. C., Chim. Oggi, 1987, 6, 21. 172 Gottfries, J., Depui, H., Fransson, M., Jogeneelen, M., Josefson, M., Langkilde, F. W., and Witte, D. T., J. Pharm. Biomed. Anal., 1996, 14, 1495. 173 Drennen, J. K., and Lodder, R. A., J. Pharm. Sci., 1990, 79, 622. 174 Jönsson, U., Eriksson, E. H., and Sellberg, B., presented at the International Conference of Pharmaceutical Applications of NIR Spectroscopy, Stockholm, Sweden, 1996. 175 Jensen, R., Peuchant, E., Castagne, I., Boirac, A. M., and Roux, G., Spectrosc. Int. J., 1988, 6, 63. 176 Han, S. M., and Faulkner, P. G., J. Pharm. Biomed. Anal., 1996, 14, 1681. 177 Dean, T., and Isaksson, T., NIR News, 1993, 4(2), 8. 178 Dean, T., and Isaksson, T., NIR News, 1993, 4(4), 14. 179 Shenk, J., and Westerhaus, M. O., NIR News, 1993, 4(5), 13. 180 Bouveresse, E., and Massart, D. L., Vib. Spectrosc., 1996, 11, 3. 181 Shenk, J., and Westerhaus, M. O., US Pat. 4 866 644, 1989. 182 Shenk, J., and Westerhaus, M. O., Crop Sci., 1991, 31, 1694. 183 Dardenne, P., Biston, R., and Simmaeve, G., in Near Infra-red Spectroscopy (Bridging Gap between Data Analysis and NIR Applications), ed. Hildrum, K. I., Isaksson, T., Næs, T., and Tandberg, A., Ellis Horwood, Chichester, 1992, ch. 72, pp. 453–458. 184 Bouveresse, E., Massart, D. L., and Dardenne, P., Anal. Chim. Acta, 1994, 297, 405. 185 Bouveresse, E., and Massart, D. L., Anal. Chem., 1995, 67, 1381. 186 Mark, H., and Workman, J. J., Jr., Spectroscopy, 1988, 3(11), 28. 187 Wang, Y., Veltkamp, D. J., and Kowalski, B. R., Anal. Chem., 1991, 63, 2750. 188 Wang, Y., and Kowalski, B. R., Anal. Chem., 1993, 65, 1301. 189 Wang, Y., and Kowalski, B. R., Appl. Spectrosc., 1992, 46, 764. 190 Bouveresse, E., Hartmann, C., Massart, D. L., Last, I. R., and Prebble, K. A., Anal. Chem., 1996, 68, 982. 191 Jones, J. A., Last, I. R., MacDonald, B. F., and Prebble, K. A., J. Pharm. Biomed. Anal., 1993, 11, 1227. 192 Forina, M., Drava, G., Armanino, C., Boggia, R., Lanteri, S., Leardi, R., Corti, P., Conti, P., Giangiacomo, R., Galliena, C., Bigoni, R., Quartari, I., Serra, C., Ferri, D., Leoni, O., and Lazzeri, L., Chemom. Intell. Lab. Syst., 1995, 27, 189. 193 Forina, M., Armanino, C., and Giangiacomo, R., in Near Infra-red Spectroscopy (Bridging Gap between Data Analysis and NIR Applications), ed. Hildrum, K. I., Isaksson, T., Næs, T. and Tandberg, A., Ellis Horwood, Chichester, 1992, ch. 14, pp. 91–96. 194 Blank, T. B., Sun, S. T., Brown, S. D., and Monfre, S. L., Anal. Chem., 1995, 68, 2987. 195 Reeve, J. B., III, Appl. Spectrosc., 1995, 49, 181. 196 Reeve, J. B., III, Appl. Spectrosc., 1995, 49, 295. 197 Reeve, J. B., III, J. AOAC Int., 1996, 76,741. 198 Reeve, J. B., III, Appl. Spectrosc., 1996, 50, 154. 199 Lin, J., and Brown, C. W., Appl. Spectrosc., 1993, 47, 239. 200 Lin, J., and Brown, C. W., Anal. Chem., 1993, 65, 287. 201 Lin, J., and Brown, C. W., J. Near Infrared Spectrosc., 1993, 1, 109. 202 Application note A3-987, Guided Wave. 203 Martens, H., Næs, T., and Bjorsvik, H. R., Wave Guide, 1988, 1(1), 4. 204 Mackison, R., Brinkworth, S. J., Belchamber, R. M., Aries, R. E., Cutler, D. J., Deeley, C., and Mould, H. M., Appl. Spectrosc., 1992, 46, 1020. Paper 8/02531B Received April 2, 1998 Accepted June 9, 1998 150R Analyst, August 1998, Vol. 123

ISSN:0003-2654    

DOI:10.1039/a802531b

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

AGANTG: a Microsoft EXCEL 5.0–Visual Basic routine for the analysis of dose–response data Peter J. Zielinski*a, and Steve Bucknerb a Abbott Laboratories, D9P9, AP32-3, Abbott Park, IL 60064-3500, USA. E-mail: zielinp@megsinet.net b Abbott Laboratories, D47C, AP9, Abbott Park, IL 60064-3500, USA A Microsoft EXCEL 5.0 program was developed to evaluate data from biochemical and functional bioassays, an important step in drug discovery. The program accommodates both agonist and antagonist data.The program, written entirely in Visual Basic, is compatible with both Macintosh and PC platforms. Data are conveniently entered into a worksheet following only a few simple rules. The program performs complex data analysis and outputs calculated and graphic results to EXCEL worksheets. A set-up routine with a convenient dialog box offers the user controls regarding data analysis and results formats. After determining if the data are from an agonist or antagonist assay, the program automatically performs the analysis and outputs results in the proper format.Calculations support Schild analysis for antagonists. An agonist and antagonist were analyzed to illustrate program usage and results generated by the analsis. EXCEL–Visual Basic is a useful and convenient tool for evaluating bioassay data. Data entry is greatly simplified and custom reports can be generated with relative ease. Data are stored in a format that allows for easy editing re-analysis.Keywords: Dose–response data analysis; AGANTG; Microsoft EXCEL 5.0; Visual Basic; bioassay Rapidly improving hardware and software for desktop computers continue to open doors to what was once the exclusive domain of large computers. Transferring applications to smaller platforms has obvious benefits for the user in terms of cost and convenience, provided that performance is acceptable. Competition among desktop software manufacturers also provides choices to the user.In particular, a number of spreadsheet programs share a basic look, feel and wide range of functionality so that switching between spreadsheet products entails relatively little difficulty. To be sure, significant differences in capabilities exist between software products, but familiar spreadsheet interfaces flatten the so-called learning curve for users. Consequently, software developers who take advantage of spreadsheet interfaces expect a higher level of acceptance from users. However, developers must also be able to perform complex and diverse tasks in their applications.For the developer of such applications, a product that combines the spreadsheet user interface with a powerful programming language is highly desirable. One popular product that has successfully combined these features is Microsoft EXCEL 5.0. This version of EXCEL has inherited the many advantages of EXCEL 4.0 cited by others,1 especially the familiar feel, ease of use, flexibility and wide range of features.Moreover, EXCEL 5.0 combines extensive spreadsheet functionality with a rich programming language, Visual Basic (VB). This superior combination is useful for many applications and is well suited for the important tasks of evaluating dose–response drug binding and in vitro functional data. Briefly, many drugs work in a lock and key fashion. When tissue containing receptors, the locks, are exposed to a solution containing active drug molecules, the keys, the receptors typically change their shape upon binding with the drug.This change causes a chain of other physiological changes to occur, determining what effects the drug ultimately have. Drugs are categorized as either agonists or antagonists. When an agonist binds, it initiates some physiological reaction. When antagonists bind, they block the effects of agonists, thereby reducing their activity. A common test for agonist activity is to immerse receptorbearing tissue in a solution containing some known concentration of the drug and then to measure how the shape of the tissue changes, for example, by monitoring changes in tension that a tissue suspended between two wires exerts before and after addition of the drug.The change in tension is a direct measure of drug activity. A more complicated scheme is required for antagonists, but the essential relationship between drug activity and observed response remains. Here, the tissue is exposed to solutions having various concentrations of a reference agonist besides the antagonist being tested. The activity of the antagonist is determined by how well it prevents the agonist from performing its function.For example, a more active antagonist may mean that more agonist is required to achieve a similar tension. By repeating this measurement for a series of drug concentrations, a dose–response curve may be constructed and the activity of either an agonist or antagonist characterized by mathematically analyzing the curve.2–4 An EXCEL 5.0 VB computer program for analyzing dose– response data is described in this paper.EXCEL 5.0 worksheets are used for data input and the output of results. Data are input through a worksheet following a few simple rules. A dialog box provides control over data analysis and reports. The dialog box contains check boxes and entry fields that may be modified at any time before or after data analysis. In this way, the same data can be analyzed in different ways.Computed results and graphs are produced from the data through VB modules that run on either PC and Macintosh platforms. Program usage Starting the program and entering data The program is organized into a number of subroutines, the main routines being AGANTG (AGonist–ANTaGonist analysis) and AGANTG_SetUp. Table 1 gives a short description of the most important of these. AGANTG and AGANTG_SetUp are accessed by the user as EXCEL add-ins under the Tools menu.Like any other add-in, these add-ins may be configured to load automatically upon starting EXCEL or may be loaded at some later time using the Tools menu. When the user chooses AGANTG_SetUp from the Tools menu, the Data Input worksheet of the currently open workbook is automatically formatted for data entry according to previously stored parameters. Subsequently, the dialog box illustrated in Fig. 1 displays a series of choices to the user.With Analyst, August 1998, Vol. 123 (1661–1668) 1661the dialog box, the user may override previous format parameters and save new ones. There are currently 12 such parameters. Dialog box entries ‘Last Input Row’ and ‘Last Input Column’ define the range of worksheet cells containing input data. In Fig. 1, these entries are 20 and 10, respectively. The program color highlights these cells on the Data Input worksheet, illustrated in Fig. 2, to indicate where data to be processed are located.Data outside this region and empty columns within the region are ignored. The field labeled ‘Maximum Iterations’ limits attempts by the non-linear curve fitting routine in refining parameter estimates of dose–response curves. Another check box directs the AGANTG routine to perform pA2 calculations if the data set is an antagonist run or to perform pD2 calculations in the case of agonist data. The user is not required to specify whether the data are of type agonist or antagonist as the program is able to distinguish these data types, as will be explained below.The user may choose that standard errors rather than standard deviations be displayed on reports and chart error bars. Charts are generated by checking the appropriate boxes. Another check box permits the automatic printing of results. Finally, one box labeled ‘CHOOSE AGANTG DEFAULT SETTINGS’ will automatically set default choices for an AGANTG analysis. Specific settings, such as the ‘Last Input Column’ can be overridden by clearing the AGANTG default settings box and entering the desired number of columns.Numerical entries are made with either spinner buttons or by direct entry into the appropriate field. In this way, the AGANTG routine may be configured for purposes other than dose–response curve analysis. For example, it may be used as a general non-linear curve fitting routine. Data may be entered on to the first worksheet either before or after invoking the SetUp routine.The SetUp routine does not erase this data and may be called from the Tools menu. The role of the SetUp routine is to highlight input data and tell AGANTG what tasks to perform. However, AGANTG expects the data to conform to a few simple rules. While explaining these rules, it will be helpful to refer to Figs. 2 and 3. The top row of the Data Input sheet is reserved for what shall be referred to as ‘titles’. A title identifies information stored in the cells below it, in the same column within the data region selected.The leftmost column of data is assumed to contain values of the independent, or x, variable. If we are analyzing dose–response curves, the leftmost column would typically contain agonist concentrations, as illustrated in Fig. 2. The title of this first column, shown as [ ] in this example, is optional and has no effect on how the data are analyzed. The titles of the remaining columns with the data input region do have an impact on how data are analyzed, as explained below.The data themselves need not be in ascending or descending order as they will be sorted by the program. Empty rows or columns within the data entry region are ignored. The next non-empty column of data is assumed to be that of a vehicle, a control containing only an agonist. Data are Table 1 Important routines and their description Routine Description AGANTG Main routine for data analysis stage.Directly or indirectly calls nearly all subroutines GET_XDATA Retrieves concentrations from Input Data and puts into an array for processing GET_YDATA Retrieves responses from Input Data and puts into an array for processing Probit Performs probit analysis TransGraphData Collates data for graphing and statistical analysis. Calls many subroutines to complete its ‘messy’ tasks AGANTG_SetUp Main routine for set-up stage. Displays the Input Dialog box and stores set-up parameters CallMrq Entry point for non-linear analysis.Calls several numerical routines Estimated_C Performs initial estimates for b and c parameters used in 4PLC routine Fig. 1 1662 Analyst, August 1998, Vol. 123identified as a control when its title consists of a single string of alphanumeric characters, such as ‘veh’ in Fig. 2. The title of the next column, ‘ant 123 3e 2 9’, has two parts to it. The first part, ‘ant 123’ , is an antagonist identifier. The program knows that this sample is an antagonist because the title has a second, numerical part, ‘3e 2 9’.If an antagonist is found in any title Fig. 2 Fig. 3 Analyst, August 1998, Vol. 123 1663within the data input region, the program assumes that the data represent an antagonist experiment. Otherwise, an agonist experiment is assumed. Fig. 2 shows the data set for an antagonist and Fig. 3 illustrates an agonist data set. Information to be processed may be entered into one other area besides the data input region.The area is in the second column just below the data input area, as illustrated in Fig. 3. Headings in the first column indicate information that may be entered to aid in assay identification. Entry of report time and date information by the analyst is not required as these cells are filled automatically. Agonist, tissue, sample and date information are copied on to all charts by the program. The Agonist Test cell, which is next to that labeled ‘Agonist Test?’, is filled by the analyst to indicate whether or not the assay is a test for an agonist.This cell is used later as a check to help determine if a mistake has been made entering column titles. All of these assay identification fields and headings are copied to all worksheets containing final results. An illustration of one of the graphs output for a typical antagonist assay is given in Fig. 4. Treatment of data Data validation and analysis When a workbook is opened and the AGANTG routine started by selecting RUN AGANTG from the Tools menu, the program immediately erases any previously existing charts and worksheets in the workbook, except for the Data Input worksheet.The program then tests the data for errors. The first column in the data region of Data Input, the independent x values, must contain at least two distinct values for any analysis to continue. Titles without correct formats are highlighted and the program is halted. The same is true if characters are accidentally placed into regions reserved for numeric data.Additional data checks insure that the response data is valid. For example, any response column (the dependent y values) not associated with at least two distinct agonist concentrations, or not varying with the x values, will be ignored after appropriate alerts are issued since further calculations on these data are pointless. In any case, a cleaned up version of the data is stored on the worksheet ‘Processed Data’ and, if it was found acceptable, used for all future calculations.Regression is then attempted, as described next. When a y column is associated with at least four distinct x values, the x–y data are passed to a non-linear regression routine that finds parameters a, b, c and d according to the four parameter logistic model (4PLC): y = d + (a 2 d)/[1 + (x/c)b] (1) where y represents the observed response and x its associated dose. Parameter a is the response at x = 0, b a ‘slope factor’describing the steepness of the curve, c the EC50 and d the response for infinite dose.For a drug, the EC50 corresponds to the concentration of the drug necessary to produce 50% of the maximum possible response. Eqn. (1) is a non-linear model equation that describes sigmoidal dose–response curve data.4–6 To find the four parameters for a particular dose–response curve, the response values in a given y column are regressed against the dose values in the x column using eqn.(1). Non-linear regression requires initial estimates of the four parameters described above. AGANTG automatically performs these estimations without user intervention and then attempts regression on each valid x–y data set using eqn. (1). Our method of choice for this purpose is the Levenberg–Marquardt algorithm.7,8 Once the calculation is complete, the parameters returned are tested for reasonableness. If the parameters fail any Fig. 4 1664 Analyst, August 1998, Vol. 123of these tests, a linear regression is attempted using the probit model eqn. (2) rather than the 4PLC model: G21 (y/ymax) = m + b logx (2) where G21 is the inverse Gaussian cumulative probability function. For a given probability value p in the interval [0,1], G21 (p) returns the value along the x-axis that yields that probability. Thus, G21 (0) = 2 ° , G1 (0.5) = 0 and G21( 1) = ° . Once regression is complete, the EC50 is given by EC50 = 102m/b (3) A particular dose–response curve is abandoned if both the 4PLC and probit regression are unsuccessful.The parameters found, predicted y values and residues are stored in a worksheet labeled ‘Results’ and may be viewed, if desired. Parameter values are later retrieved from this worksheet for further calculations and reports. When all regressions have been completed and the results stored, the program determines whether the data represent an agonist or antagonist assay by checking for the presence of a second numerical component in the title, as described above.If this determination contradicts what is entered by the Agonist Test cell, a warning is issued and the analyst given an opportunity to abort the analysis. If the analysis is not aborted, the entry in the Agonist Test cell is ignored. Subsequent calculations and reports depend on the nature of the assay as decided based upon data titles. For antagonist, the objective is the calculation of pA2.According to Schild, ‘pAx is defined as the negative logarithm to base 10 of the molar concentration of an antagonistic drug which will reduce the effect of a multiple dose (x) of an active drug to that of a single dose’. This calculation involves the quantity known as the dose-ratio (DR), given by3 DR = EC50Antagonist/EC50Control (4a) where DR is the EC50 of the agonist in the presence and absence of antagonist. The quantity DR 2 1 is given by DR 21 = EC50Antagonist/EC50Control 21 (4b) = (EC50Antagonist 2 EC50Control)/EC50Control The pA2 is computed by performing a linear regression using the model eqn.(5) and then using the fitted slope, b, and intercept parameter, m, in eqn. (6): log(DR 2 1) = m + blogx (5) pA2 = 2 m/b (6) When the data contain more than one level of antagonist concentration, AGANTG produces a plot of the data transformed according to eqn. (5) together with the fitted line. If only one antagonist concentration is used, a constrained regression is performed in which the slope in eqn.(5) is held to the value 19; m is then found and again used in eqn. (6), but no pA2 chart is generated. Examples of default chart type that AGANTG generates are illustrated in Fig. 4. The objective is different for agonist analysis. The first measure of agonist potency is the pD2 value. This quantity is simply pD2 = 2 log(EC50) (7) For an agonist, the Index is the measure of potency. The Index is defined as the ratio of the effective concentration of the agonist control divided by the effective concentration of the sample agonist: Index = EC50control agonist/EC50sample agonist (8) A large Index indicates a potent sample agonist relative to the control agonist.Relative efficacy, the percentage of a sample’s maximum response relative to that of the control, is reported in the Summary Report as illustrated in Fig. 5. For agonists, key individual results are illustrated in Table 2. Fig. 5 Analyst, August 1998, Vol. 123 1665Statistics AGANTG furnishes various report sheets. Of these, the Run and Summary reports include statistics. These two sheets display several types of statistics that are described next. Consider the agonist run data shown in Table 3 and the Run report given in Table 4. Simple average values are given on the Run report in the columns containing Test Values. These values are the mean response values observed for a given experiment at each of the agonist concentrations used during the experiment.For example, If we look at the columns in Table 3 labeled ‘REF’ at a concentration of 1E 2 05, the average of the values 175, 175, 165 and 145 is given as 165 at that same concentration in Table 4. Values labeled %Resp in Table 4 represent responses after they have been self-normalized and averaged. The self normalization for yi, the response associated with the ith concentration, is performed using the equation. ynorm = 100(yi 2 ymin)/(ymax 2 ymin) (9) where ynorm is the self-normalized response value for a particular concentration and ymin is the smallest and ymax the largest response observed in the experiment in question.Values in Table 4 under the heading‘Norm%Resp’ are computed using eqn. (10), which is similar to eqn. (9) except that the maximum and minimum y values in the denominator are those of the control immediately preceding the column of interest. These are labeled ycmax and ycmin: ycnorm = 100(yi 2 ymin)/(ycmax 2 ycmin) (10) To obtain an idea of how values fluctuate around raw or normalized averages, a column labeled ‘+ 2 SEM’ or ‘+ 2 SD’ is provided, depending upon whether the standard error of the mean or standard deviation was chosen during the SetUp dialog.These values are computed using either eqn. (11) or eqn. (12), with n being the number of data points. Table 4 indicates that the standard error was chosen for the illustrated analysis. SD = ave S( ) ( ) / y y n i i - - 2 1 (11) SEM ave = - - S( ) ( ) / y y n n i i 2 1 (12) Eqns.(11) and (12) are also used in the beginning section of the Summary report sheet to compute results labeled ‘+ 2 SEM’ or Table 2 Individual results for agonist run data in Table 3 Table 3 Agonist run data [ ] REF SAMPLE REF SAMPLE REF SAMPLE REF SAMPLE 1E 2 10 55 48 49 46 49 50 48 45 3E 2 10 55 48 49 46 49 50 48 45 1E 2 09 55 49 49 47 49 51 48 50 3E 2 09 55 155 49 78 49 128 48 118 1E 2 08 55 180 49 120 49 150 48 140 3E 2 08 58 180 49 135 49 190 48 140 1E 2 07 59 53 135 55 190 63 3E 2 07 83 83 75 100 1E 2 06 120 125 110 125 3E 2 06 175 175 165 145 1E 2 05 175 175 165 145 3E 2 05 1666 Analyst, August 1998, Vol. 123‘+ 2SD’.However, the SE values of the pA2 results, in the pA2 section, are computed differently, as explained by eqn. (5.4) of Tallarida and Murray.10 Likewise, the SE of the slope is computed using eqn. (5.2) of Tallarida and Murray.10. Values under the headings EC50 and so on refer to average values having the number of points listed under the heading ‘N’.In all estimates of confidence intervals, limits are obtained from eqn. (5.3) of Tallarida and Murray.10 All intervals in the Summary report are estimated at 95% confidence. The coefficient of linear correlation, r, follows a standard equation: 11 r x x y y y y x x i i i i = - - - - S S [( )( )]/ ( ) ( ) ave ave ave ave 2 2 x (13) Program validation The ALLFIT program is the reference of choice for validating new programs.Table 5 summarizes final results produced by the AGANTG and ALLFIT programs using the same data. Intermediate results, including some of the fit parameters, are not reported by ALLFIT and are not compared here. Table 5 contains data for both agonists and antagonists. Each agonist experiment was repeated four times and each antagonist experiment was repeated six or eight times for the purpose of this comparison.The data were then analyzed by both programs using a maximum of 200 iterations for the curve fitting step. The two methods agree well, as illustrated by a comparison of two of the most important results contained in Table 5. Linear regression of equally weighted AGANTG EC50 results for agonists against the ALLFIT results produces a line with a slope of approximately 0.995, an intercept of approximately 0.000 000 7 and a correlation of 1.000 00. A similar analysis of antagonist pA2 results produce a slope of approximately 1.001, an intercept of 20.026 and a correlation of 0.999.The small differences observed between the results of these programs are most likely due to the non-linear regression step required for analysis. Because non-linear regression is more complicated than ordinary linear regression, some differences are expected. The many implementation details of non-linear algorithms may produce differences in results even when using the same method, such as the Levenberg–Marquardt algorithm.Some likely sources of the observed differences in results may include differences in starting parameter estimates and differences in the implementation of regression algorithms themselves. Conclusions EXCEL 5.0 combines a user interface and programming language that benefits both software users and developers. EXCEL–VB is a powerful tool for developing complex applications such as AGANTG. Graphical objects, worksheets, charts, menus and dialog boxes can be easily created and manipulated by the developer through VB.VB is a structured language with an extensive library of classes and functions to aid developers in their task. Users benefit because applications developed with EXCEL have the familiar look and feel of spreadsheets. Users face less needless anxiety since they can concentrate on completing their work instead of learning software tools. Best of all, the user obtains all the functionality of EXCEL on top of the particular application.In the present example, this means that users can cut, paste, import or export data and objects to and from other spreadsheets and applications. Such flexibility can greatly reduce tasks such as entering data, writing reports and transferring results to a database. The AGANTG program is an important application that illustrates the power of the EXCEL–VB combination. Armed with a few simple rules, the user may enter agonist or antagonist experimental data into a worksheet and perform complex analysis with the click of a mouse.Data or analysis options can be edited and the data re-analyzed. One may, for example, analyze intermediate data then later add to those data and reanalyze. Analysis options are edited whenever necessary using a convenient dialog box. During analysis, the data are scrutinized and any detected errors highlighted. This step prevents wasting time spent looking for subtle mistakes.Additional messages and message boxes are used to alert the user to the status of the analysis and to issue warnings when appropriate. Table 4 Run report for data in Table 3 [Agonist] REF Test Val + 2 S.E.M. REF %Resp + 2 S.E.M. Norm %Resp + 2 S.E.M. 1.00e 2 10 50.3 1.6 0.0 0.0 3.00e 2 10 50.3 1.6 0.0 0.0 1.00e 2 09 50.3 1.6 0.0 0.0 3.00e 2 09 50.3 1.6 0.0 0.0 1.00e 2 08 50.3 1.6 0.0 0.0 3.00e 2 08 51.0 2.3 0.6 0.6 1.00e 2 07 57.5 2.2 6.8 2.9 3.00e 2 07 85.3 5.3 31.6 7.4 1.00e 2 06 120.0 3.5 61.6 6.2 3.00e 2 06 165.0 7.1 100.0 0.0 1.00e 2 05 165.0 7.1 100.0 0.0 3.00e 2 05 [Agonist] SAMPLE Test + 2 S.E.M.SAMPLE %Resp + 2 S.E.M. Norm %Resp + 2 S.E.M. 1.00e 2 10 47.3 1.1 0.0 0.0 0.0 0.0 3.00e 2 10 47.3 1.1 0.0 0.0 0.0 0.0 1.00e 2 09 49.3 0.9 2.0 1.1 1.9 1.1 3.00e 2 09 119.8 16.0 62.4 10.4 64.3 13.7 1.00e 2 08 147.5 12.5 88.6 7.0 88.2 11.0 3.00e 2 08 161.3 13.9 100.0 0.0 99.8 10.8 1.00e 2 07 162.5 27.5 100.0 0.0 95.7 25.0 3.00e 2 07 1.00e 2 06 3.00e 2 06 1.00e 2 05 3.00e 2 05 Analyst, August 1998, Vol. 123 1667The AGANTG program was written in a manner that minimizes the number of inputs required of the user. For example, the program is able to decide if the experiment is that of an agonist or an antagonist from the format of column data headers and perform the correct analysis. In the case of antagonists, the program automatically determines if the pA2 uses multiple concentrations of antagonists and whether or not a pA2 plot should be generated.Although this may seem a trivial task, such a hands off approach can save the user both time and aggravation when other seemingly trivial tasks are likewise handled. A few other examples of simple but valuable automated tasks performed by AGANTG that eliminate user inputs include avoiding the entering of the number of data points used in an experiment, the sorting of concentration– response data into the proper order and deciding if data are inhibitory or excitatory.The authors express their thanks to Guennadi Safronov for his help in debugging the AGANTG program. References 1 Tack, G., Roselli, K. A., Overhoser, A., and Harris, T. R., Comput. Biomed. Res., 1995, 28, 24. 2 Arunlakshana, O., and Schild, H. O., Br. J. Pharmacol., 1959, 14, 48. 3 Waud, D. R., and Parker, R. B., J. Pharmacol. Exp. Ther., 1971, 177. 4 Rodbard, D., and McClean, S. W., Clin. Chem., 1977, 23, 112. 5 Rodbard, D., Clin. Chem., 1974, 20, 1255. 6 Delean, A., Munson, P.J., and Rodbard, D., Am. J. Physiol., 1978, 4, e97. 7 Press, W. H., Flannery, B. P., Teukolsky, S. A., and Vetterling, W. T., Numerical Recipes, The Art of Scientific Computing, Cambridge University Press, Cambridge, 1986. 8 Dahlquist, G., and Bjorck, A., Numerical Methods, Prentice-Hall, Englewood Cliffs, NJ, 1974. 9 Furchgott, R. R., in Handbook of Experimental Pharmacology, ed. Blaschko, H., and Muscholl, E., Springer, Berlin, 1972, vol. 33, pp. 283–335. 10 Tallarida, R. J., and Murray, R. B., Manual of Pharmacologic Calculation with Computer Programs, Springer, New York, 2nd edn., 1987. 11 Snedecor, G. W., and Cochran, W. G., Statistical Methods, Iowa State University Press, Ames, IA, 7th edn.,1980. Paper 8/04438D Accepted June 12, 1998 Table 5 Comparison of ALLFIT and AGANTG results Runs Model EC50 SEM/EC50 Lower 95% Upper 95% % Max. Method 4 VAS ala 2 7 2.09e 2 6 1.56e 2 7 1.84e 2 6 2.37e 2 6 89.1 ALLFIT 1.98e 2 6 2.68e 2 7 1.24e 2 6 2.73e 2 6 89.1 AGANTG 4 SPLEEN alb 2 8 2.11e 2 4 2.80e 2 5 1.70e 2 4 2.62e 2 4 77.4 ALLFIT 2.10e 2 4 3.01e 2 5 1.27e 2 4 2.94e 2 4 77.4 AGANTG 4 VAS ala 2 7 9.20e 2 8 1.35e 2 8 7.45e 2 8 1.14e 2 7 92.7 ALLFIT 9.36e 2 8 1.44e 2 8 5.37e 2 8 1.34e 2 7 92.7 AGANTG 4 PROST ala 2 17 1.22e 2 7 5.04e 2 8 7.12e 2 8 2.08e 2 7 126.7 ALLFIT 1.20e 2 7 5.18e 2 8 2.34e 2 8 2.64e 2 7 126.7 AGANTG 4 AORTA ald 2 10 1.70e 2 6 5.70e 2 7 9.83e 2 7 2.93e 2 6 62.3 ALLFIT 1.82e 2 6 6.45e 2 7 3.18e 2 8 3.61e 2 6 62.3 AGANTG SEM/pA2 Runs Model pA2 SEM/slope Correlation r Slope Low/up 95% Method 6 PROST 71 6.72 0.16 0.98 1.03 6.30/7.14 ALLFIT 0.09 6.67 0.15 0.99 1.04 6.38/6.96 AGANTG 0.06 6 URETHRA 5.57 0.30 0.91 1.05 4.74/6.41 ALLFIT 0.24 5.56 0.13 0.03 1.12 5.20/5.92 AGANTG 0.14 6 AORTA 8.22 0.24 0.98 1.18 7.57/8.88 ALLFIT 0.11 8.13 0.67 0.96 1.18 6.26/9.99 AGANTG 0.17 8 VAS 8.47 0.24 0.96 1.22 7.89/9.05 ALLFIT 0.14 8.59 0.36 0.92 1.16 7.70/9.47 AGANTG 0.15 8 AORTA 9.46 0.19 0.99 1.06 9.00/9.92 ALLFIT 0.05 9.40 0.22 0.99 1.09 8.87/9.93 AGANTG 0.05 1668 Analyst, August 1998, Vol. 123

