ISSN:0003-2654    

DOI:10.1039/AN99217FX023

出版商:RSC    

年代:1992

数据来源: RSC

ISSN:0003-2654    

DOI:10.1039/AN99217BX025

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 933 Relationship Between Content Limits and Assay Methods: An lnterlaboratory Statistical Evaluation Jacques 0. De Beer Institute for Hygiene and Epidemiolog y, Section of Drug Analysis and Pharmacopoeia, J. Wytsmanstraat 14, B- 1050 Brussels, Belgium Bart M. J. De Spiegeleer" S. C. Federa C. V., B- I030 Brussels, Belgium Jos Hoogmartens and Isabelle Samson Ka th o lie ke U n ive rsiteit Leu ven, Institute for Pha rmaceu tical Sciences, Laboratory for Pharmaceutical Chemistry, Van Evenstraat 4, B-3000 Leuven, Belgium Desire L. Massart and Martine Moors Vroe Universiteit Brussel, Pharmaceutical Institute, Laboratory for Pharmaceutical and Biomedical Analysis, Laarbeeklaan 103, B- 1090 Brussels, Belgium It has been statistically demonstrated, by means of interlaboratory acquired experimental results, that the prescribed content limits in a pharmacopoeial monograph should be critically reconsidered if one volumetric assay method is systematically substituted by another.Three different volumetric methods, prescribed as assays in phenothiazine monographs and ensuring the same content limit intervals, reveal different repeatabilities or reproducibilities. Analysis of variance and box-plot presentations confirm that for some of the three methods examined, interlaboratory variations contribute significantly t o the total variances. For one particular method, however, this contribution is obviously less pronounced. Important contributions of method-to-method variations t o the total variance are also established for three of the four participating laboratories.Keywords: Statistics; content limits; volumetric assay methods; analysis of variance; precision For an increasing number of volumetric assays of hydro- chloride salts of drugs described in the European Pharmaco- poeia (Ph. Eur. 11), the generally used non-aqueous titration with perchloric acid in anhydrous acetic acid medium in the presence of mercury(I1) acetate is substituted by non-aqueous titration in acetic anhydride or by titration in ethanol with 0.1 mol dm-3 sodium hydroxide. Phenothiazines such as chlorpromazine hydrochloride (Ph. Eur. 475-1986), levomepromazine hydrochloride (Ph. Eur. 505-1986), promethazine hydrochloride (Ph. Eur. 524-1989), thioridazine hydrochloride (Ph. Eur. 586-1988) and trifluoper- azine hydrochloride (Ph.Eur. 59-1980) are involved in this adaptation. As a consequence, phenothiazine assays are performed by different volumetric methods. This means that there should be no differences in the systematic (method) and random errors between the different volumetric methods applied, if identical (narrow) content limits are prescribed. The aim of this work was to examine, by means of internationally recognized statistical tests, whether substitu- tion of one volumetric method by another in a pharmacopoeial assay is acceptable without critically reviewing the fixed content limits. Experimental Working Procedures Four laboratories (A, B, C and D) were asked to participate and three different chlorpromazine hydrochloride batches were selected.Each participating laboratory followed a previously approved working scheme with instructions on the method sequences, to ensure final comparable results. * Present address: Alcon Couvreur, Rijksweg 8. B-2870 Puurs, Belgium. Tables were made available to allow detailed reporting of all the experimental data involved in each titration such as masses and volumes. Volumetric Methods Three European Pharmacopoeia1 volumetric assay methods, currently, or previously, used for the determination of phenothiazine and assuming the same content limits (99.0- 101 .O%), were applied in each participating laboratory to each of the three chlorpromazine hydrochloride batches. Method A This assay method is described in the chlorpromazine hydrochloride monograph of the Ph. Eur. I.* Instead of 0.400 g, 0.250 g was analysed to match the weighed amount of substance in the assay method of the actual monograph of chlorpromazine hydrochloride in the Ph.Eur. 11. The substance was dissolved in 100 cm3 of acetone and titrated with 0.1 rnol dm-3 perchloric acid. Method B The second assay method examined is that described in the trifluoperazine hydrochloride monograph of the Ph. Eur. 11.2 Instead of 0.300 g, 0.250 g was analysed. The substance was dissolved in 50 cm3 of anhydrous acetic acid and 10 cm3 of a mercury(i1) acetate solution were added. The titration was performed with 0.1 rnol dm-3 perchloric acid. Method C The third assay method examined is that described in the chlorpromazine hydrochloride monograph of the Ph. Eur. 11.3 The substance was dissolved in a mixture of 5.0 cm3 of 0.01 mol dm-3 hydrochloric acid and 50 cm3 of ethanol and titrated with 0.1 mol dm-3 sodium hydroxide.In this method the934 ANALYST, JUNE 1992, VOL. 117 Table 1 Summary of analytical equipment Laboratory A Titrator Metrohm E535 Dosimat , E536 potentiograph Burette volume/cm3 20.00 Burette precision/cm3 0.01 Glass electrode Combined Metrohm 6.0202.100(KF) Reference electrode - Balance * Used for titration in anhydrous medium. Sartorius A200S 0.1 mg B Burette 25 .O 0.1 Corn bined Corning 476530 - Mettler AT200 0.1 mg C Metrohm 672 Ti troprocessor 20.00 0.01 Metrohm 6.0102.100 (ID) 6.0104.100 (KB)* Calomel 6.0702.100 Sartorius Type 1712 0.1/0.01 mg D Burette 50.0 0.05 Com bined Orion - Sartorius A 200s 0.1 mg Table 2 Values and statistical parameters of the different volumetric solutions used for the chlorpromazine hydrochloride assays in each participating laboratory Laboratory A B C Titration of anhydrous Na2C03 with 0.1 rnol dm-3 HCI- X/mol dm-3 (n = 10) 0.1008 0.1004 0.0984 mol dm-3 3.38 3.31 3.41 s2/10-7 1.14 1.1 1.16 s/x x loo(%) 0.335 0.33 0.346 95% CI 0.1005-0.1010 0.1001-0.1006 0.0982-0.0987 X/mol dm-3 s / ~ O - ~ rnol dm-3 3.33 0.872 2.52 s2/10-8 1.11 0.760 6.36 s/x x loo(%) 0.336 0.086 0.25 95% C I 0.0988-0.0993 0.1005-0.1006 0.0998-0.1001 Titration of potassium hydrogen phthalate with 0.1 rnol dm-3 HClO4- X/mol dm-3 s / ~ O - ~ rnol dm-3 1.12 2.25 7.81 s2/10-8 1.25 5.06 61.2 s/x x loo(%) 0.114 0.22 0.78 95% CI 0.0981-0.0983 0.1008-0.1011 0.0995-0.1006 Titration of0.l rnol dm-3 HC1 with 0.1 mol dm-3 NaOH- (n = 10) 0.0991 0.1001 0.0999 (n = 10) 0.0982 0.1009 0.1Ooo D 0.1007 2.71 0.734 0.27 0.10054 1009 0.0993 2.17 4.72 0.21 0.0992-0.0995 0.1005 3.01 9.07 0.29 0.1003-0.1OO7 volume added between the two potentiometric inflection points was read. Potentiometric end-point detection (Ph.Eur. V.6.14.) was required to be performed in each method. By convention, the assay results were expressed with reference to the substance as such. Apparatus A description of the titration equipment and its characteris- tics, for each participating laboratory, is given in Table 1. Samples The three different sample batches were divided in the same laboratory (Institute for Hygiene and Epidemiology) and distributed to the other participating laboratories in well- closed and sealed containers.These samples were analysed and found to comply with the Ph. Eur. requirements for identity and purity. The loss on drying was less than 0.5%. Reagents The reagents and reference substances, used in the different volumetric methods, complied with the purity requirements of the Ph. Eur. (sections VII. 1. l., VII.2.1. and VII.2.2.). The exact concentrations of the volumetric solutions were establi- shed as described in section VII.2.2. Statistical Methods Each prescribed experiment was required to be repeated ten times in order to estimate statistical parameters, e.g., the mean value X, the variance s*, the standard deviation s and the 95% confidence interval (CI) for the mean, p, according to a preliminary draft of the European Pharmacopoeia1 Guide for the Technical Content of Monographs.4 The 95% CI for p was calculated with the equation5 t0.975 with nine degrees of freedom (df) is 2.262.The repeatability and reproducibility for the assay methods compared were calculated from the interlaboratory results obtained following the IS0 5725-1986 document .6ANALYST, JUNE 1992, VOL. 117 935 The same results were used for computing the analysis of variance (ANOVA) to estimate differences firstly between the laboratories for each method and secondly between the assay methods in each laboratory. The ANOVA computa- tions were performed with the Statgraphics version 5.0 software,-/ which can display the ANOVA table with plots of, e.g., the sample means with the 95% CI for the variables, interaction plots of the variables and multiple box-and- whisker plots, among others.Box-and-whisker plots were drawn to provide complemen- tary visual information on data groupings. Results and Discussion The mean values, standard deviations, variances, relative standard deviations and the 95% CI of the mean, obtained in each laboratory for the concentrations of the volumetric solution involved, are summarized in Table 2. In laboratory B one outlier with Dixon's test638 was detected in the titration of anhydrous sodium carbonate with hydro- chloric acid. Maintaining this outlier, a mean value of 0.1008 mol dm-3, a standard deviation of 1.39 x 10-3 mol dm-3, a relative standard deviation of 1.37% and a 95% CI for the mean between 0.0998 and 0.1018 mol dm-3 were calculated.The results from this laboratory, given in Table 2, are those after rejection of the outlier. As far as these results are concerned, it is interesting to compare the volumetric pre- cision obtained in each of the four participating laboratories. From the data given in Table 2 it is clear that in each participating laboratory sufficient precision can be achieved for the precise assay of chlorpromazine hydrochloride. By means of Bartlett's test for the comparison of more than two variances,g.'" it can be shown that in the four laboratories the recovered precision for the titration of hydrochloric acid is not statistically different. For the titration of sodium hydrox- ide or perchloric acid, this is true for three of the four laboratories. Repeatability and Reproducibility of the Assay Methods The whole interlaboratory experiment was organized so that the reported assay results were suitable for statistical treat- ment as described in the I S 0 5725-1986 document for the computation of the repeatability value r and the reproducibil- ity value R for a standard test method by interlaboratory tests.6 The repeatability (reproducibility) value r ( R ) is the value below which the absolute difference between two single test results, obtained under repeatability (reproducibility) condi- tions, may be expected to lie with a probability of 95%.Repeatability conditions are those where mutually indepen- dent test results are obtained with the same method on identical test material in the same laboratory by the same operator using the same equipment within short intervals of time.Reproducibility conditions are those where these test results are obtained in different laboratories with different operators using different equipment. Further reference was made to the IS0 5725-1986 document with respect to the field of application, the statistical model, the design of the precision experiment and the analysis of the data. In the ideal case, the results of an experiment with p laboratories and q levels will give rise to a table with pq cells each containing n replicate results, that can all be used for computing r and R . The computational procedure for r and R used here is that for a uniform-level experiment with a constant n >2 (=lo) replicates per cell. Each combination of a participating laboratory and a level (which here is constant) is referred to as a cell.The assay results with their basic statistical parameters for each sample, in each laboratory, are summarized in Tables 3-5 including the r and R values. According to Dixon's test there were no outliers. The computational equations for r and R are: number of laboratories, p = 4; and number of replicates, n = 10 TI = CXj, T2 = 2Xi2, T3 = XS? s,2 = T3Ip = repeatability variance p x T2- Ti2 ~ , 2 SL2 = _ - - - between-laboratory variance P X b - 1 ) n sR2 = sL2 + s,2 = reproducibility variance Mean m = T,/p Repeatability r = 2 . 8 c 2 Reproducibility R = 2 . 8 m Table 3 Chlorpromazine hydrochloride assay results and their statistical parameters obtained with volumetric method A Laboratory Sample 1- x(%)(n=lO) s (Yo) S2 95% CI r = 2.90 R = 4.26 Sample 2- x (%) ( n = 10) s (Yo) S2 95% CI r = 2.80 R=4.53 Sample 3- X(%) ( n = 10) s (Yo) S2 95% CI r = 2.51 R = 4.21 A 99.31 0.29 0.082 99.10-99.5 1 99.41 0.61 0.369 98.98-99.85 98.76 0.49 0.240 98.40-99.1 1 B 99.70 1.30 1.700 98.77-100.63 99.59 0.59 0.352 99.16-100.01 99.62 0.91 0.824 98.97-100.27 C 101.35 1.25 1.554 100.46-102.24 101.75 1.39 1.923 100.76-102.74 101 .oo 1.12 1.260 100.20-101 .80 D 98.61 0.98 0.954 97.91-99.30 98.68 1.11 1.221 97.89-99.47 98.12 0.95 0.894 97 .&98.79936 ANALYST, JUNE 1992, VOL.117 Table 4 Chlorpromazine assay results and their statistical parameters obtained with volumetric method B Laboratory A B C D Sample 1- x (Yo) ( n = 10) s (Yo) S* 95% CI r = 1.56 R = 1.88 Sample 2- k (Yo) ( n = 10) s (Yo) S2 95% CI r = 1.87 R = 1.87 Sample 3- x (%) (n = lo) s (Yo) 52 r = 1.69 R = 1.98 95% CI 100.64 0.40 0.163 100.35-100.93 100.38 0.71 0.498 99.88-1 00.89 99.75 0.45 0.204 99.42-1 00.07 100.63 0.59 0.351 100.21-101.06 100.53 0.25 0.061 100.35-100.7 1 100.31 0.80 0.640 99.74-100.88 100.24 0.91 0.827 99.59-100.89 100.08 0.49 0.241 99.73-100.43 100.48 0.24 0.057 100.3 1-100.65 99.62 0.88 0.772 98.99-100.25 99.58 0.53 0.277 99.20-99.96 99.88 0.59 0.347 99.46-100.3 1 Table 5 Chlorpromazine hydrochloride assay results and their statistical parameters obtained with volumetric method C Laboratory A B C D Sample 1- x (Yo) (n = 10) s (Yo) S* 95% CI r = 1.92 R = 3.11 Sample 2- x(Y0) ( n = 10) S (%) S2 95% CI r = 2.91 R = 4.32 Sample 3- x ( Y O ) (n = 10) s (Yo) S* 95% CI r = 1.95 R = 3.21 100.20 0.23 0.051 100.03-100 .36 100.35 0.79 0.625 99.78-100.92 102.16 1.03 1.063 101.42-102.90 100.63 0.37 0.135 100.36-100.89 100.13 0.105 0.01 1 100.06-1 00.2 1 99.80 0.71 0.498 99.29-100.3 1 102.46 1.91 3.643 101.09-103.83 100.68 0.41 0.167 100.39-100.98 100.65 0.31 0.094 100.43-1 00.87 99.40 1.14 1.310 98.58-100.22 101.68 0.70 0.486 101.1&102.18 100.80 0.25 0.062 100.62-100.98 The I S 0 recommendations assume that for a properly standardized test method, variance differences between labor- atories should be small and that it is justifiable to establish a common value of within-laboratory variance for all the participating laboratories. This common value, which is the average of all the within-laboratory variances taken over all the laboratories taking part in the precision experiment, is called the repeatability variance. The within-laboratory variances are often not always the same in real-world situations, owing to differences such as equipment and skills of operators. This problem also arises with these interlaboratory experimental results.However, any laboratory can, by carrying out a series of tests under repeatability conditions, arrive at an estimate of its own particular repeatability for the test method, and check it against the common standard value. In this context, calcula- tion of a common repeatability can be considered as meaning- ful, even with within-laboratory variance differences. Once determined, the values of r and R can also serve to compare methods, e.g., to verify if a new experimental technique is comparable to a standard technique, or to assess the suitability of rival test methods.Here, the conclusions obtained from the repeatability and reproducibility