Simultaneous Determination of Cobalt and Manganese in Urine by Electrothermal Atomic Absorption Spectrometry. Method Development Using a Simplex Optimization Approach I Journal of I Analytical 1 Atomic 1 Spectrometry 1 BENT SCHACK IVERSEN ANTONIA PANAYI JUAN PABLO CAMBLOR AND ENRICO SABBIONI Commission of the European Communities Joint Research Centre Via E. Fermi 21 020 Ispra (VA) Italy A rapid method for the simultaneous determination of Co and Mn in urine was developed using a multielemental Perkin- Elmer SIMAA 6000 graphite furnace atomic absorption spectrophotometer. A simplex optimization was used to establish the pyrolysis and atomization temperature and the ramp and hold time for the pyrolysis step. The sample preparation was a simple 1 + 1 dilution using a diluent of 2% v/v HN03 and 0.05% m/v Triton X-100 in MilliQ-water.The LODs were 0.18 pg 1-' for Co and 0.09 pg 1-l for Mn. The recovery studies to check for bias have shown acceptable accuracy for the procedure (104% for Co and 100% for Mn with 15 and 13% RSD respectively; 2 pg 1-' Co added to 10 different samples at 0.46-1.91 pg l-l 2 pg 1-' Mn added to 10 different samples at 0.11-1.09 pg 1-l). Keywords Simplex optimization; simultaneous multielement analysis; cobalt; manganese; urine; electrothermal atomic absorption spectrometry Cobalt is an essential element for mammals being an integral component of vitamin BIZ but it is also recognized as a toxic metal in workplace exposure. In particular occupational expo- sure to cobalt in the metals industry diamond polishing and the porcelain chemical and pharmaceutical industries is a potential health risk to humans causing or exacerbating diseases such as lung asthma and fibrosis which are probably all immunotoxic reactions.',2 Cobalt dust has also been impli- cated as an etiological agent in contact dermatitis in metal workers.Consideration of these sensitization reactions leads to the question of exposure to low doses of Co as a possible environmental problem. Manganese is also recognized as an essential and neurotoxic element with known exposures from Fe alloys dry cells oxidizers and organomanganese com- p o u n d ~ . ~ Abnormally high amounts of Mn in the brain of Parkinsonian subjects suggests a possible connection between the element and this di~ease.~ Biomonitoring of Co and Mn is therefore strongly recommended to minimize health risks arising from environmental and occupational exposure to these elements.As part of the EURO-TERVIHT project (Trace Element Reference Values in Human Tissues) which aims to establish and compare trace metal reference values from inhabi- tants of the European Union,' background levels of trace elements including Co and Mn in biological fluids and tissues from non-exposed individuals are being determined to establish baseline values for biomonitoring strategies. ETAAS has for many years been the method of choice for biomonitoring of trace elements. It is a reliable and easy to handle method with low costs per sample. A significant draw- back is the single-elemental character which can be time consuming if more than one element has to be measured.With the introduction of new multi-elemental instrumentation the analytical potential of ETAAS has increased. The aim of this paper is to improve the simultaneous determination of Co and Mn in urine. In order to optimize the ETAAS analytical system the modified simplex method as described by Aberg and Gustavsson6 was used. Simplex methods have previously been used to optimize analytical parameters for chromatographic procedures. Optimization is fast and reliable if there is an existing knowledge of the significant factors of the chemical measurement process. The simplex optimization is characterized by changing one or more factors for a new experiment as a result of the response of the previous experiments. This is different from the classical and more time consuming experimental approach where one factor is changed at a time while holding the other factors constant.MATERIALS AND METHODS Equipment An AA spectrometer SIMAA 6000 from Perkin-Elmer Norwalk USA was used for simultaneous multielemental analysis. Zeeman-effect background correction employed a built-in 0.8 T magnetic field oriented longitudinally to