ISSN:0003-2654    

DOI:10.1039/a804438d

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

Extraction and determination of the Mitins sulcofuron and flucofuron from environmental river water P. M. Hancocka, M. Walsha, S. J. G. Whitea, D. A. Catlowb and P. J. Baugh*a a Department of Chemistry and Applied Chemistry, University of Salford, Salford, UK M5 4WT b Zeneca Pharmaceuticals, Hurdsfield Industrial Estate, Macclesfield, Cheshire, UK SK10 2NA Flucofuron and sulcofuron, both examples of Mitins, were employed as the active ingredients in mothproofing formulations for the protection of textile fabrics by the dyeing industry.Monitoring of their presence in components of the river ecosystem is a regulatory requirement so precise extraction techniques, combined with sensitive detection systems, are required to obtain valid data concerning the levels of target pollutants present. This study continued the development of liquid chromatography combined with negative-ion electrospray ionization tandem mass spectrometry (LC–ESI-MS–MS) operated in the multiple reaction monitoring mode for the determination of these analytes in complex matrices.The paper describes the development of liquid–liquid extraction (LLE) and solid-phase extraction (SPE) techniques for the determination of the analytes in environmental river water. The methods employed an internal standard, trichlorocarbanilide (TCC), to check the extraction efficiencies but not to correct environmental data. The extraction efficiencies obtained with LLE were 73.2 ± 6.7, 112.4 ± 8.6 and 96.4 ± 14.3% (n = 5) compared with 74.3 ± 8.4, 115.9 ± 3.1 and 112.7 ± 4.5% (n = 4) employing SPE for sulcofuron, flucofuron and TCC (100 ng l21 matrix fortification level), respectively.The SPE results are consistent with those obtained for LLE, although the precision of the SPE method was better than that of the LLE method. These methods were then successfully applied to samples obtained from a contaminated ecosystem. Keywords: Mitins; sulcofuron; flucofuron; mothproofing agents; environmental river water; liquid–liquid extraction; solid-phase extraction; liquid chromatography–electrospray ionization tandem mass spectrometry The pattern of mothproofing agent use has altered considerably since the end of the 1970s.The discovery that dieldrin was highly toxic to mammals and very persistent in the environment led to a decline in its use and replacement by formulations based on substituted ureas, commonly grouped under the term urons. The term mothproofing describes the treatment of wool or wool-based fabrics to prevent damage by the larvae of a number of insect pests1,2 that are capable of digesting keratin.Each mothproofing formulation contains an active ingredient, or combination of active ingredients, solvent, surfactant and water. Sulcofuron and flucofuron, both examples of urons, exert their toxic effect on the target organism by inhibiting the synthesis of the enzyme required to break down keratin.1 Mothproofing agents form chemical bonds with wool fibres in the same way as dyes and, therefore, the preferred method of application is during the dyeing process.The requirements for an active ingredient are restrictive, and consequently few general formulations are suitable for mothproofing. For example, in addition to being toxic to the insects at low levels of application, it must be stable to the application conditions, resistant to washing and light and effective for prolonged periods on the textile.2 Mitin is the registered tradename for mothproofing agents produced by Ciba-Geigy (Basle, Switzerland). Mothproofing agents previously marketed in the UK include Mitin FF High Conc containing sulcofuron (80%) and Mitin LP containing flucofuron (7.6%).1 Regulations designed to prevent the pollution of surface waters by sulcofuron and flucofuron were implemented in the UK on January 1, 1993.3 Contamination of freshwater ecosystems by sulcofuron and flucofuron occurs directly owing to the discharge of industrial effluents or indirectly through the discharge from sewage treatment works.In tests involving activated sewage sludge, both analytes were strongly adsorbed on particulate matter, so it is therefore not unreasonable to suggest that they are persistent in the environment.1 The environmental quality standards (EQSs) for sulcofuron and flucofuron in fresh water required to support fish are 25.0 and 1.0 mg l21, respectively.1,3 In this paper, the reported use of liquid chromatography combined with negative-ion electrospray ionization tandem mass spectrometry (LC–ESI-MS–MS) for the determination of sulcofuron and flucofuron is continued.4 We describe the development of liquid–liquid extraction (LLE) and solid-phase extraction (SPE) techniques for the determination of the analytes in environmental surface waters.The methods employ an internal standard, trichlorocarbanilide (TCC), to check the extraction efficiencies.The methods reported were successfully applied to samples obtained from a contaminated ecosystem. Experimental Sulcofuron {sodium 5-chloro-2-[4-chloro-2-(3,4-dichlorophenyl) ureido]phenoxybenzenesulfonate} [3567-25-7] was obtained from Ciba-Geigy (Macclesfield, Cheshire, UK), flucofuron [1,3-bis(4-chloro-a,a,a-trifluoro-m-tolyl)urea] [370-50-3] was custom synthesized at the University of Salford, from 4-chloro-3-(trifluoromethyl)phenyl isocyanate [16588-69-5] and 5-amino-2-chlorobenzotrifluoride [320-51-4],5 and TCC [1-(4-chlorophenyl)-3-(3,4-dichlorophenyl) urea] [101-20-2], the internal standard, was purchased from Aldrich (Gillingham, Dorset, UK).HPLC grade dichloromethane and methanol and analytical-reagent grade acetone, ethyl acetate and anhydrous sodium sulfate were all obtained from Fisons (Loughborough, UK). The variety of solid-phase cartridges were purchased from Jones Chromatography (Hengoed, Mid-Glamorgan, UK).The eluent was delivered by an isocratic system employing a Gilson Model 302 HPLC pump (Anachem, Luton, UK). LC separations were performed using a C8/C18 silica fully endcapped HiChrom HiRPB column (250 3 2.1 mm id, 5 mm particle size) (HiChrom, Reading, Berkshire, UK) with methanol- water (9 + 1 v/ v) as the mobile phase at a flow rate of 0.2 cm3 min21. Samples were introduced on to the column employing manual injection into a six-port Rheodyne (Cotati, CA, USA) Model 7125, injector fitted with a 20 ml loop.The LC system was connected to a VG Quattro mass spectrometer (VG Biotech, Manchester, UK) equipped with a Analyst, August 1998, Vol. 123 (1669–1674) 1669Megaflow electrospray probe and operated in the negative-ion mode. After separation, the sample flow was split so that 25% of the injection volume entered the source through a 150 cm3 100 mm id fused-silica capillary line (SGE, Milton Keynes, UK) together with nitrogen nebulizing gas (CP grade) (Air Products, Walton-on-Thames, UK) which flowed coaxially through the probe tip at 40 l h21.A nitrogen bath gas, flow rate approximately 150 l h21, was also employed to assist the desolvation process. Mass spectra were collected in full scan (m/z 100–1200 in 2 s) and multiple-reaction monitoring (MRM) (dwell time = 0.1 s, span = 0.02 u) modes. The source temperature was maintained at 120 °C and the sampling cone voltage was 35 V. MRM was performed while employing argon gas (CP grade, Air Products) in the collision cell at a nominal pressure (1.03 1024 mbar) set to induce a 50% reduction in the precursor ion intensity.The collision energy was 70 eV. Instrument control and data processing included the use of the supplied VG MassLynx 2.1 application software. Stock standard solutions of the reference compounds for long-term storage were prepared at concentrations of 100 mg l21. These solutions were serially diluted to prepare mixed working standard solutions of sulcofuron, flucofuron and TCC in the mobile phase in the required concentration range (0.5–500 mg l21).These standard solutions were analysed by LC–ESI-MS–MS operating in the MRM mode. External calibration was employed for the quantification of the analytes and the internal standard. Linearity was observed for both analytes (1–500 mg l21) with correlation coefficients of 0.999 obtained routinely. Peak areas were obtained from the mass chromatograms for the monitored reaction of each analyte and calibration curves generated from plots of peak area against analyte concentration.The instrumental limit of detection (LOD) was calculated from the analyte peak height/peak-topeak baseline noise ratio and was defined as a 3 : 1 signal-tonoise ratio (S/N). The procedures were developed and modified from US Environmental Protection Agency (USEPA) Method 625, Base/ Neutrals and Acids,6 and Methods for the Examination of Waters and Associated Materials7 for analysing the target analytes in spiked and real samples.The recommended method makes use of tetraethylammonium bromide (TEAB) to induce ion-pair formation.7 When applied to environmental matrices, this method is known to be unreliable.8 The spiking solutions, consisting of a mixture of sulcofuron and flucofuron in acetone (100 mg l21), were added to doubly distilled, de-ionized water for the recovery experiments, generating a matrix fortification level of 0.1 mg l21.The flask was vigorously shaken manually with frequent venting to release pressure and was then allowed to equilibrate overnight. The contaminated samples were collected from seven sites located at Meltham (near Huddersfield), West Yorkshire, UK, a catchment of the River Calder. The region has previously been highlighted as having a concentration of industrial mothproofing agent activity amongst the highest in the world.3 Table 1 lists the National Grid References of the seven sampling sites. Water samples were collected in glass bottles (2500 cm3) fitted with PTFE lined caps such that a headspace was avoided.The bottles were fully immersed to collect sub-surface water. After collection, samples were transported back to the laboratory, where they were stored in the dark at 4 °C. Further processing was undertaken within 48 h of collection. For LLE, a 1000 cm3 aliquot of water was placed in a clean, dry separating funnel, 0.1 mg of internal standard in acetone was added into the water and the funnel was shaken vigorously to ensure homogeneity.Approximately 30 g of sodium chloride were added to the aliquot to prevent the formation of emulsions and then the solution pH was adjusted to pH 2 with the addition of 1.0 m sulfuric acid. Dichloromethane (50 cm3) was added and the funnel was shaken vigorously for 5 min with frequent venting. When the two layers had separated sufficiently, the organic layer was drained and filtered through a sintered glass filter capped with anhydrous sodium sulfate, prepared by heating at 500 °C for 4 h, and collected in a rotary evaporation flask.The same procedure was repeated twice more with fresh dichloromethane and these extracts plus the additional rinses were combined. The volume of the solvent extracts was reduced to 0.1 mg in 5 cm3 using a rotary evaporator under reduced pressure. For SPE, a 1000 cm3 aliquot of water was placed in a clean, dry glass container, 0.1 mg of internal standard in acetone was added and the container was shaken vigorously to ensure homogeneity.The 500 mg C18 cartridge was activated and conditioned with 5 cm3 of acetone, followed by 5 cm3 of methanol and 5 cm3 of water. The sample was applied at a flow rate of approximately 10 cm3 min21. Once the sample had passed through, the cartridge was dried for approximately 40 min. The analytes were eluted with 3 3 2 cm3 aliquots of methanol.The concentrated extracts from the described procedures were transferred in to a 1.0 cm3 graduated conical vial and evaporated to less than 1 cm3 using a gentle stream of clean, dry nitrogen gas directed on to the surface of the solvent. The vial was rinsed several times with methanol–water (9 + 1, v/v) before concentration to 0.1 mg cm23. Reconstitution of the extract in the HPLC mobile phase is reported to minimize disruption of the column equilibrium following injection and to Table 1 National Grid References of the seven sampling sites at Meltham (near Huddersfield), West Yorkshire, a catchment of the River Calder, and the concentrations of sulcofuron and flucofuron in river water, employing LLE in October 1995 and SPE in June 1997.Concentration/ng l21 LLE SPE National Sampling Grid TCC site Reference Sulcofuron Flucofuron Sulcofuron Flucofuron recovery (%) Blank N.d.* N.d. N.d. N.d. 105 BDI SE 102098 N.d N.d. N.d. N.d. 107 BDII SE 102100 N.d N.d.N.d. N.d. 100 BDIII SE 103101 4.1 N.d. 6.0 4.2 95 BR SE 111112 2.0 1.8 6.9 4.3 102 HR SE 110113 6.7 3.8 2.7 3.8 93 STW SE 114117 3.1 N.d. N.d. 4.5 99 HHB SE 118122 8.5 N.d. N.d. 5.7 110 * N.d. = not detected (less than the LOD). 1670 Analyst, August 1998, Vol. 123O SO3Na Cl N C N Cl Cl Cl H O H Sulcofuron N C N CF3 Cl H O H F3C Cl N C N Cl H O H Cl Cl Flucofuron 3,4,4¢-Trichlorocarbanilide improve the precision of the method.9 The extracts were stored at 4 °C prior to instrumental analysis. Results and discussion Sulcofuron and flucofuron (Fig. 1) were selected by the National Rivers Authority (NRA) because of their history of use as textile mothproofing agents in the UK. Monitoring of their presence in environmental water is a regulatory requirement and involves reliable identification and quantification of trace levels at or below the environmental quality standards (EQSs). The use of LC–ESI-MS–MS operated in the MRM mode has been reported previously.4 This approach provided a highly sensitive and specific method for determining sulcofuron and flucofuron in the mid-picogram range.The instrumental LODs in the MRM mode were calculated to be 10 and 2.5 pg for sulcofuron and flucofuron, respectively. Sulcofuron and flucofuron were extracted from doublydistilled, de-ionized water with dichloromethane as described above without the addition of TEAB. Owing to the sulfonic acid functionality on sulcofuron, the effect of pH on extraction efficiency was investigated.This was performed by adjusting the pH of the spiked water samples at the 100 ng l21 (1000 cm3) matrix fortification level. Quantification was conducted by comparison of the analyte peak areas with those of external standards analysed in the same manner. The results (Table 2) illustrate the suspected pH dependence of sulcofuron on the extraction efficiency from water. The best extraction efficiencies for both sulcofuron and flucofuron extracted concurrently were obtained at pH 2 and were 73.2 ± 6.7 and 112.4 ± 8.6% (n = 5), respectively.The recovery results are consistent with those obtained for the standard method.7 It is difficult to compare the precision of this method with that reported, as the published method reports duplicate results, whereas this study employed five replicates, in accordance with USEPA Method 625.6 Having attained a suitable extraction procedure from doubly distilled, de-ionized water, the study moved to the more complex environmental water matrix.Environmental water samples were collected from seven sites located at Meltham, West Yorkshire, UK (Fig. 2). The samples (1000 cm3 aliquots) were adjusted to pH 2 and then treated identically to the spiked samples. Analysis of the extracts by negative-ion LC– ESI-MS–MS operated in the MRM mode was performed. The results for the environmental water samples are summarized in Table 1.Sulcofuron was detected at all the active sites, in the highest concentrations at sites downstream of the sewage treatment works (STW, 6.7 ng l21) and the textile mill (BDII, 8.5 ng l21). Flucofuron was only detected at sites downstream of the sewage treatment works (STW, 3.8 ng l21, and HHB, 1.8 ng l21). These results were as expected because the most likely sources of contamination are due to the direct discharge of industrial effluents or indirectly through the outfall from sewage treatment works, where it is suggested that sulcofuron and flucofuron are accumulated by the surrounding biosphere.1 Flucofuron was not detected at any other sampling sites.This result was considered to reflect the pattern of mothproofing agent use by the local textile industry rather than a failure of the sampling strategy, the extraction procedure or the quantitative technique employed. Flucofuron is not currently being commercially marketed in any mothproofing agents, whereas sulcofuron- based formulations are still being manufactured under the principal trade name Mitin.10 The BDI and HR sites were employed as control blanks and helped to confirm that the analytical method was not subject to interference, as these sites are not known to have a history of contamination and were, therefore, expected to be clean.The levels of sulcofuron and flucofuron monitored in the environmental water samples from all the active sites were well below the EQSs of 25.0 and 1.0 mg l21 for sulcofuron and flucofuron, respectively.Having obtained a selective, sensitive, precise method for determining sulcofuron and flucofuron in river water samples, an internal standard was required to check for recovery in environmental matrices. 3,4,4A-Trichlorocarbanilide (TCC, Fig. 1 Structures of flucofuron, sulcofuron and TCC. Table 2 Dependence of LLE efficiency on water sample pH at 100 ng l21 (n = 5) Recovery ± s (%) Adjusted pH Sulcofuron Flucofuron Blank N.d.* N.d. 1 44.5 ± 5.8 42.5 ± 4.5 2 73.2 ± 6.7 112.4 ± 8.6 4 51.8 ± 8.2 117.0 ± 10.1 7 51.9 ± 16.9 114.1 ± 11.6 * N.d. = not detected (less than the LOD). Fig. 2 Location of the seven sampling sites at Meltham, West Yorkshire, UK. Analyst, August 1998, Vol. 123 1671Fig. 1) was initially selected from three suitable alternatives (results not shown) on the basis of its elution proximity by reversed-phase HPLC to that of sulcofuron and flucofuron (Fig. 3) and its low probability of environmental occurrence.A negative-ion electrospray ionization mass spectrum was collected in full scan mode for TCC (Fig. 4). Owing to the soft nature of electrospray ionization, the pseudo-molecular ion, [M–H]2, is easily observed in the mass spectrum. This ion (m/z 313, 315, 317, 319) shows the correct chlorine isotope ratio (27 : 27 : 9 : 1) for the presence of three chlorine atoms. The [M 2 H]2 ion undergoes cleavage of the nitrogen–carbon bond to give the two amines (m/z 126, 128 and m/z 160, 162, 164) after the ejection of the neutral isocyanate moieties.These fragments contain one and two chlorine atoms, as shown by the 3:1 and 9:6:1 isotope ratios, for m/z 126 and 160, respectively. Fig. 3 Typical LC–ESI-MS–MS MRM reconstructed ion trace and TIC for sulcofuron, flucofuron and TCC after extraction, at a matrix fortification level of 100 mg l21. Fig. 4 Negative-ion ESI mass spectrum of TCC using a 35 V cone voltage. 1672 Analyst, August 1998, Vol. 123A product-ion spectrum was obtained for TCC from m/z 313 (Fig. 5). The product-ion spectrum for TCC confirms the pseudo-molecular ion cleaving at the nitrogen–carbon bond to give the respective amines (m/z 160 and 126). Owing to the intensity of the product ion, m/z 160, compared with m/z 126 under these MRM conditions, the fragmentation of m/z 313 to give m/z 160 is monitored; a response at the correct retention time enables TCC to be quantified. A typical MRM mass chromatogram obtained for sulcofuron, flucofuron and TCC at 100 mg l21 (Fig. 3) illustrates that this method of detection gives a sufficiently intense response at the level of interest (mid-picogram range) for TCC. Linearity was observed for TCC (5–100 mg l21) with correlation coefficients > 0.998 obtained routinely. The absolute instrumental LOD, defined as an S/N of 3 : 1, was 3.0 mg l21 for TCC. The extraction of TCC employed methodology identical with that for sulcofuron and flucofuron.The internal standard was spiked at the 100 ng l21 (1000 cm3) matrix fortification level with pH adjustment to pH 2. The LLE efficiency from water for TCC was 96.4 ± 14.3% (n = 5). TCC was used to check the efficiencies of all subsequent methods but not to correct environmental data. SPE is now well established in the analytical chemistry laboratory and has largely replaced classical LLE. Having developed a reliable, precise LLE method for quantifying sulcofuron and flucofuron in river water, a method was required to utilize the well documented benefits of SPE over LLE,9,11,12 which for this application were speed, less use of solvents and reduction in costs.Sulcofuron, flucofuron and TCC were extracted from water employing a variety of 500 mg solid phases, including octadecyl, octyl, cyclohexyl and specialty environmental phases. Once again, owing to the sulfonic acid functionality on sulcofuron, the effect of pH on extraction efficiency was investigated. This was performed by adjusting the pH of the spiked water samples.The analytes were eluted with methanol– ethyl acetate (1 + 1 v/v). The doubly distilled, de-ionized water samples were spiked at the 1 mg l21 (100 cm3) matrix fortification level. The results (Table 3) illustrate that, like LLE, the SPE efficiency of sulcofuron is dependent on pH but with SPE the maximum efficiencies are obtained at pH 7. This is probably due to the suspected instability of the solid phase at low pH as the efficiencies of flucofuron and TCC are also affected.These analytes do not contain ionizable groups and so should largely be unaffected by pH changes in the matrix. This effect was revealed while performing LLE. The solid-phase composition also affected the SPE efficiency dramatically. The specialty environmental phases gave no extraction efficiency for sulcofuron, which was not proven to be 100% breakthrough or 100% retention. The fact that the ENVICARB cartridge, based on a carbon black phase, did not extract any of the analytes is surprising because Di Corcia and Marchetti13–15 described a rapid, sensitive method for 14 urons in aqueous samples using carbon black cartridges.The extraction efficiencies were reported to be 95–104 ± 4.6% (n = 6). The ENV+ cartridge, based on a styrene–divinylbenzene phase, extracted flucofuron and TCC with similar efficiencies to those of CH, C8 and C18 phases, the three phases that gave the maximum recoveries for the three analytes. The use of SPE was continued with the C8 and C18 phases employing a variety of eluting solvents with differing polarities and desorption capabilities. The results (Table 4) reveal that the extraction efficiency is also dependent on the elution solvent.The maximum efficiencies (74.3 ± 8.4, 115.9 ± 3.1 and 112.7 ± 4.5% for sulcofuron, flucofuron and TCC, respectively) are obtained with a C18 cartridge using methanol as the eluting solvent.The effect of increasing the sample volume to 1000 cm3 was investigated. No significant variation was observed, with extraction efficiencies of 76.3, 108.6 and 112.2% for sulcofuron, flucofuron and TCC, respectively. The SPE recovery results are consistent with those obtained with the LLE method, but the precision of the SPE method ( < 8.4%) is better than that obtained with the LLE method ( < 14.3%). Moore et al.,9 Junker-Buchheit and Witzenbacher11 and Patsias and Papadopoulou-Mourkidou16 described rapid methods for determining urons in field water samples using a variety Fig. 5 TCC. Product-ion scan from m/z 313. Analyst, August 1998, Vol. 123 1673of solid phases. Moore et al.9 reported extraction efficiencies for chlorotoluron, isoproturon, diuron and linuron of 101–118 ± 10% (n = 5). Junker-Buchheit and Witzenbacher11 compared the extraction efficiency of a new polymeric sorbent and classical C18 cartridges for 10 urons.The extraction efficiency and precision with the C18 cartridge, 95–121 ± 4% (n = 6), was similar to those with the polymeric sorbent, 96–108 ± 4% (n = 3). Patsias and Papadopoulou-Mourkidou16 reported extraction efficiencies for fluometuron, linuron, metobromuron and monolinuron of 78–95 ± 11% (n = 3). The mean recoveries and the extraction precision for sulcofuron, flucofuron and TCC are consistent with those reported for the selected urons.Once again, environmental water samples were collected from the seven sites in the Meltham area of the River Calder catchment. Samples (1000 cm3 aliquots) were extracted identically with the spiked samples. Analysis of the extracts by negative-ion LC–ESI-MS–MS operated in the MRM mode was performed. The results for the environmental water samples are given in Table 1. Sulcofuron was detected in the highest concentrations at sites downstream of the sewage treatment works (HHB, 6.9 ng l21) and the textile mill (BR, 6.0 ng l21).Flucofuron was detected in the highest concentrations at sites downstream of the textile mill (BDII, 5.7 ng l21, and BDIII, 4.5 ng l21). These results are different from those obtained by employing LLE, but this is to be expected as there was an 18 month period between the sampling dates. The BDI and HR sites were used to confirm that the analytical method was not subject to interference as these sites are not known to have a history of contamination and were therefore expected to be clean.The extraction efficiency of TCC was excellent for all the sites (93–110%) and illustrated that the method was reliable for the extraction of the analytes. The levels of sulcofuron and flucofuron monitored in environmental water were, again, well below the respective EQSs. In conclusion LC–ESI-MS–MS operated in the MRM mode has continued to provide a highly sensitive and specific method of determination for sulcofuron and flucofuron in the environmental water matrix.LLE and SPE techniques with good reproducibility and efficiency have been successfully developed for the extraction of the analytes from environmental water. These extraction techniques, combined with the instrumental method of detection, allow the analytes to be determined at levels three to four orders of magnitude below their respective EQSs. The methods successfully employ an internal standard,TCC, to check extraction efficiency of the target analytes in quality control analysis.These methods have also been successfully applied to samples obtained from a contaminated ecosystem. We would especially like to thank Zeneca Pharmaceuticals (Macclesfield, UK) for making their facilities available and R. Hayes (University of Salford, UK) for the synthesis of flucofuron. We acknowledge R. Armitage (NRA, West Yorkshire Division, UK), G. Bonwick (MRIC, Immunology Group, North East Wales Institute, UK) and D.Davies (Department of Biological Sciences, University of Salford, UK) for their involvement in the early stages of this study (NRA Project No. 319). P. Hancock thanks the Engineering and Physical Sciences Research Council (EPSRC) and S. White and M. Walsh thank the European Social Fund (ESF) for financial support. References 1 Zabel, T. F., Seager, J., and Oakley, S. D., Proposed Environmental Quality Standards for List II Substances in Water, Marlow, 1988. 2 Mehra, R.H., Mehra, Anil R., and Mehra, Arun R., Colourage, 1991, 38, 69. 3 Shaw, T., J. Inst. Water Environ. Manage., 1994, 8, 393. 4 Hancock, P. M., White, S. J. G., Catlow, D. A., Baugh, P. J., Bonwick, G. A., and Davies, D. H., Rapid Commun. Mass Spectrom., 1997, 11, 195. 5 Hancock, P. M., PhD Thesis, University of Salford, 1997. 6 USEPA Method 625, Base/Neutrals and Acids, in Guidelines Establishing Test Procedures for Analysis of Pollutants Under the Clean Water Act. Rules and Regulations, vol. 49, USEPA, Washington, DC, 1984, pp. 153–161. 7 Standing Committee of Analysts, Chlorophenylid, Flucofuron and Sulcofuron in Water (Tentative Methods), in Methods for the Examination of Waters and Associated Materials, HM Stationary Office, London, 1992. 8 Daniels, M., West Yorkshire Division, NRA, UK, personal communication, 1994. 9 Moore, K. M., Jones, S. R., and James, C., Water Res., 1995, 29, 1225. 10 The Pesticide Manual, ed. Tomlin, C. D. S., 10th edn., Royal Society of Chemistry, Cambridge, and British Crop Protection Council, London, 1995. 11 Junker-Buchheit, A., and Witzenbacher, M., J. Chromatogr. A, 1996, 737, 67. 12 Hopper, M. L., in Emerging Strategies for Pesticide Analysis, ed., Cairns, T., and Sherma, J., CRC Press, Boca Raton, FL, 1992, pp. 39– 50. 13 Di Corcia, A., and Marchetti, M., J. Chromatogr., 1991, 541, 365. 14 Di Corcia, A., and Marchetti, M., Environ. Sci. Technol., 1992, 26, 66. 15 Di Corcia, A., and Marchetti, M., Anal. Chem., 1991, 63, 580. 16 Patsias, J., and Papadopoulou-Mourkidou, E., J. Chromatogr. A., 1996, 740, 83. Paper 8/02528B Received April 2, 1998 Accepted June 5, 1998 Table 3 SPE efficiencies for sulcofuron, flucofuron and TCC extracted from water using different extraction cartridges and pH values at 1 mg l21 Recovery (%) Cartridge pH Sulcofuron Flucofuron TCC CH 2 49.7 63.6 N.d.* CH 7 74.0 129.1 83.1 C8 2 47.3 124.6 N.d. C8 7 74.3 127.3 59.3 C18 2 49.3 86.1 50.7 C18 7 70.7 113.1 86.3 ENV+ 2 N.d. 120.2 115.7 ENV+ 7 N.d. 105.8 68.5 ENVI-CARB 2 N.d. N.d. N.d. ENVI-CARB 7 N.d. N.d. N.d. * N.d. = not detected (less than the LOD). Table 4 SPE efficiencies for sulcofuron, flucofuron and TCC extracted from water using C8 and C18 cartridges with different eluting solvents at 1 mg l21 (n = 4) Recovery ± s (%) Cartridge Elution solvent Sulcofuron Flucofuron TCC C8 MeOH 72.7 ± 2.5 115.1 ± 6.8 112.6 ± 6.3 MeOH–EtOAc 62.7 ± 10.4 76.4 ± 7.9 98.5 ± 7.4 MeOH–acetone 61.8 ± 6.1 65.1 ± 5.1 88.7 ± 11.1 C18 MeOH 74.3 ± 8.4 115.9 ± 3.1 112.7 ± 4.5 MeOH–EtOAc 62.6 ± 5.7 80.1 ± 4.9 89.4 ± 5.6 MeOH–acetone 62.3 ± 5.1 54.2 ± 5.8 93.3 ± 12.8 1674 Analyst, August 1998, Vol. 123