calcula- tions, even with unequal within-laboratory variances, can be confirmed by other statistical methods, which prove to be much more rugged versus within-laboratory variance differ- ences. The results presented in Tables 3-5 generally show that within each applied assay method, almost analogous repeat- abilities r and reproducibilities R are achieved for each sample batch. For method A, the repeatability and reproducibility are rather poor (r >2%; R >4%). In contrast for method B both parameters r (1.56-1.87%) and R (1.87-1.98%) are not only <2%, but are also concordant.ANALYST, JUNE 1992, VOL.117 937 Table 6 Summary of variances (s2) on the assay measurements for each sample in the four participating laboratories Laboratory Method A- Sample 1 Sample 2 Sample 3 Method B- Sample 1 Sample 2 Sample 3 Method C- Sample 1 Sample 2 Sample 3 A 0.08 0.37 0.24 0.16 0.06 0.06 0.05 0.01 0.09 B 1.70 0.35 0.82 0.50 0.64 0.77 0.64 0.53 1.19 C 1.55 1.92 1.26 0.20 0.83 0.28 1.06 3.64 0.49 D 0.95 1.22 0.89 0.35 0.24 0.35 0.13 0.17 0.06 Table 7 Two-way ANOVA by method, on the assay results, obtained for the three sample batches by the four participating laboratories Source of Sumof Degrees Mean variation squares of freedom square F-ratio Method A- Main effects 142.205 5 28.441 30.01 Laboratories 136 394 3 45.631 48.15 Samples 5.310 2 2.655 2.80 interactions Lab-Sample 2.103 6 0.351 0.37 2-Factor Residual 102.353 108 0.948 - Method B- Main effects 12.473 5 2.495 6.76 Laboratories 7.515 3 2.505 6.78 Samples 4.958 2 2.479 6.71 2-Factor interactions Residual 39.878 108 0.369 - Lab-Sample 4.081 6 0.680 1.84 Method C- Main effects 82.452 5 16.490 24.74 Laboratories 81.863 3 27.288 40.95 Samples 0.589 2 0.294 0.44 2-Factor interactions Residual 71.971 108 0.666 - Lab-Sample 7.467 6 1.244 1.87 Finally, method C gives an acceptable repeatability ( r = 2%) but less reproducibility ( R >3%) within the four participating laboratories.This means, however, that if a content of 100.0% is found after the first titration with this method, in 95% of cases after a second titration, a content between 98.0 and 102.0% will be found.The reproducibility interval R , obtained with two results from two different laboratories on the same sample, will be even larger. The precision of assay methods is further impaired when the results need to be calculated with reference to the dried or anhydrous substance. In this context, one can ask how to interpret the content limits, defined for a raw material in a pharmacopoeia1 monograph. It is stated in the Ph. Eur. under ‘Limit values’ (section IV.2.), that the substance under investigation only satisfies the requirements of the mono- graphs if no further deviation from the fixed limit values is established. Evaluation of Sources of Variation. Analysis of Variance (ANOVA) The object of ANOVA is to compare systematic errors with the random error obtained for the replicates, i.e., the precision of the laboratories o r of the procedures.It is clear that, in this study, the assumption of variance homogeneity between the participating laboratories is, according to Bartlett, not always fulfilled (Table 6). However, ANOVA is a very robust statistical technique, insensitive to moderate deviations of variance homogeneity.9 Box-and-whisker plots were also produced as they can be used to compare visually groups of data in order to investigate whether some groups of data are very different from others or whether the variance homogeneity is inadequate for ANOVA computations. Box-and-whisker plots divide the data into four areas of equal frequency. The box encloses the middle 50%.The median is drawn as a vertical line inside the box. The lower whisker is drawn from the first quartile to the smallest data point within 1.5 interquartile ranges from the first quartile. The other whisker is drawn from the third quartile to the largest data point within 1.5 interquartile ranges from the third quartile. Data points beyond the whiskers are plotted indivi- dually. Two-way ANOVA (Sample X Laboratory) for each Method Two-way ANOVA, for the multiple comparison of analytical results from more than two samples in more than two laboratories, using the same assay method, allows the determination of whether: there is a significant contribution to the total variance from laboratory-to-laboratory variations with the same method; and whether there is a significant contribution to the total variance caused by sample differ- ences, within the same method.The ANOVAs for each of the three methods are given in Table 7. It is clear from this table that laboratory-to-labora- tory variations contribute significantly to the total variance in all three assay methods. Method B is the most precise method in each laboratory, giving the smallest 95% CIS (F-ratio = 6.7 > F(0.05)3/114 = 2.7). With this method, assay differences between the four participating laboratories are rather small, with respect to the other methods. Sample-to-sample variations only contribute to the total variance in method B. In none of the three methods is a significant interaction between sample and laboratory found, so that interaction and residual variances may be pooled.The box-and-whisker plots in Figs. 1-3 show comparable group positions of the data to those obtained with ANOVA, and show clear differences in results from the laboratories applying the same method. The better reproducibility of method B is confirmed. Two-way ANOVA (Sample X Method) for each Laboratory Two-way ANOVA for the multiple comparison of analytical results from more than two samples from the same laboratory allows the determination of whether: there is a significant contribution to the total variance from method-to-method variations in each participating laboratory; and whether there is a significant contribution to the total variance caused by sample differences within the same laboratory. The results are reported in Table 8.These ANOVAs reveal that in three of the four participating laboratories a significant difference between the three assay methods is obtained. Only in laboratory B do method-to-method variations not contribute significantly to the total variance. This does not mean that the similar assay precisions, obtained in that particular labora- tory, justify maintainance of assay limits of 99.0-101.0%.ANALYST, JUNE 1992, VOL. 117 938 I I I I I 1 1 I I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P . . . . . . . . . . . . . . . . . . . . . . . . c! ' . . 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 I . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . +- . . . . . . . . . . . . . -0. . . . . . . . . . . . + 'i . . . . . . . . . . Q 1 . . . . . . _ : . . . . . . . . . . . . : . . . . . . . : . . . . . . . : . . . . . . . . . . . . . . . b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I I I 1 A 6 C D Laboratory I I 1 I I A B C D Laboratory Fig. 3 medians, ranges and extreme values in each laboratory Table 8 Two-way ANOVA by laboratory, on the assay results, obtained for the three sample batches by the three assay methods Multiple boxplot of the results with method C, comparing the Fig, 1 medians, ranges and extreme values in each laboratory Multiple boxplot of the results with method A , comparing the I I I I .. . . . . . . . . . . . . . . . . . . . . . Degrees Mean of freedom square . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Source of variation Laboratory A- Main effects Samples Methods 2-Factor interactions Method- Sample Residual Sum of squares 33.752 0.131 33.622 4.068 10.458 9.028 5.846 3.182 2.782 64.030 86.518 8.013 78.504 0.248 101.971 84.163 2.005 82.158 3.083 39.971 F-ratio 65.35 0.51 130.20 7.88 - 2.85 3.70 2.01 0.88 - 17.33 3.21 31.45 0.05 - 43.58 2.08 85.09 1.60 - . . . . . . . . . . . . . . . . . . . . . . . 4 8.438 2 0.065 2 16.811 . . . . . . . . . . . 4 1.017 81 0.129 Q 0 0 . . . . .+ 0 . . . . . . . . . . . . 0 Laboratory B- Main effects Samples Methods 2-Factor interactions Method- Sample Residual . . . . B . . . . . . .. . . . . . . . . . . . 4 2.257 2 2.923 2 1.591 4 0.695 81 0.790 . . . . Laboratory C- Main effects Samples Methods 2-Factor interactions Method- Sample Residual 4 21.629 2 4.007 2 39.252 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I I 1 I A B C D Laboratory 4 0.062 81 1.248 Fig. 2 medians, ranges and extreme values in each laboratory Multiple boxplot of the results with method B, comparing the Laboratory D- Main effects Samples Methods 2-Factor interactions Method- Sample Residual 21.041 4 2 1.002 2 41.079 Sample-to-sample variations do not contribute significantly to the total variance for each of the participating laboratories. The box-and-whisker plots in Figs. 4-7 visually confirm the ANOVA conclusions, that the measurements obtained with the three methods in each laboratory are different, with an exception being laboratory B .4 0.771 81 0.483ANALYST, JUNE 1992, VOL. 117 I I I ................................................................. - 939 107.0 104.8 s 102.6 I 4- - 3 2 2 100.4 > v) 98.2 96.C I I I I 1 I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................................................................ . .- P .:.. ............. :.. .............................. .n.. ............... 107.0 104.8 A 8 102.6 a 2 6 - CI - 2 100.4 98.2 96.0 I : . . . . . . . . ........... 1 ............ ........... ............ . . . . . . . . . . . . . . . . . . . . . . . . ............. 0 e ... ........ . .. . . . + I + . . . . . . . . . . . . . . . . 9.. . . . . 9 . . . . . . . 0 ........... . . . . . . . . . ................. . . . . . . . . . . . . . . . . . . . ................ .............. ............... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I I I A B C Method I I I A B C Method Fig. 4 Multiple boxplot of the results in laboratory A, comparing the medians, ranges and extreme values obtained with each assay method Fig. 6 Multiple boxplot of the results in laboratory C, comparing the medians, ranges and extreme values obtained with each assay method 107.0 104.25 I s 2 2 - c.’ - 101.5 > a 98.75 96.0 I I t ............................................................................ ...... ..... . . . . . ......... .......... (j .. . . . . . . ........... ........... ........... ........ .............. ........ .............. .... ..... .... ........ .... .... . . . 0 ......... + ... ..Q.. .. P .... .+. ... ............... ................... + ........................ ..................... ~~ ............. + ............ 0 .:.. .... . - . ... :. .......................... ... ............ .......................................................... I I I I A B C Method Fig. 7 Multiple boxplot of the results in laboratory D, comparing the medians, ranges and extreme values obtained with each assay method I I I A B C Fig. 5 Multiple boxplot of the results in laboratory B, comparing the medians, ranges and extreme values obtained with each assay method Method Conclusions The primary conditions for an assay method to be appropriate as a pharmacopoeia1 method are a small reproducibility value R and a fair agreement between the repeatability value r and the reproducibility value R, as a widespread application in many different laboratories should provide similarly precise results for the same sample. Applying the IS0 guidelines for the determination of repeatability and reproducibility for a standard test method by interlaboratory tests, it has been demonstrated experimen- tally that important differences in precision exist between the volumetric assay methods studied.The three assay methods examined are assumed to measure content levels of phenothiazines between 99.0 and l O l . O % , with reference to the dried or anhydrous substance. This also means that supplementary experimental errors, produced by the analytical measurement of the loss on drying, are included in the content limits, together with the experimental error of940 the assay method.Nowhere in the Ph. Eur. is it specified for volumetric methods what the standard deviation s should be, when it is applied in a certain laboratory, performing a certain number of replicates. Considering the assay with method C, a 95% probability was found that the difference between two intralaboratory measurements may reach 2% (r) and between two interlaboratory measurements even 3.2% ( R ) . Appar- ently, this method can, in at least one laboratory, provoke a systematic error, the origin of which can be due to different causes, which are now under investigation.Only one method (B) showed concordant reproducibility and repeatability values, R and r, smaller than 2%. These statistical intervals, however, contradict the imposed narrow content interval of 99.0-101.0% given in the monograph, which may never be exceeded, according to the statements under ‘Limit values’ in section IV.2. of the Ph. Eur. These conclusions are confirmed by the two-way ANOVA results, which reveal that for each of the three assay methods examined, significant contributions from laboratory-to-labor- atory variations to the total variances are measured. With method B, the laboratory-to-laboratory variations are much less important so that even small sample-to-sample differences are detected. It must, therefore, be concluded that method B is the most appropriate method and that it is questionable whether its substitution by method C is feasible, without changing the content limit requirements. Analogously, a two-way ANOVA applied to the assay results of each participating laboratory demonstrates that in three of the four laboratories method-to-method variations contribute significantly to the total variance. As a conse- quence, adjusted content limit intervals should be introduced ANALYST, JUNE 1992, VOL. 117 in a monograph, after substitution of assay method B by assay method C . The authors thank E. Deblay, R. Schoenaers and J . Hoebus for skilful technical assistance during the experimental work. 1 2 3 4 5 6 7 8 9 10 References Pharmacopke Europkenne I , Conseil de I’Europe, Maisonneuve, France, 1975. vcl. 111, p. 194. Pharmacopke Europkenne I I , Conseil de 1’Europe. Maison- neuve, France, 1980, p. 59. Pharmacopke Europkenne II, Conseil de I’Europe. Maison- neuve, France, 1986, p. 475. Guide for the Technical Content of Monographs. European Pharmacopoeia Commission, Council of Europe, Document PA/PH/SG(85)14,4R, 1991. Wernimont. G. T., in Use of Statistics to Develop and Evaluate Analytical Methods, ed. Spendley W., Association of Official Analytical Chemists. Arlington, VA. 1985, pp. 25-28. Precision of Test Methods, Determination of Repeatability and Reproducibility for a Standard Test Method by Interlaboratory Tests, International Standard IS0 5725-1986. Statgraphics version 5.0 software, STSC, Rockville, MD, USA. Dixon. W. J . , Biometrics. 1953, 9, 4. Massart, D. L., Vandeginste, B. G. M., Deming, S. N., Michotte. Y ., and Kaufman, L., Chemometrics: a Textbook, Elsevier, Amsterdam, 1988, pp. 70-71. Sokal, R. R., and Rohlf, F. J . , Biometry, Freeman, New York, 2nd edn., 1981, p. 402. Paper 1106376F Received December 20, 1991 Accepted February 17, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700933