the optical path. Pyrolytically coated graphite furnaces with plat- form were used. The samples were injected from an AS70 autosampler with an 80-position tray and with a micro- dispenser selectable in increments of 0.1 pl. The wavelengths were Co 242.5 nm at 30 mA; Mn 279.5 nm at 20 mA. The injected volume was 20 p l ( l 0 p1 of urine and 10 pl of diluent). The procedure was controlled by the AA Winlab software version 1.1 (Perkin-Elmer Norwalk CT USA).The lamps used were hollow cathode lamps (HCL) from Perkin-Elmer. Reagents The Co and Mn 1.000 g I-' reference solutions were certified AA standards from Fisher Scientific Company New Jersey USA. Throughout the procedure Suprapure double sub-boiling distilled HNO (Romil Loughborough UK) was used. The Mg(N0,)2 -6H,O chemical modifier and Triton X-100 were from Merck Darmstadt Germany. The Pd chemical modifier was an Atomic Absorption Standard solution from Aldrich Milwaukee WI USA. All dilutions were made using Milli-Q ultrapure water (Milli-Pore Molsheim France). Pipette tips (Eppendorf Hamburg Germany) and sampler cups ( Perkin-Elmer) were made of polypropylene. Procedure Samples stored at - 20 "C in screw-capped polypropylene tubes then being allowed to defrost and mixed well were diluted 1+1 with a diluent of 2% HNO and 0.05% Triton X-100 in MilliQ-water; 20 pl of the solution were injected into the furnace.From the reference solutions containing 1 g 1-' of Co and 1 g 1-1 of Mn 100 pl volumes were pipetted into a calibrated flask and Milli-Q water was added to 50ml. From this stock solution (2 mg 1-' Co; 2 mg 1-l Mn) 100 pl were pipetted into a calibrated flask and Milli-Q water was added to 50ml. The resulting solution was used to spike a urine sample to give matrix matched calibration standards with the concentrations of Co and Mn shown in Table 1. Journal of Analytical Atomic Spectrometry August 1996 Vol. 11 (591 -594) 591Table 1 Calibration scheme for Co and Mn in urine t B Calibration point Addition of Co/pg I-’ Addition of Mn/pg 1-1 1 0 0 2 1.20 1.20 3 2.40 2.40 4 3.60 3.60 Simplex Optimization System A simplex is a geometric model with n+ 1 vertices in a space with n dimensions.The n dimensions correspond to the n factors that need to be optimized. The first step is to construct the initial simplex choosing n + 1 combinations of the n factors. For the ETAAS system used four recognized important factors were optimized by the procedure pyrolysis temperature; ramp time to reach this temperature; hold time at this temperature; and atomization temperature. All other steps in the furnace programme were chosen in agreement with existing procedures employed in this laboratory for ETAAS determinations of urine samples.The constraints or the boundaries for each factor and for the response of the system were decided a priori. To check if an acceptable optimum for the system had been reached two criteria were used firstly as a convergence criterion the standard deviation of the responses of all vertices of the simplex should be less than a pre-set value; secondly characteristic masses for Co and Mn should be acceptably low indicating a sufficiently sensitive optimum for the system. To establish the initial simplex five vertices were calculated according to Table2;7 vertex no. 1 has coordinates of zero and indicates the starting levels chosen by the analyst. The vertices 2-5 are then constructed by multiplying the chosen step size for each factor by the values in the table and the results are added to the value of vertex 1.The vertices are illustrated for a two-dimensional system in Fig. 1 with the initial simplex (W-NW-B) in solid lines and the possible new simplexes in dashed lines. The new coordinates are calculated by using the following equations n+l C K - V w (1) T/ - i = l C - n where i is the number of vertices n is the number of factors and V 6 Vw and V are the coordinates for the vertex of the centroid vertex number worst vertex and new vertex respect- ively. V is calculated and then substituted into eqn. (2). New vertices V = AVc + Vw (2) The value of A is 2 for the normal reflection 3 for the expanded reflection 1.5 for positive contracted and 0.5 for negative contracted vertices. The responses from the ETAAS system are examined and the