ISSN:0003-2654    

DOI:10.1039/a802528b

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

CRM 601, A stable material for its extractable content of heavy metals Jos�e Ferm�ýn L�opez-S�anchez*a, Angels Sahuquilloa, Haidi D. Fiedlera, Roser Rubioa, Gemma Raureta, Herbert Muntaub and Philippe Quevauvillerc a Universitat de Barcelona, Departament de Qu�ýmica Anal�ýtica, Av. Diagonal 647, E-08028 Barcelona, Spain b European Commission, Joint Research Centre, Environment Institute, I-21020 Ispra, Italy c European Commission, Standards, Measurements and Testing Programme, Rue de la Loi 200, B-1049 Brussels, Belgium Sequential extraction schemes provide information on the mobility of heavy metals from polluted sediments. The lack of uniformity of these schemes, however, does not allow results to be compared nor the procedures to be validated.In 1987 the Community Bureau of Reference, (BCR, now Standard, Measurements and Testing Programme) of the European Commission launched a programme to harmonise single and sequential extraction schemes for soils and sediments.As a result certified reference material CRM 601 (lake sediment) was prepared and certified for the extractable contents of some trace elements, following a standardised sequential (three-step) extraction procedure. In order to verify the long-term stability of the material an intercomparison exercise was designed with the participation of eight expert laboratories. The results confirm the stability of the extractable contents of Cd, Cr, Ni, Pb and Zn in this type of material, and demonstrate that the preparation of this kind of CRM can be achieved.Keywords: Extractable trace metals; sediments; sequential extraction; quality assurance; certified reference material; stability Since the 1970s sequential extraction schemes have been applied by a large number of laboratories to obtain information on the mobility of heavy metals from polluted sediments.1–3 The empirical character of this approach has led to proposals of a large number of sequential extraction schemes, all of them based on the same strategy, which is to dissolve, as selectively as possible, the fractions of heavy metal associated with well defined phases.The schemes used a variety of extracting agents, and so the metal distribution patterns produced when different schemes were applied to the same sediment were not comparable. As such there was a need to agree on the procedure that would yield comparable results. In 1987 the Community Bureau of Reference, (BCR, now Standard, Measurements and Testing Programme) of the European Commission launched a programme to harmonise single and sequential extraction schemes for soils and sediments.The development of this programme was discussed at a workshop4 organised by BCR in 1992, in which forty representatives of leading laboratories in soil and sediment analysis agreed on a three-steps sequential extraction procedure to study metal partitioning in sediments. The procedure was defined and carefully written and several interlaboratory trials were carried out to validate the method.5 Finally a sediment reference material (CRM 601) was prepared and certified for its extractable heavy metals by the validated procedure.6 The certification exercise ended in January 1995 and the technical discussion meeting was held in March 1995.Since there was a wide dispersion of the results for some elements in some fractions some values were stated only as indicative values.One topic discussed during the certification process was the long-term stability of extractable fractions. When preparing the material the stability of the extractable trace metals was tested for twelve months at several temperatures including 20 and 40 °C. At 40 °C a possible instability of Cu was observed and so the extractable contents of this element were not certified. Since CRM 601 was the first material certified for its extractable metal contents, and no previous experience of this kind of exercise was available, a study on the long-term stability of the material was proposed. Moreover, lyophilised biological material with a moisture content of 3.5%, similar to the CRM 601, could be unstable.7 Further removal of moisture may affect the extractability of heavy metals.Thus, in order to confirm the stability of the material an intercomparison exercise was designed with participation of eight expert laboratories, all of which had participated in the previous certification exercise two years ago.Here we describe this interlaboratory study and the results thereof. Experimental Description of the sample The BCR Certified Reference Material 601 is lake sediment collected at several sites in Lake Flumendosa (Italy). The sample pre-treatment, the homogeneity and stability studies, the certification campaign and the certified values are reported elsewhere.3 The CRM samples for the interlaboratory exercise were supplied by the Institute for Reference Materials and Measurements (IRMM, Geel, Belgium), Management of Reference Material Unit, where the CRM 601 is stored at 20 °C.This material can be purchased from the IRMM, and each bottle is accompanied by a certificate and a report describing the work performed. Intercomparison exercise Each laboratory that took part in the intercomparison exercise received the following items: four bottles of CRM 601, the BCR three-step protocol for sequential extraction,4 the forms for reporting the results and some recommendations to improve their quality control systems.The participants were requested to perform at least five independent determinations for each element (Cd, Cr, Cu, Ni, Pb and Zn) from at least two different bottles of the CRM on different days, and to follow the sequential extraction protocol strictly. The intercomparison exercise ended in October 1997 and the technical discussion meeting was held in December 1997.Analytical techniques The techniques used to determine metal concentrations in the extracts were generally FAAS (flame atomic absorption Analyst, August 1998, Vol. 123 (1675–1677) 1675spectrometry) or ETAAS (electrothermal atomic absorption spectrometry with or without Zeeman background correction). ICP-AES (inductively coupled plasma atomic emission spectrometry) was also used by some laboratories. Technical discussion The sources of error and the precautions taken to overcome them were discussed at the technical meeting.The participating laboratories applied their determination methods only when the method was under control, i.e., the standard deviations observed in the laboratory were in accordance with the normal practice of the laboratories. Most of the errors detected were due to the calibration rather than to the application of the extraction procedure. It was recommended that the participants should use pure metal solutions (Cd, Cu, Ni, Pb and Zn or NIST certified K2Cr2O7) as calibrants.If these standards were not available, the two standard solutions of different origin were to be used, and their concentration was to be checked periodically. The participants agreed that the data obtained in this exercise might be useful to the people purchasing the material and agreed to advise the Scientific Evaluation Group for the Certification of Reference Materials of the Standards, Measurement and Testing Programme to include such data as an appendix to the certification report.Results and discussion The sets of results found acceptable on technical and statistical grounds were presented in the form of bar-graphs. Some examples of the results obtained are given in Figs. 1–5. In the bar-graphs, the length of a bar corresponds to the 95% confidence interval of the mean of laboratory means. The interlaboratory means were calculated as the arithmetic mean of laboratory means.This value and the certified value are featured as a vertical dotted line and a solid vertical line, respectively; the uncertainty on mean values is given by the half width of the 95% confidence interval of the mean of laboratory means. The results and the certified values and their uncertainties (half width of the 95% confidence intervals) are given in Table 1 as mass fractions of the respective extracts obtained at the first, second and (based on dry mass) in mg kg21.For an easy comparison of the data the table also includes the percentage of recovery as well as the associated uncertainty obtained. The results for the first step compare well with the certified values for Cd, Cr, Ni and Zn showing recoveries ranging from 98 to 108%. For Pb the amount extracted is lower (77% recovery) but the two confidence intervals overlap due to the widerange of data. For Cu the amount extracted is significantly higher (125% recovery) than the indicative value.This element was not certified because of the suspicion of long-term instability inferred from the stability data at 40 °C, probably due to small changes in Mn oxides and organic matter phases of the sediment. In the second step the recoveries range from 92 to 113%. Cr and Zn presented good agreement with the certified/indicative values. For Cd, Ni and Pb larger standard deviations were observed, but they were also in agreement with the certified/ indicative values.For Cu the amount extracted is higher but the wide range of the results obtained made it difficult to draw conclusions. From examination of the raw data, two laboratories seem to be responsible for the high standard deviation. Both laboratories attributed this to the higher pH values of their final extracts, which produced a lower extraction of the metal in the solution and consequently lower amounts of metal were detected. For the third step, the amounts extracted for Cd, Cr, Cu, Pb and Zn compare well with the certified/indicative values obtaining recoveries ranging from 91 to 106%.For these elements the uncertainties were of similar magnitude to those obtained in the certification exercise. For Ni a lower recovery is obtained (79%), but the wide dispersion of data makes it difficult to draw conclusions. Conclusions The results of the study confirm the stability of the extractable contents of Cd, Cr, Ni, Pb and Zn in this type of material, and Fig. 1 Bar-graph for extracted zinc in the first step. The plotted values and their uncertainties (half width of the 95% confidence intervals) are as mass fractions, in mg kg21, of the respective extracts. Fig. 2 Bar-graph for extracted zinc in the second step. The plotted values and their uncertainties (half width of the 95% confidence intervals) are as mass fractions, in mg kg21, of the respective extracts. Fig. 3 Bar-graph for extracted zinc in the third step. The plotted values and their uncertainties (half width of the 95% confidence intervals) are as mass fractions, in mg kg21, of the respective extracts.Fig. 4 Bar-graph for extracted nickel in the first step. The plotted values and their uncertainties (half width of the 95% confidence intervals) are as mass fractions, in mg kg21, of the respective extracts. 1676 Analyst, August 1998, Vol. 123demonstrate that the preparation of CRMs for extractable trace metal contents in similar matrices can be achieved.The following laboratories participated in the intercomparison on long-term stability of CRM 601: Agricultural Research Centre, Institute of Soils and Environment (Jokioinen, Finland); Bund. für Materialforschung und Prüfung (Berlin, Germany); Department of Analytical Chemistry, University of Barcelona, (Barcelona, Spain); Department of Pure and Applied Chemistry, University of Strathclyde (Glasgow, United Kingdom); Environment Institute, Joint Research Centre (Ispra, Italy); Estación experimental del Zaidin, CSIC (Granada, Spain); Institut National de Recherche Agronomique (Villenave d’Ornon, France); The Macaulay Land Use Research Institute (Aberdeen, United Kingdom). J.F.L.-S, A.S., R.R.and G.R. also thank CEC, DGICYT and CIRIT for the financial support of this work. References 1 Salomons, W., and Förstner, U., Environ. Technol. Lett., 1980, 51, 506. 2 Pickering, W. F., Ore Geol. Rev., 1986, 1, 83. 3 Kersten, M., and Förstner, U., Trace Element Speciation: Analytical Methods and Problems, ed.Batley, G. E., CRC Press, Boca Raton, FL, 1989, ch. 8. 4 Procedings of the Workshop on the Sequential Extraction of trace metals in soils and sediments, ed. Albaigés, J., Rauret, G., and Quevauviller, Ph., Int. J. Environ. Anal. Chem., 1993, 51. 5 Quevauviller, Ph., Rauret, G., Muntau, H., Ure, A. M., Rubio, R., López-Sánchez, J. F., Fiedeler, H. D., and Griepìnk, B., Fresenius’ J. Anal. Chem., 1994, 349, 808. 6 Quevauviller, Ph., Rauret, G., López-Sánchez, J. F., Rubio, R., Ure, A. M., and Muntau, H., Sci. Total Environ., 1997, 205, 223. 7 Pauwels, J., personal communication. Paper 8/02720J Received April 9, 1998 Accepted May 21, 1998 Table 1 Results obtained in the interlaboratory study on long-term stability of the extractable metal contents in CRM 601 Certified Long-term Recovery value/ Uncertainty/ value/ Uncertainty/ Recovery uncertainty mg kg21 mg kg21 p* mg kg21 mg kg21 p* (%) (%) First step Cd 4.14 0.23 11 4.46 0.63 7 108 14 Cr 0.36 0.04 12 0.37 0.087 6 103 24 Cu 8.32† 0.46 9 10.4 0.43 6 125 3 Ni 8.01 0.73 10 8.22 0.83 7 103 10 Pb 2.68 0.35 11 2.07 0.49 7 77 24 Zn 264 5.0 12 259 13 7 98 5 Second step Cd 3.08 0.17 10 3.05 0.96 6 99 31 Cr 1.43† 1.0 13 1.42 0.83 7 99 58 Cu 5.69† 3.2 10 6.37 3.56 6 112 56 Ni 6.05 1.1 11 5.55 1.5 7 92 27 Pb 33.1 10 9 37.3 19 7 113 51 Zn 182 11 12 175 15 7 96 9 Third step Cd 1.83 0.20 11 1.80 0.17 5 98 9 Cr 18.3† 4.47 14 19.4 0.91 6 106 5 Cu 116† 26 15 116† 9.0 7 100 8 Ni 8.55 1.04 9 6.75 0.86 7 79 13 Pb 109 13 12 108 19 7 99 18 Zn 137† 30 14 124 17 7 91 7 * p = number of data sets. † Values obtained in the certification campaign, but finally not certified. Fig. 5 Bar-graph for extracted chromium in the third step. The plotted values and their uncertainties (half width of the 95% confidence intervals) are as mass fractions, in mg kg21, of the respective extracts. Analyst, August 1998, Vol. 123 1677

ISSN:0003-2654    

DOI:10.1039/a802720j

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

Comparison of microwave-assisted extraction and Soxhlet extraction for phenols in soil samples using experimental designs A. Egizabal, O. Zuloaga*, N. Etxebarria, L. A. Fern�andez and J. M. Madariaga Kimika Analitikoaren Saila, Euskal Herriko Unibertsitatea, 644 P.K., E-48080 Bilbao, Spain A study and comparison of microwave-assisted extraction (MAE) and Soxhlet extraction for five phenols (phenol, 2-chlorophenol, 2-methylphenol, 2-nitrophenol and 2,4-dichlorophenol) in soils was carried out in order to establish a comparative analysis between the two techniques under their optimum conditions.The optimisation of both the MAE and Soxhlet extraction methods was achieved using experimental designs. The parameters studied were pressure inside the extraction vessel, extraction time, percentage of microwave power, solvent volume and percentage (v/v) of acetone in an acetone–hexane mixture for MAE and extraction time and percentage (v/v) of acetone in an acetone–hexane mixture for Soxhlet extraction.The simplex method was also used for obtaining the optimum conditions in some cases. The extracts were analysed by GC–MS or GC–FID. Keywords: Microwave-assisted extraction; Soxhlet extraction; gas chromatography–mass spectrometry; gas chromatography-flame ionization detection; phenols; experimental designs; simplex method; soil samples Phenol and its derivatives are widely used in the chemical industry for the manufacture of polymers, textiles, drugs, resins, dyes, detergents, explosives, stabilisers and antioxidants.1 Further, phenolic compounds have substantial applications in agriculture as herbicides, insecticides and fungicides, thus becoming potential pollutants of soils and of surface and underground waters owing to their highly hydrophilic nature.2 However, phenolic compounds are not only generated by human activity, they are also formed naturally, e.g., during the decomposition of leaves or wood.2 Phenols have a highly toxic character and it is well known that these substances exhibit properties that are hazardous to human health.3 Owing to their toxicity and presence in the environment, the US Environmental Protection Agency (EPA) includes 11 phenols among the main environmental pollutants. 4 The extraction of organic pollutants from soil samples requires the use of organic solvents, which compete in the release of the analytes retained owing to the high activity of the matrix. Traditional methods employ large volumes of solvents under aggressive shaking and/or temperature conditions.The most frequently used method for the extraction of organic pollutants from soils is extraction at the boiling-point of the solvent (Soxhlet extraction) or the use of an ultrasonic bath. Soxhlet extraction is the most conventional of all methods and consists of a simple distillation process repeated a number of times. This method is particularly suitable for organic pollutants strongly adsorbed on soil matrices but requires long extraction times and the use of large volumes of frequently toxic solvents. Moreover, it can degrade the analytes.5,6 In the last few years, the number of procedures using extraction of organic compounds from environmental matrices by microwave energy has increased.7–11 Microwave energy is a non-ionizing radiation that causes molecular motion by migration of ions and rotation of dipoles, without changing the molecular structures if the temperature is not too high. Nonpolar solvents, such as hexane and toluene, are not affected by microwave energy and, therefore, it is necessry to add polar additives.7 Microwave extraction greatly reduces solvent consumption and extraction times.Many of these studies7–9 made use of the experimental design approach for the optimisation of microwave-assisted extraction (MAE) procedures. The variables considered differ from one publication to another, but the most commonly used combinations are temperature, extraction time and solvent volume,7 volume of solvent, percentage of acetone and extraction temperature8 of different types of extracting solvents, extraction temperature, extraction time and solvent volume.9 In this work, MAE was used for the extraction of five phenolic compounds, phenol, 2-chlorophenol, 2-methylphenol, 2-nitrophenol and 2,4-dichlorophenol and the results were compared with those obtained by Soxhlet extraction.The parameters studied for MAE were pressure inside the extraction vessel, extraction time, percentage of microwave power, percentage (v/v) of acetone in the acetone–hexane mixture and volume of solvent.For Soxhlet extraction, the percentage (v/v) of acetone in the acetone–hexane mixture and extraction time were studied. Factorial and central composite designs were mainly used for approaching the response surface but the simplex method12 implemented in the MultiSimplex program13 was also used in some cases for achieving the optimum conditions.Experimental Microwave-assisted extraction MAE experiments were performed with an MDS-2000 closed microwave solvent extraction system (CEM, Matthews, NC, USA) equipped with a 12-sample tray and pressure feedback/ control. A 0.5 g aliquot of spiked soil was accurately weighed, mixed with 2 g of anhydrous sodium sulfate and transferred quantitatively to the Teflon-lined extraction vessel. The acetone –hexane mixture was then added to each sample and the extraction vessel was closed, after ensuring that a new rupture membrane was used for each experiment. In this work, the volume and percentage (v/v) of acetone in the acetone–hexane mixture and the power, pressure and time in which extractions were performed were indicated by the experimental design.The extraction conditions were programmed in two stages. In the first stage, the system was allowed to reach the required pressure using full power; in the second stage, the pressure previously reached was kept constant and the power and the time were established according to the experimental design.In this way, the samples stayed at the previously set pressure during the whole extraction process. When the irradiation period was complete, samples were removed from the microwave cavity and allowed to cool to room temperature before opening. The supernatant was filtered through glass-wool prewashed with hexane–acetone and then combined with 7–10 ml Analyst, August 1998, Vol. 123 (1679–1684) 1679acetone–hexane rinses of the extracted sample. The extract was concentrated to approximately 1 ml using nitrogen evaporation with a Turbovap LV Evaporator (Zymark, Hopkinson, MA, USA). An equivalent amount of a stock standard solution of phenols followed both the filtration and concentration steps in order to estimate possible losses occurring in these steps. The concentrated extracts, together with 200 ml of a stock standard solution of 4-methoxyphenol (internal standard), were transferred into 5 ml calibrated flasks and dissolved in hexane.The extracts were transferred into injection vials for GC–FID or GC–MS analysis. Soxhlet extraction Soxhlet extractions were performed using 1 g of soil analytically weighed and mixed with 10 g of anhydrous sodium sulfate. Samples were extracted under reflux with 200 ml of acetone– hexane in the percentages and the times indicated by the experimental design.After cooling to room temperature, the extracts were concentrated to approximately 5 ml using a rotary evaporator at 40 °C and reduced pressure. To estimate possible losses occurring during the concentration stage, an amount of a stock standard solution of phenols was treated in the same manner. The concentrated extracts, together with 200 ml of a stock standard solution of 4-methoxyphenol (internal standard), were transferred into 10 ml calibrated flasks and dissolved in hexane.The extracts were transferred into injection vials for GC–FID or GC–MS analysis. Reagents and chemicals Phenol standards were supplied as follows: phenol and 4-methoxyphenol by Merck (Darmstadt, Germany), 2-chlorophenol and 2,4-dichlorophenol by Aldrich (Gillingham, Dorset, UK) and 2-nitrophenol and 2-methylphenol by Fka (Buchs, Switzerland). All compounds were analytical-reagent grade and the purities were stated to be higher than 99%.A stock standard solution (40 mg ml21) of phenol, 2-chlorophenol, 2-methylphenol, 2-nitrophenol and 2,4-dichlorophenol was prepared by weighing an appropriate amount of the standard and dissolution in 50 ml of hexane. This standard solution was used only for the sample spiking. A stock standard solution (1000 mg ml21) of 4-methoxyphenol (internal standard) was prepared by weighing an appropriate amount and dissolving it in 20 ml of hexane. From the stock standard solution of the internal standard a more dilute working standard solution (40 mg ml21) was prepared daily and then used in all calibration standards and sample extracts.For calibration, the mixture number 5 of 18 semivolatile phenols (2000 mg ml21) supplied by Supelco (Bellefonte, PA, USA) was used. Hexane and acetone were purchased from LabScan (Dublin, Ireland) and the purities were stated to be better than 95%. Anhydrous sodium sulfate (purissimum) was purchased from Panreac (Barcelona, Spain).All solutions were stored at 5 °C in the dark. All volumetric glassware was grade A and was calibrated at laboratory temperature. Optimisation experiments were performed using a clay soil from an industrial downfall (Metalqu�ýmica del Nervi�on, Bilbao, Spain). The sample was ground, sieved and dried following the ISO 11464 Norm14 for homogeneity. A 40 g portion of soil, which was checked to be free of phenols, was analytically weighed and 35 ml of 40 mg ml21 stock solution were added.Subsequently, acetone was added to cover the soil and the mixture was stirred for at least 24 h. The soil was then dried in an oven at 40 °C. Finally, the soil was homogenised with a mortar, bottled and stored in a refrigerator. The concentrations in the soil, on the basis of added amounts, were 34.4, 40.6, 36.3, 35.2, and 35.6 mg g21 for phenol, 2-chlorophenol, 2-methylphenol, 2-nitrophenol and 2,4-dichlorophenol, respectively. Analysis of extracts Some extracts were analysed on a Hewlett-Packard (Avondale, PA, USA) Model 5890 Series II gas chromatograph interfaced Table 1 Design matrix and corrected recovery values in the fractionated factorial design for MAE experiments Design matrix Recovery (%) Run Pressure/ Time/ Power Acetone Volume/ 2-Chloro- 2-Methyl- 2-Nitro 2,4-Dichloro- No.psi min (%) (%) ml Phenol phenol phenol phenol phenol 1 20 20 30 30 30 65 58 50 60 56 2 30 20 30 70 15 99 84 73 102 89 3 20 40 30 70 15 88 77 68 93 71 4 30 40 60 30 30 84 80 61 103 81 5 20 20 60 70 30 80 69 59 84 68 6 30 20 60 30 15 26 26 22 30 25 7 20 40 60 30 15 90 85 67 94 84 8 30 40 60 70 30 151 140 128 133 128 9 25 30 45 50 20 109 96 85 86 93 10 25 30 45 50 20 96 79 71 72 79 11 25 30 45 50 20 77 66 58 58 65 Fig. 1 Sheward plots for phenol by (a) GC–MS and (b) GC–FID. Drf = average relative response factor. 1680 Analyst, August 1998, Vol. 12380 80 70 70 60 60 40 50 50 10 15 20 25 30 35 40 time/min time/min power (%) power (%) recovery (%) recovery (%) recovery (%) (a) (b) (c) 50 45 40 35 30 40 50 60 70 80 15 20 25 30 pressure/psi pressure/psi 45 40 35 30 10 15 20 25 30 35 40 15 20 25 30 to a Hewlett-Packard 5989 MS Engine mass spectrometer MS/ DOS ChemStation and equipped with a Hewlett-Packard Model 6890 Series autosampler.The samples were analysed on a 30 m 3 0.32 mm id 3 0.25 mm film thickness HP-5 fused-silica capillary open-tubular column. The column temperature was held at 40 °C for 2 min and then increased at 30 °C min21 to 80 °C, where it was held for 1 min and subsequently programmed at 10 °C min21 to 200 °C, where it was held for 2 min.The carrier gas was helium (N-50) at a flow rate of 3.45 ml min21 and a pressure at the column head of 10 psi. The injection volume was 1–3 ml and the injection temperature was 250 °C. The injector was set in the splitless mode with split vent closed for 1 min after injection. For the mass spectrometer, the electron energy was set at 70 eV and spectral data were acquired at a rate of 0.8 scan s21 (scanning range 50–550 u).Quantitative data were acquired by operating the MS detector in the selected ion monitoring mode (SIM) and the m/z values monitored were phenol 94, 66, 2-chlorophenol 128, 64, 2-methylphenol 108, 107, 2-nitrophenol 139, 65, 4-methoxyphenol 124, 109 and 2,4-dichlorophenol 164, 162. The instrument was tuned daily with perfluorotributylamine (PFTBA) using the EI Sensitivity Tune (EISENS) internal program since this program allows better sensitivities than those obtained by the usual Automatic Tune (ATUNE).In addition, to increase sensitivity, an optimum 400 V overpotential was applied to the electronic multiplier. A five-point internal standard calibration using standards at 0.32, 1.65 and 8 mg ml21 and injecting different volumes (1–3 ml) of the standards was performed daily to establish the GC–MS calibration curve, which ranged from 0.6 to 9.2 ng. 4-Methoxyphenol (internal standard) was added to every calibration Table 2 Design matrix and corrected recovery values in the central composite design for MAE experiments Design matrix Recovery (%) Run Power Time/ Pressure/ 2-Chloro- 2-Methyl- 2-Nitro- 2,4-Dichloro- No. (%) min psi Phenol phenol phenol phenol phenol 1 44 16.5 16 89 55 51 37 51 2 70 16.5 16 71 44 37 54 47 3 44 33.5 16 106 63 59 44 58 4 70 33.5 16 121 76 69 71 80 5 44 16.5 26 122 82 67 105 95 6 70 16.5 26 108 70 52 86 82 7 44 33.5 26 125 81 68 110 96 8 70 33.5 26 78 44 42 32 45 9 80 25 21 96 64 42 87 74 10 35 25 21 105 71 56 91 82 11 57 40 21 106 66 57 58 68 12 57 10 21 99 63 53 87 77 13 57 25 30 59 37 31 45 43 14 57 25 12 93 60 49 64 64 15 57 25 21 87 51 38 46 54 16 57 25 21 87 54 45 47 58 17 57 25 21 105 52 55 95 55 Table 3 Values of the parameters obtained for central composite design for MAE experiments Compound b0 b1 b2 b12 b13 b23 b11 b22 b123 r2 Phenol 113 23.32 26.01 0.22 0.14 0.31 —* — 20.010 0.753 2-Chlorophenol 201 25.80 27.84 0.17 0.09 0.25 0.03 0.05 20.008 0.858 2-Methylphenol 112 22.13 26.45 0.13 0.07 0.18 — 0.04 20.006 0.742 2-Nitrophenol 308 29.95 216.22 0.27 0.12 0.67 0.07 0.09 20.015 0.805 2,4-Dichlorophenol 225 26.96 211.28 0.23 0.12 0.42 0.04 0.06 20.010 0.857 * Dashes indicate that the parameter was eliminated from the model.Fig. 2 Response surface for the central levels of (a) pressure (b) time and (c) power for the central composite design.Analyst, August 1998, Vol. 123 1681standard and sample extract that was analysed by GC–MS. Each calibration standard and sample extract was injected in triplicate. Chromatographic peak areas were fitted by linear regression and the correlation coefficients ranged from 0.998 to 0.999. For quantification, the average relative response factors obtained from the multilevel calibration was used. In addition, the relative response factors obtained daily were checked following Sheward representations.15 As can be deduced from Fig. 1(a) for phenol, all data are under statistical control. All the other phenols presented a similar pattern. Other extracts were analysed on a Hewlett-Packard Model 5890 Series II gas chromatograph equipped with a flame ionisation detector. The chromatographic conditions were as given above. The detector was kept at 300 °C. A five-point internal standard calibration using standards of 1, 2, 4, 6 and 8 mg ml21 and always injecting the same volume (1 ml) of the standards was performed daily to esablish the GC–FID calibration curve. 4-Methoxyphenol (internal standard) was added to every calibration standard and sample extract that was analysed by GC–FID. Each calibration standard and sample extract was injected in triplicate. Chromatographic peak areas were fitted by linear regression and the correlation coefficients ranged from 0.983 to 0.999. For quantification, the average relative response factors obtained from the multilevel calibration were used and their statistical control was checked through the Sheward plot [Fig. 1(b)]. Results and discussion Fractionated factorial design for MAE The variables considered in the fractionated factorial desiere pressure of extraction, percentage of microwave power, extraction time, volume of solvent and percentage (v/v) of acetone in the acetone–hexane mixture. The aim of this design was to evaluate which of the variables were factors, that is, which of the variables had an influence on the extraction process and which did not.Owing to the high number of experiments for a two-level factorial design (25), two levels of fractionality were introduced and the percentage of acetone and the volume of the solvent were defined as a combination of the other three variables (pressure, power and time) as shown in the equations. percentage of acetone M pressure 3 time 3 power (1) volume M pressure 3 time (2) The percentage of acetone and solvent volume were chosen for the fractionality since they were a priori the variables which least interacted with the other three (pressure, time and power).A two-level fractionated factorial design involving eight runs plus three central points was chosen. The design matrix and the results for the 11 runs (which were measured by GC–MS and whose percentage recoveries were corrected with the nitrogen blowdown evaporation recovery) are given in Table 1.The data analysis of the results given in Table 1 was performed using the non-linear regression analysis program NLREG.16 Response surfaces (Y) were taken as a function of the considered variables (xi) using polynomials of different degree depending on the experimental design followed. The general polynomial function is Y x x x i i ij i j ij i = + +Â Â b b b 0 (3) where Y is the recovery, xi are the variables considered for the optimisation of the extraction and bi and bij are the parameters to be calculated.The estimation of the parameters (bi and bij) was achieved by the minimisation of the square sum of errors (U) as given by the equation U Y Y i n = - Â( ) exp calc 2 (4) The analysis of the output was based on the evaluation of the prob(t) parameter associated with each bi parameter, since prob(t) indicates the probability of bi being zero. Those parameters whose probability of being zero was greater than 10%, i.e., prob(t) > 0.1, were systematically eliminated.The most general function allowed for the fractionated experimental design is Y = b0 + b1x1 + b2x2 + b3x3 + b4x4 + b5x5 + b13x1x3 + b23x2x3 (5) where x1 is the pressure in the extraction vessel, x2 is the extraction time, x3 is the percentage of microwave power, x4 is the percentage (v/v) of acetone and x5 is the solvent volume. b12 and b24 were not considered because they were equivalent to b5 and b13, respectively. The reason for this equivalence lies in the way variables x4 and x5 were defined (x5 M x1x2; x2x4 M x1x3).From the values obtained for the parameters for the fractionated factorial design it could be concluded that all variables [pressure, time, power, percentage (v/v) of acetone and volume] have an influence in the extraction process. For all phenols the parameters were of the same sign and magnitude with the exception of 2-methylphenol, for which pressure seemed to have no effect (parameter b1 = b13 = 0).Central composite design for MAE From the fractionated factorial design, none of the variables could be excluded because all of them [pressure, time, power, percentage (v/v) of acetone and volume] seemed to have an influence on the extraction process. Because of this, the percentage of acetone in the acetone–hexane mixture and the volume were fixed to at 50 : 50 (v/v) and 15 ml, respectively, and a central composite design was built using the other three Table 4 Design matrix and corrected recovery values in the complete factorial design for Soxhlet experiments Design matrix Recovery (%) Time/ Acetone 2-Chloro- 2-Methyl- 2-Nitro- 2,4-Dichloro- Run No.min (%) Phenol phenol phenol phenol phenol 1 8 30 51 68 14 57 98 1 8 30 72 68 14 61 85 1 8 30 65 68 15 40 101 2 16 30 61 77 11 42 90 3 8 70 107 69 15 52 94 4 16 70 95 84 14 60 80 1682 Analyst, August 1998, Vol. 123variables: pressure inside the extraction vessel, extraction time and percentage of microwave power. A two-level factorial design plus star orthogonal composite design involving 14 runs plus three central points was chosen.Table 2 gives the design matrix for this experiment and the net recoveries obtained in each run. The data in Table 2 were analysed by NLREG following the methodology mentioned above. The most general function for central composite design is Y = b0 + b1x1 + b2x2 + b3x3 + b12x1x2 + b13x1x3 b23x2x3 + b11x2 1 + b22x2 2 + b33x2 3 + b123x1x2x3 (6) where x1 is the percentage of microwave power, x2 is the extraction time and x3 is the pressure inside the extraction vessel.The results obtained by NLREG are given in Table 3. From the output data the extraction recoveries were not directly affected by pressure (parameters b3 = b33 = 0). Nevertheless, this does not mean that pressure presented no effect during the extraction since interaction parameters b13 (power–pressure). b13 (time–pressure) and b123 (power–time–pressure) were different from zero.As for the fractionated factorial design, all parameters showed the same sign and magnitude for the five phenols. Three-dimensional representations keeping one of the variables at the central point value are presented in Fig. 2 for 2-chlorophenol. At constant pressure (P = 21 psi) a maximum around 35–40% for power and 10–15 min for the extraction time could be observed [Fig. 2(a)]. On the other hand, at time = 25 min [Fig. 2(b)], the maximum was located at power = 40– 50% and pressure = 15–20 psi.Finally, for power = 57% [Fig. 2(c)], a maximum appeared for pressure = 25–30 psi and time = 15–25 min. It could be concluded that the optimum conditions were a low percentage of microwave power (approximately 35–40%), a high pressure inside the extraction vessel (approximately 25–30 psi) and a moderate extraction time (approximately 15–25 min). Nevertheless, these 3D plots do not consider the variation of the third variable, so the common optimum conditions for the five phenols might be estimated by the simplex method implemented in the Multi- Simplex program.For this purpose, the parameters given in Table 3 were introduced into an Excel spreadsheet, where the extraction recoveries for each of the phenols were calculated under each set of experimental conditions proposed by MultiSimplex. Thus, it was possible to obtain the ranges in which the variables should change in order to obtain a recovery as close to 100% as possible. Eventually, despite the simulation of the extraction for a high number of experiments, only phenol, 2-nitrophenol and 2,4-dichlorophenol could reach approximately 100% recovery.For the rest of the phenols recoveries of approximately 88% (2-chlorophenol) and 75% (2-methylphenol) were obtained. These recoveries were obtained for a percentage of microwave power of 44%, a time of extraction of 16.5 min and a pressure inside the extraction vessel of 26 psi.Complete factorial design for Soxhlet extraction The variables considered in the complete factorial design for Soxhlet extraction were the percentage of acetone in the acetone–hexane mixture and the extraction time. A two-level complete factorial design involving four runs was chosen. The design matrix and the results for the six runs (the first run of the experiment was carried out in triplicate) are given in Table 4. Samples were measured by GC–FID and losses that occurred during evaporation were corrected.The most general function allowed for the complete factorial design carried out is Y = b0 + b1x1 + b2x2 + b12x1x2 (7) where x1 is extraction time and x2 is percentage (v/v) of acetone in the acetone–hexane mixture. The results obtained by NLREG did not show a similar trend that could be observed (sign or magnitude) among the parameters calculated. For this reason, the optimum conditions were fixed taking into account the experimental data in Table 4.It could be observed that increasing the percentage (v/v) of acetone in the solvent gave better recoveries but an increase in the extraction time did not yield much higher recoveries. Hence optimum conditions for the extraction of the five phenols in soils by Soxhlet extraction were 200 ml of a acetone–hexane mixture (70 + 30 v/v) under reflux for 8 h. Table 5 Recoveries for 15 samples of 2-chlorophenol extracted for 3 days by MAE Day No. Sample No. 1 2 3 1 67 69 82 2 78 66 76 3 78 73 75 4 70 69 87 5 69 71 80 Table 6 Relative standard deviations for the five phenols using MAE RSD (%) Within Among Compound Analysis days days Total Phenol 3.03 4.84 4.84 7.48 2-Chlorophenol 2.16 5.40 6.76 8.92 2-Methylphenol 1.97 4.08 10.20 10.99 2-Nitrophenol 4.18 30.00 18.89 30.80 2,4-Dichlorophenol 1.88 12.50 7.69 14.80 Table 7 Relative standard deviations for the five phenols using Soxhlet extraction RSD (%) Within Compound Analysis days Total Phenol 2.45 17.06 17.23 2-Chlorophenol 2.32 0.85 2.47 2-Methylphenol 10.47 7.82 13.07 2-Nitrophenol 20.91 21.16 29.75 2,4-Dichlorophenol 1.91 8.96 9.16 Fig. 3 Recoveries of the five phenols under optimum conditions for both MAE and Soxhlet extraction methods. Analyst, August 1998, Vol. 123 1683Evaluation of reproducibility for MAE For evaluating the reproducibility of the measurements for MAE, five aliquots of the sample were extracted each day and this procedure was repeated three times under the optimum conditions mentioned above [15 ml of acetone–hexane (1 + 1 v/v) at 26 psi with a microwave power of 44% for 16.5 min).In other studies,7–11 the reproducibility of MAE among days was not studied. The results obtained for 2-chlorophenol during the three days are given in Table 5. The results were analysed by means of analysis of variance (ANOVA) of the set of experimental data in Table 5. It could be observed, for a degree of confidence of 95%, that there were no significant differences among samples extracted within a day (Fexp = 0.18 < Fcrit = 3.84), but there were significant differences for samples extracted among days (Fexp = 5.47 > Fcrit = 4.46).This behaviour was similar for phenol, 2-chlorophenol and 2-methylphenol. 2-Nitrophenol showed significant differences for samples extracted within a day but not among days (the RSDs were too large and all values among days were not significantly different) and 2,4-dichlorophenol showed no significant differences either within or among days, the RSD being similar in both cases.As a consequence of these results, it was decided to express the total variance of the measurements as the sum of the variance due to the analysis, the variance within days and the variance among days as indicated in the equation stot 2 = sa 2 + sw,d 2 + sa,d 2 (8) where stot 2 is the total variance, sa 2 is the variance due to the analysis (which was calculated considering error propagation and making use of the spreadsheet method17), sw,d 2 is the variance within days and sa,d 2 is the variance among days. Table 6 gives the relative standard deviations for the analysis, for samples within a day, samples among days and total relative standard deviations estimated from eqn.(8) The RSD values were good for all phenols except 2-nitrophenol. Evaluation of reproducibility for Soxhlet extraction For studying the reproducibility of Soxhlet extraction, three samples were extracted under the same conditions (run No. 1 in the experimental design, Table 4). The variance associated with each measure was expressed as in eqn. (8). In this case the variance among days was not considered in order to estimate the total variance of the results. Table 7 gives the relative standard deviations corresponding to the analysis and samples extracted within a day for Soxhlet extraction. Comparison of Soxhlet and MAE methods The mean values for the extraction carried out by MAE and Soxhlet extraction under the optimum conditions [15 ml of acetone–hexane (1 + 1 v/v) at 26 psi with a microwave power of 44% and for 16.5 min and 200 ml of acetone–hexane (70 + 30 v/v) under reflux for 8 h, respectively] were compared.Fig. 3 shows the recoveries for the five phenols under the optimum conditions for both the MAE and Soxhlet extraction methods. It can be concluded that for phenol, 2-chlorophenol and 2,4-dichlorophenol comparable results were obtained by both methods.For 2-methylphenol and 2-nitrophenol better results were obtained by MAE. In both methods, the recoveries obtained for 2-methylphenol were low, but other workers have reported low recoveries for 2-methylphenol18 and other alkylphenols.19 These low recoveries are attributed to interactions between the alkyl groups and the soil matrix. The relative standard deviations for MAE are higher than those obtained for Soxhlet extraction but it must be taken into account that the variance among days was considered for MAE but not for Soxhlet extraction.This research was supported by the Basque Government (project PIGV96/74). Olatz Zuloaga is grateful for a Scholarship granted by the Basque Government. References 1 Herberer, T., and Stan, H., Anal. Chim. Acta, 1997, 341, 21. 2 Ashraf-Khorassani, M., Gidanian, S., and Yamini, Y., J. Chromatogr. Sci., 1995, 33, 658. 3 Llompart, M., Lorenz, R., and Cela, R., J. High Resolut. Chromatogr., 1996, 19, 207. 4 Bosch, F., Campins, P., and Verdu, J., J. Liq. Chromatogr., 1995, 18, 2229. 5 Heemken, O. P., Norbert, T., and Wenclawiak, B. W., Anal. Chem., 1997, 69, 2171. 6 Snyder, J., Grob, R., McNally, M., and Oostdyk, T., Anal. Chem., 1992, 64, 1940. 7 Barnabas, I. J., Dean, J. R., Fowlis, I. A., and Owen, S. P., Analyst, 1995, 120, 1897. 8 Llompart, M. P., Lorenzo, R. A., Cela, R., and Par�e, J. R. J., Analyst, 1997, 122, 133. 9 Chee, K. K., Wong, M. K., and Lee H. K., J. Chromatogr. A, 1996, 723, 259. 10 Onuska, F. I., and Terry, K. A., J. High Resolut. Chromatogr., 1995, 18, 417. 11 Pastor, A., V�azquez, E., Ciscar, R., and de la Guardia, M., Anal. Chim. Acta, 1997, 344, 241. 12 Nelder, J. A., and Mead, R., Comput. J., 1965, 7, 308. 13 Bergstr�om, and � Oberg, MultiSimplex (1.0 Version), Gullberna Park, Karlskrona, 1997. 14 ISO/DIS 11464, Soil Quality–Pretreatment of Samples for Physicochemical Analyses, US EPA, Washington, DC, 1991. 15 Kateman, G., and Buydens, L., Quality Control in Analytical Chemistry, Wiley, New York, 2nd edn., 1993. 16 Sherrod, P. H., NLREG—Nonlinear Regression Analysis Program, Nashville, TN, 1995. 17 Kragten, J., Analyst, 1994, 119, 2161. 18 Lopez-Avila, V., and Young, R., Anal. Chem., 1994, 66, 1097. 19 Dean, J. R., Santamaria-Rekondo, A., and Ludkin, E., Anal. Commun., 1996, 33, 413. Paper 8/02117A Received March 11, 1998 Accepted May 29, 1998 1684 Analyst, August 1998,