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 941 Bias and Measurement Errors in Radioactivity Data From Four European Radiation Research Laboratories E. J. McGee and P. A. Colgan Nuclear Energy Board, 3 Clonskeagh Square, Dublin 14, Ireland M. Keatinge Department of Zoology, Trinity College, Dublin 2, Ireland A. D. Horrill and V. H. Kennedy Institute of Terrestrial Ecology, Merlewood Research Station, Cumbria LA? I 6JU, UK K. J. Johanson Swedish University of Agricultural Sciences, Uppsala, Sweden A. Aarkrog and S. P. Nielsen Risa National Laboratory, Roskilde, Denmark Four radiation research laboratories participated in an interlaboratory comparison exercise t o assess precision in the measurement of radionuclide concentrations. Each laboratory was required t o report results for 137Cs, 134Cs and 40K in accordance with their usual procedures.A normal probability plot of the pooled data showed ten of the 161 readings t o be obvious outliers and seven of these were attributed to one of the laboratories. One laboratory consistently overestimated relative t o the global mean, another laboratory consistently underestimated while the other two showed positive and negative bias dependent on isotope. Both bias and measurement errors were found t o increase in the order 137Cs, 134Cs, 40K. A need for greater standardization of analytical techniques was identified. Keywords: Intercomparison; radionuclide; measurement error; bias; gamma spectrometry Samples collected from semi-natural ecosystems are seldom included in multinational intercalibration exercises arranged by international organizations.As part of a research pro- gramme evaluating radiocaesium behaviour in such eco- systems, an intercomparison exercise using samples of soil and vegetation containing both natural and artificially produced radionuclides was organized. The participating laboratories were: Risq) National Laboratory (RISQ)), Roskilde, Denmark; Nuclear Energy Board (NEB), Dublin, Ireland; Institute of Terrestrial Ecology (ITE), Cumbria, UK; and the Swedish University of Agricultural Sciences (SLU), Uppsala, Sweden. A total of four samples was analysed by each of four radiation research laboratories, each using its existing proce- dures. Each sample was analysed for 137Cs,134Cs and MK by each laboratory. Two of the samples were soil samples, another was of Pteridium aquilinum, and the fourth was of Calluna vulgaris.The data produced by each of the laboratories were assessed for instrument measurement precision. Comparison of the results from different laboratories facilitated the investigation of consistent bias in results from individual laboratories. This made it possible to compare measurement systems and to determine if observed bias was consistent for each of the radionuclides investigated. The relative importance of instru- ment precision was then assessed and compared with errors produced by bias in systems. The importance of different activity concentrations relative to these two sources of error was also determined. The four participating laboratories are denoted by the letters A, B, C and D and are not further identified by name.The coding sequence is different from that of the authorship and of paragraph 1 above. Experimental Samples of soil and vegetation were chosen to give a range of activities typical of those found in the participating countries following the Chernobyl accident. Two soil samples of differing densities and composition were collected in Ireland. These soil samples are referred to as soil 1 and soil 2. Above ground growth of Pteridium aquilinum was collected in the Lake District of the UK, where some of the radiocaesium is likely to be derived from the Sellafield reprocessing works. This sample is referred to as plant 1. Calluna vulgaris litter was collected near Loch Laggan in the UK from an area where Chernobyl contamination was relatively heavy, and this sample is referred to as plant 2.Plant samples were dried to constant mass at 85 "C and ground in a rotary mill. Samples were homogenized by tumbling in a rotary mixer for 20 min. No further testing was undertaken prior to distribution of sub-samples to the other participants. Soils were prepared by removing any plant material, roots and stones. They were then dried at 105 "C to constant mass and ground to less than 1 mm particle size in an agate rotating mill. Homogeneity in both samples was tested at laboratory A. An aliquot was removed and analysed after which this aliquot was remixed with the sample and a second aliquot removed. The second aliquot was then counted and remixed with the sample. The procedure was repeated up to six times for each sample.Different aliquots were counted on different detec- tors. The resulting observed variation in values recorded for different samples was, therefore, a combination of errors due to sample homogeneity and instrumentation. The number of aliquots measured for each sample is given in Table 1 with the 95% confidence interval for each sample. Relative standard deviations associated with replicate measurements on aliquots for each isotope were all less than Table 1 Mean values and 95% confidence intervals of counts carried out on soil samples by laboratory A . Results are presented in Bq kg-1 dry mass at sampling date; tz is the number of measurements made in each instance Sample n 137C' *34cs 40K Soil 1 5 35 k 1.1 3 k 0.2 89 t- 5.4 Soil 2 6 220 t- 7.3 13 k 0.7 112 t- 3.2942 ANALYST, JUNE 1992, VOL.117 Table 2 Analytical criteria applied at each of the four participating laboratories. Standard laboratory procedures are listed for each laboratory Criteria Detection limits Determination limits Standards used Are corrections made for sample matrices? Cascade summing for 134Cs? Allowances made for handling errors? Counting times of intercomparison samples Labora- tory Procedure A Given by Currie” B C D A Use detection limits only B Use detection limits only C D As for detection limits A Amersham QCY48, mixed gamma B C Amersham QCY44, mixed gamma D A No B Yes C Yes D No A No B Yes C Yes D No A Yes,3% B No C No D No A 54000-83900s B 61 000-258 128 s c 8OOoos D 56 800-252 000 s 4.66 times the standard deviation of the background count Determined by Canberra Apogee software Usually only count high activity samples which are well above detection limits Determined by Canberra Apogee software Amersham, mixed gamma, standards diluted to fill the various geometries Amersham CDZ.72 for 13’Cs, CCZ.72 for l34Cs and normal KCI for -10K Counting geometries of intercomparison samples A Soils: 500cm3, vegetation: 200cm3 B 200cm3 C 150cm3 D 180,90 and 35 cm3 vials Method of counting error determination * Ref.5. A B C D Error provided by Canberra Nuclear Data software plus 3% Twice the standard error of the mean value of repeated measurements That provided by Canberra Apogee software One standard deviation of the mean of repeated measurements 20% and therefore satisfy the homogeneity criterion of Ballestra et al.1 and Veglia et a1.2 When homogeneity had been confirmed, samples were sub-divided into aliquots and dispatched to laboratories B, C and D. Laboratories were requested to quote results decay- corrected to the date of collection. It was requested that each laboratory analyse their samples using standard procedures. An assessment of differences in procedures between labora- tories was a primary objective of the work and no attempt was made to standardize analytical procedures for the analyses. Laboratories were requested to complete a questionnaire giving details of laboratory practices. The results of this questionnaire are presented in Table 2. Differences in sample preparation procedures, calibration, counting equipment, computer software, calibration stan- dards, counting geometries, counting times and methods of error calculation are shown to exist between laboratories.No two laboratories employ the same procedures, and from the outset it was probable that there would be differences in the results generated by the participants. Three of the four participants use commercially available software, which is regularly updated, for automatic peak fitting and determination of radionuclide concentrations. Laboratory B has developed and written its own software packages. Each laboratory also uses a different source for nuclear data. Laboratories A and C use the databank supplied with their software, while laboratory B refers to Ewbank and Schmorak3 and laboratory D uses the data of Heath.4 For setting detection and determination limits, laboratory A refers to Currie,s while Reus et a1.6 is used as a basic reference text by laboratory C.Laboratory B uses only detection limits, calculated as 4.66 times the standard deviation of the background count. Because it normally analyses only high activity samples, laboratory D has not found it necessary to set detection or determination limits. Results The results from each laboratory were initially assessed for over-all agreement. Inspection of the raw data shows the mean for a given sample (the aliquots dispatched to each laboratory are referred to as ‘samples’ in the remainder of the text) and isotope to be positively correlated with the variance; greater mean values are accompanied by greater variances. All data were log transformed prior to the analysis in order to remedy this situation and make the variance independent of the mean .7 Data analysis was a two-factor, ‘isotope’ (137Cs, 134Cs and 4°K) and ‘sample’ (soil 1, soil 2, plant 1 and plant 2) design.The experiment repeatedly measures the ‘isotope’ factor.8 Both ‘sample’ and ‘isotope’ were considered to be fixed effects. Laboratories are considered as ‘subjects’ in the design,ANALYST, JUNE 1992, VOL. 117 -0.30 943 - W I I I and ‘subject’ is a nested (in ‘sample’) random factor. The performance of a given laboratory was assessed under all levels of the ‘isotope’ factor, once a sample had been prepared results were obtained for all three isotopes, but only under one level of the ‘sample’ factor. For the purposes of analysis each sample is considered to be unique; this allowed us to distinguish between errors introduced in the preparation of a sample and errors systematically introduced by a particular laboratory.Repeated analyses of the same sample, as in this experiment, are considered to be the same ‘subject’, giving repeated ‘subject’ measurements. The experimental design is explained in Table 3. The design allowed testing of the hypothesis that reported results ‘between subjects within groups’ were not different. This hypothesis was rejected at the 5% level (F12,113 = 2.836 and p = 0.002, where p is the probability of the type 1 error). These findings could not be attributed to any single isotope or sample but were found to arise from a combination of these two factors as indicated by a significant interaction (F6,24, p = 0.0001).This analysis is based on the assumption of normally distributed residuals, and a normal probability plot of the residuals generated from the complete data set is shown in Fig. 1. Traditionally, the ith normal score for a sample of size n has been the mean of the sampling distribution of the ith statistic in a sample of n values drawn from a standard normal distribution. Here, the medians of the sampling distribution of the order statistics have been approximated according to the standard normal distribution.9 Fig. 1 shows that ten out of 161 data points deviate considerably from the straight line formation of the data points; these outliers are highlighted in the figure. Seven of these points were attributed to laboratory D and five of the seven outliers were determinations of 134Cs by that laboratory.Mean values for activities in each sample, with the associated measurement errors for the three isotopes deter- mined at all the laboratories, are presented in Table 4. The Table 3 The experimental data may be represented as follows: where subject 1 indicates, for example, laboratory Ahoil 1 and subject 2 indicates laboratory B/soil. 1 The subscripts refer to ‘sample’, ‘isotope’, ‘subject’ and repeated analyses of the same sample by a given laboratory Subject 137Cs 134cs 40K 1 Xllll XI211 XI311 1 XI112 XI212 XI312 Sample 1- 1 XI 1 I n X121n X131n 2 x1121 XI221 X132 1 2 XI 122 X1222 XI322 2 X112n X122n X132n 3 etc. etc. Sample 2- = =I confidence intervals generated from the homogeneity test carried out at laboratory A (Table 1) can be compared with the measurement errors presented in Table 4.The confidence intervals are of similar magnitude to the measurement errors, showing that errors due to inhomogeneity were approximately equivalent to errors associated with instrumentation. These data provide further evidence that samples were adequately homogenized prior to the distribution of aliquots. Examination of the data which gave rise to the outliers shows that repeated measurements on the same samples yielded differences that exceeded the presented measurement errors calculated from the standard procedure by laboratory D; observed precision was poorer than the standard proce- dures predicted. The inconsistency of laboratory D data suggests a possibility of poor cross-instrument calibration and/or a lack of defined laboratory procedure.In order to assess bias (accuracy), each result was compared with a grand mean calculated from the results from all four laboratories for that sample and isotope. This latter value is the best estimate of the ‘true’ or bias free result provided by assessment of the data from all the laboratories. The mean percentage deviation from the grand mean was then calculated for each laboratory, isotope and sample. Data were pooled by isotope for all four samples so that absolute percentage bias between laboratories could be investigated (Table 5 ) . This showed whether bias was negative or positive, and whether or not it was dependent on isotope.Box plots10 of these data (Fig. 2) graphically illustrate the percentage bias on a percentile basis in relation to the grand mean which lies at zero. Table 4 Mean values of counts carried out on each sample within each of the four laboratories. Results are presented in Bq kg-1 dry mass at sampling date. Counting errors are mean values calculated from the independent estimates furnished by participating laboratories (for method of calculation see text; values <1 are recorded as 1). ‘n’ is the number of measurements made at each laboratory 137Cs 134cs WK Laboratory n Mean k Mean k Mean k Data for soil 1 sample- A 5 35 (1) 3.1 (1) 89 (3) B 3 35 3.0 88 C 3 41 (1) 3.7 (1) 114 (12) D 2 36 (1) 2.5 (1) 147 (8) A 6 220 (3) 13 (1) 112 (3) B 4 233 16 112 C 3 256 (5) 16 (1) 128 (9) D 2 240 (1) 15 (1) 111 (3) A 3 79 (2) 13 (1) 160 (8) B 4 82 15 146 C 3* 78 (2) 13 (1) 194 (27) D 4 83 (1) 10 (1) 147 ( 5 ) A 3 3839 (74) 503 (11) 81 (8) B 4 4063 600 70 C 3 3863 (74) 519 (6) 135 (23) D 2 3868 (0) 570 (4) 67 (7) * Two counts for 10K.Data for soil 2 sample- Data for plant 1 sample- Data for plant 2 sample*- Table 5 Mean estimated bias (%) for each isotope and laboratory; SD = standard deviation 137Cs 13-Q MK Laboratory Mean SD Mean SD Mean SD A -5 (4) -6 (8) -10 (10) B -1 (4) 7 (9) -16 (11) D 1 (3) -50 (105) -8 (26) C 3 (6) 5 (9) 14 (18)944 ANALYST, JUNE 1992, VOL. 117 10 40 5 T 30 i 0 -5 -10 -15 - 20 c 2 4 a 30 40 t T 1': 137cs 'WCS 40K isotope Fig. 3 Absolute percentage measurement error by isotope 50.0 I I -10 t L I - s 37.5 Y I Fig.2 Estimated percentage bias by isotope and by laboratory. (a) Laboratory A; (6) laboratory B; (c) laboratory C; and (d) laboratory D 8 25.0 Table 6 Percentiles of pooled (i) bias (YO) and (ii) measurement error (YO), by isotope. Results are rounded to 1 significant figure 137cs 134cs Measure- Measure- ment ment Bias error Bias error Percentile (Yo) (YO) (Yo) (Yo) 4OK Measure- ment Bias error 4 3 12 6 24 10 (%) (%) 137cs 134cs OOK Isotope Fig. 4 Absolute percentage bias by isotope mates, are shown in Table 6. There appeared to be little correlation between measurement error and the magnitude of the count to which it applied (Pearson product moment correlation coefficient = -0.268). Discarding two abberant results for 134Cs produced by laboratory C, plots of the percentage measurement errors (Fig.3) and of percentage bias estimates (Fig. 4) were obtained. A strong association was found to exist between