best vertex (B) the worst (W) and the next-to-worst (NW) are established.The worst vertex W is reflected through the centroid to find R (reflected vertex) and further experiments N W b Factor 1 Fig. 1 Possible new vertices for a two factor system using the modified simplex method. Original simplex in solid lines. C = Centroid B =best vertex W = worst vertex NW = next to worst vertex R = reflected vertex E =expanded vertex PC =positive contracted vertex and NC =negative contracted vertex are undertaken to evaluate the response from this new vertex illustrated in Fig. 2. RESULTS AND DISCUSSION The initial simplex was constructed using the a priori con- sidered best levels and combination of the factors as vertex 1 and vertices 2-5 were calculated according to Table 2.The step sizes were chosen in order to cause a new response sufficiently different from the previous also taking into account experimental error. The ramp and the hold time were approxi- mated to the nearest integer and the temperature settings were approximated to the nearest 10 degrees the smallest increment allowed by the instrument. Table 3 shows the constraints and step sizes and Table 4 the vertices of the initial simplex. In order to obtain a large signal for the optimization procedure the urine was spiked with 4 pg I-’ Co and 4 pg 1-l Mn and measured in duplicate for each combination of factors (vertex). In order to take into account the different sensitivity of the ETAAS system for Co and Mn the over-all response evaluated was a modified summation of the absorbances from each element Corrected sum = 2.7 x integrated absorbance for Co +integrated absorbance for Mn The factor 2.7 is the ratio of recommended characteristic mass for Co to the recommended characteristic mass for Mn under single element conditions with the characteristic mass defined as the mass of analyte (normally expressed in pg) which causes an integrated absorbance of 0.0044.When evaluating the differences between the responses from the five different vertices of the simplex decisions to accept or Table 3 Constraints and step sizes for the system to be optimized Factor Constraints Step sizes Pyrolysis temperature/”C 1100-1600 200 Ramp time/s 1-30 20 Hold time/s 1-40 20 Atomization temperature/”C 1700-2600 200 Table 2 Values used to establish the initial simplex Table 4 Vertices for the initial simplex Factors Factors Vertex Pyrolysis Atomization no.temperature Ramp time Hold time temperature 1 0 0 0 0 2 1 .000 0 0 0 3 0.500 0.866 0 0 4 0.500 0.289 0.817 0 5 0.500 0.289 0.204 0.791 Vertex Pyrolysis Atomization no. temperature Ramp time Hold time temperature 1 1300 5 10 2400 2 1500 5 10 2400 3 1 400 22 10 2400 4 1400 11 26 2400 5 1400 11 14 2560 592 Journal of Analvtical Atomic Svectrometrv. AuPust 1996 voz. 11B,NW,Win I Fig. 2 Flow chart of the decision process of the simplex optimization method ‘>’ indicates better than Y is the direction to take if ‘Yes’ is the answer to the comparison and N if the answer is ‘No’. XZY indicates that X substitutes Y (1) the control is not passed and a new optimization experiment is carried out.(2) the control procedures are fulfilled and the optimization is finished. Abbreviations as in Fig. 1 reject the responses were based on the differences of the responses without testing whether they were statistically sig- nificant. As a final criterion for the evaluation of the responses the shapes of the absorbance curves had to be acceptable from experimental experience and the RSD less than 10% for each element. If not the vertex was considered to be the worst. A problem in optimizing Co and Mn together is a difference of 500°C in atomization temperatures according to the rec- ommended settings for the instrument by the manufacturer. Because of this problem the criteria for accepting the optimiz- ation must be chosen carefully.The two criteria for the control procedure for an eventual acceptable optimum were the RSD of the responses from the five vertices of the simplex should be lower than 5% and the calculated characteristic masses for system settings according to the best vertex should not exceed the recommended charac- teristic masses for single element conditions by more than 20%. The characteristic masses were established by