ISSN:0003-2654    

DOI:10.1039/a802117a

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

Solid-phase reactors as high stability reagent sources in flow analysis: selective flow injection spectrophotometric determination of cysteine in pharmaceutical formulations M. Catal�a Icardo,* L. Lahuerta Zamora* and J. Mart�ýnez Calatayud*b a Departamento de Qu�ýmica, Colegio Universitario de Farmacia C.E.U. San Pablo., Edificio Seminario, 46113 Moncada, Valencia, Spain b Departamento de Quimica Analitica, Universitat de Val`encia, Doctor Moliner, 50, 46100 Burjassot, Valencia, Spain The flow injection spectrophotometric determination of cysteine was carried out by reaction with cobalt(II) ions entrapped in a polymeric material and filling a packed-bed reactor; the released cobalt(II) complexed with the amino acid was monitored at 360 nm.The method worked with a high repeatability, even with independent reactors, days and solutions. Selectivity of the procedure was tested with twenty different foreign compounds found in pharmaceutical formulations containing cysteine, parent amino acids included; no serious interferences were observed.The calibration graph for cysteine was linear over the range 1–90 mg ml21 with a relative standard deviation of 0.8% at 60 mg ml21 (n = 158). The calculated sample throughput was 90 h21. The method was applied to determine the content of cysteine in pharmaceutical formulations. Keywords: Flow injection; solid-phase reactors; spectrophotometry; pharmaceuticals; cysteine; Simplex optimisation.Solids have been used as analytical reagents for virtually as long as dissolved reagents, even though their scope of application1 has been usually restricted to oxidations or reductions prior to some redox titrimetries. This restriction was mainly due to the time consuming experimental procedures as well as the lack of control of the undesirable side-reactions and the reaction extension. Solid-phase reactors have become one of the most interesting trends in continuous-flow methodologies in spite of those cited shortcomings, mainly due to the advantages offered over dissolved reagents, namely, simpler manifolds, saving reagents, increased sensitivity and injection rate.Moreover, and bearing in mind that in a flow injection analysis (FIA) manifold the time of contact sample and solid reagent can be controlled with precision, employing of solid-phase reactors as reagent sources could improve the kinetic selectivity ahead of their counterparts in homogeneous systems.The use of solid-phase reactors in FIA has been reviewed2,3 and also its application to pharmaceutical analysis.4,5 This research group has proposed and is developing an original, quick and easy strategy on reagent immobilization for use in continuous-flow methodologies. Polymerisation reactions from linear non-saturated polyester chains allow the preparation of reactive beds with a high mechanical, chemical and microbiological stability,6, and also with the desired preselected configuration and dimensions for the column and the suitable size of the particle.This strategy has allowed the rescue, through their immobilisation, of several water insoluble reagents all of them commercially available as a fine powder, which have made possible indirect determinations of a number of drugs, mainly in pharmaceutical formulations; namely: manganese dioxide for isoniazid,7 lead dioxide for metamizol8 and ondansetron,9 copper carbonate for amino acids6 and tetracyclines,10 iron(iii) phosphate for phenothiazines11 and silver iodide for cyanide. 12 On the other hand, solutions of cobalt(ii) have been employed as a suitable reagents to determine a variety of drugs. Usually, the drug acts as a complexing agent to yield a complex which is monitored through direct (UVIVIS,13–19 fluorescence,20 polarography21) or indirect (chemiluminescence22,23) procedures. To continue with the development of this strategy of immobilisation and bearing in mind the good results obtained with copper carbonate,9,10 the immobilisation of a new solid, fine-powdered commercial reagent, cobalt carbonate, and its application to selective analysis of cysteine in pharmaceutical formulations is proposed.Moreover, this is a pioneer article dealing with the use of solid-phase reactors as a source for controlled reproducible solutions avoiding the tedious experimental process of preparing the reagent solution. Experimental Reagents Reagents were obtained as follows: Polyester resin solution Al-100 (Reposa, Barcelona, Spain); ethyl methyl ketone (Akco, Barcelona, Spain); l-Cysteine, Nacetyl- l-cysteine, l-threonine, l-glutamine, l-serine, l-isoleucine, l-proline, l-aspargine monohydrate, l-histidine, benzalkonium chloride, pyridoxine hydrochloride, saccharin (all from Guinama, pure); l-arginine, l-phenylalanine, dl-tryptophan, l-lysine hydrochloride (Merck, Darmstadt, Germany); glycine, l-cystine (UCB, analytical-reagent grade); l-tyrosine, l-leucine (UCB, pure); l-methionine (Scharlau, pure); lactose 1-hydrate, potassium iodide, sodium thiosulfate pentahydrate (Panreac, pure); cobalt chloride hexahydrate (Panreac, analytical- reagent grade); nitroso R-salt, dl-penicillamine (Fluka, Buchs, Switzerland); alanine, glutamic acid, aspartic acid and valine.Procedures and apparatus Preparation of the bed reactor CoCO3·6H2O (30 g) was added to 30 g of the polyester resin solution. The mixture was homogeneized by manual stirring and then 1.2 g of ethyl methyl ketone were added and stirring was continued until the polymer became too rigid.The solid was dried for 2–3 h at room temperature, then broken with a hammer and ground in a coffee grinder, a suitable size (between Analyst, August 1998, Vol. 123 (1685–1689) 1685100 and 200 mm) subsequently being selected by sieving. Finally, the selected particles were washed, dried at 80 °C, sieved again and stored.The solid-phase reactor was prepared by introducing the particles via a mini-funnel with stirring into a PTFE tube of 1.5 mm id. Flow injection procedure Aliquots of sample (182 ml of cysteine in 5 3 1024 m sulfuric acid) were injected by means of a Rheodyne (Cotati, CA, USA) Model 5041 injection valve into a pure water stream, at a flow rate of 2.3 ml min21 [see Fig. 1(a)]. The inserted acidic solution released CoII when flowing through the solid-phase reactor (14 cm3 1.5 mm id) packed with the reagent (entrapped CoCO3 by a polymeric resin; ratio of reagent/resin, 1 : 1; particle size, 100–200 mm).Then the inserted solution merged with a sodium tetraborate buffer solution (0.04 m, pH 9.5; flow rate, 1.8 ml min21) and the mixture was led to the detector flow-cell (reactor length 117 cm and 0.5 mm id) that monitored the absorbance of the reaction product at 360 nm. The peristaltic pump was a Minipuls 2 from Gilson (Villiers-le-Bel, France).The PTFE tubing was of 0.5 and 1.5 mm id for the manifold and the bedreactor column, respectively. A Hewlett-Packard (Avondale, PA, USA) diode array spectrophotometer (8452A) was used and was provided with a flow-cell (from Hellma, Jamaica, NY, USA) 18 ml volume. A Perkin Elmer (Offenbach, Germany) atomic absorption spectrometer (Zeeman 5000) was also used to test for the released CoII by the acid solution. Results and discussion Preliminary tests We carried out a preliminary qualitative study in tube tests.A volume of 1 ml of a 1023 m aqueous solution of CoII was mixed with 1 ml of solutions of various amino acids and other drugs at a 500 mg ml21 concentration in order to examine the formation of coloured complexes at different pH values from 2 to 10.5. The compounds tested included cysteine, histidine, lysine, penicillamine, penicillin, leucine, N-acetyl-l-cysteine, phenformin, arginine, proline, serine, tyrosine, alanine, methionine, glycine, chloramphenicol, paracetamol, erythromycin, doxycycline hydrate, terramycin, gentamicin sulfate, ampicillin trihydrate, sulfanilamide, sulfamethoxypyridazine, sulfamethoxazole, sulfamerazine, streptomycin sulfate, ascorbic acid and phenylalanine. The reaction was positive for pH values above 6 for cysteine and penicillamine, both of which formed a yellow-coloured complex.Histidine, lysine and N-acetyl-l-cysteine formed complexes of ar yellow colour at pH values over 8.The pH range in which the complexes are formed are according to the reported protonation constants in the analytical literature;29 pH ranges resulted in the anionic form of the ligand that starts to be the predominant concentration. The behaviour of the system under flow conditions was studied by using the assembly depicted in Fig. 2. Experiments were carried out at room temperature and a flow rate of 2.1 ml min21 in each manifold channel. The tested compounds were cysteine, penicillamine, N-acetyl-l-cysteine, lysine and histidine, all at a 200 mg ml21 concentration.Cobalt was used in 0.01 m sulfuric acid because preliminary experiments with the reactor resulted in this medium releasing the greatest amount of metal. The third channel was used to circulate a buffer solution in order to establish the alkaline medium required for optimal formation of the cobalt complexes. The buffers tested were HPO4 22–H2PO42 at pH 7.6 and NH4 +–NH3 at pH 8.8.Under these working conditions, only cysteine and penicillamine were found to react with the cobalt; the highest absorbances were measured at pH 8.8, with absorption maxima at 288 and 296 nm for cysteine and penicillamine, respectively. However, in order to use as distant as possible a wavelength from the UV region and minimize the resulting interferences, we chose to make measurements at 360 nm, where the absorption band exhibited a plain shoulder. The calibration curves thus obtained for cysteine and penicillamine exhibited similar linear ranges; however, that for cysteine had a slightly greater slope, so it was chosen for subsequent experiments.We then carried out a preliminary study of potential interferents by using solutions containing cysteine and the other 20 primary protein-forming amino acids, all of which were employed at a 200 mg ml21 concentration. Only histidine was found to interfere with the signal at 360 nm. The experiment was conducted in the manifold of Fig. 2, using an NH4 +–NH3 buffer of pH 8.8 and a 7x1024 m solution of cobalt in 0.01 m H2SO4. Each of the above reported experiments were conducted in a freshly prepared reactor that was 14 cm long and 1.5 mm id and contained the reagent and resin in a 1 : 1 ratio; the particle size was 300–400 mm. Study and optimisation of chemical and FIA variables The influence of the acid circulated through the reactor was examined by using the flow assembly depicted in Fig. 3.Once released by an acid solution, the CoII was neutralized with sodium hydroxide and merged with a solution of nitroso R salt24,25,26 to form a strongly coloured complex in an acetic– Fig. 1 Tested FIA manifolds. The selected manifold is depicted in 1(a). For details see text. P, pump; B-R, solid-phase reactor; D, detector; and, W, waste. Q1 is a stream of H2O at flow-rate of 2.3 ml min21; Q2, borate buffer pH 9.5 at 1.8 ml min21. V, 182 ml of cysteine in 5 x 1024 m sulfuric acid; L, 117 cm.Fig. 2 Continuous-flow manifold for the study of the reaction between CoII and some drugs. 1686 Analyst, August 1998, Vol. 123acetate medium. The tested acids were HCl, HClO4, HNO3, HAcO and H2SO4, all at a 0.05 m concentration and a 1.4 ml min21 flow-rate. With all the acids, the signal remained constant for at least 1 h after an activation period of 20 min. Cobalt release was maximal with sulfuric acid. Next, we studied the effect of the H2SO4 concentration by using 0.05, 0.025, 0.01, 0.005 and 0.002 m solutions of the acid.The lowest concentration was used to examine the stability of the reactor at 40 °C, conditions under which the signal decreased with time, with no clear-cut zone of increased stability. We also examined the influence of the flow-rate on the stability of the reactor. As in previous experiments, the reactor remained stable for at least 1 h at the three flow rates studied, viz., 1.0, 1.4 and 1.8 ml min21.The highest value was adopted because it resulted in the shortest stabilisation time for the solidphase reactor (15 min). In all subsequent experiments, the solid-phase reactor was prepared by passing 0.01 m H2SO4 at a flow-rate of 1.8 ml min21 for 15 min. The effects of FIA variables were analysed by using the three types of assembly depicted in Fig. 1. In all, Q2 was NH4 +–NH3 buffer circulated at pH 8.8 at 2.1 ml min21, L1 = 14 cm, L2 = 65 cm and V = 200 ml of a 130 mg ml21 solution of cysteine in 0.01 m H2SO4 (the cysteine used in the third assembly was prepared in the buffer solution).The first assembly [Fig. 1(a)] was studied under two sets of conditions that differed solely in the nature of carrier Q1 (water and 0.01 m H2SO4, respectively). The second assembly [Fig. 1(b)] was used with 0.01 m H2SO4 as carrier; however, the injection valve was placed behind the reactor in order to avoid passage of the analyte through it. The sample used in the third assembly [(Fig. 1c)] was prepared in the buffer and inserted into channel Q2; the distance to the merging point was 25 cm and carrier Q1 was 0.01 m H2SO4. The analytical signal obtained was identical under the two sets of conditions used with the first assembly, 10% lower with the third and 50% lower with the second. The first assembly was adopted for further work because it resulted in the longest reactor lifetime. The chemical variables and those related with the solid-phase reactor were optimised by using the univariate method.A calibration was performed at each experimental value (n = 5) and that leading to the highest signal and/or largest slope was selected. The assembly used was that of Fig. 1(a), with the shortest possible L1 and L2 = 75 cm. We initially studied the effects of the nature, concentration and pH of the buffer, and then examined the influence of the H2SO4 concentration used to release the cobalt from the reactor.Table 1 shows the chemical variables examined, the ranges studied and the values selected. After this chemical pre-optimisation, we investigated the influence of the reactor variables, which are shown in Table 1 alongside the ranges examined and values selected. The choice of the solid-phase reactor was dictated not only by the quality of the analytical signal, but also by the fact that some conditions leading to increased cobalt release might also produce a ‘blank’ signal by partial precipitation of the metal in the alkaline medium used.The H2SO4 concentration was also re-optimised with this aim; the concentration adopted, 1023 m, resulted in a linear range from 40 to 140 mg cysteine ml21. All other FIA variables were optimised by the multivariate method, using the modified simplex.27,28 Table 2 shows the ranges studied and values selected for each parameter; the program was centered after 15 apices. In order to obtain more precise results, a second simplex was conducted over more narrow ranges that was centred after 15 apices; the results are also shown in Table 2.Once all FIA variables were optimised, chemical variables were re-optimised. The ranges studied and values selected are given in Table 2. Study of the stability and repeatability of the solid-phase reactor The next step in this work involved examining the stability and repeatability of the reactor under the working conditions used. Solid–liquid reactivity can be influenced for various reasons and by different factors, either separately or combined: the composition of the solution, additives in the solid, illumination of the solid, etc. When a solid reacts with a solution the products may form a layer on the solid surface, even in continuous flow where the reagent stream is continuously ‘washing’ the solidphase reactor.On the other hand, when a solution is continuously forced through the reactor, its reactivity can be altered by the combined influence of hydrodynamic and chemical action.In this work the reproducibility and life-span of the reactor is more relevant because the solid-phase reactor is acting Fig. 3 FIA manifold for testing the life-span of the solid-phase reactor. Q1 (acid stream) and Q2 (sodium hydroxide stream), 1.4 ml min21; Q3, reagent in acetic–acetate buffer at pH 5.8 at 1.3 ml min21; L1, 24.5 cm; L2, 241.5 cm; absorbance at 526 nm. Table 1 Study of chemical variables and solid-phase reactor features Selected Variable Tested values value Buffer solution Monohydrogen Tetraborate– (0.2 m; pH 8.6) phosphate–dihydrogen sodium phosphate hydroxide Ammonium–ammonia Tetraborate–sodium hydroxide Buffer solution 0.025; 0.040; 0.070; 0.200 0.04 concentration/m pH 8.0; 8.6; 9.0; 9.3; 9.5; 10.0 9.5 Sulfuric acid 0.001; 0.003; 0.006; concentration/m 0.010; 0.015; 0.020 0.010 Internal diameter/mm 0.8; 1.5; 2.0 1.5 Particle size/mm 300–400; 200–300; 100–200 100–200 Concentration reagent: 1 : 1; 1 : 2; 1 : 3; 1 : 4; 1 : 5 1 : 1 resin (g : g) Column length/cm 5; 10; 14; 20; 25; 30 14 Re-optimization of 2.0 3 1023; 1.0 3 1023; 5 3 1024 sulfuric acid 8 3 1024, 5 3 1024 concentration/m Re-optimization of buffer 0.02; 0.04; 0.06; 0.08 0.04 solution cocentration/m Re-optimization of the 8.5; 9.0; 9.3; 9.5; 9.8 9.5 buffer pH Table 2 Multivariate optimisation of FIA parameters: simplex method Simplex 2 Simplex 1 Studied Studied Studied Variable range range value Q1/ml min21 1.0–2.6 1.4–2.4 2.3 Q2/ml min21 0.8–5.0 1.5–2.8 1.8 V/ml 112–309 151–289 182 L/cm 20–300 80–250 117.1 Analyst, August 1998, Vol. 123 1687H2N S NH2 S COII OH2 OH2 CH2 CHCOO– H2C –OOCHC –2 as a diluted reagent source.29 Bearing in mind all those questions, an overall 158 insertions of 60 mg ml21 cysteine in 5 3 1024 m H2SO4 produced a relative standard deviation (RSD) of 0.8%. We also performed calibrations of cysteine over the range 10–90 mg ml21 (five points) using the same reactor in 15 consecutive sessions; it was observed that the signal remained constant throughout.The mean slope obtained was 3.693 1023 mlmg21, with an RSD of 1.6%. The between-day reproducibility study was carried out by using seven reactors and freshly prepared solutions; seven different calibration graphs were performed (range 10–90 mg ml21) and the calculated mean slope was 3.703 1023 ml mg21, with an RSD of 1.9% (n = 7). Study of CoII released from the reactor The amount of CoII released from the reactor was quantified by using a single-channel system through which 50 mg ml21 cysteine in 5 3 1024 m H2SO4 was circulated at the usual working flow-rate; the line was fitted to an AAS detector.The signal remained constant for at least 130 min after the activation period, outputs were recorded each 10 s interval; the 780 outputs recorded allowed a mean absorbance of 0.1819 to be obtained with an RSD of 1.0%. Calibration for CoII was performed at the same flow-rate after the reactor was removed in order to determine the amount of analyte released from the bed, which was found to be 2 3 1024 m.The same experiment was conducted in an assembly such as that depicted in Fig. 1(a) with the injection valve excluded; sulfuric cysteine solution was continuously circulated through a UV/VIS detector. The outputs were recorded at each 0.5 s interval; the total number of recorded outputs was 7200. The experiment was monitored over three working sessions on as many consecutive days.For the first and second days, the signal remained constant for 3 h after the activation period; on the third, it started to decrease slowly with time (the decrease was 2.8% at 42 min) after 1 h of constant measurements. Bearing in mind the flow-rate, 2.3 ml min 21 and the sample volume, 182 ml, the forced sample volume through the solid-phase reactor was equivalent to 5080 samples. Study of the CoII–cysteine complex To study the Co-cysteine complex obtained we compared the spectra of the resulting solution when the acid solution was forced through the solid-phase reactor versus the spectra of the solutions resulting from mixing CoII and cysteine solutions, both at the same pH (9.5) with the aid of the reported buffer solution.Spectra were identical from both procedures a maximum absorbance at 296 nm and a ‘shoulder’ around 360 nm. Then we studied the stoichiometry of the complex formed between cysteine and cobalt. It was determined by using two different methods, viz., the continuous-variation or Job method, and the mole-ratio method.In both cases, the cysteine and drug solutions were mixed and the absorbance of the resulting product at 360 nm was measured after 1 min of reaction. With the Job method, the combined concentration of the drug and CoII was kept constant at 8.0 3 1024 m at a variable mole ratio between the two (0.16–0.98 for cysteine).The cysteine-to- CoII stoichiometric ratio thus calculated was 2.1 : 1. With the mole-ratio method, the CoII concentration was kept constant at 4 3 1024 m while the drug concentration was varied between 131024 and 14.531024 m. The cysteine-to- CoII ratio thus obtained was 2.2 : 1. Based on the previous results, the cysteine–cobalt complex studied is the Co atom bonded to S and NH2 groups of two cysteine molecules30,31 and two H2O molecules as is depicted in Fig. 4. No other chelate was formed since the chelate rings involving S and N should be very stable and it is not possible to convert one into another using the same conditions in which the chelates were originally formed.Analytical applications The linear range comprised the values between the limit of detection, which was experimentally determined as the concentration yielding a signal more than three times the baseline width (1 mg ml21, RSD = 3%), and 90 mg ml21 cysteine. The average insertion rate for 30 peaks was 90 insertions h21 The RSD thus obtained was 0.8%.Interferences were sought among all the primary proteinforming amino acids studied, some of the usual vitamins in commercially available formulations and their most common excipients. The potential interferents were added to 60 mg ml21 cysteine solutions in 5 3 1024 m H2SO4. The relative errors obtained are given in Table 3; no interferences were observed. The interference of benzalkonium chloride and vitamin B12 was only apparent provided the amount found in pharmaceutical formulations with respect to 60 mg ml21 of cysteine exceeded 0.38 mg ml21 and 0.096 mg ml21 for benzalkonium chloride and vitamin B12, respectively. It should be emphasized that penicillamine (as was stated in the preliminary test) gives the same reaction as cysteine, nevertheless, as far as we know, both drugs are not present simultaneously in pharmaceutical formulations.The proposed method was used to analyse various commercially available pharmaceutical preparations.The results were compared with those stated by the manufacturer and those Fig. 4 Structure of the complex CoII-cysteine.31 Table 3 Influence of foreign compounds. All solutions contained 5 3 1024 m sulfuric acid and 60 mg ml21 of cysteine Concentration/ Foreign compound mg ml21 Er (%) Aspartic acid 500 23.0 Glutamic acid 1000 22.7 Alanine 500 23.0 Arginine 100 23.0 Asparagine 1000 21.0 Benzalkonium chloride 8 23.0 Caffeine 1000 1.5 Cystine Saturated 22.4 Sodium chloride 1000 21.1 Phenylalanine 1000 2.2 Glycine 1000 3.0 Glutamine 1000 23.0 Histidine 100 2.7 Isoleucine 1000 20.04 Lactose 1000 0.9 Leucine 1000 20.8 Lysine 1000 22.6 Methionine 1000 2.0 Pyridoxine 1000 1.7 Proline 1000 21.9 Saccharin sodium 1000 0.4 Sucrose 1000 0.16 Serine 1000 21.3 Tyrosine 350 3.0 Threonine 1000 1.5 Tryptophan 1000 21.1 Valine 1000 22.5 Vitamin B1 1000 0.11 Vitamin B12 6 1.6 1688 Analyst, August 1998, Vol. 123provided by the British Pharmacopoeia’s recommended method32 in (see Table 4). References 1 Pickering, W. F., Chem. Anal., 1964, 53, 91. 2 Garc�ýa Mateo, J. V., and Mart�ýnez Calatayud, J., Chem. Anal. (Warsaw), 1993, 38, 1. 3 Mart�ýnez Calatayud, J., Flow Injection Analysis of Pharmaceuticals. Automation in the Laboratory, Taylor and Francis, Cambridge, 1996. 4 Garc�ýa Mateo, J. V., and Mart�ýnez Calatayud, J., Pharm. Technol. Int., 1992, 4, 17. 5 Garc�ýa Mateo, J.V., and Mart�ýnez Calatayud, J., Pharm. Technol. Int., 1992, 4, 30. 6 Garc�ýa Mateo, J. V., and Mart�ýnez Calatayud, J., Anal. Chim. Acta, 1993, 274, 275. 7 Lahuerta Zamora;a Mateo, J. V., and Mart�ýnez Calatayud, J., Anal. Chim. Acta, 1992, 265, 81. 8 Lahuerta Zamora, L., and Mart�ýnez Calatayud, J., Talanta, 1993, 40, 1067. 9 Lahuerta Zamora, L., and Mart�ýnez Calatayud, J., Anal. Chim. Acta, 1994, 280, 145. 10 L�opez Paz, J. L., and Mart�ýnez Calatayud, J., J.Pharm. Biomed. Anal., 1993, 11, 1093. 11 Catal�a Icardo, M., Lahuerta Zamora, L., and Mart�ýnez Calatayud, J., LRA, 1998, 10, 33. 12 L�opez G�omez, A. V., Garc�ýa Mateo, J. V., and Mart�ýnez Calatayud, J., Analyst, in the press. 13 Alwarthan, A. A., and Al-Lohedan, H. A., Talanta, 1994, 41, 225. 14 Amer, M. M., El-Tarras, M. F., Fattah, S. A., and Metwally, F. M., Egypt J. Pharm. Sci., 1989, 30, 1. 15 Sen, A. K., and Das, T. K., Indian Drugs, 1989, 27, 124. 16 Kumar, Y., Rathore, Y.K. S., Mathur, S. C., Murugesan, N., and Sethi, P. D., Indian Drugs, 1992, 29, 416. 17 Hern�andez-Mart�ýnez, J., Mart�ýnez, P. J., Guti�errez, P., and Mart�ýnez, M. I., Talanta, 1992, 39, 637. 18 Besada, A., Anal. Lett., 1988, 21, 435. 19 Chatterjee, P. K., Jain, C. L., and Sethi, P. D., Indian J. Pharm. Sci., 1987, 49, 34. 20 Jie, N. Q., Yang, J. H., and Zhan, Z. G., Anal. Lett., 1993, 26, 2283. 21 Khodari, M., Ali, A. M., and El-Maali, N. A., Anal. Lett., 1993, 26, 1099. 22 MacDonald, A., and Nieman, T. A., Anal. Chem., 1985, 57, 936. 23 Alwarthan, A. A., Al-Tamrah, S. A., and Akel, A. A., Anal. Chim. Acta, 1994, 292, 201. 24 Feigl, F., An�alisis cualitativo mediante reacciones a la gota. Aplicaciones inorg�anicas y org�anicas, Paraninfo, Madrid, 1949. p. 131. 25 Willard, H., and Kaufmann, S., Anal. Chem., 1947, 19, 505. 26 Shipman, W. H., and Lai, J. R., Anal. Chem., 1956, 28, 1151. 27 Morgan, S. L., and Deming, S. N., Anal. Chem., 1974, 46, 1170. 28 Nelder, J. A., and Mead, R., Comput. J., 1965, 808. 29 Martinez Calatayud J., and Garc�ýa Mateo, J. V., Recent research developments in pure and applied analytical chemistry, Transworld Research Network, Trivandum, India, in the press. 30 Bjerrum, J., Schwarzenbach, G., and Sillen, L. G., Stability constants of metal-ion complexes, The Chemical Society, London, 1964, p. 348. 31 Martell, A. E., and Calvin, M., Chemistry of the Chelate Compounds, Prentice-Hall, Englewood Cliffs, NJ, 1956. 32 British Pharmacopoeia, Her Majesty’s Stationery Office, London, 1993. 33 Martindale, The Extra Pharmacopoeia, The Pharmaceutical Press, London, 30th edn. 1993. Paper 8/01787E Received March 4, 1998 Accepted June 12, 1998 Table 4 Pharmaceutical formulations analysed British Declared Pharmaceutical Proposed Parmacopoeia value formulation method method (manufacturer) Pil-food (capsule)/mg 81 ± 2 81 ± 2 80 Policolinosil 12.7 ± 0.3 12.75 ± 0.09 12.5 (pill)/mg Galenic preparation*/ g 100 ml21 1.618 ± 0.014 1.581 ± 0.008 1.576 * Prepared according to ref.