isotope and percentage measurement error (Fig. 4). Isotopes follow the same order for increasing percentage bias as they do for increasing percentage measurement error. Hence, for example, the larger measurement errors relating to 40K are likely to be further compounded with the greater bias that is also found for this isotope. The data presented in Figs. 3 and 4 are summarized in Table 6. From these data it is noted that the percentage bias is approximately twice the percentage measurement error. Hence the measurement errors tend to provide for overlap of individual data points with the global mean. Consideration might be given to the possibility that when, for example, a laboratory consistently shows negative bias then the measure- ment errors be always read as positive in order to compensate.This work has concentrated principally on investigating bias and measurement error between laboratories. However, the performance of the participating laboratories in international intercalibration exercises is also of interest. Laboratories A, B and C have taken part in International Atomic Energy Agency (IAEA) intercomparison exercises. l ~ 1 1 Each laboratory con- fidentially provided results of performance in relation to recommended values and 5% confidence intervals (Table 7). Results produced by laboratory A in these studies were in excellent agreement with recommended values.All data from this laboratory (without considering measurement errors) were within the recommended 5% confidence interval. Two further unpublished studies by the IAEA (IAEA-367 and IAEA-368) were carried out in 1990, the same year in 25 1 1 3 2 50 4 2 7 3 75 7 3 12 5 Laboratories A and C both show a consistent trend across the three isotopes. Laboratory A tends to underestimate consistently, while laboratory C tends to overestimate consist- ently. These results suggest systematic bias within these two laboratories. Results from laboratory B show errors of similar magnitude to those from A and C, but do not display a consistency of sign; errors are both positive and negative and bias appears to be isotope specific. Laboratory D yields median values close to the grand mean (bias is neither consistently positive or negative) but the range of the measurement errors is large [Fig.2 ( d ) ] , especially when compared with those produced by the other laboratories. These large deviations can be traced to the outliers highlighted in the normal probability plot (Fig. 1). Given the lack of precision in results from this laboratory it becomes difficult to distinguish between bias and measurement error in this instance. The fact that outliers affect the calculation of the grand mean must also be noted as this might cause problems in the assessment of the data from other laboratories. Bias analysis also reflects an increase in the magnitude of errors from 137Cs to 134Cs to 4% (Fig. 2). The final part of the analysis looked at the percentage measurement error furnished by three of the four participating laboratories and compares measurement error (precision) with estimated bias (accuracy). Independent measurement errors (Table 4) were expressed as a percentage of the mean and pooled by isotope, then expressed as absolute values to produce absolute measure- ment error values by isotope.Percentiles of pooled percentage measurements, presented with pooled percentage bias esti-ANALYST, JUNE 1992, VOL. 117 945 Table 7 Results of IAEA intercomparison exercises for laboratories A, B and C. Recommended values (Bq kg-I) and 5% confidence intervals are presented for 137Cs, 1 3 % ~ and jOK Laboratory Study Nuclide IAEA-306 13JCs 137Cs 10K IAEA-307 134Cs 137Cs ?OK 137cs 3% IAEA-308 134Cs Recommended value 53 201 785 1.6 4.9 1.6 5.6 150 1381 Confidence interval (5%) 50-54 194-206 757-827 1.5-1.9 4.5-5.2 141-161 1.5-1.8 5.3-6.0 1320-1456 A 53 f 1 196 k 2 808 f 12 1.5 rt 0.2 5.1 -t 0.3 151 k 5.0 1.7 k 0.3 5.5 & 0.3 1396 k 10 B C 56k 1 50 -t 2 190 k 4 199 rt 1 766 k 8 - 1.3 f 0.1 - 4.4 -t 0.3 139 rt 7 - 1.6 f 0.2 - 5.3 _+ 0.3 1325 k 66 - 4.7 f 0.4 6.5 k 0.6 which samples used in this work were analysed.Results from laboratory A are again in good agreement with a preliminary report of recommended values, while laboratories B and C also emerge favourably from the intercomparison. These data are summarized in Table 7 and are presented as evidence that laboratories A, B and C produce results consistent with stringent international standards. Nevertheless, this study has shown that there was a statistically significant difference in the results produced by laboratories, and also presents evidence of consistent bias in results. It is therefore proposed that the performance of all laboratories could be further improved.It is clearly shown in Table 2 that there is a diversity of analytical procedures between laboratories, and these different practices cause problems with comparability of results. The sources of inconsistency in the results presented cannot be readily identified, but greater consistency in analytical procedures is clearly required and should be considered on an international level. Conclusions Initial analysis showed a significant difference in the results produced by the four laboratories. A normal probability plot showed that only ten out of 161 data points were obvious outliers and most of the outliers were attributed to laboratory D.Data from this laboratory were found to show poor precision when the same sample was measured repeatedly and the measurement errors calculated by standard laboratory procedure underestimated the observed measurement errors. Measurement errors were small in the other three labora- tories, although bias in results was identified as a problem. Comparison of laboratory means with the grand mean for all laboratories identified consistent negative bias in results from laboratory A and consistent positive bias in those from laboratory C, but both of these laboratories had relatively small observed measurement errors. Laboratory B was found to show measurement errors of similar magnitude to those of laboratories A and C, but bias was found to be both positive and negative depending on isotope.Median values from laboratory D lay close to the global mean for all isotopes, but the range of measurement errors was large, particularly for 134Cs; it is therefore difficult to separate bias from measurement errors in this instance. The large measurement errors from this laboratory may also have been detrimental to accuracy in the calculation of the grand mean. Bias errors (accuracy) and measurement errors (preci- sion) were found to increase in magnitude from 137Cs to 134Cs to 40K. An assessment of bias errors compared with measure- ment errors shows the former to approximate to twice the value of the latter; hence, measurement errors were, in general, sufficient to allow for inaccuracy due to bias.The authors thank Hugh Synnott who provided technical assistance and Dr. Ann McGarry for her helpful comments on the text. Both are employees of the Nuclear Energy Board. All participating laboratories receive funding through the Radiation Protection Research Programme of the European Communities under contract B17-0044. 1 2 3 4 5 6 7 8 9 10 11 References Ballestra. S . , Vas, D . , Lopez, J . J . , and Noshkin, V., Intercomparison of Radionuclide Measurements in Mixed Sea- weed Sample IAEA-306, IAENALl013, IAEA, Vienna, 1989. Veglia, A., Ballestra, S., and Vas, D . , Report on the Intercom- parison of IA EA-307 Radionuclides in a Sea Plant, IAEAIALI 014. IAEA, Vienna, 1989. Ewbank, W. B., and Schmorak, M. R., Evaluated Nuclear Structure Data File: A Manual for the Preparation of Data Sets, Riso National Laboratory, Denmark, 1978. Heath, R. L., Gamma-Ray Spectrum Catalog, ANCR-1000-2, 1974. Currie. L. A., Anal. Chem., 1968,40, 586. Reus, U., Westrneier, W., and Warnecke. I . , Gamma-Ray Catalog, GSI-Report 79-2, Gesellschaft fur Schwerionenfor- schung, Darrnstadt, 1979. Sokal, R. R. E., and Rohlf, F. J . , Biometry, Freeman, San Francisco, 1981. Winer, B . J . , Statistical Principles in Experimental Design, McGraw-Hill, New York, 1971. Velleman, P. F., Data Desk 111, Odesta Corporation, 1990. Vellernan. P. F., and Hoaglin, D. C., Applications, Basics and Computing of Exploratory Data Analysis, Duxbury Press, Boston, MA, 1981, p. 354. Ballestra, S . , Vas, D . , Lopez, J . J . , and Noshkin, V., Intercomparison of Radionuclide Measurements in Mixed Sea- weed Sample IA EA-308, IAENALlO15, IAEA, Vienna, 1989. Paper 1 I0341 1 A Received July 8, 1991 Accepted January 17, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700941

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 947 Windowing Technique for Determining the Composition of Organic Samples by Near-infrared Reflectance Lorna S. Aucott School of Mathematics and Statistics, Lancashire Polytechnic, Preston, Lancashire PR I ZTQ, UK Stephen T. Buckland Scottish Agricultural Statistics Service, ML URI, Aberdeen A59 ZQJ, UK Paul H. Garthwaite Department of Mathematical Sciences, University of Aberdeen, Aberdeen A59 2W, UK Ian M. Nevison Scottish Agricultural Statistics Service, RRI, Aberdeen A52 SSS, UK Ian Murray Scottish Agricultural College, Aberdeen AB9 IUD, UK A method of windowing regions of interest in near-infrared reflectance spectra is described. Windows are bounded by wavelengths at which spectra from all samples are re-scaled to zero. These bounds are determined numerically to maximize the correlation between the scaled spectral values within a window and chemical composition, as determined by wet-chemical methods, of a calibration set of samples.It is shown that the method yields predictions of chemical composition of forages that are, on average, more accurate than comparable methods based on principal components regression. Keywords: Near-infrared reflectance spectroscopy; principal components regression Near-infrared reflectance (NIR) spectroscopy has been applied successfully to the rapid analysis of organic materials such as foods, plastics, pharmaceuticals and petrochemicals. Speed of analysis and minimal sample preparation are obvious advantages of NIR over conventional methods of chemical analysis.Physical methods such as NIR are also more portable, more precise methods of analysis capable of on-line or fibre optic application in vivo or remote from the laboratory. Near-infrared reflectance spectra are measurements of the log reciprocal reflectance (log[ 1/R]) relative to a non-absorb- ing standard tile at discrete (usually 2 nm) wavelength intervals. Such measurements behave like absorbance measurements in transmission-mode spectroscopy. The amount of light absorbed depends on the pathlength (i.e., the amount of absorbing material the light passes through). In reflectance, this is not easily controlled, and the assessment of the concentration of any one constituent relies on calibrating the NIR spectra to results found by ‘wet-chemical’ methods, for example, the Kjeldahl technique for determining nitrogen.Hence, NIR methods are usually applied by first acquiring a calibration set of samples for which wet-chemical results are known. A calibration equation is derived and used to predict the composition of samples for which wet-chemistry results are unknown. The calibration is usually achieved by multi- variate methods such as stepwise multiple linear regression,l-3 or principal components analysis, followed by regression on the components.4 Such processes are highly dependent on the precision of the wet-chemical measurement relative to the spread in composition of the calibration samples, and on their nature and pre-treatment (drying, milling, residual moisture content, etc.). Scatter correction prior to calibration, which is robust across different populations and monochromators, is currently the subject of statistical research in NIR spectro- scopy.The above multivariate methods perform well if the calibration set is representative (ideally a random selection) of future samples to be analysed, and if spectral readings for a single sample at a given wavelength are invariant with time and between monochromators. In practice, this is not the case. An attempt was made in this work to develop a procedure that ( i ) is at least as effective as methods currently favoured, such as principal components regression; (ii) requires relatively simple computer software; and (iii) requires that log[llR] is measured only at a limited number of wavelengths (typically between 3 and 20), so that the methods can be used on fixed-filter photometers.The procedure requires an initial calibration against wet-chemistry results, so is limited by the accuracy of those results, but thereafter, it relies on only the optical data. For the purposes of this paper, the composition of forages is discussed, although the procedures could be applied to any organic substances. Transformation of Spectra Prior to analyses, spectra can be transformed by using the method recommended by Aucott et al.5 This transformation ( i ) helps to identify analytically useful wavelength regions that provide information on composition, and (ii) reduces the effects of scatter resulting from variation in particle size and, possibly, refractive index. Let ui represent the untransformed log[llR] value at wavelength j .An increment of one in j generally corresponds to 2 nm. The second derivative can then be calculated as uy = Ui-8 - 2u; + u;+s (1) The recommended transformation of Aucott et al.5 is A plot of xi versus wavelength shows a more complex pattern than does a plot of untransformed u, versus wavelength. However, use of xi largely eliminates the effects of particle size from the spectra.5 If the user wishes to stay in the log[llR] domain, rather than in the second-derivative domain, the xi above can be replaced by v; = log, (1 + u;) (3) Suppose two wavelengths, A, and Ah, are defined as bounds to a region that potentially contains spectral information on a constituent of interest. We define such a region as a ‘window’.948 ANALYST, JUNE 1992, VOL.117 0*050 * 2066 2100 2144 Wavelengthlnm Fig. 1 Transformed spectra z of 20 grass samples within a window identified for predicting nitrogen. Note that the spectra pass through zero at the two nodes, located at 2066 and 2144 nm, and separate out within the window. Samples showing the highest values within the window have the greatest nitrogen content The following procedure removes particle size effects in the log[ 1/R] domain, provided that samples that are identical, apart from particle size, generate parallel spectra within the window. Use of v, instead of uj generally ensures that this proviso is closer to the truth. Creating Nodes In the following, spectral values transformed by use of eqn. (3) are considered. Where v, appears below, x, or u, could be used equally as well, although the use of the untransformed log[llR] values, u,, is not recommended.For a given window, a linear transformation is applied to create nodes at h, and hb, so that the transformed spectrum of every sample in the calibration set passes through both points (see Fig. 1). These nodes are, therefore, artificial isosbestic cross-overs, obtained by transforming vj to zj as follows: z; = v, - (bo + blj) (4) where and bo = vha - blha, for j = h,, --., hh For every sample, z, equals zero at j = h, and j = hb. A procedure is required for relating the zjvalues to chemical composition. A single wavelength h at which the correlation between the zj and the wet-chemistry measurements is highest can be selected for making predictions. Alternatively, a function of the zj values might be used, for example, a simple average, or a linear combination determined by principal components analysis.In practice, selection of a