relating the differences in integrated absorbance with the differences in mass of the elements from the same urine sample spiked with 2 ppb Co and 2 ppb Mn and with 4 ppb of each element. An optimum which was acceptable according to the two control criteria was reached as vertex number 17 see Table 5.In total 13 experiments were carried out and four vertices were ignored because one or more factors were outside their constraints. The two criteria for an acceptable optimum were fulfilled because the RSD for the corrected sums of the final five vertices was 3.0% and the characteristic masses calculated for the optimum were for Co 20.9 pg; and for Mn 4.1 pg. For Mn the result is better than the value published by the manufacturer and for Co the value published by the manu- facturer is less than 20% better than the value determined for this system. It is important to stress that the responses should be independent of time. In an ETAAS optimization procedure the graphite tube will change during use therefore the firings should be limited in number so that the tube can be considered unchanged throughout the optimization.The optimization was not improved by adding 5pg of Pd-15 pg of Mg(NO,) as chemical modifier to each sample as addition the modifier increased the risk of sample contami- nation. A control of 20 calibration curves for Co and Mn was made by the multivariate procedure suggested by Mestek et a1.* The Mahalanobis distances are calculated according to the Journal procedure and drawn in the control charts as shown in Fig. 3. The control limits in the charts correspond for ICL (Inner Control Limit) to the distance from the base line to T0.95452 from the Hoteling’s T2 distribution and for OCL (Outer Control Limit) to the distance G,99732. The limits are compar- able to the well known limits +2s and +3s from Shewart control charts.Fig. 3 demonstrates that all 20 calibration curves for Co and Mn show good stability without disturbances. Recovery Study Ten different urine samples with and without addition of analyte were measured with 2 repetitions. As shown in Table 6 the results were satisfactory. The RSDs were quite high but taking the low concentrations into consideration this is acceptable. Certified Reference Materials The analytical usefulness of the proposed method could not be assessed by analysis of standard reference materials because they do not exist for Co and Mn in urine. Instead the Seronorm Trace Elements in Urine with a recommended content of 10.2 pg 1-’ (173 nmol 1-’) of Co was used. It was 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Calibration curves Fig.3 in urine Control charts of 20 calibration curves for (a) Co and (b) M n of Analytical Atomic Spectrometry August 1996 Vol. 1 1 593Table 5 The vertices and their responses (corrmm) and the action to be taken for the optimization procedure Vertex 1 2 3 4 5 6 77 8 9 107 11 1271 13 14 157 16 177 I' Pyrolysis 1300 1500 1400 1400 1400 1550 1360 1430 1420 1290 1190 1230 1090 1240 1360 1220 1270 /"C Atomiza- Ramp/s Hold/s tion/"C 5 10 2400 5 10 2400 22 10 2400 11 26 2400 11 14 2560 20 20 2480 9 13 2420 12 5 2470 22 11 2530 31 12 2590 27 14 2560 38 12 2630 29 13 2400 16 13 2520 15 16 2620 17 14 2560 13 -3 2490 Corr.* 0.05 18 0.0576 0.0379 0.0544 0.0127 0.0623 3 0.0470 0.0637 9 0.0644 6 5 6 $ 0.0661 0.0673 Action? R NC R PC R E R E R NC R PC * Corrected sum (see text).t Actions taken for the new vertex R = reflected E = expanded PC = positive contracted NC = negative con- tracted. $ Absorbance peaks not accepted. $ Factors outside the con- straints. 1 Last five vertices. 11 Optimum. Table 6 Recovery* of 2.0 pg 1-' Co and 2.0 pg 1-' Mn added to urine samples ~ ~~~ ~ Range/g 1-' Recovery RSD (%) (%) Wth addition Wthout addition co 0.46-1.91 2.60-3.9 5 104 15 Mn 0.11-1.09 1.76-2.28 100 13 * n= 10. diluted 4 times (2.55 pg 1-l) to be within the calibration range and the value found was 2.72 pg 1-1 (10.88 pg 1-l for the undiluted sample) with an s of 0.16 pg 1-l (n= lo) so the accuracy is acceptable. Comparison With Other Analytical Techniques Radiochemical neutron activation analysis RNAA is recog- nized as an outstanding analytical technique for many elements at trace levels because it has multielement capability good selectivity and accuracy with low LODs. However due primar- ily to the difficulty of obtaining access to irradiation and special radiochemical facilities RNAA as a routine method has lost ground to other powerful technique^.