ISSN:0003-2654    

DOI:10.1039/a801787e

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

Flow injection determination of anionic surfactants with cationic dyes in water bodies of central India Rajmani Patel and Khageshwar Singh Patel* School of Studies in Chemistry, Pt. Ravishankar Shukla University, Raipur-492010, MP, India A new, simple and specific flow injection analysis (FIA) procedure for the determination of anionic surfactants, viz., sodium lauryl sulfate (SLS), sodium dodecyl sulfonate, sodium hexadecyl sulfonate and sodium dodecyl benzenesulfonate, with cationic dyes, viz., Brilliant Green (BG), Malachite Green, Methylene Blue, Ethyl Violet and Crystal Violet, in water bodies, viz., ponds, tube wells, rivers and municipal wastes, of central India (east Madhya Pradesh) is described.It is based on the precipitation of the cationic dyes with the anionic surfactant due to formation of an ion-associate species within the pH range 5.5–8.0. The apparent molar absorptivity of the ion-associate species formed with various anionic surfactants and cationic dyes is in the range (0.60–1.50) 3 104 l mol21 cm21 at lmax 590–665 nm.Among them, the pair BG+–LS2 was selected for detailed investigation. The detection limit (amount causing absorbance > 3s) of the method with BG is 100 ppb SLS and the sample throughput is 50 h21. Optimization of FIA and the analytical variables in the precipitation and determination of SLS with BG is described. The method is free from interferences from almost all ions which are commonly present with the surfactant.The proposed method was applied to the mapping of SLS pollution levels in the various water bodies. All surface waters and municipal waste waters and some ground waters lying near the sources were found to be contaminated with SLS beyond permissible limits. Keywords: Flow injection analysis; spectrophotometry; cationic dyes; anionic surfactants; sodium lauryl sulfate pollution level; water bodies Anionic surfactants have been reported as pollutants and their permissible limit in drinking water prescribed by the WHO is 1.0 ppm.1–3 The main sources of the commonly used anionic surfactant sodium lauryl sulfate (SLS) and others, viz., sodium dodecyl sulfonate (SDS), sodium hexadecyl sulfonate (SHDS) and sodium dodecyl benzenesulfonate (SDBS), in water bodies are household commodities and personal care products, e.g., detergents, soaps, shampoos and fabric and cosmetic materials.The level of surfactants in water bodies of densely populated countries such as India is increasing owing to the changes in lifestyle similar to western countries.Flow injection analysis (FIA) procedures for the determination of anionic surfactants based on the extraction of ion-pair species with dyes, e.g., Methylene Blue and Ethyl Violet, have been reported.4–14 Large ions (viz., I2, IO32, SCN2, RCOO2, ArO2, RNH3 +, R2NH2 +, R3NH+, etc.) form extractable ion-associate species with these dyes and interfere in the determination of anionic surfactants.In this work, a new, simple and specific FIA procedure for the determination of anionic surfactants, viz., SLS, SDS, SHDS and SDBS, based on precipitation of a cationic dye, viz., Brilliant Green (BG), is proposed. The present method is simple and reliable and overcomes most of the drawbacks of the established methods based on the extraction of the ion-associate CD+–AS2 (CD+ = cationic dye and AS2 = anionic surfactant) with comparable sensitivity.Experimental Apparatus A Tecator (H�og�anas, Sweden) Model 5012 flow injection analyser equipped with an ALPKEM (Wilsonville, OR, USA) Model 510 UV/VIS spectrophotometer with a 0.55 cm flow cell was employed. A Systronics Model 106 spectrophotometer with 1 cm quartz cells and a Systronics (Ahmedabad, India) Model 361 m-pH meter were used. The FIA configuration used in this work is shown in Fig. 1. Reagents All chemicals were of analytical grade reagents (Merck, Darmstadt, Germany).A stock standard solution (1000 ppm or 3.5 31023 mol l21) of SLS was prepared by dissolving 1.0 g of SLS in 1 l of doubly distilled water and working standard solutions were prepared by appropriate dilution of the stock standard solution. A solution of BG (0.002% m/v or 4.0 3 1025 mol l21) was prepared by dissolving 0.010 g of BG in 0.2 ml of ethanol (95% v/v) and diluting to 500 ml with doubly distilled water. Similarly, 1000 ppm (2.8 3 1023–3.7 3 1023 mol l21) solutions of other surfactants, viz., SDS, SHDS and SDBS, and 0.002% m/v (4.0 3 1025–6.3 3 1025 mol l21) solutions of other dyes, viz., Malachite Green (MG), Methylene Blue (MB), Ethyl Violet (EV) and Crystal Violet (CV), were prepared.Deionized, doubly distilled water (pH 6.0, adjusted with 0.1 mol l21 acetic acid) containing 0.001% m/v (3.4 3 1025 mol l21) potassium dichromate and 0.001% m/v (1.3 3 1024 mol l21) thiourea was employed as the carrier. Potassium dichromate and thiourea were used to oxidize and to mask ions, viz., SO3 22 and CuII, respectively.All solutions employed were filtered and de-gassed before use. Sample collection Surface, ground and municipal waste water samples from central India (east Madhya Pradesh, between latitude 18 and 23° N and longitude 80°17A and 84°11A E) were collected in 100 ml polyethylene bottles during July and November 1997 as prescribed in the literature.15 The sampling points for these samples were rural, urban, semi-urban and industrial areas.The Fig. 1 Schematic diagram of FIA configuration: C = carrier, de-ionized doubly distilled water (pH 6.0, adjusted with 0.1 mol l21 acetic acid) containing 0.001% m/v (3.4 3 1025 mol l21) potassium dichromate and 0.001% m/v (1.3 3 1024 mol l21) thiourea; R = 0.002% m/v (4.0 3 1025 mol l21) Brilliant Green in 0.04% v/v ethanol. Analyst, August 1998, Vol. 123 (1691–1695) 1691samples were filtered with Whatman (Maidstone, Kent, UK) No. 42 filter-paper and stored in a refrigerator at 4 °C. Procedure for the precipitation of the ion-associate aggregate Five aliquots of 2.5 ml of BG (0.02% m/v or 4.0 3 1024 mol l21) were taken in 50 ml beakers, mixed with 0, 5.0, 10.0, 15.0 and 20.0 ml of SLS (25 ppm) and diluted to 25 ml with doubly distilled water at room temperature (22 ± 2 °C). The pH was adjusted to 6.0 with dilute acetic acid (0.1 mol l21). All solutions were filtered with Whatman No. 42 filter-paper. Their absorbances were measured spectrophotometrically against a reagent blank (doubly distilled water, pH 6.0) at lmax 625 nm.Procedure for flow injection determination of anionic surfactant For the FIA determination of anionic surfactants, two silicon tubes of bore size 0.51 and 0.64 mm to propel doubly distilled water and dye solution, respectively, were used. A smooth baseline having zero absorbance was plotted under the optimum FIA conditions (Fig. 1). A 200 ml aliquot of the analyte solution containing up to 20.0 ppm SLS (or 7.0 3 1025 mol l21 anionic surfactant) was injected.The decrease in colour of the flowing stream was plotted at lmax. The filtered water sample was injected in the same way. The concentration of the anionic surfactant in terms of SLS in the sample was determined from the calibration curve obtained. Results and discussion Precipitation and composition of ion-associate species The cationic dyes BG, MG, MB, EV and CV form an ionassociate precipitate with the anionic surfactants SLS, SDS, SHDS and SDBS when they are present at higher concentration levels in the aqueous solution.16,17 The colour intensity decreases quantitatively when the anionic surfactant SLS is added within the pH range 5.5–8.0 owing to formation of a slightly soluble ion-associate species (Fig. 2).The decrease in colour of the dye was found to be linear up to 20 ppm SLS with a slope, intercept and correlation coefficient of 20.028, 0.988 and 0.999, respectively.Their stoichiometry was determined by the FIA technique using the curve-fitting method, plotting log (heq./hmax 2 heq) versus log (molar concentration of SLS injected). The results indicated that the cationic dye (CD+) reacted with the anionic surfactant (AS2) in 1 : 1 molar ratio. The decrease in colour of the dye is due to precipitation of the ion-associate species, CD+sh; AS2, in a similar fashion as described in the literature.16,17 Optimization of analytical variables A dilute acetic acid solution (0.1 mol l21) was employed to maintain the pH of the carrier (C).The optimum acidity range for the determination of SLS was found to be in the pH range 5.5–8.0, hence further experimental work was carried out at pH 6.0. Similarly, the optimum pH range of the sample solutions injected is 6.0–8.0. Potassium dichromate and thiourea were used in the carrier to enhance the tolerance limits of diverse ions, viz., SO3 22 and CuII, by oxidizing and masking them, respectively.An increase in concentration of BG up to 0.004% m/v (8.0 3 1025 mol l21) enhanced the peak height but a 0.002% m/v (4.0 31025 mol l21) solution was used for detailed studies as no smooth baseline was recorded beyond this concentration. The use of an alcohol, viz., ethanol, enhanced the peak height with a smoother baseline owing to the higher solubility. A 0.04% v/v concentration of ethanol in the dye solution was found to be adequate for the determination of the surfactant.The effect of temperature on the peak height was examined and a maximum and constant peak height was obtained when the coil was dipped into a water-bath maintained between 15 and 40 °C. The effect of the nature of dye and of the surfactant on the peak height was examined. The replacement of BG in the ion-associate species with other dyes, viz., MG, MB, EV and CV, decreased the sensitivity of the reaction considerably owing either to a lower +I effect (MG), lower conjugation (MB) or greater steric hinderance (EV, CV).In the case of surfactants, the sensitivity of the reaction increased as the molecular mass of the surfactants increased from SDS to SDBS (Fig. 3), which may be due to an increase in either the 2I effect or the carbon chain, or both. Among them, the pair BG+– LS2 was selected for detailed investigation. Optimization of FIA variables The effect of FIA variables, viz., bore size of silicon tubes, length of reaction coils, volume size of the solution injected and residence time, on the peak height were examined.The optimum bore size of the silicon tubes used is shown in Fig. 1. A PTFE tube of bore size 0.5 mm was used throughout the work. The peak height increased with increase in the reaction coil length up to 180 cm but after 120 cm no smooth baseline was recorded. A constant and maximum peak height was recorded over the volume range 200–370 ml. Injection and residence times of at least 25 and 45 s, respectively, were found to be adequate to obtain a maximum and constant peak height Fig. 2 Absorption spectra of Brilliant Green in sodium lauryl sulfate solutions. BG, 0.002% m/v (4.0 3 1025 mol l21) in 0.04% v/v ethanol; concentration of SLS: A, 0; B, 5; C, 10; D, 15; E, 20 ppm. Fig. 3 Signals of various surfactants with brilliant green. 1 = sodium laurylsulfate (SLS), 2 = sodium dodecylsulfonate (SDS), 3 = sodium hexadecylsulfonate (SHDS), 4 = sodium dodecylbenzenesulfonate (SDBS). 1692 Analyst, August 1998, Vol. 123with better resolution. Silicon tubes of bore size 0.51 and 0.64 mm, a PTFE coil of size 120 cm 3 0.5 mm and a volume of sample solution of 200 ml were selected for detailed investigation. The sample throughput was determined and found to be 50 h21 at a flow rate of 1.6 ml min21. Optimum concentration range, detection limit, sensitivity and statistics The anionic surfactant SLS was used as a representative surfactant and a calibration curve was obtained using standard solutions of SLS.The curve was linear up to 20.0 ppm of SLS at a gain factor of 1 with a slope, intercept and correlation coefficient of 21.0, 0.005 and 0.999, respectively (Fig. 4). The apparent molar absorptivity of the ion-associate precipitate formed (calculated using the molar concentration of surfactant injected and a pathlength of the flow cell of 0.55 cm) with five dyes and four anionic surfactants was found to be in the range of (0.60–1.50) 3104 l mol21 cm21 at the absorption maximum, Fig. 4 Signal peak height recorded for standard solution of SLS, 1.0–20.0 ppm. Table 3 Comparison of the present method with Methylene Blue methods5,18 for samples collected during November 1997 Present method FIA method5 Manual method18 SLS SLS SLS Sampling point found (ppm) RSD* (%) found (ppm) RSD* (%) found (ppm) RSD* (%) Ponds— Budha 13.8 0.5 14.2 1.3 12.7 1.0 Kankali 12.6 0.5 13.5 1.4 Raja 21.8 0.6 23.2 1.2 Katora 18.1 0.5 18.9 1.2 Handi 12.0 0.6 12.5 1.3 Rivers— Kharoon 1.3 0.8 1.7 1.6 2.6 1.1 Hasdo 3.0 0.7 3.6 1.7 Shivnath 2.2 0.7 2.8 1.6 Arpa 2.8 0.6 3.4 1.8 Ground waters— Budhapara 2.8 0.7 2.0 2.0 1.5 1.3 Kankalipara 2.2 0.8 2.4 1.9 Raja talab 4.2 0.6 4.5 1.8 Katora talab 3.5 0.7 3.8 1.6 Handipara 4.6 0.6 5.3 1.5 Municipal waste waters— Raipur 34.4 0.5 36.0 1.4 32.0 1.0 Bhilai 42.2 0.4 43.8 1.2 Durg 32.0 0.5 33.2 1.2 Rajnandgaon 25.6 0.6 27.0 1.3 Raigarh 23.2 0.6 25.8 1.3 Bilaspur 29.3 0.5 32.0 1.4 * Six replicate measurements were made. Table 1 Effect of various dyes and surfactants on the absorptivity Surfactant Dye lmax/ nm Apparent e/ 104 l mol21 cm21 SLS [CH3(CH2)11OSO3Na] BG 625 1.30 MG 630 0.60 MB 665 0.90 EV 595 0.70 CV 590 0.60 SDS [CH3(CH2)11SO3Na] BG 625 1.20 SHDS [CH3(CH2)15 SO3Na] BG 625 1.40 SDBS [CH3(CH2)11C6H4SO3Na] BG 625 1.50 Table 2 Effect of diverse ions on the determination of 5.0 ppm of SLS Ions added Tolerance limit* (ppm) Na, K, SCN2, S22, I2 500 NO32 400 SO4 22 250 Br2 , CaII 200 Cl2 , MgII, TX-100, TX-300, Briz-35 100 AlIII, NiII, MnII, CoII 50 FeIII, PO4 32 25 F2, VV,CrVI, ZnII, thiourea 20 BiIII, CuII, CPC, CTAB, SO3 22 10 * Causing an error < ±2%.Analyst, August 1998, Vol. 123 1693590–665 nm (Table 1). The detection limit (the amount causing an absorbance or peak height more than three times the standard deviation of the blank) was determined and found to be 100 ppb SLS with BG.The relative standard deviation for the determination of 10.0 ppm of SLS (n = 6) was found to be 0.5%. Effect of diverse ions The effect of various diverse ions on the determination of 5.0 ppm of SLS was examined using the proposed procedure. Only the tolerance limit of the ions CuII and SO3 22 was found to be critical, but it could be increased by adding potassium dichromate (0.001% m/v or 3.4 3 1025 mol l21) and thiourea (0.001% m/v or 1.3 3 1024 mol l21) solutions, respectively, to the carrier.Cationic surfactants, viz., cetylpyridinium chloride (CPC) and cetyltrimethylammonium bromide (CTAB), could be tolerated up to at least a twofold mass excess and thereafter they caused a positive effect (decrease peak height). The tolerance limits of various diverse ions tested in the determination of 5.0 ppm of SLS are summarized in Table 2. Application of the method The validity of the method was checked with the conventional manual and FIA Methylene Blue methods5,18 and its precision was found to be better than those of the established methods (Table 3).The application of the present method to the determination of anionic surfactants in terms of SLS was extended to various water bodies, viz., 53 ponds, 11 rivers, 53 tube wells and six municipal waste waters from central India (see Experimental). All pond, river and municipal waste waters were found to be contaminated with SLS in the ranges 6.0–22.6 (mean 13.6, median 13.2, standard deviation 4.3), 1.0–3.0 (mean 1.9, median 1.8, standard deviation 0.6), and 23.2–42.2 (mean 31.1, median 30.6, standard deviation 6.8) ppm, respectively (Tables 4–6).In addition, the shallow tube well waters lying near stationary surface and municipal waste water reservoirs were also found to be contaminated with SLS in the range of 1.2–4.6 (mean 2.6, median 2.5, standard deviation 0.8) ppm (Table 7). The highest level of the surfactant of up to 42.2 ppm of SLS was observed in municipal waste water and the lowest level of down to 1.0 ppm SLS in the mobile surface and ground waters.The level of SLS in the pond water decreased by approximately half in the rainy season whereas it increased at least threefold in the tube well ground water (Table 8). The high values of the standard deviation for the SLS level in the pond water (4.3 ppm) and municipal waste water (6.8 ppm) show that the nature and strength of the primary anthropogenic sources, viz., use of detergents, soaps, shampoos, cosmetics, etc., and meteorological, geographical and geological factors differ from Table 4 Determination of SLS in stationary surface waters collected during November 1997 Sampling point SLS (ponds) Place found (ppm)* Budha Raipur 13.8 Kankali 12.6 Raja 21.8 Katora 18.1 Handi 12.0 Bendri 10.7 Guma 7.6 Tandua 6.0 Kukurbeda 10.2 Labhandi 9.4 Chherikhedi 8.0 Kota 16.2 Jora 9.2 Dharampura 7.8 Lalpur 7.5 Mandir Hasaud 8.4 Kalyansagar, Bhatapara 13.2 Baloda Bazar 12.5 Arang 16.1 Mahasamund 11.7 Dhamtari 17.3 Kumhari Durg 18.4 Charoda 14.2 Borsi 19.4 Kugda 17.6 Suryakund, Bhilai 13.5 Nehru Nagar 15.0 Luchki 20.2 Hudco 22.6 Sarkanda Bilaspur 17.4 Juna Bilaspur 19.5 Ratanpur 11.5 Bheema, Janjgir 14.3 Sardha, Lormi 8.3 Lormi 11.4 Mungeli 10.3 Pandaria 9.2 Panchavati 15.0 Torva Naka 18.2 Chandrapur 12.7 Korba 19.3 Champa 16.2 Dalpatsagar Bastar 18.8 Bhanupratappur 10.4 Narayanpur 11.5 Kanker 12.4 Keshkal 10.3 Bal Samund Raigarh 10.2 Mithumuda 18.5 Tarapur 6.9 Sarangarh 14.0 Rajnandgaon Rajnandgaon 18.3 Chichola 16.2 * Mean 13.6, median 13.2, range 6.0–22.6, SD 4.3 ppm.Table 5 Determination of SLS in river waters collected during November 1997 Sampling point SLS (rivers) Place found (ppm)* Kharoon Raipur 1.3 Shivnath Durg 2.2 Hasdo, Korba Bilaspur 3.0 Hasdo, Champa 2.4 Mahanadi, Sheorinarayan 1.0 Arpa 2.8 Kelo Raigarh 1.8 Mand 1.2 Indravati Bastar 1.7 Dudh, Kanker 2.1 Maitrayani Rajnandgaon 1.5 * Mean 1.9, median 1.8, range 1.0–3.0, SD 0.6 ppm.Table 6 Determination of SLS in municipal waste waters during November 1997 SLS Sampling point Place found (ppm)* Raipur Raipur 34.4 Bhilai Durg 42.2 Durg 32.0 Rajnandgaon Rajnandgaon 25.6 Raigarh Raigarh 23.2 Bilaspur Bilaspur 29.3 * Mean 31.1, median 30.6, range 23.2–42.2, SD 6.8 ppm. 1694 Analyst, August 1998, Vol. 123place to place in central India. The lower standard deviations for river water (0.6 ppm) and tube well water (0.8 ppm) water show that they will be polluted by secondary sources, viz., polluted ponds, municipal waste water reservoirs, etc.Conclusion Proposed method is simple, specific and precise and is applicable to the determination of anionic surfactants in terms of SLS in various water bodies of central India. Almost all the surface waters of rural, urban, semi-urban and industrial areas and shallow tube well waters lying near the sources were found to be contaminated with SLS beyond the permissible limit ( > 1.0 ppm).The main sources of anionic surfactants in water bodies of central India are assumed to be household commodities and personal care products, viz., detergents, soaps and fabric and cosmetic materials, etc. The authors are grateful to Pt. Ravishankar Shukla University, Raipur, the Ministry of Environment and Forests, New Delhi, and the Alexander von Humboldt Foundation, Bonn, for providing financial support for this work.References 1 Potokar, M. S., Surfactant Sci. Ser., 1980, 10, 87. 2 Kaestner, W., Surfactant Sci. Ser., 1980, 10, 139. 3 Valdiya, K. S., Environmental Geology: Indian Context, Tata McGraw-Hill, New Delhi, 1987, pp. 116–117. 4 Liu, J., Anal. Chim. Acta, 1997, 343, 33. 5 Fan, S. H., and Fang , Z. L., Fresenius’ J. Anal. Chem., 1997, 357, 416. 6 Masadome, T., and Imato, T., J. Flow Injection Anal., 1996, 13, 120. 7 Motomizu, S., Oshima, M., and Goto, N., J.Flow Injection Anal., 1993, 10, 255. 8 Liu, H., and Dasgupta, P. K., Anal. Chim. Acta, 1994, 288, 237. 9 Motomizu, S., and Kobayashi, M., Anal. Chim. Acta, 1992, 261, 471. 10 Hirai, Y., and Tomokuni, K., Anal. Chim. Acta, 1985, 167, 409. 11 Kawase, J., Nakae, A., and Yamanaka, M., Anal. Chem., 1979, 51, 1640. 12 Motomizu, S., Hazaki, Y., Oshima, M., and Toei, K., Anal. Sci., 1987, 3, 265. 13 Motomizu, S., Oshima, M., and Kuroda, T., Analyst, 1988, 113, 747. 14 Gallego, M., Silva, M., and Valcarcel, M., Anal.Chem., 1986, 58, 2265. 15 Pandey, S. P., Narayanswamy, V. S., and Hasan, M. Z., Indian J. Environ. Health, 1979, 21, 35. 16 Oshima, M., Motomizu, S., and Doi, H., Analyst, 1992, 117, 1643. 17 Aggregation Processes in Solution, ed. Wyn-Jones, E., and Gormally, J., Elsevier, Amsterdam, 1983, p. 241. 18 Standard Methods for the Examination of Waters and Waste Waters, American Public Health Association, Washington, DC, 17th edn., 1989, pp. 5-59–5-63. Paper 8/02945H Received April 20, 1998 Accepted June 1, 1998 Table 7 Determination of SLS in ground waters collected during November 1997 Sampling point SLS (tube wells) Place found (ppm)* Budhapara Raipur 2.8 Kankalipara 2.2 Raja talab 4.2 Katora talab 3.5 Handipara 4.6 Bendri 2.5 Guma 1.2 Tendua 1.4 Kukurbeda 2.0 Labhandi 1.8 Chherikhedi 1.6 Kota 3.2 Jora 1.7 Dharampura 2.1 Lalpur 2.4 Mandir Hasaud 2.7 Kalyansagar, Bhatapara 3.1 Baloda Bazar 2.4 Arang 1.9 Mahasamund 2.6 Dhamtari 2.8 Kumhari Durg 3.9 Charoda 2.6 Borsi 1.7 Kugda 2.2 Suryakund, Bhilai 2.5 Nehru Nagar 2.2 Luchki 3.5 Hudco 2.4 Sarkanda Bilaspur 3.7 Juna Bilaspur 2.8 Ratanpur 3.1 Bheema, Janjgir 3.8 Sardha, Lormi 1.8 Lormi 2.3 Mungeli 2.5 Pandaria 2.0 Panchavati 3.2 Torva Naka 2.2 Chandrapur 1.8 Korba 4.5 Champa 3.5 Dalpatsagar Bastar 4.2 Bhanupratappur 2.2 Narayanpur 2.8 Kanker 2.4 Keshkal 2.4 Bal Samund Raigarh 2.2 Mithumuda 3.2 Tarapur 1.8 Sarangarh 2.8 Rajnandgaon Rajnandgaon 3.5 Chichola 2.1 * Mean 2.6, median 2.5, range 1.2–4.6, SD 0.8 ppm. Table 8 Seasonal variation of SLS levels in various water bodies (location, Raipur) SLS Sampling point Date of collection found (ppm) Raja pond 21 July 1997 11.8 Raja pond 15 November 1997 21.8 Raja tubewell 22 July 1997 11.2 Raja tubewell 15 November 1997 4.2 Analyst, August 1998, Vol. 123 1695