single wavelength from each window works at least as well as a more sophisticated approach, and has the advantage that, having selected a wavelength with use of a monochromator, a fixed-filter photometer can be used for determining the composition of further samples: if determinations are based on v, values [eqn. (3)], reflectance values at only three wavelengths per window, ha, h and hb, are required, whereas if xj values [eqn. (2)] are used, reflectance values at nine wavelengths per window are needed so that the second derivative can be measured at ha, h and kb. 2.5 2.0 1.5 1 .o 0.5 0 7.0 6.0 L 5.0 0, ?! P 4.0 c 3.0 CI 0 E 2.0 1 .o 0 I 6.0 5.0 4.0 3.0 2.0 1 .o n c 4 8 12 16 20 4 8 12 16 20 Number of terms Fig.2 Root mean square error of calibration [ ( a ) , ( c ) and (e)] and of prediction [(b). ( d ) and cf)] by number of terms for principal components regression using CLG (solid curve) and SDL (broken curve) transformations, and for the windows method (dotted curve). Data are Australian hays and measured attributes are crude protein [(a)and (b)], NDF[(c)and ( d ) ] and IVD [(e) and m] For the analyses described in this paper, potential windows were identified by carrying out a numerical search on all (A,, hb) pairs with hh - h, 2 18 nm and with ha b 1100 nm and hh 6 2300 nm. Only those that yielded a correlation between zh and wet-chemistry results of at least pmin were retained.A different value for pmin was selected for each wet-chemistry result, with the aim of selecting approximately 50 windows. If the wavelengths A, at which correlation was maximized for any pair of selected windows, were less than 10 nm apart, only the window giving the larger correlation was retained. The approach is computer intensive, but only needs to be carried out once for a given constituent. Further research into methods for identifying appropriate windows could prove worthwhile.ANALYST, JUNE 1992, VOL. 117 949 3.0 2.5 2.0 1.5 1 .o L 0.5 Q) !?! = o 5: c 6.0 m E * 0 5.0 a 4.0 3.0 2.0 1 .o 0 4 8 12 16 20 4 8 12 16 20 4 8 12 16 20 Number of terms Fig. 3 Root mean square error of calibration [(a) and (d) and of prediction [(b), (c), (e) and (f)] by number of terms for principal components regression using CLG (solid curve) and SDL (broken curve) transformations, and for the windows method (dotted curve).Data are EEC grasses and measured attributes are crude protein [(a)-(c)] and IVD [(d)-(f)] Calibration Equation Suppose a single window is defined. The log[l/R] values are transformed to z values for each of the n samples in the calibration set, and a simple linear regression is carried out to provide the calibration equation. The dependent variable is the constituent of interest as measured by wet chemistry, yi, i = l , . . . , n . For sample i(i = l , . . . , n ) , the independent variable wi may be the value of z, for a selected wavelength j between ha and hh, or the average of the z values, or the score on a principal component.The calibration equation is then given by j = & + @ w ( 5 ) where n c yi(wi - ii)) 2 (Wi - i+)* p = i7-1 i = I and & = 9 - pi+, with i+ = Zwi/n, 9 = Zyi/n When a new sample is scanned, the spectra are transformed to z values, and the value of w , which is the z corresponding to the selected wavelength, or the chosen function of zj values, is substituted into the above equation to provide an estimate of chemical composition. The above method is univariate in the sense that the original multivariate data matrix, typically of size 700 x n, where n is the number of samples in the calibration set, has been reduced to a single vector of length n. For some purposes, this vector may contain sufficient information to provide satisfactory predictions.In practice, however, other regions of the spectrum are likely to contain useful and different information for predicting the constituent. In this circumstance, additional windows can be included in the analysis by using either stepwise multiple linear regression or principal components regression. Hence, if an instrument is used that yields spectral values for at least six wavelengths and the transformation of eqn. (3) is applied, there is no loss from using two windows; each additional window requires spectral values from a further three wavelengths. Summary of the Method Defining and Selecting the Windows These steps, which require a monochromator, only need to be carried out once for a given constituent.(1) Transform the log[ 1/R] values ui by using eqn. (2) or (3). (2) Define a window bounded by wavelengths ha and hb. (3) Further transform the spectra within this window to create nodes at A, and Ab, and obtain z, values by using eqn. (4) Find the wavelength h within the window for which the correlation between z, and the wet-chemistry results is a maximum. (5) Carry out a numerical search of all possible windows to extract those that yield high correlations with the wet- chemistry results. (6) If only one window is required, select the one that provides the largest correlation with wet-chemistry results and obtain a calibration equation by using simple linear regression [eqn. (5)]. For more than one window, use stepwise multiple regression to select windows or principal components regres- sion to extract information from all windows.In each instance, the regression provides the calibration equation. (4). Analysing New Samples These steps are carried out when samples of unknown composition are analysed, and can be implemented on either a monochromator or a fixed-filter photometer.ANALYST, JUNE 1992, VOL. 117 950 4.0 3.5 3.0 2.5 2.0 1.5 1 .o 8 0.5 Q) ?? = o 5 ~6.0 0 K 5.0 4.0 3.0 2.0 1 .o 0 4 8 12 16 20 4 8 12 16 20 Number of terms transformations, and for the windows method (dotted curve). Data are Norwegian grasses and measured attributes are crude protein [(a) and (b)] and NDF [(c) and (d)] (1) Transform the log[l/R] values uj by using eqn. (2) or (3). The same transformation must be used as in step (1) above.(2) Further transform the data, using eqn. (4), to obtain the value of z j at the wavelength j selected at step (4) above. ( 3 ) Substitute the z j values into the calibration equation from step (6) above to obtain the predicted content of the constituent. Assessing the Method Suppose a validation data set consists of m samples for which wet-chemistry results are available. Let: yi = wet-chemistry estimate for the calibration set, i = l,-.-,n; y k = wet-chemistry estimate for the validation set, k = l;-.,m; jji = value predicted by calibration equation for the calibration set, i = l , - . . , n ; j k = value predicted by calibration equation for the validation set, k = 1 ,'--,rn; and r = number of windows (stepwise regression) or components (principal components regression) selected by analysis.Root mean square errors were used to compare the window method with methods based on principal components, as they measure both bias and precision. If the root mean square error is small, then a method yields estimates of organic composi- tion with low bias and high precision. The root mean square error of calibration, qc, is given by I " and the root mean square error of prediction, qp, is given by rn In addition, bias in the predictions for the validation set is measured as 1 m bias = C (jjk - y k ) mk=1 Bias in the predictions for the calibration set as determined in this equation is always zero as a consequence of the statistical technique used. In order to assess the windows method, z values were obtained by using the logarithmic transformation of eqn.(3), and selecting the calibration equation by stepwise regression. For comparison, principal components regression was applied to the same data sets by using a centred logarithmic transform (CLG), which is comparable to using the transformation of eqn. ( 3 ) on windows data. In order to compare this with a transformation found to perform best by Aucott et aZ.,5 the principal components regression was repeated by using the second-derivative logarithmic (SDL) transformation of eqn. (2). (This transformation could be used in the windows method, but would involve a new search for potential windows. ) For each method, selection of terms is based on testing for significance at the 1% level. Assessments of three different properties of forages were considered: nitrogen (and hence crude protein), neutral detergent fibre (NDF) and in vitru digestibility (IVD).The minimum acceptable correlation pmin for retaining a window was set at 0.6 for nitrogen and 0.5 for both NDF and IVD. In total, 55 potential nitrogen windows, 58 NDF windows and 48 IVD windows were identified in the range 1100-2300 nm by numerical search from analyses of a set of 157 Australian hays.6 Stepwise regression was used to select from these windows for each of the analyses presented here. In each analysis and for each method, the calibration equation was obtained, using stepwise regression on the calibration set listed below. This equation was then carried across to predict the composition of the validation set(s). (i) The Australian hays+ 53 samples analysed in 1980 were used as the calibration set, and 104 samples analysed in 1981 were taken as the validation set. (ii) A set of EEC silages, separated into a calibration set of 40, a validation set of 20 (the first validation set) and a blind set of 20 (the second validation set).7 (iii) Ninety-six samples of Norwegian grasses were ran- domly sub-divided into a calibration set of 50 and a validation set of 46.8 Results The performance of the methods is indicated in Figs.2 4 and summarized in Table 1. The left-hand side of each figure corresponds to the calibration data set. For all three methods, the root mean square error of calibration (qc) becomes smaller as the number of windows or components increases. If terms are tested for inclusion at the 1% significance level, the number of terms selected is shown by a vertical line in each figure.If the selected calibration equations are now carried over to the validation set, the root mean square errors ofANALYST, JUNE 1992, VOL. 117 95 1 Table 1 Performance of the windows method relative to principal components regression (PCR) with use of a centred log transformation (CLG) and the transformation of Aucott et al.5 (SDL). All values are in percentages Data set Method hays crude protein PCR, CLG PCR, SDL Windows Australian Australian hays NDF PCR, CLG PCR, SDL Windows Australian hays IVD PCR, CLG PCR, SDL Windows EEC silages crude protein PCR, CLG PCR, SDL Windows EEC silages IVD PCR, CLG PCR, SDL Windows Norwegian grasses crude protein PCR. CLG PCR.SDL Windows Norwegian grasses NDF PCR, CLG PCR, SDL Windows Mean crude protein PCR, CLG PCR, SDL Windows Mean NDF PCR, CLG PCR, SDL Windows Mean IVD PCR. CLG PCR, SDL Windows * Root mean square error. Calibration set Validation set 1 Validation set 2 Number of terms 7 2 4 3 4 5 4 2 2 7 9 8 3 2t 2 7 6 5 7 4 6 7.0 5.7 5.7 5 .O 4.0 5.5 3.5 2.0 2.0 RMSE* 0.44 0.56 0.47 2.10 1.84 1.75 2.56 2.77 2.17 0.65 0.55 0.45 3.36 4.33 3.34 1.15 1.14 0.92 1.56 1.97 1.46 0.75 0.75 0.61 1.83 2.06 1.53 2.96 3.55 2.76 RMSE 0.91 1.10 1.04 2.92 2.21 2.40 4.00 3.96 2.79 0.98 0.84 0.85 3.28 4.84 3.37 1.18 1.48 1.41 1.83 1.80 2.02 0.85 1.06 1 .oo 2.38 2.00 2.21 3.70 4.47 3.25 Bias RMSE Bias -0.13 0.13 0.49 - - 0.37 0.98 0.05 0.49 0.67 0.80 0.32 0.18 0.25 0.78 -0.51 0.55 0.19 0.37 0.27 -0.39 -0.17 -0.14 0.77 0.80 0.72 3.82 4.60 3.60 0.34 0.12 0.04 2.46 2.07 2.47 0.18 0.20 0.26 -0.01 -0.02 0.18 1.24 0.74 1.27 - - - - - - t Stepwise regression selected no terms in this instance, by increasing the size of the test slightly above 1% , two terms were selected and the tabulated results obtained.prediction (qp) are as shown in the right-hand side of Figs. 2-4. Table 1 shows that principal components regression, using the SDL transformation, required marginally fewer terms, on average, than the windows method, but its average root mean square error was larger. Principal components regression, using the CLG transformation, required more terms, on average, than the windows method, and its average root mean square error was also larger. The windows method yielded higher bias, on average, indicating that, for these data, it is more precise but less accurate (Le., more biased) than the methods based on principal components regression.The figures indicate that the windows method generally yields better predictions of the wet-chemistry results ( i . e . , smaller root mean square errors) if a small number of terms is used. Over-fitting leads to poor performance of the windows method for predicting IVD (see Figs. 2 and 3), but provided a stopping rule is used that guards against over-fitting, the windows method seems to offer precise predicted composi- tion, on average, relative to methods based on principal components regression. The performance of the windows method is particularly encouraging as the windows were identified from just one of the data sets, indicating that windows d o not need to be identified from scratch for new data sets.Further details of the method are given by Aucott.9 Discussion Information on a given constituent is known to be concen- trated in certain regions of the NIR spectra. The windowing method identifies these regions, creating nodes at their bounds, and extracts the relevant information in a parsimoni- ous way. It attempts to reduce the data required to a minimum, while retaining the full predictive potential of the spectra. When data are recorded from different instruments or over long periods of time, some form of standardization is advisable. For example, suppose two stable samples of known composition are available, one high in the constituent of interest and one low.Readings can then be standardized over time or across instruments by first scanning these test samples whenever new samples are to be analysed. Suppose at the time of calibration, the two test samples yield for each window z values of zh, for the sample with a high level of the constituent and zli for the sample with a low level, j = h,,-.-,hh.952 Subsequently, during analyses of new samples, possibly on a different instrument, these test samples yield z * ~ and z*hj. Let a single new sample of interest have corresponding z values of z*;. In order to make the new sample comparable to those of the calibration set, the z* values might be transformed to standardized z values by using Z; = ZI; + b(z*; - z*O) where ( z h j - ZQ) (z*hj - z*l;) b = This standardization would be carried out between steps (2) and (3) defined under Analysing New Samples. We are grateful to the researchers who kindly supplied the data sets used in these examples: Peter Flinn of the Pastoral Research Institute, Hamilton, Victoria (Australian hays), Dr. Peter Marum, Loken, Norway (Norwegian forages) and Dr. Christian Paul, FAL Braunschweig, Germany (EEC silages). ANALYST, JUNE 1992, VOL. 117 We also thank the referees for their constructive comments on the method. References Norris, K. H., Barnes, R. F., Moore, J. E.. and Shenk, J . S., J. Anim. Sci., 1976. 43, 889. Murray, I . , and Hall, P. A., Anal. Proc ... 1983, 20, 75. Wetherill, G. Z., and Murray, I.. J. Agric. Sci., Camb.. 1987, 109, 539. Cowe. I. A.. and McNicol. J. W., Appl. Spectrosc.. 1985, 39. 257. Aucott, L. S., Garthwaite. P. H., and Buckland, S. T.. Analyst, 1988, 113, 1849. Flinn. P. C., M.Sc. Thesis, University of Aberdeen, 1990. Wetherill. G. Z . , in Forage Quality Analysis by NIR Reflectance Spectroscopy, ed. Paul, C . , CEC's Directorate General for Agriculture, Brussels, 1987, pp. 20-47. Marum, P., personal communication, 1987. Aucott. L. S., Ph.D. Thesis, University of Aberdeen, 1990. Paper 1102746H Received June 10, 1991 Accepted January 27, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700947