^.'^ Its role as a reference technique is still very important in the analysis of biological samples particularly in developing spectrochemical techniques such as ETAAS and ICP-MS.The method for Mn was compared with an independent analytical technique. Twelve spot urine samples were collected from subjects living in Ispra (Varese Italy) and submitted to RNAA which included the isolation of induced 56Mn and its counting by computer-based high resolution y-ray spec- trometry.A linear regression procedure with no special assumptions regarding the distribution of the samples and measurement errors was used for the comparison of the two methods." The statistical procedure of Passing and Bablok'l gave the following regression line for the relationship between RNAA ( X ) and ETAAS (Y) Y = -0.24+1.40X with 95% confidence interval [ - 0.80 0.221 for the intercept and [0.91:2.00] for the slope. According to the theory of the procedure the hypothesis of an intercept of 0 is accepted because the confidence interval contains 0 and the hypothesis of a slope of 1 is also accepted because the confidence interval contains 1. Table 7 The furnace programme for the Co-Mn method Temperature/ Vertex "C 1 110 2 130 3 800 4 1270 5 2560 6 2600 Gas flow/ Ramp/s Hold/s ml min-' Read 1 20 250 5 30 250 5 25 250 17 14 250 1 5 0 X 1 5 250 Analytical Performance The calibration curves showed linearity up to 6 pg 1-l for both elements.No attempt was made to check if the analytical ranges could be extended. The LOD (3s n=20) was 0.18 pg 1-l for Co and 0.09 pg l-' for Mn. Owing to its significant influence on the very low levels of Co and Mn great care must be taken in establishing the blank values for autozero. Conditions Selected for the Final Optimized Method The optimized temperature programme is shown in Table 7. CONCLUSION The SIMAA 6000 instrument can improve the determination of Co and Mn in urine at the ultra-trace level. The advantages of this instrument such as simultaneous multielement determi- nation and fast and reliable automatic handling of the samples are fully used in the Co-Mn method.The practical LOD are sufficiently low (Co 0.18 pg 1-I; Mn 0.09 pg 1-') for the determination of the metals from non-occupationally exposed subjects. The procedure is fast allowing more than 100 determi- nations per day and requires only a minimum of sample preparation thereby reducing the risk of contamination. In any case the use of ultra-pure acids and ultra-clean room conditions is still essential for accurate determination of Co and Mn at these levels. The optimization described employed four important factors pyrolysis temperature ramp time hold time and atomization temperature. However it cannot be excluded that further optimization including e.g.the sample volume injected or the concentration of HN03 or Triton X-100 could lead to further improvement. The study also confirms that the simplex method could be used successfully in atomic absorption optimization. REFERENCES 1 2 3 4 5 6 7 8 9 10 11 Lauwerys R. and Lison D. Sci. Tot. Environ. 1994 150 1. Christensen J. M. Sci. Tot. Environ. 1995 166 89. Sekt P. K. and Chandra S . V. in Metal Neurotoxicity eds. Bondy S. C. and Prassad K. N. CRC Press Inc. Boca Raton Florida 1988 pp. 19-33. Mena I. in Disorders of Mineral Metabolism eds. Bronner F. and Coburn J. W. Academic Press New York 1981 vol. I Sabbioni E. Minoia C. Pietra R. Fortaner S. Gallorini M. 4nd Saltelli A. Sci. Tot. Environ. 1992 120 39. Aberg E. R. and Gustavsson A. G. T. Anal. Chim. Acta 1982 144 39. Long D. E. Anal. Chim. Acta 1969 46 193. Mestek O. Pavlik J. and Suchanek M. Fresenius J. Anal. Chem. 1994 350 344. Pietra R. Sabbioni E. Gallorini M. and Orvini E. J. Radioanal. Nucl. Chem. Articles 1986 102 69. De Goeij J. J. M. and Woittiez J. R. W. J. Radioanal. Nucl. Chem. Articles 1993 168 429. Passing H. and Bablok W. J. Clin. Chem. Clin. Biochern. 1983 21 709. Paper 610095 7C Received February 9 1996 Accepted June 4 1996 pp. 233-270. 594 Journal of Analytical Atomic Spectrometry August 1996 Vol. 11