ISSN:0003-2654    

DOI:10.1039/a802945h

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

Effect of high salt concentrations on the determination of arsenic and selenium by flow injection hydride generation electrothermal atomic absorption spectrometry Robert I. Ellisa, Nils G. Sundina, Julian F. Tyson*a, Susan A. McIntoshb, Christopher P. Hannab and Glen Carnrickb a Department of Chemistry, University of Massachusetts, Box 34510, Amherst, MA 01003-4150, USA b Perkin-Elmer Corporation, 761 Main Avenue, Norwalk, CT 06897, USA In the determination of arsenic and selenium by flow injection hydride generation ETAAS, the presence of up to 20% sodium chloride enhanced the signals for 20 mg l21 arsenic and selenium by up to 28%.The enhancement was obtained with a variety of gas–liquid separators. A systematic study of the possible causes of the signal enhancement in the determination of selenium was undertaken, from which it was concluded that the effect originated in the processes responsible for the distribution of the hydrogen selenide between the solution and gas phases.Processes related to the transport of the analyte from the gas–liquid separator and the trapping of the analyte on the interior of the atomizer were not affected by the presence of dissolved salts. As sodium was found to be transported to the atomizer, it was deduced that aqueous aerosol was deposited in the atomizer, although the quantities were irreproducible. The enhancement could be eliminated by increasing the borohydride concentration. However, with the small volume gas–liquid separator, this latter approach was limited because of carry-over of liquid to the atomizer.The effect could be compensated for by adding up to 40% m/v of salt to the borohydride reagent. Keywords: Flow injection; hydride generation; electrothermal atomic absorption spectrometry; selenium; arsenic; electrolytes; salting-out Flow injection hydride generation electrothermal atomic absorption spectrometry (FI–HG–ETAAS) is an emerging technique for the determination of several trace elements, including selenium and arsenic.1,2 The procedure combines the high precision and sample throughput of flow injection, the ability to separate the analyte from the matrix by hydride generation, and the high sensitivity and low detection limit of ETAAS.A discrete volume of acidified sample solution is merged with sodium borohydride solution, and the volatile hydride of the trace element which is formed is transferred into the gas phase by (a) the evolution of hydrogen from the decomposition of the excess borohydride and (b) the merging of an argon stream.Following bulk gas–liquid separation, the hydride is trapped on the interior surface of the graphite tube atomizer of an electrothermal atomic absorption spectrometer, which has been pre-treated with a trapping agent (e.g., iridium chloride solution) prior to atomization. A possible area of application of this methodology is in the analysis of solutions containing high concentrations of dissolved salts, such as sea-waters, industrial brines and digests of solids with predominantly inorganic matrices such as sludges and sediments.Initial studies3 of the determination of selenium in solutions containing up to 20% sodium chloride by FI–HG– ETAAS indicated that the presence of salt caused signal enhancements (up to 127%). To our knowledge, this effect has not been reported previously. In this paper, the results of further studies of this effect are presented.The high recoveries are confirmed and a number of possible causes are proposed. The results of a systematic investigation of these possible causes are discussed, from which it is concluded that the diminished solubility of gases in the presence of dissolved salts (the so-called ‘salting-out’ effect) was responsible. Two possible procedures for overcoming this interference were investigated. Experimental Reagents and supplies Working standard SeIV and AsV solutions were made by diluting an aliquot of the respective stock standard solution (Perkin- Elmer, Norwalk, CT, USA) to produce a concentration of 1000 mg l21.The carrier stream was 10% v/v hydrochloric acid (Fisher, Pittsburgh, PA, USA) and the reductant stream was 0.2% m/v NaBH4 (Fisher) prepared daily and stabilized with 0.05% m/v NaOH (Fisher). Samples containing selenium were prepared by placing 10 ml of 10% v/v HCl in a 100 ml calibrated flask followed by the appropriate amount of salt where required.Approximately 30 ml of purified water were added before adding an aliquot of selenium standard solution and dilution to volume. Samples containing arsenic were prepared by placing 10 ml of 10% v/v HCl, 10 ml of 1 % m/v KI (Fisher) and 10 ml of 1% m/v ascorbic acid (Fisher) in a calibrated flask followed by addition of an aliquot of arsenic standard solution. These samples were stoppered and left at room temperature for 1 h to allow reduction of AsV to AsIII prior to addition of salt when required and dilution to volume with high purity water.In the investigation into the cause of the increased signals, solutions of SeIV, prepared as above, of concentrations from 0 to 20 mg l21 containing either 0 or 20% of sodium chloride were used unless specified otherwise. Instrumentation A multi-line flow injection manifold4 was used in which the sample was injected into an acid carrier, then merged with the alkaline borohydride reagent.After passage through a reaction coil, the reaction zone was merged with an argon stream, passed through a stripping coil and delivered to the gas–liquid separator. Drainage from the gas–liquid separator was controlled by a peristaltic pump channel. The manifold was used in conjunction with a Perkin-Elmer Model 4100ZL Zeeman corrected electrothermal atomic absorption spectrometer interfaced with a Digital 316sx workstation, and controlled using Analyst, August 1998, Vol. 123 (1697–1701) 1697Perkin-Elmer Gem software (version 7.2.1). The manifold was constructed from PTFE manifold tubing (1 mm id) and a Perkin- Elmer Chemifold and gas–liquid separator. The carrier stream flow rate was 4 ml min21 and that of the borohydride reagent was 6 ml min21. The waste line from the gas–liquid separator could be pumped at up to 15 ml min21. The length of tubing between the confluence point and the addition of argon was 110 cm and the length between this point and the gas–liquid separator was 300 cm.A Permapure Nafion dryer was fitted to the gas transfer line which connected the gas–liquid separator to the autosampler probe of the spectrometer. This removed moisture from the transferred gases.4 The FIAS program is given in Table 1. The sample loop volume was 500 ml. The argon stripping gas flow rate was 130 ml min21. The hydrides were trapped on a transversely heated graphite tube pre-treated with 120 ml of 0.1% m/v iridium chloride solution (Perkin- Elmer).5 Tubes which are pre-treated with iridium may be used for up to 300 firings provided that the temperature does not exceed 2300 °C.6 Perkin-Elmer System II electrodeless discharge lamps, operated at 260 mA (selenium) and 350 mA (arsenic), were used, with detection at a wavelength of 196.0 nm for selenium and 193.7 nm for arsenic.Peaks were quantified by area. The furnace programs are given in Table 2. Experiments with quartz tube atomization were performed with a Perkin- Elmer Model 3100 atomic absorption spectrometer fitted with a quartz tube atomizer positioned within an air–acetylene flame and a Perkin-Elmer selenium hollow cathode lamp operated at 15 mA.All other relevant parameters were as described above. Two other gas–liquid separators were studied; one was supplied by PS Analytical (Orpington, Kent, UK) for use in a continuous flow hydride generator and the second was constructed in-house from borosilicate glass and is shown schematically in Fig. 1. The use of the first of these devices has been described numerous times (see, for example, refs. 8, 10 and 11). Procedures Effect of salts on signal Solutions of SeIV (1–10 mg l21) and AsIII (5 and 10 mg l21) were prepared as described above, either with or without salts, to determine the percentage increase in signal compared with that obtained when no salt was present. It was considered that the most plausible explanations for the effect were either an increase in (a) the efficiency of trapping of the hydride within the graphite tube (due to the presence of salt transported as aerosol), (b) the efficiency of the gas–liquid separation behaviour of the manifold components in the presence of salt or (c) the extent of formation of the hydride.Contamination Blank solutions containing salt were analyzed to ensure that the increase in signal did not result from contamination of the salts by trace concentrations of selenium or arsenic.Trapping efficiency It is possible that some of the dissolved salt may be transferred to the graphite tube in the form of an aerosol. Several workers have speculated that this occurs,7–9 but it has been conclusively demonstrated recently by Wickstrom et al.10 for a continuousflow hydride generation system with a Thompson-type11 gas– liquid separator. If the trapping efficiency is below 100%, there may be an improvement in trapping efficiency in the presence of the transferred salt aerosol.The trapping temperature of the furnace was varied between 20 and 800 °C and untreated graphite tubes or tubes treated with Pd were also used. A sodium hollow cathode lamp was fitted to the spectrometer and the amount of sodium in the graphite tube accumulated during the hydride generation of selenium from solutions with and without sodium chloride was measured. A blank solution containing 20% m/v of sodium chloride was injected into the manifold and carried through the procedure, omitting the furnace firing stage.The furnace was held at the trapping temperature and a sample without salt was injected. The furnace was then fired in the normal manner. If salt in the furnace had the effect of increasing the trapping efficiency, there would be an increase in signal in the latter determination. To determine whether the effect was exclusive to electrothermal atomization, solutions of between 0 and 30 mg l21 Se, containing either 0 or 20% m/v of NaCl, were analyzed using quartz tube atomization. Kinetic effects An increase in the rate of the hydride forming reaction in the presence of dissolved salts could be responsible for the enhancement.If this was the case, increasing the length of the manifold tubing would allow a longer time for reaction and hence could eliminate the increase in signal. To investigate the effect of changing manifold parameters, a systematic study which involved varying the length of manifold tubing was carried out.The effect of including a stopped-flow period was also studied. Transfer between gas–liquid separator and atomizer To investigate whether a greater proportion of the analyte was being transferred to the graphite tube by an aerosol formed in Table 1 FIAS parameters Step Time/s Pump 1/rpm Pump 2/rpm Valve Comments Pre-fill 10 100 0 Fill Fill loop and preheat atomizer 1 20 100 80 Fill Fill loop 2 8 0 0 Inject Insert probe to atomizer 3 30 0 80 Inject Generate hydride and trap 4 8 0 0 Inject Remove probe from atomizer 5 5 0 80 Fill Begin atomization Table 2 Furnace parameters Element Step Temperature/ °C Ramp Hold/s Gas flow rate/ml min21 Read/s Se Hold 250 1 15 250 Atomize 2000 0 5 0 5 Clean 2300 1 3 250 As Hold 400 1 15 250 Atomize 2100 0 5 0 5 Clean 2300 1 3 250 Fig. 1 ‘Spray chamber’ design of gas–liquid separator, constructed inhouse. 1698 Analyst, August 1998, Vol. 123the presence of the salt, the Nafion dryer was removed from the gas transfer line and replaced with either a PTFE tube or a drying tube containing magnesium perchlorate.The composition of aerosol generated from concentrated salt solutions can be different from that generated from low ionic strength solutions, an effect known as aerosol ionic redistribution.12 Batch versus flow A batch system was constructed as shown in Fig. 2 to check whether the increase in signal was specific to flow injection systems.Gas–liquid separation To investigate the efficiency of gas–liquid separation, a second gas–liquid separator was coupled to the waste line from the first gas–liquid separator as shown in Fig. 3. Should there be an increase in efficiency in the presence of salt, there would be a lower signal from the second gas–liquid separator compared with that obtained in the absence of salt. The signals for both a blank and a standard in the presence and absence of 20% of sodium chloride were obtained for the other two gas–liquid separator devices.The device from PS Analytical was operated with an additional hydrostatic head on the waste in order to compensate for the higher back-pressure of the narrow tubing connecting the device to the graphite furnace atomizer than the normal back-pressure from a quartz tube atomizer and associated connecting tubing. The optimized argon flow rate was 150 ml min21. The ‘spray chamber’ device was also operated at an optimized gas flow rate of 150 ml min21.Overcoming the effect Because the increase in signal in the presence of high electrolyte concentrations would require matrix matching of standards to obtain accurate analyses, two ways to overcome the problem were investigated: variation of the borohydride concentration and the addition of salt to the reagents. The effect of changing the borohydride concentration over the range 0.2–2.0% was investigated. At high concentrations of sodium borohydride, foaming in the gas–liquid separator may cause liquid to rise through the gas transfer tube and be deposited on the furnace, thereby impeding the trapping ability.Therefore, a second waste line was fitted to the gas–liquid separator and a 10 mm 3 8 mm id acrylic tube was inserted into the gas transfer line to prevent liquid droplets from reaching the graphite tube. The tube was used in place of the Nafion dryer and was removed and thoroughly cleaned between measurements.Salt was added to the sample in the manifold from an extra line which merged with the sample carrier stream upstream of the merging with borohydride. The effect of adding salt to the borohydride reagent was also investigated. Results and discussion Effect of salts on signal enhancement Typical recoveries for solutions of AsIII and SeIV are given in Table 3. The effect was observed for both arsenic and selenium in the presence of sodium chloride but was most pronounced in the determination of selenium in the presence of potassium sulfate.Contamination There was no measurable amount of selenium or arsenic in any of the salts used. Trapping efficiency Although sodium was observed in the furnace in amounts in excess of the background, leading to the conclusion that aerosol droplets are transported to the furnace, the sodium signals were highly irreproducible. As the increase in signal in the presence of sodium chloride is precise, it is unlikely that the variable amounts of sodium on the furnace would be a part of the mechanism responsible for this effect.High signals were not observed for a solution run subsequently to a high-salt blank for which the furnace had not been fired. It was concluded that the increased signals in the presence of salt were not due to processes inside the graphite tube, such as increased trapping efficiency. For the analyses with quartz tube atomization, signal enhancements of between 121 and 127% were obtained for the analysis of the solutions containing 30 mg l21 Se and 20% m/v of NaCl.It is therefore unlikely that the increase is specific to graphite tube atomization. Fig. 2 Batch hydride generation manifold. Fig. 3 Flow injection manifold with second gas–liquid separator. Table 3 Typical recovery data Analyte Salt Salt concentration (%) Signal* relative to that for a solution with no salt (%) Se (5 ng ml21) NaCl 0.62 107 ± 4 1.25 109 ± 4 2.5 112 ± 4 5 115 ± 4 10 120 ± 5 20 126 ± 5 K2SO4 0.31 104 ± 4 0.62 108 ± 4 1.25 110 ± 4 2.5 112 ± 4 5 112 ± 4 As (5 ng ml21) NaCl 20 128 ± 5 K2SO4 5 104 ± 3 * The ± values are standard deviations of the ratios calculated as the square root of the sum of the squares of the standard deviations of the numerator (n = 4) and denominator (n = 4).Analyst, August 1998, Vol. 123 1699Kinetic effects Neither lengthening the manifold nor incorporating a stoppedflow stage into the FIAS program affected the signal enhancement, although the sensitivity was affected. This suggests that the kinetics of the reaction are not significantly influenced by the presence of dissolved salt.Transfer between gas–liquid separator and atomizer When the Nafion dryer was replaced with either a PTFE tube or a magnesium perchlorate drying tube, the enhancement was unaffected, although the magnesium perchlorate drying tube reduced the sensitivity. This suggests that even if aerosol is transported through the gas transfer line, the transport of the analyte is unaffected by the presence of aerosol.Batch versus flow The batch experiment using the apparatus shown schematically in Fig. 2 did not result in a higher signal in the presence of dissolved salt, so the phenomenon was considered to be related to the flow procedure. The increase in signal in the presence of dissolved salt with the FI system was therefore due to greater transfer of the hydride to the vapor phase.Gas–liquid separation In the presence of salt, the signal for the selenium from the second gas–liquid separator (see Fig. 3) was 75% of that for solutions which did not contain salt, indicating that less of the analyte was pumped out of the first gas–liquid separator (to waste) in the presence of salt. That is, there was an increase in the proportion of the analyte which was transferred into the gas phase in the presence of dissolved salts. The performance of the various separator devices is shown in Table 4, from which it is clear that the effect is not a unique feature of the Perkin-Elmer device.Manifolds incorporating the other two devices also produced enhanced signals in the presence of salt, although both gave reduced peak area in comparison with peaks obtained with the Perkin-Elmer device, which is designed to maximize sensitivity by minimizing the dispersion in the headspace. High concentrations of electrolytes in aqueous solution lower the solubility of dissolved gases in that solution.This effect is known as ‘salting-out’.13–15 The extent of the decrease in gas solubility is increased when both the charge on and size of the ions are increased, and with increasing ionic concentration.16 This phenomenon is commonly utilized in headspace sampling in gas chromatography.17 Salts are added to sample solutions in order to decrease the solubility of gases and vapors in solution and increase the partial pressure of the analyte species in the headspace.In a FI system the process of formation of the hydride and removal from solution are initiated once the acidified sample and borohydride stream mix, as hydrogen gas (produced by the decomposition of excess borohydride) forms a bulk gas phase in contact with the bulk liquid phase. The process of removal of the hydride from the liquid phase continues with the addition of a stream of argon gas which changes the ratio of gas to liquid in the flowing stream.It is possible that further transfer occurs in the gas–liquid separator, the role of which is primarily that of separating bulk phases. Hydrogen selenide, unlike some other hydrides (including arsine), has a solubility in aqueous solution which can involve the formation of the corresponding anions (Se22 and HSe2) and hence the acidity of the solutions is a factor which affects solubility. Clearly, the goal of any procedure for the trace determination of selenium via the formation of hydrogen selenide is to maximize the amount of hydrogen selenide removed from the solution.Although many previous hydrogen selenide generation procedures, both batch and FI, have been optimized for just this situation, the approaches used have always been empirical. Reports of such studies contain little or no explicit discussion of the role of the factors which affect the distribution of hydrogen selenide gas between the aqueous and vapor phases.In an FI system, as opposed to a batch system, a description of this distribution is complicated further by kinetic effects. Kinetic effects would seem to be relevant in the case of the salting-out effect, as there appear to be no reports of enhanced signals for procedures based on batch generation and separation of the hydrogen selenide. The results obtained in the present study indicate that the effect is not observed when a batch reactor is used. Presumably the action of bubbling argon through the bulk liquid (combined with the evolution of hydrogen—see the following section) provides a mechanism for the removal of the same proportion of the analyte regardless of the dissolved salt content.Previous work18 with this system showed that the overall efficiency for the transfer of selenium into the furnace by HG inatomizer trapping was 75%. Sturgeon et al.19 found the processes to be 79% efficient. It is therefore feasible that, if the missing 20–25% of the selenium is not removed from the solution, the salting out of this fraction could account for an increase in signal of 25–33%.Overcoming the effect As can be seen in Fig. 4, increasing the concentration of sodium borohydride increased the sensitivity and reduced the magnitude of the difference in signal between samples which contained salt and samples which did not. A concentration of 1.5% m/v of sodium borohydride was sufficient to remove any significant differences between the signals for up to 20% of sodium chloride in solution.The use of sodium borohydride concentrations above this value resulted in violent reactions, which caused large amounts of solution to be transferred to the graphite tube, adversely affecting the results. As the concentra- Table 4 Performances of different gas–liquid separator devices Integrated absorbance for selenium/s Device Blank, no salt Blank, with salt* 10 ng ml21, no salt 10 ng ml21, with salt* Perkin-Elmer 0.007 0.006 0.446 0.589 PS Analytical 0.003 0.002 0.204 0.385 In-house 0.007 0.007 0.259 0.499 * Salt concentration was 20% (m/m) of sodium chloride. Fig. 4 Variation of signal for 5 ng ml21 Se with different borohydride concentrations. 8, No salt in sample; <, 20% of salt in sample. 1700 Analyst, August 1998, Vol. 123tion of borohydride is increased, the distribution of the hydride between the two phases must be shifted towards the gas phase to such an extent that the salting-out process is insignificant.Therefore, one possible strategy for overcoming the salting-out effect would be the use of higher concentrations of borohydride. However, there are practical problems associated with excess foaming and bubble formation in the gas–liquid separator that may make this approach problematic in practice. In addition, interferences from transition metals are increased as the borohydride concentration is increased. However, the other two designs of gas–liquid separator were more tolerant to the use of increased borohydride concentrations and with 2.0% of borohydride gave signals which were around 94% of that of the Perkin- Elmer device operated with 0.2% of borohydride.The addition of salt solutions through an extra line added to the manifold lowered the sensitivity. However, as can be seen in Fig. 5, the addition of salt to the borohydride solution produced an increase in sensitivity for solutions without salt such that there was no significant difference due to the presence of salt in the sample.The ‘salting-out ability’ (i.e., concentration, charge, size) of the salt in the reductant stream must be sufficiently high to eliminate any differences between samples and standards. Conclusion It is concluded that the presence of high salt concentrations produces a positive interference in the FI–HG–ETAAS determination of arsenic and selenium due to increased transfer of the hydride into the vapor phase. It is proposed that this is a thermodynamic rather than a kinetic effect, that is, the presence of the dissolved salt affects the equilibrium distribution of the hydrogen selenide or arsine between the gas and liquid phases.Although evidence was found for the transport of aerosol to the atomizer, this was not related to the signal enhancement. The enhancement effect could be compensated for by adding salt to the borohydride reagent or eliminated by increasing the borohydride concentration.This latter strategy was not suitable for the FI system used because the design of the gas–liquid separator was such that bulk liquid was carried to the atomizer, adversely affecting the trapping ability of the furnace coating. For FI–HG with continuous atomization, the design of the gas– liquid separator is governed by the need to minimize dispersion effects. As dispersion increases there is a consequent decrease in sensitivity and a decrease in throughput.However, when inatomizer trapping is used, the shape of the peak is, to a large extent, independent of the kinetics of any of the processes of generation and separation of the hydride. Hence the requirement for minimizing the dispersion processes in the gas–liquid separator could be relaxed and designs (such as those of the other two devices used in this study) which allowed a greater tolerance to borohydride concentration could be used. Several designs of gas–liquid separator have been evaluated.20,21 The authors thank Perkin-Elmer for financial support for Nils Sundin and Robert Ellis and the provision of equipment, and Walter Slavin, Bonaire Technologies, for many helpful discussions.References 1 Matusiewicz, H., and Sturgeon, R. E., Spectrochim. Acta, Part B, 1996, 51, 377. 2 Fang, Z., Flow Injection Atomic Absorption Spectrometry, Wiley- Interscience, Chichester, 1996. 3 Carnrick, G., personal communication, 1995. 4 Sundin, N. G., Tyson, J. F., McIntosh, S. A., and Hanna, C. P., Spectrochim. Acta, Part B, 1995, 50, 369. 5 Perkin-Elmer, Manufacturer’s Instruction Manual, Part Number B050-9907, Publication B3212.10, Perkin-Elmer, Norwalk, CT, 1993. 6 Hanna, C. P., Carnrick, G. R., McIntosh, S. A., Guyette, L. C., and Bergemann, D. E., At. Spectrosc., 1995, 16, 82. 7 Welz, B., and Schubert-Jacobs, M., J. Anal. At. Spectrom., 1986, 1, 23. 8 Tyson, J. F., Offley, S. G., Seare, N. J., Kibble, H. A. B., and Fellows, C., J. Anal. At. Spectrom., 1992, 7, 315. 9 Bax, D., Agterdenbos, J., Worrell, E., and Beneken Kolmer, J., Spectrochim. Acta, Part B, 1988, 43, 1349. 10 Wickstrom, T., Lund, W., and Bye, R., Analyst, 1996, 121, 201. 11 Thompson, M., Pahlavanpour, B., Walton, S., and Kirkbright, G. F., Analyst, 1978, 103, 568. 12 Cresser, M., in Sample Introduction in Atomic Spectrometry, ed. Sneddon, J., Elsevier, Amsterdam, 1990, ch. 2. 13 Atkins P. W., Physical Chemistry, Freeman, New York, 3rd edn., 1986. 14 Dack, M. R. J., Techniques of Chemistry Vol VII—Solutions and Solubility, Part 1, Wiley, New York, 1986. 15 Battino, R., and Clever, H. L., Chem. Rev., 1966, 66, 395. 16 Glasstone, S., Physical Chemistry, Macmillan, London, 2nd edn., 1948. 17 Seto, Y., J. Chromatogr. A., 1994, 674, 25. 18 Tyson, J. F., Sundin, N. G., Hanna, C. P., and McIntosh, S. A., Spectrochim. Acta, Part B, 1997, 52, 1773. 19 Sturgeon, R. E., Willie, S. N., Sproule, G. I., Robinson, P. T., and Berman, S. S., Spectrochim. Acta, Part B, 1989, 44, 667. 20 Hanna, C. P., Haigh, P. E., Tyson, J. F., and McIntosh, S., J. Anal. At. Spectrom., 1993, 8, 585. 21 Brindle, I. D., and Zheng, S., Spectrochim. Acta, Part B, 1996, 51, 1777. Paper 8/02944J Received April 20, 1998 Accepted June 9, 1998 Fig. 5 Variation of signal for 5 ng ml21 Se for samples with different NaCl concentrations for different concentrations of NaCl in the borohydride solution. 8, No NaCl in borohydride; <, 20% of NaCl in the borohydride solution; and -, 40% of NaCl in the borohydride solution. Analyst, August 1998, Vol. 123 1701