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 953 Cholesterol Immobilization via Ether-linked Sepharose Gels Rajiv K. Satsangi" and Glen E. Mottt Department of Pathology, University of Texas Health Science Center, San Antonio, TX 78284, USA Cholesterol has been immobilized on Sepharose-6B via oxyether linkages to the 3- or 25-position. The 3- or 25-hydroxysterol methanesulfonates were coupled with epoxy-Sepharose-6B at 80 "C for 24 h. Approximately 2% of the ligand was incorporated into the gel. These types of affinity columns may be useful in purifying proteins that specifically bind or metabolize cholesterol. Keywords: Cholesterol linked affinity gels; ~-tert-butyldimethylsilyloxycholest-5-en-25-one; 3-tert- butyldimeth ylsilyloxycholest-5-en-25-ol; 3fLtert- butyldimeth ylsilyloxycholest-5-en-25-yl methanesulfonate Affinity chromatography is an effective method for exploiting the biological specificity of protein-protein and enzyme- substrate interactions for the purification of either a protein or its ligand.Recent reports have described the coupling of androgens' and estrogens2 through thioether linkages to Sepharose gels, for purification of their respective receptors. This paper describes the immobilization of cholesterol on Sepharose via oxyether linkages to ring A or to the side-chain. Experimental Materials Epoxy-activated Sepharose-6B was purchased from Phar- macia LKB Biotechnology (Piscataway , NJ, USA). Choles- terol, sodium tetrahydroborate(iI1) and ethylenediamine- tetraacetic acid (EDTA) were purchased from Sigma (St. Louis, MO, USA). [4-"T]-Cholesterol was acquired from Amersham (Arlington Heights, IL, USA). Methanesulfonyl chloride, toluenesulfonyl chloride, 2-mercaptoethanol, y-pi- coline (4-methylpyridine), 2-aminoethanol and boron tri- fluoride-diethyl ether were purchased from Aldrich (Mil- waukee, WI, USA).Tetrahydrofuran (THF) and l ,4-dioxane were dried by heating under reflux over sodium pellets, followed by distillation under a nitrogen atmosphere. Com- mercial y-picoline was redistilled before use. Formamide was purified by passing it through a column of mixed-bed resin and was stored over activated molecular sieves (3 A). Dry silylation-grade solvents: acetonitrile, N-N-dimethylform- amide (DMF) , pyridine and N-(tert-buty1dimethylsilyl)-N- methyltrifluoroacetamide were obtained from Pierce (Rock- ford, IL, USA).De-ionized water was de-oxygenated by bubbling nitrogen through it for 30 min. All other solvents used were of analytical-reagent grade unless specified other- wise. Tritium-labelled sodium tetrahydroborate(iI1) was acquired from ICN Radiochemicals (Irvine, CA, USA). 3~-Hydroxycholest-5-en-25-one was purchased from Steral- oids (Wilton, NH, USA). Tetrabutylammonium fluoride- THF reagent was obtained from Applied Science (State College, PA, USA). Thin-layer chromatographic (TLC) analyses were performed on 250 pm pre-coated silica gel G plates ( 5 x 20 cm), obtained from Analtech (Newark, DE, USA). The chromatogram was developed with hexane- diethyl ether (80 + 20) and the spots were visualized by charring with 50% v/v H2S04.Instrumentation All melting-points (m.p.) were taken with a Hoover-Unimelt capillary melting-point apparatus (Thomas Scientific, Swedes- * Present address: Department of Genetics and Immunology, Sanjay Gandhi Post-Graduate Institute of Medical Sciences, P.O. Box 375, Lucknow-226 001, India. t To whom correspondence should be addressed. boro, NJ, USA) and are uncorrected. Infrared (IR) spectra were measured with a Perkin-Elmer 283 B infrared spectro- photometer (Perkin-Elmer, Norwalk, CT, USA), using poly- styrene as a standard. The compounds were dissolved in CCI4 and IR peaks for CCI4 were subtracted from the spectrum of each sample. Proton and 13C nuclear magnetic resonance (NMR) spectra were recorded on a Jeol FX-90Q Fourier transform NMR spectrometer (Jeol, Palo Alto, CA, USA) at 90 and 22.5 MHz magnetic frequencies, respectively.All of the NMR spectra were recorded with reference to CHC13, which was found to resonate at aH 7.25 at 90 MHz (1H NMR) and at aC 77.00 at 22.5 MHz (13C NMR) relative to tetramethylsilane as internal reference. Only salient reso- nances which prove the identity of the respective compound are reported. Gas chromatographic analyses were performed on a Hewlett-Packard 5710 gas chromatograph (Hewlett- Packard, Avondale, PA, USA) with a packed column of 3% OV-17 at an oven temperature of 270 "C and a nitrogen carrier gas flow rate of 30 cm3 min-1. Radiolabelled compounds were monitored by silica gel TLC, by scraping 1 cm bands of the gel into scintillation vials. Radioactivity was assessed by liquid scintillation spectrometry on an LKB-Wallac-1219 Rackbeta liquid scintillation counter (LKB, Gaithersburg, MD, USA).Procedures Swelling and washing of epoxy-activated Sepharose-6B The gel was prepared for coupling by the method of Andersson et al.3 [4-14C]-Cholest-5-en-3-y1 methanesulfonate [4-'4C]-Cholest-5-en-3-y1 methanesulfonate was synthesized by the method of Helferich and Gunther,4 using [4-14C]- cholesterol in place of unlabelled cholesterol. Coupling of epoxy-activated Sepharose-6B with 2-mercapto- ethanol (Fig. 1) Epoxy-activated Sepharose-6B (250 mg) was suspended in buffer (0.3 mol dm-3 sodium hydrogen carbonate and 1.1 mmol dm-3 EDTA; pH adjusted to 11.0 with 1 mol dm-3 NaOH). 2-Mercaptoethanol (50 mm3) was added to the suspension and the mixture was shaken at 45 "C for 18 h.The coupling mixture was filtered, washed with water (30 cm3) and acetone (30 cm3) and air-dried to produce P-hydroxyethylthio- activated Sepharose-6B ( Gel-SCH2CH20H). Coupling of [4-"C]-cholest-5-en-3-yl methanesulfonate with hydroxyethylthio-activated Sepharose-6B (Fig. 1 , reaction 2) The Gel-SCH2CH20H (40 mg of dry gel) was suspended in DMF (3 cm3) and [4-14C]-cholest-5-en-3-yl methanesulfonate (10 mg; specific activity 18041 counts min-1 pmol-1) dissol- ved in DMF (500 mm3) was added. The reaction mixture was954 ANALYST, JUNE 1992, VOL. 117 1 H - Fig. 1 Synthetic pathway for coupling the 3-hydroxy group of cholesterol to epoxy-activated Sepharose gently shaken at 75°C for 20 h and then filtered and the residue was washed with 1,4-dioxane (15 cm3), water (30 cm3), 1,4-dioxane (15 cm3) and finally with acetone (15 cm3) and air-dried.The residual radioactivity bound to the gel, measured by liquid scintillation spectrometry, was 1.7% of the radioactivity or 13 pmol g-1 of dry gel. Coupling of [4-"T]-cholest-5-en-3-yl methanesulfonate with epoxy-activated Sepharose-6B (Fig. 1, reaction 1) [4-"C]-Cholest-5-en-3-yl methanesulfonate (9.5 mg, 20 pmol, 18041 counts min-1 pmol-1) was reacted with washed epoxy-activated Sepharose-6B (40 mg) in DMF under the reaction conditions described above. Only 0.7% of the radioactivity was incorporated into the gel. 3~-tert-Butyldimethylsilyloxycholest-5-en-25-one I1 (Fig. 2) 3P-Hydroxycholest-5-en-25-one I (15.4 mg, 40 pmol) was dissolved in silylation-grade DMF (200 mm3) in a 5 cm3 screw-cap Reactivial which was immediately flushed through with nitrogen.N-(tert-ButyldimethyIsi1yl)-N-methyltrifluoro- acetamide (100 mm3, excess) was added and then the tube was purged with nitrogen and the reaction mixture was incubated at 80 "C for 2 h. The liquid was evaporated under a stream of nitrogen and the residue obtained was partitioned between 2% d v K2C03 (500 mm3) and diethyl ether (500 mm3). The aqueous layer was extracted a second time with diethyl ether (500 m m . The combined ethereal extract was washed twice with water (2 X 500 mm3), and dried under a stream of nitrogen and then under vacuum overnight to obtain 19 mg (95% yield) of compound 11. The compound was sufficiently pure to allow further reaction as analysed by gas-liquid chromatography (GLC) and TLC.A portion of the compound was recrystallized by dissolving it in dichloromethane (30 mm3) and adding methanol (270 mm3) slowly down the side of the tube so that the separate layers initially formed were not allowed to mix. The mixture was stored in a refrigerator overnight to obtain crystals of compound 11; m.p. 120-122 "C, RF = 0.68, GLC (retention time 6.65 min); m/z 443 (M - But)+; Y,,, (CCI4)/cm-1 1720 (s, C=O), 1690 (C=C), 870,890 and 940 (Si-0-C); 6~ (CDCI3) 0.03 [s, 6 H, (CH3)2SiO], 0.87 [s, 9 H, (CH3)3CSiO], 2.11 (s, 3 H, terminal CH3COCH2), 2.34 (distorted t, 2 H, CH2COCH3), 3.44 (m, 1 H, SiOCH) and 5.28 (m, 1 H, GCH); aC (CDCI3) -4.4 [(H3C)2Si], 26.02 [(CH3)3CSi], 29.65 (terminal CH2COCH3), 42.54 (quaternary CSi), 44.33 (terminal CH2COCH3), 72.77 [CH(OTBDMS)], 121.09 (C=CH) and 141.79 (C=O).313-tert- Butyldimethylsilyloxycholest-5-en-25-ol I11 (Fig. 2) Compound I1 (17.5 mg, 35 pmol) was treated with NaBH4 (25 mg, excess) in ethanol (1 cm3). The reaction mixture was stirred with a magnetic stirrer at room temperature for 3 h. Then, acetone (1 cm3, excess) was added to the reaction mixture which was stirred at room temperature for 30 min to destroy the excess of NaBH4. The solvents were removed under an atmosphere of nitrogen. The residue was partitioned between water (1 cm3) and chloroform (1 cm3). The aqueous layer was extracted with chloroform (1 cm3) and the combined chloroform extracts were washed twice with 2% ammonium sulfate (2 x 1 cm3) and finally with water (1 cm3).The chloroform layer was dried (Na2S04). The solvent was evaporated under a stream of nitrogen to obtain 7 mg of compound I11 (96.7% of the theoretical yield). The compound was sufficiently pure to be used for the next reaction as analysed by GLC and TLC. A portion was recrystallized from light petroleum (b.p. 3040°C) at 0°C; m.p. 162-163°C; RF 0.25, GLC (retention time 6.18 min); m/z 445 (M - Bur)+; Y,,, (CCI4)/cm-' 3630, 1370, 1090 (s, secondary OH), 1670 (C=C, trisubstituted), 870,890 and 940 (Si-0-C); aH (CDCI3) 0.03 [s, 6 H, (CH3)2SiO], 0.87 [s, 9 H, (CH3)3CSiO], 3.47 (m, 1 H, SiOCH), 3.74 [m, 1 H, terminal CH2CH(OH)CH3] and 5.28 (m, 1 H, C=CH); 6~ (CDC13) -4.4 [(CH3)2Si], 23.52 [terminal CH2CH(OH)CH3], 25.97 [(CH3)3CSi], 36.10 [ter- minal CH2CH(OH)CH3], 42.54 (quaternary CSi), 68.27 [terminal CH2CH(OH)CH3], 72.77 [CH(OSi)] and 121.09 (C=CH) .[25-3H]-3P-tert- Butyldimethylsilyloxycholest-5-en-25-ol [25-~H]-3~-tert-Butyldimethylsilyloxycholest-5-en-25-01 was prepared by treating compound I1 (1.7 mg, 3.5 pmol), dissolved in 1,4-dioxane (250 mm3), with NaB3H4 (1.85 x 108 Bq, specific activity 5.18 X 107 Bq pmol-1) suspended in ethanol (250 mm3). The reaction was performed in a closed system. Two traps in series, with excess amounts of unsatu- rated compounds, were connected to the outlet of the reaction vessel, so that most of the excess of tritium gas evolved was trapped. The reaction mixture was stirred at room tempera- ture for 2 h. In order to consume unchanged compound 11, an excess of NaBH4 (approximately 2 mg) in ethanol (200 mm3) was added, and the reaction mixture was again stirred at room temperature for 30 min.The final work-up procedure was the same as that described above for unlabelled compound 111. The purity and authenticity of the tritium-labelled analogue of compound I11 was checked by comparison with unlabelled compound I11 by TLC. The specific activity of the product was 8.03 x 106 Bq pmol-1.ANALYST, JUNE 1992, VOL. 117 955 t v 1 (Bug4N+F-/DMF epharose Fig. 2 Synthetic pathway for coupling cholesterol at the 25-position via an oxyether linkage to Sepharose 3~-tert-Butyldimethylsilyloxycholest-5-en-25-y1 methanesul- fonate IV (Fig. 2) Compound I11 (15 mg, 30 pmol) was placed in a 5 cm3 Reactivial with a conical magnetic bar. Dry pyridine (500 mm3) was added, the vial was sealed with a Teflon-lined screw cap and cooled in ice for 15 min.Methanesulfonyl chloride (20 mm3, excess) was added in the cold, and the reaction mixture was stirred magnetically at 0-5°C for 1 h and then at room temperature for 5 h. The reaction mixture was evaporated under a stream of nitrogen. The residue was dissolved in diethyl ether (1 cm3) and extracted with 1% m/v K2CO3 (500 mm3) (prepared with oxygen-free water). The aqueous phase was again washed with diethyl ether (500 mm3). The combined ethereal phase was washed twice with oxygen-free water (500 mm3 each) and then dried (Na2S04). The solvent was dried under a stream of nitrogen. The solid residue remaining was recrystallized from light petroleum (b.p. 30-60 "C) at 0 "C overnight to obtain 16 mg (92% of theoretical yield) of compound IV; m.p.144-145 "C; RF 0.38; GLC (retention time 3.22 min); mlz 523 (M - But)+, 427 [M - (Buf + CH3S03H)]+ and 353 [M - (TBDMS + CH,SO,H)]+; Y,956 ANALYST, JUNE 1992, VOL. 117 (CCI4)/cm-1 1670 (C=C, trisubstituted), 1375, 1350, 1180 (ROS020CH3), 870,890 and 920 (SiOC); aH (CDC13) 0.04 [s, 6 H , (CH3)2SiO], 0.87 [s, 9 H , (CH3)3CSiO], 2.98 (s, 3 H , S02CH3), 3.47 (m, 1 H , SiOCH), 4.80 [m, 1 H , terminal CH2CH(OS02CH3)CH3] and 5.29 (m, 1 H , C=CH), aC (CDC13) -4.4 [(CH3)2Si], 24.39 [terminal CH2CH- (OS02CH3)CH3], 25.96 [(CH3)3CSi], 36.80 [terminal CH2CH(OS02CH3)CH3], 42.54 (quaternary CSi), 60.80 (S020CH3), 72.77 [CH(OSi)], 80.19 [terminal CH2CH- ( OS02CH3)CH3] and 121.10 (C=CH).[ 25-3H1-3P- tert- Butyldimethylsilyloxycholest-5-en-25-yl methanesulfonate The tritium-labelled analogue of compound IV was prepared by treating [25-3H]-3~-tert-butyldimethylsilyloxycholest-5- en-25-01 (20 mg, 0.04 mmol, 33 000 disintegrations min-1 pmol-1) with methanesulfonyl chloride (50 mm3) by the same procedure as described above for unlabelled compound IV. The yield of tritium-labelled product IV was 20 mg (88%). The purity and authenticity of this product was checked by a comparison with unlabelled compound IV by TLC. Coupling of [25-3H]-3(3-tert-butyldimethylsilyloxycholest-5- en-25-yl methanesulfonate with epoxy-activated Sepharose-6B V (Fig. 2) [25-3H]-3~-terf-B~tyldimethyl~ilyloxycholest-5-en-25yl methanesulfonate IV (26 mg, 45 pmol; 33 000 disintegrations min-l pmol-l) dissolved in diethyl ether was dried under a stream of nitrogen and stored under vacuum overnight in a 50 cm3 standard-jointed Erlenmeyer flask.Washed and dried epoxy-activated Sepharose-6B (210 mg, approximately 45 pequiv of epoxy sites) and dry pyridine (20 cm3) were added to the flask followed by powdered KOH (50 mg). The atmos- phere in the flask was purged with nitrogen for 5 min. The flask was sealed with a standard-joint stopper and the reaction mixture was incubated at 80 "C with continuous gentle shaking for 24 h. The resulting mixture was filtered and the solid gel was washed by slow dripping filtration with 1,4-dioxane (20 cm3), water (30 cm3) and then with 1,4-dioxane (20 cm3). Finally, the gel was washed with acetone (10 cm3) and dried under reduced pressure.Based on the measurement of residual radioactivity bound to the gel, 2.1% of compound IV was coupled with the gel. Subsequently, epoxy-activated Sepharose-6B (3 .O g) was coupled with unlabelled compound IV (368 mg), using powdered KOH (700 mg) in pyridine (280 cm3). The reaction was performed under the same conditions as described above. Removal of tert-butyldimethylsilyloxy (TBDMS) group from compound I1 Compound I1 (2 mg) was added to a screw-cap tube with DMF (1 cm3). The atmosphere above the reaction mixture was purged with nitrogen ; te trabu t ylammonium fluoride-THF reagent (100 mm3; excess) was added and the mixture was stirred at room temperature for 8 h. The solvent was removed under a stream of nitrogen at room temperature and the residue was partitioned between hexane (1 cm3) and water (1 cm3).The product in the hexane phase was found to be homogeneous and co-migrated with standard 3P-hydroxy- cholest-5-en-25-one I when analysed by TLC. Removal of TBDMS group from the coupled Sepharose gel VI (Fig. 2) The Sepharose gel (3.0 g), coupled with the 38-tert-butyl- dimethylsilyloxycholest-5-en-25-oxy ligand, was suspended in a 250 cm3 Erlenmeyer flask, and covered by an air-tight rubber septum. N,N-Dimethylformamide (65 cm3) was injected into the flask through a syringe and the gel was allowed to swell for 2 h. Subsequently, tetrabutylammonium fluoride-THF reagent (2 cm3) was injected into the flask and the reaction mixture was incubated at room temperature for 8 h. The mixture was filtered, washed with water (500 cm3) and air-dried.Blocking of excess of active epoxy groups of Sepharose gel The Sepharose gel, coupled with 3P-hydroxycholest-5-en-25- oxy ligand, was suspended in 1 mol dm3 aqueous 2-amino- ethanol and incubated at 40°C overnight. The mixture was filtered and the residue was washed with water (500 cm3) until the pH of the filtrate was neutral. Finally, the gel was washed with acetone (100 cm3) and air-dried. The gel can be stored at 4°C in a 0.05% m/v aqueous NaN3 suspension. Results and Discussion Before attempting the coupling reaction between epoxy- activated Sepharose-6B and 3~-tert-butyldimethylsilyloxy- cholest-5-en-25-olII1, a model study was performed, based on the work of Anderson et a1.3 These workers reported the immobilization of polyhydroxy ligands on epoxy-activated Sepharose-6B, using various catalysts, and solvents of differ- ent polarities.As a model study, an attempt was made to couple cholesterol with epoxy-activated Sepharose-6B. The reaction medium was one of the major limitations. Some of the solvents suggested such as dimethyl sulfoxide (DMSO), formamide and water, which effectively swell the gel, and have been used with Lewis acid catalysts (boron trifluoride, zinc chloride and zinc perchlorate), were not good solvents for solubilization of cholesterol and other steroids. Hence, coupling could not be achieved. Various combinations of reaction conditions, using DMF, y-picoline, pyridine or 1,4-dioxane as solvent; and zinc chloride, boron trifluoride- diethyl ether, NaOH or 4-N,N-dimethylaminopyridine (DMAP) as catalyst;5 and varying the reaction temperature between 25 and 80°C, also did not promote coupling of the model compound.Direct coupling between the epoxy groups of this gel and cholest-5-enyl chloride in pyridine was also attempted and was also found to be unsuccessful. A sugges- ted5 use of toluene-p-sulfonic acid as a catalyst for the coupling reactions of alcohols with epoxy groups could not bring about the required coupling because the epoxy-activated Sepharose- 6B completely decomposed into a brown, jelly-like semi-solid material with toluene-p-sulfonic acid. The activation of the 3P-hydroxy group of cholesterol by forming its sodium salt with sodium hydride in dry THF and treating this sodium salt with epoxy-activated gel in solvents such as THF, y-picoline, pyridine, 1,4-dioxane and mixed solvents at various tempera- tures promoted coupling of up to only 0.23% of the ligand.Possible steric interference between the gel matrix and the 3(3-hydroxycholest-5-enyl group could be eliminated by extending the spacer arm at the distal end of the gel matrix. Because formation of a thioether linkage is favoured ther- modynamically compared with producing an ether linkage, epoxy-activated Sepharose-6B was treated with 2-mercaptoe- thanol under basic conditions similar to those described by Ray et al.1 Hence, the epoxy ring was opened to form a P-hydroxythioether as a spacer arm for the gel. The hydroxy group at the distal end of the new extended spacer arm was finally coupled with the cholest-5-enyl radical by an ether linkage.In order to achieve this goal and to determine the optimum reaction conditions, cholest-5-enyl methanesulfo- nate and cholest-5-enyl toluene-p-sulfonate were treated with epoxy-activated Sepharose-6B and its 2-mercaptoethanol derivative under different conditions. It was found that cholest-5-enyl methanesulfonate gave better coupling (1 .O- 1.7%) than did cholest-5-enyl toluene-p-sulfonate (0.6-1.37% coupling of ligand). The amount of coupling of the ligand to mercaptoethanol-modified Sepharose gel was better (1.7%) than that obtained by epoxy-activated Sepharose gel (0.7%) under the same conditions. The same reaction in basic medium did not improve the yield. The temperature ofANALYST, JUNE 1992, VOL.117 957 reaction was also found to be a limiting factor for the extent of coupling. For temperatures below 45 "C for 24 h, no coupling could be detected; however, incubation at 75 "C for 20 h gave the optimum yield of coupled product. An increase in reaction temperature to 90°C led to a decrease in the amount of coupling and at 120 "C (oil-bath) the gel was decomposed to a black residue. Based on swelling of the gel, solubility of ligand and the yield of coupled product, DMF was found to be the best reaction medium tested. In the absence of a chemical or physico-chemical analytical method to monitor the reaction and to determine the extent of coupling of ligands to gels of this type, the extent of binding of the ligand to the gel was determined using radioactive labelling of the ligand. The synthesis of the affinity ligand of 3~-hydroxycholest-5- en-25-01 started with the commercially available 3P-hydroxy- cholest-5-en-25-one I, its 3-hydroxy group being protected as a tert-butyldimethylsilyloxy (TBDMS) derivative I1 by treating it with N-tert-bu t y ldime thylsil yl-N-me thy1 trifluoroacetamide .The procedure of Mawhinney and Madsons was adopted with some variations. Hence, compound I1 was obtained as a pure homogeneous compound in almost quantitative yield. Nearly quantitative reduction of the carbonyl group at position-25 of compound I1 was achieved by its reaction with NaBH4. The amount of ligand coupled to the gel matrix in subsequent steps was monitored by tritiating the ligand molecule at this step. Reduction of compound I1 with NaB3H4 generated the 3H-labelled analogue of compound 111.Finally, compound 111 was converted into its methanesulfonate IV. The results of the coupling reaction between Gel- SCH2CHZOH and compound IV were surprising. Although cholest-5-enyl methanesulfonate reacted with Gel- SCH2CH20H to an extent of 1.7%, 3P-tert-butyldimethylsilyl- oxycholest-5-en-25-yl methanesulfonate IV did not react with Gel-SCH2CH20H for unknown reasons. However, a direct reaction between compound IV and epoxy-activated Sepha- rose-6B in pyridine in the presence of powdered KOH at 80 "C resulted in 2.1% incorporation of ligand as ascertained by measurement of the radioactivity incorporated into the gel. Removal of the TBDMS group from 3P-tert-butyldimethyl- silyloxycholest-5-en-25-one I1 to obtain 3P-hydroxycholest-5- en-25-one I was achieved by deprotection with tetrabutyl- ammonium fluoride-THF reagent with some variations from the standard procedure.' The same conditions were used for the removal of the TBDMS group from the Sepharose gel, coupled with the 3~-tert-butyldimethylsilyloxycholest-5-en-25- oxy ligand. Finally, the excess of active epoxy groups was blocked with 2-aminoethanol as recommended by the manu- facturer.8 The technology described in this paper should have general application to coupling of non-polar compounds to epoxy- activated gel matrices for use in protein purification.Choles- terol-Sepharose beads may be useful for isolation of proteins that specifically bind or metabolize cholesterol or its deriva- tives. We thank Dr. P. N. Rao of the Southwest Foundation for Biomedical Research, San Antonio, TX, and Dr. James T. Slama, University of Texas Health Science Center at San Antonio (UTHSCSA) for discussions and advice during the course of this work. We also thank Dr. Susan T. Weintraub, UTHSCSA, for performing the mass spectral analyses. This work was supported by grant HL-28972 from the National Heart, Lung, and Blood Institute. References Ray, S., Salman, M., Ruiz, A. A., Stotter, P. L., and Chamness, G. C., J. Steroid Biochem., 1986. 24, 1111. Greene, G. L., and Jensen, E. V., J. Steroid Biochem., 1982, 16, 353. Andersson. K., Bywater, R., Cmogorcevic, G., McKenzie, R., and Ottosson, T., React. Polym. Ion Exch. Sorbents, 1983, 1, 273. Helferich, B., and Gunther, E., Ber. Dtsch. Chem. Ges., 1939, 72B, 338. Sexton, A. R., and Britton, E. C., J. Am. Chem. SOC., 1948,70, 3606. Mawhinney, T. P., and Madson, M. A., J. Org. Chem., 1982, 47, 3336. Corey, E. J., and Venkateswarlu, A., J. Am. Chem. SOC., 1972, 94,6190. Affinity Chromatography, Principles and Methods, Pharmacia LKB Biotechnology, Ljungforetagen, A.B., Uppsala, Sweden, 1988, p. 30. Paper 1105392 B Received October 23, 1991 Accepted January 21, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700953

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 959 Determination of Blood Lead in Dried Blood-spot Specimens by Zeeman-effect Background Corrected Atomic Absorption Spectrometry Stephen T. Wang and Helen P. Demshar Chemistry Section, Laboratory Service Branch, Ontario Ministry of Health, 81 Resources Road, Etobicoke, Ontario, Canada M9P 3TI A simple procedure for the determination of blood-lead levels in dried blood-spot filter-paper specimens is described. A $ inch dried blood spot was analysed by a method involving extraction of the lead into 1.25% (NH4)2HP04-0.5% Triton X-100 solution. A matrix-based calibration was used for analysis, with use of a Zeeman-effect background corrected atomic absorption spectrometer. The within-run precision (% relative standard deviation) of the method at the low end of the analytical range was 19% at 0.36 pmol dm-3 and 14% at 0.60 pmol dm-3.The accuracy of the method was verified by recovery tests and by comparison with a routine whole-blood method. The mean blood lead level of 425 Toronto newborns was 0.19 pmol dm-3 (range 04.75 pmol dm-3). Keywords: Blood lead; newborn dried blood spot; Zeeman-effect background corrected atomic absorption spectrometry Lead poisoning in children has been recognized for several years.'-3 Recent studies have shown that lead exposures as low as 0.50 pmol dm-3 (10 pg per 100 cm3) can be toxic to newborn babies and can impair their brain development permanently.4-8 Hence, a sensitive analytical system with precise and accurate blood-lead measurements is imperative at the lower cut-off point of lead exposure.Our involvement in the blood-lead screening of Ontario children in 1984 and 1987 enabled us to establish a simple analytical method, with a microsampling technique, for determining the concentration of blood lead by Zeeman-effect background corrected atomic absorption spectrometry for lead poisoning in children.9-10 This method of sampling is, however, not suitable for newborns because of the difficulty in obtaining the blood specimen from an infant. As a result, a dried filter-paper blood-spot specimen has been established as the means of sample collection for newborn screening," e.g., for phenyl- ketonuria and hypothyroidism. It is also possible to use dried blood-spot specimens as the sampling technique for the determination of lead in the blood of newborns.Indeed, the dried blood-spot specimen has been used for the determination of lead in blood by the Delves cup technique. 12-14 The technique of flame atomic absorption spectrometry is not considered to be sufficiently sensitive at the lower level of lead exposure. The technique of electrother- mal atomic absorption spectrometry (ETAAS) with a L'vov platform has been established as one of the most sensitive methods for trace element analysis over the last decade. Therefore, it is a logical choice to use this sensitive ETAAS method in undertaking the investigation of blood-lead deter- mination for newborns, using the dried blood-spot specimens. This paper describes the investigation of an analytical proce- dure regarding precision, accuracy and recovery studies for the determination of blood-lead contents in dried blood-spot specimens. Experimental Apparatus and Material A Hitachi (Tokyo, Japan) Model 180-80 Zeeman-effect background corrected electrothermal atomic absorption spec- trometer, equipped with a Hitachi data-processing unit (Model 180-0205), an automatic sampler (Model 180-0451) and a pyrolytic graphite coated graphite tube and pyrolytic graphite platform from Ringsdorff-Werke (Bonn-Bad Godes- berg, Germany) were used.A hollow cathode lamp (Hama- matsu Photonics K.K., Shizuoka-Ken, Japan) for lead was used at a working current of 5 mA, with a wavelength of 283.3 nm and a bandwidth of 1.3 nm. Triton X-100 was purchased from Eastman-Kodak (Rochester, NY, USA). High-purity distilled, de-ionized water with a specific resistance greater than 2 x 106 Q was used throughout.The other reagents used have been described previously. 15-16 Blood standards M1 and M2 from the Behring Institute (Marburg, Germany) and human whole-blood controls from Bio-Rad (Anaheim, CA, USA) were used as blood standards and quality controls. Newborn dried blood-spot specimens on filter-paper were obtained from the Neonatal Screening Section, Ontario Ministry of Health. Blood-lead specimens from the College of American Pathologists and the Quebec Centre of Toxicology17 were also used to verify the analytical performance. Preparation of Whole Blood on Dried Filter-paper A 50 mm3 aliquot of the blood specimen was applied slowly to the centre of the filter-paper (Schleicher & Schuell #903 No.W872; Keene, NH, USA) of our neonatal sampling sheet.18 Wet blood spots were not allowed to come into contact with any surface. The spotted sheets of filter-paper were allowed to air-dry at an ambient temperature for at least 4 h and thereafter stored in a plastic bag in a refrigerator. The quality controls and blood standards were prepared in the same way as described above. Before applying the blood standards on the filter-paper, blood standards can be prepared in-house by adding the lead standards to human whole blood containing an undetectable amount of lead to establish daily instrumental calibration graphs. However, it is difficult to obtain this human whole blood, with an undetectable amount of lead, in a sufficient amount for this purpose.Blood standards M1 and M2 from the Behring Institute were generally used to construct daily instrumental calibration graphs for this study. Analytical Procedures The analytical procedures regarding sample extraction time, furnace temperature programme, buffer and calibration graph were investigated. The blood spots of & inch were taken from the peripheral location from each spot by a & inch paper punch (McGill, Marengo, IL, USA). The blood spot was extracted960 0.10 0.08 0 0.06 2 20.04 ANALYST, JUNE 1992, VOL. 117 - - - - with 250 mm3 of a matrix buffer, consisting of 1.25% (NH4)2HP04-0.5% Triton X-100, for 30 min at room tem- perature with constant shaking in an IKA-Werk (Staufen, Germany) Vibrax-VXR vortex mixer. The solution was transferred into a polystyrene cup and placed in the auto- sampler.A 20 mm3 aliquot of the solution was delivered to a pyrolytic graphite L'vov platform via an autosampler probe. The standard dried blood spots for calibration and quality control were prepared in the same way as described above. The blood-lead content was then determined by using the optimized graphite furnace programme, as shown in Table 1. Results Calibration Graph, Sensitivity and Detection Limit The calibration graphs based on the aqueous solution and standard dried blood spot were investigated. As the aqueous solution calibration required a correction factor to provide the correct diluted concentration and to compensate for the matrix effect of the specimen, the standard blood-spot calibration was chosen to match the dried blood-spot specimen.The results obtained from the calibration graph were linear for up to 9.66 pmol dm-3. The calibration sensitivity of the method, defined as the blood-lead concentration