ISSN:0003-2654    

DOI:10.1039/a802944j

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

Speciation determination of arsenic in urine by high-performance liquid chromatography–hydride generation atomic absorption spectrometry with on-line ultraviolet photooxidation† Dimiter L. Tsalev‡, Michael Sperling* and Bernhard Welz Bodenseewerk Perkin-Elmer GmbH, Alte Nussdorfer Strasse, D-88662 � Uberlingen, Germany. E-mail: Sperlimn@perkin-elmer.com A coupled system for arsenic speciation determination based on high-performance liquid chromatography (HPLC), on-line UV photooxidation and continuous-flow hydride generation atomic absorption spectrometry (HGAAS) was built from commercially available modules with minor modifications to the electronic interface, the software and the gas–liquid separator.The best results were obtained with strong anion-exchange columns, Hamilton PRP X-100 and Supelcosil SAX 1, and gradient elution with phosphate buffers containing KH2PO4–K2HPO4. The on-line UV photooxidation with alkaline peroxodisulfate, 4% m/v K2S2O8–1 mol l21 NaOH, in a PTFE knotted reactor for 93 s ensures the transformation of inorganic AsIII, monomethylarsonate, dimethylarsinate, arsenobetaine, arsenocholine, trimethylarsine oxide and tetramethylarsonium ion to arsenate.About 32–36 HPLC–UV–HGAAS runs could be performed within 8 h, with limits of detection between 2 and 6 mg l21 As, depending on the species. The method was applied to the analysis of spot urine samples and certified urine reference materials (CRMs). Upon storage at 4 °C, reconstituted CRMs are stable for at least 2 weeks with respect to both their total arsenic content and the individual species distribution.Keywords: Arsenic speciation determination; high-performance liquid chromatography–hydride generation atomic absorption spectrometry; human urine; organoarsenic species; on-line ultraviolet photooxidation Many arsenic compounds are toxic and potentially carcinogenic, which raises much concern from environmental, occupational and nutritional points of view.However, the various inorganic and organic arsenic species exhibit large differences in their mobility, availability and toxicity in environmental and biological systems,1 and can be approximately ranked in the following decreasing order of toxicity: As2III (AsH3) >> inorganic AsIII (i-AsIII, such as arsenite, metaarsenite and As2O3) > inorganic AsV (i-AsV, such as arsenate and metaarsenate) >> monomethylarsonate (MMA) > dimethylarsinate (DMA) >> ‘fish arsenic’ [such as arsenobetaine (AB), tetramethylarsonium ion (Me4As+) and arsenocholine (AC)].Furthermore, these species can behave significantly differently during sampling, storage, sample preparation and generation of analytical signals.2,3 It is therefore generally accepted that a species-selective determination of arsenic is needed for a meaningful risk assessment and better understanding of its biological and metabolic pathways and its analytical behavior.Speciation determination of arsenic has been addressed in more than 500 original publications and discussed in depth in several recent reviews1,4–6 and book chapters.2,3,7,8 Human urine is a suitable biological specimen for monitoring recent exposure to arsenic because of its relatively fast excretion rates of this element from the organism. Misleading results, i.e., an over-estimation of occupational exposure, may be obtained owing to the contribution of dietary arsenic of low toxicity (AB, DMA, Me4As) when total arsenic is determined in urine, unless seafood is excluded from the diet for 3–5 days before sampling.9–17 Several non-chromatographic procedures have been proposed to determine ‘toxic arsenic’ as a summation parameter.As the fish-derived organoarsenic species (AB, AC, Me4As) do not form volatile hydrides, the sum of i-AsIII + V and its organic metabolites (MMA and DMA) may be determined directly by HGAAS.9–11,18 On-line wet oxidation of organoarsenic species, assisted by thermal19–21 or microwave heating13,22 –30 or by UV irradiation,18,22,31–40 is very attractive in view of automation of this non-chromatographic speciation analysis.Determining As with and without on-line oxidation gives a measure for the ‘total As’ and the ‘toxic As’. On-line conversion of non-hydride forming organoarsenic species is indispensable in HPLC speciation techniques with post-column continuous-flow hydride generation (HG) such as HPLC– HGAAS,19–21,29,30,36 HPLC–HG–ICP-OES,8,32–34 or HPLC– HG–MIP-OES.41 Interfacing HPLC with AAS via an on-line oxidation module and hydride generator does entail some sensitivity loss due to measuring the analyte as the less sensitive arsenate species and involves effects of several chemical and instrumental parameters on analytical performance that call for a thorough optimization and control of the entire analytical system.On-line pre-reduction of arsenate to the more sensitive i-AsIII species26,27,30,42 is not yet straightforward and entails dispersion problems and further complication of the chemical system.Nevertheless, for routine purposes the analysis of HPLC effluents by means of on-line UV photooxidation and continuous-flow HGAAS with a quartz tube atomizer (QTA) appears to be a more affordable instrumental approach than HPLC coupled with more expensive detectors such as ICPMS8,13 –16,40,41,43–46 or electrospray ionization MS.47 The aim of this work was to design an automated system for flow injection (FI)–UV–HGAAS and HPLC-UV-HGAAS based on commercially available modules and to evaluate the role of instrumental and analytical parameters in the determination of some important arsenic species in urine.Experimental Instrumentation A Perkin-Elmer Model 4100 atomic absorption spectrometer with an FIAS 400 flow injection system (mercury/hydride † Presented at The Third International Symposium on Speciation of Elements in Toxicology and in Environmental and Biological Sciences, Port Douglas, Australia, September 15–19, 1997.‡ On leave from the Faculty of Chemistry, University of Sofia, Sofia 1126, Bulgaria. Analyst, August 1998, Vol. 123 (1703–1710) 1703system mode), AS 91 autosampler, EDL II electrodeless discharge lamp and Epson LQ-870 printer, controlled by a Digital DECstation 425c computer with AA software Version 7.30 under GEM 3.11 and FIAS Firmware 6.1, was interfaced with a Model hwg 6808 beam boost photochemical reaction unit (ICT, International Chemie–Technik, Bad Homburg, Germany), as shown in Fig. 1(a). PTFE reaction coils (knotted reactors) of 0.5 mm id and lengths of 5, 10 and 15 m (Part Nos. IC89551, IC89552 and IC89553, respectively), providing irradiation times (tirr) of 35, 70 and 105 s, respectively, with an 8 W UV lamp (254 nm) were used. In some experiments on the effect of heating on UV photooxidation37 or on-line prereduction, 48 an extra 2 or 5 m knotted heating coil L2, placed in a thermostated bath (TB) at 100 °C, was incorporated in the manifold for heating the liquid flow.Both peak-height (Ap) and integrated absorbance (Aint) measurements were used in FI– UV–HGAAS studies. In the HPLC–UV–HGAAS mode [Fig. 1(b)], the mobile phase was delivered by means of a Perkin-Elmer Applied Biosystems (Foster City, CA, USA) Model 140 C Microgradient LC pump. The flow rates were varied between 0.7 and 1.4 ml min21. The thermostated bath and heating coil L2 were omitted.The HPLC effluent was merged with an oxidant solution (Ox) and passed through a UV irradiation coil L3. A Rheodyne (Cotati, CA, USA) Model 9725i syringe loading sample injector with a 50 ml sample injection loop made of polyetheretherketone (PEEK) (Upchurch Scientific, Oak Harbor, WA, USA) was used for sample injection. The integrated intensity for each measurement phase of the AA spectrometer was digitized using a 12 bit A/D converter synchronized to the instrument cycle using circuitry and software similar to that described by Lum et al.49 installed in an Epson AXII PC.The absorbance was calculated for eh cycle and applied to a 12 bit D/A converter after averaging. Thus a continuous analogue output of suitable range and time constant was generated for application to the output of a Perkin-Elmer Nelson (San Jose, CA, USA) Model 1022 integrator. Quantification was based on peak area measurements in the ‘valley-to-valley’ mode of baseline establishment.Optimized instrumental parameters are summarized in Table 1. Reagents Most solutions were prepared from analytical-reagent grade reagents (Merck, Darmstadt, Germany) unless stated otherwise. Stock standard solutions of inorganic and organic arsenic species containing 1000 mg ml21 As were prepared from the reagents given in Table 2. Solutions of organoarsenic species at 1000 and 10 mg ml21 As levels were stored in Teflon FEP flasks at 4 °C without adding any preservatives; more dilute working standard solutions were prepared fresh daily.Alkaline peroxodisulfate solutions containing 0.5–4% m/v K2S2O8 and 0.5–2 mol l21 NaOH were prepared daily. A stock standard solution of the reductant containing 6% m/v sodium tetrahydroborate (NaBH4) (Riedel-de Ha�en, Seelze, Germany) and 1% m/v NaOH was stored in a refrigerator for up to 2 weeks and was diluted daily to yield 0.4–1% m/v NaBH4 solutions containing NaOH and 0.08% v/v antifoaming agent (Dow Corning Antifoam 110 A Emulsion, Midland, MI, USA).Optimized concentrations and flow rates of reagents are given in Table 1. Three HPLC columns with strong anion exchangers were evaluated: CompAx PEEK Spherisorb S SAX, 120 3 4 mm id, 5 mm film thickness, with a 15 mm pre-column (Knauer, Berlin, Germany); Supelcosil SAX 1, 250 3 4.6 mm id, 5 mm film thickness (with trimethylpropylamine groups) with a 20 mm pre-column (Sigma–Aldrich Chemie, Gesch�aftsbereich Supelco, Deisenhofen, Germany); and Hamilton PRP X-100, 250 3 4 mm id, 10 mm film thickness [with poly(styrene– divinylbenzene) with trimethylammonium exchanger] with a 20 mm pre-column (CS Chromatographie Service, Langerwehre, Germany).To prepare the mobile phases, phosphate buffers with pH values between 2.22 and 8.48, containing 5–30 mmol l21 Fig. 1 Schematic diagram of the manifold and instrumental set-up for (a) FI–UV–HGAAS and (b) HPLC–UV–HGAAS.AS, autosampler; S, sample; C, carrier; P1 and P2, peristaltic pumps of the FIAS- 400 system; L1, sample injection coil (in FIAS mode only); TB, thermostated bath with a heating coil L2; UV, photochemical reaction unit with a PTFE irradiation coil L3; MHS, mercury/ hydride system manifold with an MHS reaction coil L4; Ox, oxidant; A, acid; R, reductant; W, waste; GLS, gas–liquid separator; QTA, quartz tube atomizer (for details see text and Table 1).Table 1 Optimized instrumental and analytical parameters for FI–UV– HGAAS and HPLC–UV–HGAAS Parameter Setting Lamp power EDL II, 350 mA Wavelength 193.7 nm Bandpass 2 nm (low) Deuterium background correction AA–BG mode Quartz cell temperature 900 °C Gas–liquid separator Type B (Fig. 4), fed and drained via PTFE capillaries (see text) Integration time (FIAS mode) 49 s Sample injection coil 100 ml (PTFE, L1, FIAS mode) 50 ml (PEEK, HPLC mode) UV irradiation coil (L3) PTFE knotted reactor, 10 m 3 0.5 mm id MHS reaction coil (L4) PTFE knotted reactor, 1 m 3 0.8 mm id Flow rate of oxidant (Ox), acid (A) and reductant (R) 0.5 ml min21 at 70 rev min21 of pump 2 with ‘green/yellow’ tubing, 0.44 mm id Flow rate of mobile phase 1 ml min21 Flow rate of Ar purge gas 45 ml min21 Drainage of GLS 100 rev min21 of pump 1 with ‘red/red’ tubing, 1.14 mm id Oxidant (Ox) 4% m/v K2S2O8–1 mol l21 NaOH Acid (A) 4 mol l21 HCl Reductant (R) 1% m/v NaBH4–0.17% m/v NaOH– 0.08% v/v antifoaming agent Mobile phase Phosphate buffers (see text for optimum compositions with different HPLC columns) 1704 Analyst, August 1998, Vol. 123PO4 32, were prepared fresh daily from orthophosphoric acid (H3PO4), potassium dihydrogenphosphate (KH2PO4) and anhydrous dipotassium hydrogenphosphate (K2HPO4) and were filtered through a 0.45 mm PTFE membrane, purged and pressurized with helium (99.995% v/v) (Linde, Unterschleissheim, Germany). Urine samples and Certified Reference Materials The Standard Reference Material SRM 2670 Toxic Metals in Freeze-Dried Urine ( Normal and Elevated Level) from the National Institute of Standards and Technology (NIST) (Gaithersburg, MD, USA) and Lyphochek Urine Metals Control Level 1, Lot No. 69021, and Level 2, Lot No. 69022, from Bio- Rad Laboratories (Munich, Germany) were reconstituted with de-ionized water as recommended by the manufacturers. Samples denoted ‘fish-eater’s urine’ were ‘spot’ samples of first morning urine obtained 10 h after consumption of 100 g of shrimps and 100 g of tuna fish. Urine samples and CRMs were filtered through a 0.45 mm membrane, FP 030/20 White Rim disposable filter holder (Schleicher and Schuell, Dassel, Germany) before analysis and analyzed the same day except when stability on storage at 4 °C was being studied.Results and discussion A number of instrumental and analytical parameters require optimization in speciation analysis by HPLC–HGAAS. The effects of some variables are illustrated in Figs. 2–4 and Table 3, and the recommended optimum conditions are summarized in Table 1. Among the most important considerations are matching of the flow rates (Fl) of fluids in the entire analytical system, ensuring high and reproducible responses from all inorganic and organic arsenic species, achieving baseline HPLC separation within a reasonable time and providing high signalto- noise ratios, i.e., good limits of detection (LOD).Flow rates of carrier or HPLC mobile phase Typical flow rates of the mobile phase for HPLC columns of 4.0–4.6 mm id are between 0.5 and 1.5 ml min21, which are significantly lower than the optimum flow rates of the carrier in FI–HGAAS systems, viz., 4–7 ml min21. Decreasing the flow rate of the carrier to 1 or 0.5 ml min21 [Fig. 2(a)] and correspondingly also of other reagents resulted in more pronounced signal fluctuations and dispersion due to pulsation of peristaltic pumps at low pump speeds and higher noise levels produced during gas–liquid separation in the MHS reaction coil (L4) and gas–liquid separator (GLS).The flow rate of the carrier was fixed at 1 ml min21 in FI– HGAAS and FI–UV–HGAAS experiments, and efforts were made to decrease signal fluctuations by optimizing other parameters, such as addition of an antifoaming agent, construction of GLS, Ar flow rate and length of L4. The dual-syringe solvent delivery system was found to produce much less fluctuation in HPLC–UV–HPLC measurements than the peristaltic pump in FI–UV–HGAAS experiments.Effect of antifoaming agent Preliminary experiments with the addition of an antifoaming agent to the reductant solution showed that gas–liquid separation was smoother and signal fluctuations were reduced by about fivefold [Fig. 2(b)]. This was in agreement with previous experience with direct FI–HGAAS18 and FI–on-line microwave- assisted digestion–HGAAS23,24 in the analyses of urine samples.Hence the concentration of antifoaming agent in the NaBH4 solution was kept at an optimal level of 0.08% v/v in all further experiments; increasing its concentration above 0.1% v/v entailed a decrease in sensitivity. UV Photooxidation Optimization of UV photooxidation of organoarsenic and organotin species with alkaline and acidic peroxodisulfate, respectively, in an FI–UV–HGAAS system has been discussed in more detail elsewhere.37 The concentrations of K2S2O8 and NaOH are not critical within the ranges 0.5–4% m/v and 0.4–2 mol l21, respectively, and recoveries of i-AsIII, i-AsV, MMA, DMA, AB and AC were between 91 and 101% with a 10 m reactor (irradiation time tirr = 70 s).Additional experiments performed with two other non-hydride forming species, TMAO and Me4As, resulted in recoveries of 114 ± 4 and 102 ± with 1% m/v K2S2O8–0.5 mol l21 NaOH and 103 ± 3 and 94 ± 9% with 4% m/v K2S2O8–0.5 mol l21 NaOH, respectively.In HPLC–UV–HGAAS measurements, the concentrations of peroxodisulfate and NaOH in the oxidant solution were kept at their maximum levels, 4% m/v K2S2O8–1 mol l21 NaOH, while the flow rate of the oxidant was decreased from 1 to 0.5 ml min21, providing longer irradiation/reaction times (93 versus 70 s) in the knotted reactor (10 m 3 0.5 mm id) without causing any apparent peak broadening. The thermostated bath (TB) with a heating coil L2 was omitted since it had only a minor effect on UV photooxidation of organoarsenicals but entailed a shortened lifetime of the quartz lamp wrapped with a heated coil.The overall effect of the UV photochemical reaction unit was an apparent increase in the HPLC retention times by about 2 min (consisting of tirr = 93 s + travel times in T-pieces and connections). However, UV photooxidation resulted not only in the transformation of all analyte species into a definite chemical form, i-AsV, but also in a substantial decrease in sensitivity (2–7 -fold) because hydride generation is significantly slower from i- AsV than i-AsIII.Therefore, the HG conditions require careful optimization. Table 2 Arsenic species Abbreviation Formula Compound and solvent Supplier* i-AsIII NaAsO2 As2O3 in NaOH solution A i-AsV H3AsO4 Arsonic acid in 0.5 m HCl B MMA CH3AsO3Na2·6H2O Sodium methylarsonate C DMA (CH3)AsO2Na·3H2O Dimethylarsinic acid, sodium salt D AB (CH3)3As+CH2COO2 Arsenobetaine in water D AC (CH3)3As+CH2CH2OHBr Arsenocholine bromide in water D TMAO (CH3)3AsO Trimethylarsine oxide in water E Me4As (CH3)4AsI Tetramethylarsonium iodide in water E * A, Riedel-de Ha�en, Seelze, Germany; B, Merck, Darmstadt, Germany; C, Carlo Erba, Milan, Italy; D, Community Bureau of Reference (BCR), Brussels, Belgium; and E, donated by Dr.Walter Goessler, University of Graz, Austria. Analyst, August 1998, Vol. 123 1705Length of the reaction coil L4 in the MHS manifold Hydride generation and stripping from solutions in flow systems is strongly kinetically controlled, increased reaction times on increasing the length of the coil L4 therefore had a marked effect on the i-AsIII signal [Fig. 2(c)]. The integrated absorbance doubled with a knotted reactor (KR, 100 cm 3 0.8 mm id) versus a minimum length (RC, 10 cm 3 0.5 mm id) of PTFE tubing. The integrated absorbance of the i-AsV species was increased sevenfold under the same conditions. Hence the loss of sensitivity caused by HPLC–UV–HGAAS measurements of i-AsV can be substantially alleviated by employing a long reaction coil L4.Unfortunately, using longer reaction coils also entails more pronounced signal fluctuations [see Fig. 2(c)]. Concentrations and flow rates of reagents A 1% m/v NaBH4 solution, which is a higher concentration than usual, was found to result in a better sensitivity for the i-AsV species and a better signal-to-noise ratio [Fig. 2(d)]. The effect of HCl concentration in the acidification channel on the integrated absorbance signals for i-AsIII, i-AsV, MMA and DMA species in FI– HGAAS is shown in Fig. 3. The sensitivity for i- AsV increases with increasing acid concentration, and a concentration of 4 mol l21 was found to be the optimum after UV photooxidation, whereas a 0.5 mol l21 HCl concentration provided good compromise conditions for measuring the other hydride forming species, i-AsIII, MMA and DMA [Fig. 2(d)]. Under these conditions, TMAO gives a very small response of about 10% of that for i-AsIII, whereas AB, AC and Me4As produce no hydride at all.The flow rate of the Ar purge gas affects the sensitivity and the signal stability in opposite ways. The lower the gas flow rate, the better is the sensitivity, because of the dilution effect of the purge gas. At the same time, however, signal fluctuations are more pronounced at low Ar flow rates; hence a compromise flow rate of 45 ml min21 was employed in all measurements.Using higher flow rates of 50, 60 and 70 ml min21 entailed decreases in sensitivity by 3, 23 and 32%, respectively, for i- AsIII in integrated absorbance. Construction of the gas–liquid separator Four GLSs, which are shown schematically in Fig. 4, were evaluated, aiming at improved signal-to-noise ratios, sensitivities for i-AsV and other performance characteristics given in Table 3. The mode of feeding and draining of the GLS was found to be particularly important.The gas–liquid flow emerging from the outlet of the capillary L4 forms a jet about 7–8 cm long which can be a source of excessive noise when it Fig. 2 Typical flow injection peaks illustrating optimization of hydride generation module at 10 ng As levels. (a) Effect of low flow rate of carrier; (b) effect of antifoam addition; (c) effect of the length of MHS reaction coil L4, a 100 cm knotted reactor (KR), versus a straight 10 cm reaction coil (RC); (d) signals for four hydride-forming arsenic species under optimized conditions (0.5 mol l21 HCl, 1% m/v NaBH4, 100 cm KR). Fig. 3 Effect of HCl concentration in the acidification channel on the integrated absorbance signal for four arsenic species (2.5 ng As each) in HPLC–HGAAS mode. Fig. 4 Schematic depiction of four different designs of gas–liquid separators and typical peak shapes for 5 ng of i-AsIII obtained with these separators in FI–HGAAS. 1706 Analyst, August 1998, Vol. 123is not directed in a proper manner within the GLS.More aerosol and pulsation are produced when the jet feeds the upper part of the GLS; better results are obtained when it impacts the wall close to the bottom of the GLS. The worst performance was obtained with the GLSs with a frit at the bottom (type C) or at the side (type D), in which the formation of even a small layer of liquid on the bottom because of ‘underdraining’ the GLS resulted in bubbling and excessive noise. Better results were obtained when the GLSs types B, C and D were fed and drained by means of PTFE capillaries (0.8 mm id) fitted into the sidearms of these GLSs, whereby the outlet (drain) capillary touched the lowest bottom part of the GLS while the inlet capillary was about 2 mm away from the drain capillary in GLS type B or C or touched the upper part of the side frit of the GLS type D.Smooth draining was facilitated by always using higher flow rates in the drain channel by means of an independently controlled pump 1, i.e., the GLS was working in an ‘overdraining’ mode.This arrangement minimized flooding, bubbling and aerosol formation; however, it also affected the sensitivity slightly, as part of the gaseous phase is also pumped to waste. As can be seen from the results in Table 3, GLS type B showed the best overall performance and was adopted in subsequent HPLC–UV–HGAAS work. HPLC separation conditions The composition, concentration and pH of the phosphate buffers (mobile phase) were optimized for three HPLC columns with strong anion exchangers.With a Spherisorb S SAX column and 10 mmol l21 KH2PO4–K2HPO4 at pH 7.5–7.8, the retention times for DMA, i-AsIII, MMA and i-AsV were 2.17, 2.25, 3.30 and 5.80 min, respectively; the first two peaks of DMA and i- AsIII could not be separated. With a Supelcosil SAX 1 column and a mobile phase of 20 mmol l21 KH2PO4 at pH 4.64, the retention times for AC + Me4As, i-AsIII, AB, DMA + TMAO, MMA and i-AsV were 5.0, 6.0, 6.6, 7.8, 14.5 and 15.6 min, respectively.Again, some species could not be resolved. Increasing the pH of the mobile phase (KH2PO4–K2HPO4, 20 mmol l21 PO4 32) to 6.2–7.3 resulted in faster elution of the MMA (9.5 min) and i-AsV (10.2 min) but impaired the separation of the early eluting species: i-AsIII (6.2 min), AB (6.2 min) and AC (6.2 min). Typical chromatograms with and without UV photooxidation are shown in Fig. 5. Applications to urine samples are demonstrated in Fig. 6: chromatogram A (from a person with occupational exposure) shows peaks for 39 mg l21 of i-AsIII and 15 mg l21 of As as DMA species; chromatogram B (a fish-eater’s urine) shows peaks for , DMA and i-AsV (300, 19 and 10 mg l21 As, respectively). Similar results but improved resolution of MMA and i-AsV species were obtained with a Hamilton PRP X-100 column and gradient elution with KH2PO4–K2HPO4 buffer (pH 6.22) as shown in Fig. 7. Chromatographic separation started with 12 Table 3 Comparison of four gas–liquid separators (GLS) for FI–HGAAS at carrier flow rates of 1 ml min21.For design details see Fig. 4 Parameter Type A Type B Type C Type D Total internal volume of GLS/ml 2150 4500 1200 2100 Construction material of GLS PMP* BSG† BSG† BSG† Other characteristics of GLS PTFE — Bottom frit§ Side frit§ membrane‡ Characteristic mass mo for Aint 7.4 6.7 6.5 3.9 Characteristic mass mo for Ap 114 123 105 108 RSD (%) (Aint)¶ 3.8 2.6 1.7 7.3 RSD (%) (Ap)¶ 3.4 1.5 6.2 2.9 i-AsIII/i-AsV sensitivity ratio (QA) 2.8 2.3 2.4 3.0 i-AsIII/i-AsV sensitivity ratio (Ap) 2.8 2.4 2.5 2.8 Compatibility with organic solvents· 222 +++ N.e.†† N.e.Compatibility with heating** + +++ 2 22 Signal stability ++ +++ 22 222 Stability against flooding + +++ 222 22 Washout (signal to baseline) +++ ++ ++ 22 * Polymethylpentene. † Borosilicate glass. ‡ PTFE membrane (Part No. B050-8567) on top of GLS. § Robu-Glass filter, diameter 5 mm, porosity P 1.6, No. 5 (Robuglasfilter-Ger�ate, Hattert, Germany). ¶ n = 5–6 at the 5 ng i-AsIII level. · Tested with a methanol– acetonitrile–water (40 + 30 + 30, v/v) mobile phase. ** Thermostated bath at 100 °C. †† N.e., not evaluated. Fig. 5 Chromatograms from anion-exchange HPLC–UV–HGAAS of arsenic species (2.5 ng of As each) in standard solutions with a Supelcosil SAX1 column, (a) without and (b) with oxidant addition and UV irradiation. Peaks: 1, i-AsIII; 2, DMA; 3, MMA; 4, i-AsV; 5, AB; 6, AC; 7, TMAO; and 8, Me4As.Mobile phase: 20 mmol l21 phosphate buffer (KH2PO4– K2HPO4), (a) pH 6.85 and (b) pH 4.64, at a flow rate of 1 ml min21. Analyst, August 1998, Vol. 123 1707mmol l21 phosphate buffer and continued after 6 min with a ‘stronger’ eluent, 24 mmol l21 PO4 32, in order to decrease the retention time for the strongly retained species, i-AsV, from 14.4 to 11.7 min. The system showed good long-term stability with retention times of 4.43 ± 0.02, 5.51 ± 0.08, 6.69 ± 0.24 and 11.70 ± 0.06 min (RSD 0.5–3.6%) for i-AsIII, DMA, MMA and i-AsV, respectively, during 10 successive chromatographic runs over 2 h.Arsenobetaine, the most common fish-derived organoarsenic species, is eluted together with i-AsIII but could be determined by difference from two chromatographic runs: with (AB + i-AsIII) and without UV photooxidation (i-AsIII only) as shown in Fig. 7. The limit of detection in urine samples expressed as three times the standard deviation for arsenic in a blank solution was 0.2 ng or 4 mg l21 As with UV photooxidation and 2, 4, 3 and 6 mg l21 As without UV photooxidation for i-AsIII, i-AsV, MMA and DMA, respectively.Analysis of urine CRMs Four samples of freeze-dried urine CRMs were analyzed by HPLC–UV–HGAAS with a Hamilton PRP X-100 column: the NIST SRM 2670 Normal and Elevated Levels on the days 1, 7 and 14 after reconstitution and the Bio-Rad Lyphochek Urine Metals Control Levels 1 and 2 on days 1 and 9 after reconstitution.Samples were stored at 4 °C between analyses. No significant difference in arsenic speciation was found during storage, which is in agreement with some recent reports (Ritsema and van Heerde38 found the total arsenic in urine to be Fig. 6 Chromatograms from anion-exchange HPLC–UV–HGAAS of arsenic species in 50 ml urine samples with a Supelcosil SAX1 column with oxidant addition and UV irradiation. Peaks as in Fig. 5. A, urine sample from an occupationally exposed person; and B, B urine sample from a fish eater.Table 4 Arsenic speciation determination by HPLC–UV–HGAAS in NIST SRM 2670 Toxic Metals in Freeze-Dried Urine Normal and Elevated Levels (in mg l21 As). Means ± SD from three series of measurements with three parallels in each series on days 1, 5 and 14 after reconstitution with intermediate storage at 4 °C Certified value and confidence Sum of all species Urine sample limits i-AsIII i-AsV MMA DMA AB determined Normal Level, this work (2/2)* 60‡ < 2 < 4 12.8 ± 2.0 52.4 ± 7.6 N.d.§ 65.2 Normal Level, this work (2/2)† < 4 < 4 18.4 ± 4.4 48.3 ± 2.8 35.6 ± 7.8 102.3 Normal Level, ref. 46 N.d.N.d. N.d. N.d. N.d. 62.8 (58–66) (i-As + MMA + DMA) Normal Level, ref. 43 52.6 ± 4.1 0.8 ± 0.5 9.9 ± 1.4 45.5 ± 3.5 N.d. 109 ± 6 Elevated Level, this work (2/2)* 480 ± 100 < 2 417 ± 64 12.9 ± 4.2 52.3 ± 10.2 N.d. 482.2 Elevated Level, this work (2/2)† < 4 416 ± 40 15.9 ± 3.8 48.9 ± 2.0 32.0 ± 8.4 513 Elevated Level, ref. 43 43.8 ± 9.1 406 ± 153 5.0 ± 3.6 34.8 ± 8.7 N.d. 489 ± 154 Elevated Level, collaborative N.d. N.d. 15.5 ± 4.3 52.2 ± 4.3 ( ~ 30) 430 ± 56 (total study, ref. 50 i-AsIII+V) 487 ± 54 (i-As + MMA DMA) 502 ± 50 (total As) * No UV photolysis/no oxidant. † With UV photolysis and oxidant. ‡ Information value only. § N.d., not determined. Fig. 7 Chromatograms from anion-exchange HPLC–UV–HGAAS of arsenic species with a Hamilton PRP X-100 column and gradient elution with KH2PO4–K2HPO4 buffer, pH 6.22 (0–6 min, 12 mmol l21 PO4 32; 6–13 min, 24 mmol l21 PO4 32).Peaks as in Fig. 5. Chromatograms A and C without and chromatograms B and D with oxidant addition and UV irradiation; A and B, standard solutions containing 2.5 ng of As each; C and D, urine sample (fish-eater’s urine). 1708 Analyst, August 1998, Vol. 123stable during storage for 30 d at 5 ± 4 °C, Crecelius and Yager50 found their intercomparison samples to be relatively stable during storage and shipping and M�urer et al.11 found arsenobetaine and arsenocholine to be relatively stable during storage in the dark during 21 d but observed some loss in daylight).The mean values of all determinations are presented in Tables 4 and 5. Typical chromatograms for both levels of NIST SRM 2670 samples are shown in Fig. 8 and reveal that the Normal and Elevated Levels are identical in their matrix and contents of the arsenic species except that the Elevated Level is spiked with about 410 mg l21 of i-AsV.Another finding was that all four reference materials contain AB, the contribution of which to the total content may have been underestimated during the analytical certification. Some probable reasons for such an underestimation could be the use of quantification procedures based on direct HG or solvent extraction without complete mineralization of the non-reactive AB or DMA, which gives a lower response in direct HGAAS.At least two recently published studies reported the presence of AB in NIST SRM 2670. Pergantis et al.47 found 11 ± 3 mg l21 of AB (as As) by microbore HPLC coupled on-line with electrospray ionization mass spectrometry. Another laboratory reported approximately 30 mg l21 of AB (as As) by HPLC–ICP-MS in a recent intercomparison speciation study.50 Several recent publications reported higher total contents than the information value for As in NIST SRM 2670 Normal Level, such as 69,51 81 ± 152 and 109 ± 6 mg l2143 by ICP-MS, 84 ± 2 mg l21 by FI–HGAAS after UV photolysis22 and 80 ± 6 mg l21 by direct ETAAS22 versus 62 ± 2 mg l21 by FI–HGAAS without UV photolysis.22 Accordingly, higher mean values were also published for the Elevated Level material, such as 489 ± 15443 and 491 ± 3 mg l2152 by ICP-MS, 545 ± 4 mg l21 by FI–HGAAS after off-line UV photolysis38 and 530 ± 42 and 540 ± 19 mg l21 by FI–HGAAS after off-line digestion followed by two different on-line preredon FI–HGAAS procedures.20 The results of our study as presented in Table 4 are generally in good agreement with the recent interlaboratory exercise on arsenic speciation in urine, reported by Crecelius and Yager.50 However, we were unable to find any i-AsIII in these samples, as previously reported by Heitkemper et al.43 using HPLC–ICP-MS (see Table 4).One possible explanation for this discrepancy could be that the i-AsIII and AB were not resolved in the HPLC separation by Heitkemper et al.43 and the peak for AB was attributed to i-AsIII.More work is obviously required before these coupled techniques reach maturity and can enter routine toxicological laboratories. Conclusions The coupling of anion-exchange HPLC and continuous-flow HGAAS, with and without on-line UV photooxidation by alkaline peroxodisulfate, is a suitable and affordable approach to the speciation determination of toxicologically important arsenic species in urine.Although the characteristic mass (m0 = 0.01 ng) and absolute LODs (0.1–0.3 ng) were very low, the relative LODs were impaired to 2–6 mg l21 because of the small sample volumes (50 ml) and because hydride generation was from the oxidized, less sensitive i-AsV species. Some instrumental and chemical parameters that affected analytical performance were identified and optimized. Speciation analysis of urine CRMs revealed the presence of significant amounts of arsenobetaine in these materials, resulting in higher total arsenic levels than previously recommended or certified.The authors are grateful to J. Bauslaugh and B. Radziuk for the modification of the software and interface between the AA spectrometer and the PE Nelson integrator and to W. Goessler for providing standard solutions of TMAO and Me4As. References 1 Cullen, W. R., and Reimer, K. J., Chem. Rev., 1989, 89, 713. 2 Tsalev, D. L., Atomic Absorption Spectrometry in Occupational and Environmental Health Practice, Vol. III: Progress in Analytical Methodology, CRC Press, Boca Raton, FL, 1995, ch. 3, pp. 19–31. Table 5 Arsenic speciation determination by HPLC–UV–HGAAS in Bio-Rad Lyphochek Urine Metals Control Levels 1 and 2 (in mg l21 As, n = 4–6) Certified value and Sum of all confidence species Urine sample limits i-AsIII i-AsV MMA DMA AB determined Level 1 (2/2)* 52 (42–62) < 2 55 ± 11 < 3 10 ± 2 N.d.‡ 65 Level 1 (+/+)† < 4 58 ± 4 < 4 15 ± 6 26 ± 11 99 Level 2 (2/2)* 152 (121–182) < 2 145 ± 21 < 3 35 ± 14 N.d. 180 Level 2 (+/+)† < 4 151 ± 21 < 4 13 ± 4 30 ± 11 194 * No UV photolysis/no oxidant. † With UV photolysis and oxidant. ‡ N.d., not determined. Fig. 8 Arsenic speciation in NIST SRM 2670 Toxic Metals in Freeze- Dried Urine. A and C, Normal Level; B and D, Elevated Level; A and B, without oxidant addition and UV irradiation; C and D, with oxidant and UV irradiation. Chromatographic conditions as in Fig. 7. Analyst, August 1998, Vol. 123 17093 Dedina, J., and Tsalev, D. L., Hydride Generation Atomic Absorption Spectrometry, Wiley, Chichester, 1995, ch. 8, pp. 182–245. 4 Morita, M., and Edmonds, J. S., Pure Appl. Chem., 1992, 64, 575. 5 Gonz�alez Soto, E., Alonso Rodriguez, E., Fern�andez Fern�andez, E., and Prada Rodriguez, D., Ciencia, 1996, 4, 149. 6 Burguera, M., and Burguera, J. L., Talanta, 1997, 44, 1581. 7 Irgolic, K. J., in Hazardous Metals in the Environment, ed.Stoeppler, M., Elsevier, Amsterdam, 1992, pp. 287–350. 8 Caroli, S., La Torre, F., Petrucci, F., and Violante, N., in Element Speciation in Bioinorganic Chemistry, ed. Caroli, S., Wiley, New York, 1996, ch. 13, pp. 445–463. 9 Fo`a, V., Colombi, A., Maroni, M., Buratti, M., and Calzaferri, G., Sci. Total Environ., 1984, 34, 241. 10 Chana, B. S., and Smith, N. J., Anal. Chim. Acta, 1987, 197, 177. 11 M�urer, A. J. L., Abildtrup, A., Poulsen, O. M., and Christensen, J.M., Analyst, 1992, 117, 677. 12 SAS Trace Element Laboratories, Clinical and Analytical Handbook, ed. Walker, A. W., Royal Surrey County and St. LukeAs Hospitals, Guildford, 2nd edn., 1992. 13 Le, X.-C., Cullen, W. R., and Reimer, K. J., Talanta, 1993, 40, 185. 14 Shibata, Y., and Morita, M., Anal. Sci., 1989, 5, 107. 15 Goessler, W., Schlagenhaufen, C., Kuehnelt, D., Greschonig, H., and Irgolic, K. J., Appl. Organomet. Chem., 1997, 11, 327. 16 Le, X.-C., Cullen, W. R., and Reimer, K.J., Clin. Chem., 1994, 40, 617. 17 Le, X. C., Ma, M., and Wong, N. A., Anal. Chem., 1996, 68, 451. 18 Guo, T., Baasner, J., and Tsalev, D. L., Anal. Chim. Acta, 1997, 349, 313. 19 L�opez, M. A., G�omez, M. M., Palacios, M. A., and C�amara, C., Fresenius’ J. Anal. Chem., 1993, 346, 643. 20 Stummeyer, J., Harazim, B., and Wippermann, Fresenius’ J. Anal. Chem., 1996, 354, 344. 21 Palacios, M. A., G�omez, M., C�amara, C., and L�opez, M. A., Anal. Chim. Acta, 1997, 340, 209. 22 Willie, S. N., Spectrochim. Acta, Part B, 1996, 51, 1781. 23 Tsalev, D. L., Sperling, M., and Welz, B., Analyst, 1992, 117, 1729. 24 Tsalev, D. L., Sperling, M., and Welz, B., Analyst, 1992, 117, 1735. 25 Le, X.-C., Cullen, W. R., and Reimer, K. J., Appl. Organomet. Chem., 1992, 6, 161. 26 Welz, B., He, Y., and Sperling, M., Talanta, 1993, 40, 1917. 27 Sperling, M., He, Y., and Welz, B., in CANAS’93 Colloquium Analytische Atomspektroskopie, ed. Dittrich, K., and Welz, B., Universitat Leipzig und UFZ-Umweltforschungszentrum Leipzig- Halle, Leipzig, 1993, pp. 989–1004. 28 Le, X.-C., Cullen, W. R., and Reimer, K. J., Talanta, 1994, 41, 495. 29 L�opez-Gonz�alvez, M. A., G�omez, M. M., C�amara, C., and Palacios, M. A., J. Anal. At. Spectrom., 1994, 9, 291. 30 Lamble, K. J., and Hill, S. J., Anal. Chim. Acta, 1996, 334, 261. 31 Atallah, R. H., and Kalman, D. A., Talanta, 1991, 38, 167. 32 Rubio, R., Peralta, I., Alberti, J., and Rauret, G., J. Liq. Chromatogr., 1993, 16, 3531. 33 Rubio, R., Alberti, J., and Rauret, G., Int. J. Environ. Anal. Chem., 1993, 52, 203. 34 Rubio, R., Padr�o, A., Alberti, J., and Rauret, G., Anal. Chim. Acta, 1993, 283, 160. 35 Rubio, R., Alberti, J., Padro, A., and Rauret, G., Trends Anal. Chem., 1995, 14, 274. 36 Howard, A. G., and Hunt, L. E., Anal. Chem., 1993, 65, 2995. 37 Tsalev, D. L., Sperling, M., and Welz, B., in, CANAS’97 Colloquium Analytische Atomspektroskopie, ed. Werner, G., Universitat Leipzig und UFZ-Umweltforschungszentrum Leipzig–Halle, Leipzig, in the press. 38 Ritsema, R., and van Heerde, E., Fresenius’ J. Anal. Chem., 1997, 358, 838. 39 Zhang, X., Cornelis, R., De Kimpe, J. and Mees, L., Anal. Chim. Acta, 1996, 319, 177. 40 Zhang, X., Cornelis, R., Mees, L., Vanholder, R. and Lameire, N., Analyst, 1998, 123, 13. 41 Costa-Fern�andez, J., Lunzer, F., Pereiro-Garcia, R., Sanz-Medel, A., and Bordel-Garcia, N., J. Anal. At. Spectrom., 1995, 10, 1019. 42 Hanna, C. P., Tyson, J. F., and McIntosh, S., Clin. Chem., 1993, 39, 1662. 43 Heitkemper, D., Creed, J., Caruso, J., and Fricke, F. L., J. Anal. At. Spectrom., 1989, 4, 279. 44 Larsen, E. H., Pritzl, G., and Hansen, S. H., J. Anal. At. Spectrom., 1993, 8, 557. 45 Feldmann, J., Anal. Commun., 1996, 33, 11. 46 Bavazzano, P., Perico, A., Rosendahl, K., and Apostoli, P., J. Anal. At. Spectrom., 1996, 11, 521. 47 Pergantis, S. A., Winnik, W., and Betowski, D., J. Anal. At. Spectrom., 1997, 12, 531. 48 Tsalev, D. L., Sperling, M., and Welz, B., to be published. 49 Lum, K., Naranjit, D., Radziuk, B., and Thomassen, Y., Anal. Chim. Acta, 1983, 155, 183. 50 Crecelius, E., and Yager, J., Environ. Health Perspect., 1997, 105, 650. 51 V�ollkopf, U., and Barnes, K., At. Spectrosc., 1995, 16, 19. 52 Nixon, D. E., and Moyer, T. P., Spectrochim. Acta, Part B, 1996, 51, 13. Paper 8/03044H Received April 23, 1998 Accepted June 8, 1998 1

ISSN:0003-2654    

DOI:10.1039/a803044h

出版商:RSC    

年代:1998

数据来源: RSC

摘要:

Validated determination of total arsenic species of toxicological interest (arsenite, arsenate and their metabolites) by atomic absorption spectrometry after separation from dietary arsenic by liquid extraction: toxicological applications L. Benramdaneab, M. Accominottiab and J. J. Vallon*ab a Laboratoire de Biochimie, Pharmacotoxicologie et Analyse des Traces, H�opital Edouard Herriot, Place d’Arsonval, 69437 Lyon 03, France b Laboratoire de Chimie Analytique III, Facult�e de Pharmacie, Universit�e Claude Bernard, 8 Avenue Rockefeller, 69373 Lyon 08, France A validated method for the selective extraction of total As species of toxicological interest (arsenite, arsenate and mono- and dimethylated arsenic species) from urine, followed by atomic absorption spectrometric determination, is described.The mechanisms involved in extraction were studied and the extraction method was optimized. The urine sample was acidified with concentrated HCl and KI and sodium hypophosphite were added.Under these conditions, As species were reduced to their corresponding iodide arsines, extracted with toluene and back-extracted with 1 mmol l21 NaOH solution. Only inorganic arsenic and its metabolites in humans (monomethylarsonic and dimethylarsinic acid) were extracted. Arsenobetaine of dietary origin was not extracted. This method can detect if any As increase in urine originates from inorganic As intoxication or only from dietary non-toxic As species such as arsenobetaine.Keywords: Inorganic arsenic; inorganic arsenic metabolites; arsenobetaine; arsenic reduction; arsenic extraction; atomic absorption spectrometry; urine Because of its ubiquity in the environment , arsenic is present in all environmental compartments in numerous inorganic and organic chemical forms. The distribution of As in the environment results from natural processes (e.g., marine sedimentary rocks, erosion and weathering of soils and minerals) and human activities (e.g., mining operations, smelting, pesticide manufacture). 1,2 The chemical form of As depends on its source: inorganic As from minerals, industrial discharges and pesticides and organic As from industrial discharges, insecticides and biological action on inorganic As.2,3 In humans, inorganic As [arsenite (AsIII) and arsenate (AsV)] is by far the most toxic of the arsenic species: for an adult individual, the lethal dose is probably between 100 and 200 mg of arsenious acid.2,4 In the organism, inorganic As is metabolized in its methylated forms [monomethylarsonic acid (MMA) and dimethylarsinic acid (DMA)].The ingestion of large amounts of inorganic arsenicals causes acute intoxication characterized mainly by serious abdominal, cardiovascular and neurological symptoms.5–7 Chronic intoxication by inorganic As is observed under continuous exposure even to very low doses over a few years. Chronic exposure is associated with dermatological, neurological and vascular effects.2,4,6 After some years of exposure, As causes skin cancer, lung cancer and other types of malignant neoplasms.2,8,9 From a metabolic point of view, inorganic As is methylated in the organism and the major metabolite is DMA.In workers exposed chronically to inorganic As, Farmer and Johnson10 found that the total urinary As level is increased up to hundreds of mg l21 with the following distribution of the different species: AsV 3.5%, AsIII 12.5%, MMA 16% and DMA 66.5%.Urine excretion forms the major pathway for the elimination of As compounds from the body; hence urinary As determinations are important for assessing a subject’s exposure to toxic As species, because a good correlation is observed between absorption and urinary excretion. In urines, average concentrations are between 20 and 50 mg l21.11,12 In biological analysis the total urinary As concentration is still used for the assessment of As exposure.However, the measurment of total As may not be appropriate for the biological monitoring of occupational exposure. Indeed, numerous studies have shown that total urinary As can show an important rise within the 10 h following seafood ingestion.11,13 Moreover, these elevated levels are without any consequence on health because As of dietary origin is exclusively in organic and therefore non-toxic forms.13215 The major organic form of As in marine organisms is arsenobetaine (Asbet); it represents 96% of total extracted As from the tissue, the remaining 4% being arsenocholine (Aschol), tetramethylarsonium (TMAs) and arseno sugars.15217 The As species of toxic origin are inorganic As (AsIII, AsV) and its metabolites (DMA and MMA).In unexposed subjects these species are present in the urine in small amounts and their levels remain unchanged after marine food consumption.11 These observations show that it is important to apply a method selective enough to determine the As of toxic origin separately from Asbet, which is the major As form of dietary origin.With this aim, several speciation methods have been reported. They involve the determination of the individual concentrations of AsIII, AsV, DMA, MMA, Asbet, Aschol and sometimes TMAs and arseno sugars17–19 in urine. Speciation is generally accomplished in three steps: sample preparation, species separation and their detection. HPLC is the most common separation technique used in ion-exchange19,20 and ion-pairing17 modes, coupled with various systems of detection: ICP-AES,21,22 AAS23,24 and ICP-MS.18–20 These methods are selective and sensitive but require a very fastidious pre-treatment to eliminate the matrix (proteins, salts and interferent elements), which alters the chromatographic separation and interferes in detection by ICP-MS.In this paper, we propose a useful, simple, rapid and sensitive method for the unique determination of inorganic As (AsIII, AsV) and its metabolites (DMA, MMA), based on solvent extraction and determination by AAS; organic As from dietary intake (Asbet) is not extracted.Similar methods have been reported. One method is based on reduction of AsV with KI in 40 min after acidification with HCl, extraction with toluene followed by back-extraction with dilute HNO3.20 In another modified Analyst, August 1998, Vol. 123 (1711–1715) 1711method toluene and dilute HNO3 solution were replaced with chloroform and de-ionized water, respectively.25 This paper demonstrates the mechanism of extraction of each As species and the role of each reagent implicated in the reaction preceding the extraction step.This study allowed the optimization of a very simple, fast and reproducible method, with a high yield of As species of inorganic origin; sodium hypophosphite added to the medium drastically improved the performance of the method. When total As is found at elevated levels in urine, it is important to determine the percentage of As of inorganic origin.In this case the proposed method can be applied with great facility. Experimental Reagents All reagents were of analytical-reagent grade. Stock standard solutions of each arsenic compound (1 g l21) were prepared by dissolving in water appropriate amounts of As2O3 (Sigma, St. Louis, MO, USA), Na2HAsO4·7H2O (Sigma), CH3AsO3- Na2·6H2O (Carlo Erba, Milan, Italy) and (CH3)2AsO2Na·3H2O (Sigma).A mixture of these four As species was also prepared (see Sample preparation). The Asbet solution (0.412 g l21 As) was provided by CNRS Solaize (Lyon, France). For the reduction and extraction step, 37% HCl (Merck, Darmstadt, Germany), KI (Fluka, Buchs, Switzerland), NaH2PO2·H2O (Prolabo, Paris, France) and 1 mmol l21 solution NaOH (Fluka) were used. For ETAAS, 0.05% Triton X-100 (Merck) was used for sample dilution. A 10 g l21 Ni(NO3)2·6H2O (Merck) solution was used as a chemical modifier. A 50 mg l21 working standard solution, prepared daily from a 1 g l21 As stock standard solution (Spex Industries, Edison, NJ, USA) and Lyphocheck urine controls levels 1 and 2 with As concentrations of 50 and 150 mg l21, respectively (Bio-Rad Labs., Richmond, CA, USA) were used for checking the AAS instrument. Urines selected with a low total As content ( < 10 mg l21) were spiked with the mixture of As spesee Sample preparation) for calibration of the extraction step.Instrumentation The AAS instrument used was a SpectrAA Zeeman 220 (Varian, Palo Alto, CA, USA) equipped with a pyrolytic graphite-coated graphite furnace and a Zeeman-effect background corrector. Extraction method Qualitative study Basis of the method. Our method for the extraction of AsIII, AsV, DMA and MMA from urine was adapted from an early method described by Charlot:26 AsIII is extracted with benzene from a concentrated HCl medium; AsV is extracted only in the presence of KI and hypophosphite.Our adaptation to the four As species of inorganic origin was developed as follows. Study of mechanisms involved in extraction. Concentrated solutions (0.1 g l21) of AsIII, AsV, MMA and DMA were used because of their color properties, allowing rough observation of the extraction process. Benzene was replaced with the less toxic toluene. (i) Arsenite. Arsenite is extracted by the organic solvent from a concentrated HCl medium and in the presence of I2 according to the following reaction: H3AsO3 + 3IH " AsI3 + 3H2O (1) The reaction is complete and immediate in concentrated HCl; AsI3 is yellow in water and orange–red in toluene.If I2 is absent, AsIII can also be extracted, probably by formation of AsCl3 which is an uncolored species. Back-extraction of AsIII in water is then explained by the following equation: (AsI3)toluene + 3H2O " (H3AsO3)water + 3HI (2) The equilibrium involves both the organic and aqueous phases.Its deplacement to the right is greatly facilitated by alkalinization, and the yield is greatly improved compared with an acidic or neutral medium. (ii) Arsenate. Arsenate can only be extracted if it is reduced to AsIII. I2 is responsible for this reduction together with the formation of AsI3, which is then extracted. The reactions involved are as follows: H3AsO4 + 2HI"H3AsO3 + I2 + H2O (3) H3AsO3 + 3HI"AsI3 + 3H2O (4) (iii) MMA and DMA. MMA and DMA are extracted by the same mechanism as arsenate because in these species arsenic is pentavalent (AsV).MMA and DMA are reduced with the formation of the methylated arsines CH3AsI2 and (CH3)2AsI.27 Both of these species are brown in both HCl and toluene media. The corresponding reactions are CH3AsO(OH)2 + 4HI"CH3AsI2 + I2 + 3H2O (5) (CH3)2AsO(OH) + 3HI"(CH3)2AsI + I2 + 2H2O (6) The back-extraction from toluene is performed according to eqn. (2). (iv) The role of sodium hypophosphite. During the qualitative study of extraction we observed that in the absence of hypophosphite, AsI3, CH3AsI2 or (CH3)2AsI is slowly destroyed with the appearance of a yellow species in HCl (I32) and a purple species in toluene (I2).This destruction can be avoided if the medium is added with the very reducing hypophosphite ion. If KI is also present, its reducing power makes the reaction immediate. Optimization of the extraction method Optimization of the method was studied on each species (AsIII, AsV, DMA, MMA) in 300 mg l21 solutions.Several parameters can modify the nature of extraction, e.g., concentration of HCl, KI, NaH2PO2, nature of extraction solvent, pH of the aqueous solution for back-extraction of As and reaction and mixing times. At concentrations of HCl !6.5 mol l21 (Fig. 1), KI !75 Fig. 1 Extraction yield as a function of HCl concentration in the medium. 1712 Analyst, August 1998, Vol. 123mmol l21 and NaH2PO2 !325 mmol l21 (Fig. 2), total and immediate reduction is observed.The yields are close to 100% if toluene extraction is accomplished in two steps (2 ml each) with 2 min of vortex mixing. Back-extraction from toluene is complete with 1 mmol l21 NaOH. Assays with identical conditions showed that arsenobetaine is unextractable. Validation of the method Sample preparation Calibration standards (15, 20, 30, 40, 60 and 120 mg l21) were prepared by spiking six 1 ml urine aliquots with appropriate volumes of concentrated arsenic mixture.This mixture contained AsIII, AsV, DMA and MMA, at 0.250 g l21 equivalent As (the total As concentration is 1 g l21). The volume added was always @2% (5–20 ml) of total volume of the sample in order to keep the matrix constant.28 Linearity The extraction was tested from 0 to 300 mg l21. Atomic absorption showed linearity up to 100 mg l21. The linearity of the method was confirmed using classical tests, i.e., comparison of the intercept with zero and correlation coefficients.The intraassay reproducibility was determinated for replicate calibration curves prepared on the same day (n = 6). The inter-assay calibration curves were determined on six successive days. All data were obtained using the same urine spiked with the same mixture of AsIII, AsV, DMA and MMA. Precision and accuracy The intra- and inter-day precision and the accuracy of the method were determined by performing replicate analyses of urine spiked with the As mixture (AsIII, AsV, DMA and MMA) to prepare low, medium and high concentration levels (10, 50 and 150 mg l21, respectively).The procedure was repeated using the same spiked standards on the same day (n = 6) and on different days (n = 6) to determine the intra- and inter-day repeatability, respectively. The accuracy, expressed as the percentage deviation of the observed concentration from the theoretical concentration, with the relative error, was evaluated. Recovery The extraction efficiency was determined for all compounds by comparing the signal from urine spiked with a known amount of As species, in the range of the calibration curves, assayed accordingly, versus the signal for the same concentrations prepared in the NaOH–Triton X-100 mixture.Each sample was analysed in duplicate. Limit of detection (LOD) and limit of quantification (LOQ) As the urine sample used for calibrations is not free from any arsenical, it could not be taken as a blank. Therefore, for AAS measurements, 1 mmol l21 NaOH diluted twofold in 0.05% Triton X-100 was used to determine the LOD and the LOQ.This procedure was validated by comparison of the slopes (B) of calibration curves for NaOH–Triton X-100 with curves for urine extracts. Twelve curves for each medium showed that the mean slopes were identical (B = 0.0024 ± 0.0002 and 0.0025 ± 0.0001, respectively) and that the procedure was valid. The LOD was determined by measuring 30 times the signal of the NaOH–Triton X-100 mixture and determining the concentration from the calibration curve in NaOH.The LOD was taken as three times the standard deviation of this concentration. The LOQ was taken as 10 times the standard deviation of this concentration. Results Optimized method A 1 ml aliquot of urine in a polypropylene tube containing 100 ml of 3.5 mol l21 KI, 0.5 ml of 3 mol l21, NaH2PO2, and 3 ml of 37% HCl was extracted with 2 ml of toluene. Following vigorous shaking for 2 min, the mixture was centrifuged for 2 min at 3000 rpm.The organic supernatant was removed and retained and a second extraction step was performed. The total volume (4 ml) of organic phase was then back-extracted in 1 ml of 1 mmol l21 NaOH with 5 min of vortex mixing. The aqueous alkaline phase was diluted twofold with 0.05% Triton X-100 solution and then measured by AAS. In this step, the sample and the nickel modifier were co-injected into the furnace according to the procedure given in Table 1.Sample analyses were made by reference to calibration curves. Validation Linearity The correlation coefficients (r) for the calibration curves were 0.9988 and 0.9973 for intra-assay and inter-assay linearity, Fig. 2 Extraction yield as a function of KI final concentration in the medium. Table 1 AAS instrument and furnace parameters Instruments parameters— Wavelength 193.7 nm Slit width 0.2 nm Lamp current 10 mA Tube Pyrolitic graphite coated Background correction On Measurement mode Peak area Sample volume 10 ml Sample modifier 5 ml Furnace parameters— Temperature/ Gas flow rate/ Step °C Time/s 1 min21 Read 1 85 5.0 3.0 No 2 95 40.0 3.0 No 3 120 10.0 3.0 No 4 300 10.0 3.0 No 5 1200 5.0 3.0 No 6 1200 1.0 3.0 No 7 1200 2.0 0.0 No 8 2700 0.6 0.0 Yes 9 2700 2.0 0.0 Yes 10 2600 2.0 0.5 No 11 2600 1.0 3.0 No Analyst, August 1998, Vol. 123 1713respectively. For each point of the calibration, the concentrations were recalculated from the equation of the linear regression curves (experimental concentrations) and the relative standard deviations (RSD, %) were calcualted.The results are presented in Table 2. The small percentage differences between the nominal and found concentrations of the standards on the calibration curves confirmed that the assays were linear over the concentration ranges investigated. The mean slopes, intercepts and correlation coefficients are presented in Table 3. Concentrations of toxic As in unknown samples were determined using the calibration curves.Precision and accuracy The intra- and inter-day precision and the accuracy of the method are given in Table 4. The precision around the mean values did not exceed 10% (RSD). Recovery The recoveries obtained for AsIII, AsV, DMA, and MMA were 100.8, 101.9, 99.8 and 100.3%, respectively with corresponding RSDs of 6.40, 6.85, 7.35 and 8.32% (n = 18). LOD and LOQ The LODs and the LOQs determined as previously defined were 2 and 7 mg l21, respectively.Specificity Arsenobetaine is of dietary origin and non-toxic. In order to evaluate a possible interference, we spiked the same batch of urines with Asbet. The results (Table 5) indicate that Asbet is not extractable by the proposed method, at any concentration, and in the range of concentrations of inorganic As on the calibration curves. Discussion and conclusions The study of the mechanisms of the reactions involved in the reduction and extraction of iodide arsines formed by toluene showed that reduction of pentavalent As species (arsenate, DMA, MMA) is greatly improved by hypophosphite; hypophosphite is a better reductive species than HI, which was used alone in all previous methods proposed in the literature.Addition of hypophosphite also increases the yields of reduction –extraction together with shorter reduction times, owing to its reducing power, which improves the reducing action of HI and preserves the iodide arsines from destruction by oxygen.Another improvement is obtained by back-extraction in dilute NaOH, which displaces the equilibrium of extraction towards the aqueous phase with better yield. The recovery of all As species was 100% and the LOQ was 7 mg l21. Arsenobetaine, which is the major non-toxic organic species of As, appeared unextractable by the proposed method. The described validated method is very simple, specific and very fast.It allows the rapid investigation of any inorganic As intoxication without the use of the much more expensive, long and tedious separation methods of arsenic speciation. References 1 Fitzgerald, L. D., in Arsenic, Industrial, Biomedical and Environmental Perspectives, ed. Lederer, W. H., and Fensterheim, R. J., Van Nostrand, New York, 1983, pp. 3–8. 2 Hindmarsh, J. T., and McCurdy, R. F., Crit. Rev. Clin. Lab. Sci., 1986, 23, 315. 3 Braman, R. S., Top. Environ. Health, 1983, 6, 141. 4 Clinical Toxicology, ed. Polson, C. J., and Tattersal, R. N., Pitman, London, 1969, pp. 181–184. 5 Franzblau, A., and Lilis, R., Arch. Environ. Health, 1989, 44, 385. 6 Lovell, M. A., and Farmer, J. G., Hum. Toxicol., 1985, 4, 203. Table 2 Reproducibility of the method Found/mg l21* Intra-assay Inter-assay Added/mg l21 (n = 6) (n = 6) 0 8 ± 0.7 8 ± 0.8 (8.8%) (10.0%) 15 23 ± 1.7 23 ± 1.9 (7.4%) (8.3%) 20 29 ± 1.8 28 ± 2.2 (6.2%) (7.9%) 30 38 ± 1.9 38 ± 2.5 (5.0%) (6.6%) 40 48 ± 1.9 49 ± 2.1 (4.0%) (4.3%) 60 69 ± 2.8 69 ± 3.7 (4.1%) (5.4%) 120 129 ± 3.0 128 ± 3.7 (2.3%) (2.9%) * RSD in parentheses.Table 3 Intra- and inter-assay linearity Intra-assay Inter-assay Parameter linearity linearity Slope (mean ± s) 0.0025 ± 0.0001 0.0024 ± 0.0002 Intercept (mean ± s) 0.0280 ± 0.0019 0.0293 ± 0.0013 Coefficient of linear regression (r) 0.9988 ± 0.0008 0.9973 ± 0.0019 Table 4 Precision and accuracy of the method Found/mg l21* Intra-day Inter-day Added/mg l21 (n = 6) (n = 6) 0 8 ± 0.8 8 ± 0.7 (10.0%) (8.8%) 10 (low concentration) 18 ± 1.7 18 ± 1.8 (9.4%) (10.0%) 50 (medium concentration) 58 ± 2.4 58 ± 2.9 (4.1%) (5.0%) 150 (high concnetration) 157 ± 4.1 158 ± 4.6 (2.6%) (2.9%) * RSD in parentheses.Table 5 Extraction of AsIII, AsV, DMA and MMA in the presence of arsenobetaine and analysis by AAS As concentration in urine/mg l21 AsIII + AsV + As extracted/ DMA + MMA Asbet Total mg l21 15 21 36 14.4 25 42 67 25.6 50 42 92 50.3 120 210 330 119.0 125 105 230 124.5 200 42 242 200.4 1714 Analyst, August 1998, Vol. 1237 Takahashi, W., Pfenninger, K., and Wong, L., Arch. Environ. Health, 1983, 38, 209. 8 Hopenhayn, R. C., Biggs, M. L., Fuchs, A., Bergolio, R., Telo, E. E., and Smith, A. H., Epidemiology, 1996, 7, 117. 9 Morton, W., Starr, G., Pohl, J., and Stoner, S., Cancer, 1976, 37, 2523. 10 Farmer, J. G., and Johnson, L. R., Br. Med. J., 1990, 47, 342. 11 Buratti, M., Calzaferri, G., Caravelli, G., Colombi, A., Maroni, M., and Foa, V., Int.J. Environ. Anal. Chem., 1984, 17, 25. 12 Yamato, N., Bull. Environ. Contamin. Toxicol., 1988, 40, 633. 13 Arbouine, M. W., and Wilson, H. K., J. Trace Elem. Electrolytes Health Dis., 1992, 6, 153. 14 Food Contamination from Environmental Sources, ed. Nriagu, J. O., and Simmons M. S., Wiley, New York, 1990, pp. 121–139. 15 Goessler, W., Maher, W., Irgolic, K. J., Kuehnelt, D., Schlagenhaufen, C., and Kaise, T., Fresenius’ J. Anal. Chem., 1997, 359, 434. 16 Beauchemin, D., Bednas, M. E., Berman, S. S., McLaren, M. W., Siu, K. W. M., and Sturgeon R. E., Anal. Chem., 1988, 60, 2209. 17 Mingsheng, M., and Le, X. C., Clin. Chem., 1998, 44, 539. 18 Heitkemper, D., Creed, J., Caruso, J., and Fricke, F. L., J. Anal. At. Spectrom., 1989, 4, 279. 19 Shibata, Y., and Morita, M., Anal. Sci., 1989, 5, 107. 20 Bavazzano, P., Perico, A., Rosendahl, K., and Apostoli, P., J. Anal. At. Spectrom., 1996, 11, 521. 21 Morita, M., Uehiro, T., and Fuwa, K., Anal. Chem., 1981, 53, 1806. 22 Rubio, R., Padro, A., Alberti, J., and Rauret, G., Microchim. Acta, 1992, 109, 39. 23 Rubio, R., Padro, A., and Rauret, G., Fresenius’ J. Anal. Chem., 1995, 351, 331. 24 Zhang, X., Cornelis, R., De Kimpe, J., and Mees, L., Anal. Chim. Acta, 1996, 319, 177. 25 Lebouil, A., Notelet, S., Calleux, A., Tutcant, A., and Allain, P., Toxicorama, 1997, 9, 171. 26 Charlot, G., Les Réactions Chimiques en Solution Aqueuse et Caractérisation des Ions, Masson, Paris, 7th edn., 1983, pp. 272– 276. 27 Pascal, P., Nouveau Traité de Chimie Minérale, Masson, Paris, 1958, vol. 9, pp. 474–494. 28 Shah, V. P., Midha, K. K., Dighe, S., McGilveray, I. J., Skelly J. P., Yacobi, A., Layloff, T., Viswamanda, C. T., Cook, C. E., McDowall, R. D., Pittman, K. A., and Spector, S., J. Pharm. Sci., 1992, 81, 309. Paper 8/03842B Received May 21, 1998 Accepted June 22, 1998 Analyst, August 1998, Vol. 123 1715

ISSN:0003-2654    

DOI:10.1039/a803842b

出版商:RSC    

年代:1998

数据来源: RSC