that yielded an absorbance of 0.0044, according to the slope of the calibration graph,lY was found to be 0.10 pmol dm-3. The detection limit, defined as the concentration corresponding to three standard deviations of the concentrations in the blank solutions, was found to be 0.05 pmol dm-3. A random sample of 30 blank filter-paper spots from our neonatal sampling sheet was tested for lead. No detectable levels of lead were found in any of the blank filter-paper spots tested. The sensitivity of the method is similar to that of the routine blood-lead determination with whole blood at 10-fold dilu- tion.15,16 A & inch dried blood spot represents about 6.7 mm3 of whole blood; in 250 mm3 of extraction buffer this yields approximately a 40-fold dilution of the whole blood. The sensitivity of the method could be doubled by using two 3 inch dried spots in one tube or by reducing the volume of the extraction buffer to 125 mm3. Extraction Buffer Three extraction buffer systems, 0.1% HN03-0.1% Triton X-100, 1.25% (NH4)2HP04-0.5% Triton X-100 and 0.1% HN034.1% Triton X-lOO-0.001% palladium (as PdC12), were tested. The effect of these three matrix buffer systems on the ashing temperature at a lead level of 2.41 vmol dm-3 is shown in Fig. 1. Both matrix buffers of 1.25% (NH4)2HP04- 0.5% Triton X-100 and 0.1% HN03-0.1% Triton X-100- 0.001 YO palladium were suitable for blood-lead determina- tion.There was no loss of lead up to an ashing temperature of 800 "C. The graphite furnace programme for blood-lead determination in a dried blood-spot specimen has, therefore, been established, as shown in Table 1. The matrix buffer of 1.25% (NH4)2HP04-4.5% Triton X-100 chosen was also used in the routine blood-lead determinations.15~16 A time of 30 min at room temperature, with constant shaking in a vortex mixer, was chosen to allow complete extraction of the blood lead into the matrix buffer. Precision and Accuracy The precision established for this analytical system at various blood-lead levels, based on the pair studies, was the relative standard deviation (RSD): 45% in the range 0-0.24 pmol dm-3, with a mean value at 0.12 ymol dm-3 (n = 26); 19% in the range 0.25-4.48 ymol dm-3, with a mean value at 0.36 ymol dm-3 ( n = 11); 14% at 0.60 ymol dm-3 (n = 16); 3.7% at 1.21 ymol dm-3 ( n = 36); 3.2% at 2.60 pmol dm-3 (n = 25); and 3.2% at 3.79 pmol dm-3 (n = 13).These analytical pair precisions were comparable to the routine system at blood-lead levels of 1-4 ymol dm-3.20 The over-all precision (YO RSD) for projected toxicity levels in newborns of about 0.50 pmol dm-3 is 20% at 0.60 ymol dm-3 ( n = 32). The over-all precision of two quality controls of dried blood-spot specimens was measured throughout the entire study, with an RSD of 11.8% at 1.16 pmol dm-3 ( n = 73) and 7.7% at 2.56 pmol dm-3 ( n = 51). The same two quality controls used in the routine system with whole blood yielded RSDs of 4.7% (n = 27) and 3.8% ( n = 27), respectively. The accuracy of the method was assessed by two different procedures. In order to verify the accuracy of the method, recovery studies were carried out by determining several known concentrations of blood lead, between 0.39 and 3.70 ymol dm-3, in dried blood-spot specimens.The results yielded average recoveries of blood lead at 97 k 10% ( n = 18). A linear regression analysis comparing the blood lead obtained by the routine method ( x ) with that obtained by the dried blood-spot method (y) on 30 specimens, ranging from 0.05 to 1.70 ymol dm-3 lead, is shown in Fig. 2. The regression equation is y = 0 . 9 7 ~ + 0.04 pmol dm-3, with a correlation coefficient of 0.9801 and a mean k SD Table 1 Graphite-furnace programme for blood-lead determination TemperaturePC Carrier Stage Initial Final Time/s gas Dry 50 120 50 On Dry 120 120 20 On Ash 800 800 20 On Atomization 1800 1800 7 Off Clean 2200 2200 3 On I 0 300 500 700 900 1100 TemperaturePC Fig.1 Effect of matrix systems on ashing temperature at a lead level of 2.41 pmol dm-3. A, 0.1% HN03-0.1% Triton X-100; B. 1.25% (NH&HPO4-0.5% Triton X-100; and C, 0.1% HN03-0.1% Triton 0*02 i X-1OO-O.001% PdC12 H I 0.8 : I 0 0.4 0.8 1.2 1.6 2.0 Blood lead/pmol dm-3 Regression analysis of blood-lead levels between whole blood Fig. 2 ( x ) and dried blood spots (y)ANALYST, JUNE 1992, VOL. 117 961 0 0:l 012 0:3 0:4 0:5 0:s 017 0:8 0.9 Blood lead concentration/pmol dm-3 Fig. 3 Histogram of blood-lead levels from 425 newborns (pmol dm-3), with the x method giving 0.62 k 0.47 and the y method giving 0.63 k 0.47.Blood-lead Levels in the Sampling Population The distribution of blood-lead levels in dried blood-spot specimens from 425 newborns in the Toronto area is shown in Fig. 3. The mean and SD of this population was 0.19 k 0.13 pmol dm-3 (range 04.75 pmol dm-3). Nine of the 425 newborns had a blood-lead level greater than 0.50 pmol dm-3; this represented 2% of the population studied. Discussion Since 1972, capillary blood samples collected on lead-free filter-paper have been used for childhood lead screening, by determining lead in blood with use of the Delves cup technique.12-’4 We first established here an ETAAS system with the L’vov platform for lead screening of newborns, with dried blood-spot specimens.The platform system has a sensitivity twice that of the routine system; therefore, it is very useful for the small blood samples available from newborn specimens. Table 1 shows that it took only 100 s to analyse one sample by using either this method or the routine method.16 Therefore, 80 duplicate blood-lead analyses can be screened in 1 d by using dried blood spots from newborns. The over-all precision of blood-lead determination by the dried blood-spot method is about twice the % RSD of the routine method on whole blood; the average recovery of blood lead was about 97 k 10%. This is because of the inherent variation of blood absorption by the filter-paper. The de- scribed system can, however, be used as a screening method.The effect of lot-to-lot variability in filter-paper on the determination of thyroxine, thyrotropin and phenylalanine in dried blood specimens has shown that the maximum differ- ence in mean observed values among lots ranged from 6 to 36% in all the cited analytes.” It is suggested that venous blood or cord-blood of newborns should be used for confirma- tory or diagnostic testing, if possible, to avoid this lot-to-lot variation in filter-paper and also to ensure that there is less chance of lead contamination during the sampling procedure. The inherent property of variation in filter-paper was also shown in the mean blood-lead value of the 425 newborns in this study. The mean blood-lead value of 0.19 pmol dm-3 in the population showed a slightly higher value than the 0.08 pmol dm-3 in 95 Toronto infants when cord-blood specimens were used.22 Other factors, such as external contamination during specimen collection, storage and handling, could have contributed to this slightly higher value.However, the above-mentioned factors should not affect the over-all results, as the best precision RSD at a blood-lead level o€ 0.12 pmol dm-3 is 45%. The sampling population of 425 Toronto newborns showed that nine out of 425 newborns had blood-lead levels greater than 0.50 pmol dm-3 of the projected toxic level as indicated in the USA Centers for Disease Control (CDC) statement.8 The blood-lead levels of all nine newborns were in the range 0.5-0.75 pmol dm-3, which represents only 2% of the sampling population and is below the guideline set up by the CDC for individual child intervention.Therefore, it can be concluded that the Toronto newborn sampling population does not have lead poisoning according to this guideline.8 This study also confirms that there is a much lower lead exposure in Toronto newborns than in Boston newborns.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 References Lin-Fu. J. S., in Low Level Lead Exposure: The Clinical Implications of Current Research, ed. Needleman, H. L., Raven Press. New York, 1980, pp. 5-16. Agency for Toxic Substances and Disease Registry, The Nature and Extent of Lead Poisoning in Children in the United States: A report to Congress, US Department of Health and Human Services, Atlanta, GA, 1988. Tenenbein, M., Can. Med.Assoc. J . , 1990, 142, 40. Needleman, H. L., Am. J. Public Health. 1991.81, 685. Bellinger, D.. Leviton, A., Waternaux, C., Needleman, H. L., and Rainowitz, M., New Engl. J . Med., 1987,316, 1037. Needleman, H. L., Schell, A., Bellinger, D., Leviton, A., and Allred, E. N.. New Engl. J. Med., 1990,322, 83. Lee, W. R., and Moore, M. R.. Br. Med. J., 1990, 301,504. Centers for Disease Control. Preventing Lead Poisoning in Young Children: A statement by the Centers for Disease Control, Atlanta, GA, 1991. O’Heany. J., Kusiak, R., Duncan, C. E., Smith, J. F., Smith, L. F., and Spielberg, L., Sci. Total Environ., 1988, 71, 477. Wang, S. T., Pizzolato, S., and Demshar, H. P., Sci. Total Environ.. 1989, 89, 251. National Committee for Clinical Laboratory Standards, Blood Collection on Filter Paper for Neonatal Screening Programs, NCCLS Publ. LA4-A. Villanova. PA, 1988. Joselow, N. M., and Bogden, J. D.. At. Absorpt. Newsl., 1972, 11,99. Mehkeri, K. A., Romanowski, M.. and Smallbone, B., Am. Ind. Hyg. Assoc. J . , 1976. 37, 541. Verebey, K., Eng, Y. M.. Davidow. B., and Ramon, A.. J. Anal. Toxicol., 1991, 15, 237. Wang, S. T., Strunc, G., and Peter, F., in Chemical Toxicology and Clinical Chemistry of Metals. eds. Brown, S. S., and Savory, J., Academic Press, New York, 1983, pp. 57-60. Wang, S. T., Pizzolato, S., and Peter, F., Sci. Total Environ., 1988, 71, 37. Weber, J.-P., in Biological Trace Element Research, eds. Subramanian, K. S., Iyengar, G. V., and Okamoto, K., ACS Symposium Series No. 445, American Chemical Society, Washington, DC, 1991, pp. 120-128. Wang, S. T., and Demshar, H. P., Clin. Chem. ( Winston-Salem, N . C . ) , 1991, 37, 132. Skoog, D. A., in Principles of Instrumental Analysis, eds. Skoog, D. A., and West, D. M., Saunders, Philadelphia, 3rd edn.. 1985, pp. 22-23. Wang, S. T., and Peter, F., J. Anal. Toxicol., 1985, 9, 85. Slazyk, W. E., Philips. D. L., Therrell, B. L., and Hannon, H. W.. Clin. Chem. (Winston-Salem, N. C . ) , 1988, 34, 53. Koren, G.. Chan, N., Gonen, R., Klein, J., Weiner, L., Demshar, H. P., Pizzolato, S., Radde, I., and Shime, J., Can. Med. Assoc. J., 1990, 142, 1241. Paper 1 lO5285C Received October 17, 1991 Accepted January 14, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700959

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 963 Determination of Wear-metals in Used Lubricating Oils From Marine Engines by Flame Atomic Absorption Spectrometry Maria Purificacion Hernandez-Artiga and Juan Antonio Muiioz-Leyva Analytical Chemistry Department, University of Cadiz, Puerto Real, Cadiz, Spain Ramon Cozar-Sievert Nautical High School, University of Cadiz, Cadiz, Spain A method for the determination of the total content of Fe, Cu, Pb and Cr in used lubricating oils from marine engines by flame atomic absorption spectrometry is described. The influence of the sample basicity, the addition of acids, and dilution with different solvents was studied. The presence of solid species which cause serious problems concerning reproducibility was also investigated. The procedure involves the use of mineral acid to dissolve the metal particles, and a mixture of isobutyl methyl ketone and a non-ionic surfactant as solvent.The standards were prepared with aqueous inorganic salt solutions. The method is rapid and shows good sensitivity and reproducibility. A comparison is made between the results obtained from the proposed procedure and those from other atomic absorption methods. Keywords: Wear-metal determination; used lubricating oil; flame atomic absorption spectrometry The determination of wear-metals in used lubricating oils is of great interest as it allows predictions of equipment failure to be made and prompts the appropriate preventive maintenance to be undertaken. The industry requires analytical methods that can be used to carry out this determination with speed, precision and economy.Atomic spectrometry, and particu- larly flame atomic absorption spectrometry (FAAS), is one of the most useful techniques for such determinations. Methods based on the ashing of the sample and acid dissolution of the residue to allow the determination of the total metal content, and the use of inorganic salts as standards, require many manipulations, are time consuming and incur the risk of contamination or loss. Methods involving direct dilution’-3 of the oil with an organic solvent are widely used but do not allow determina- tion of the total metal content because the larger particles are not atomized by the flame and the results obtained are lower than those from the ashing method. In addition, these methods require the use of organometallic standards which are expensive and not readily available. Several mixed solvent systems have been developed that permit the determination of metals by simple dilution of the sample with use of inorganic salts as standards.4-6 Such methods are useful for metals derived from additives.However, large metallic wear-particles produced by more severe wear are not determined and several cases have been documented in which aircraft oil-wetted component failure was not predicted.’ Several workers have developed particle size independent methods’-1 1 in which inorganic acids were added to digest the metals in situ followed by dilution with a mixed solvent able to dissolve oil and inorganic acids; however, the standards used were organometallic compounds or metal powder suspensions which showed poor sensitivity and precision.The formation of stable oil-water emulsions has also been reported.”-19 This method involves the use of aqueous inorganic standards. All of the above methods have been developed for the analysis of aircraft, railway or automotive lubricating oils but no specific research has been applied to used marine engine lubricating oils, the characteristics and behaviour of which during sample preparation appear to be different. Simple dilution, and in particular acidification followed by dilution, causes the precipitation of solid species. This solid content leads to problems concerning reproducibility which have not been reported in the literature. In this paper, a method is described for the determination of the total metal content in used marine engine lubricating oil.The method avoids filtration of the sample by using a homogenizing diluent which is able to accommodate the oil sample in addition to inorganic acids and aqueous inorganic standards. The metals to be determined were chosen with the aim of being able to predict the state of the engines by means of the oil analysis; therefore, the metals determined were Fe, Cu, Pb and Cr, which are representative of marine engine wear.20.21 Experimental Instrumentation The determinations were performed with use of a Philips PU92OOX flame atomic absorption spectrometer (Philips Analytical, Eindhoven, The Netherlands). Reagents pro analysi) . Isobutyl methyl ketone (IBMK) (Panreac, Barcelona, Spain; Isopropyl alcohol (IPA) (Panreac; pro analysi) .Tergitol Type 15-S-3 (Sigma, St. Louis, MO, USA). Hydrochloric acid, 35% (Panreac; pro analysi). Nitric acid, 70% ( Panreac; pro analysi) . Cu” standard aqueous solution (1 mg dm-3 of Cu). Prepared Pb” standard aqueous solution (1 mg dm-3 of Pb). Prepared FelIL standard aqueous solution (1 mg dm-3 of Fe). Prepared CrlI1 standard aqueous solution (1 mg dm-3 of Cr). Prepared from C U ( N O ~ ) ~ . ~ H ~ O (Merck, Darmstadt, Germany; GR). from Pb(N03)2 (Merck; GR). from iron reduced (Merck; GR). from CT(NO~)~-H~O (Merck; GR). Sampling The samples were obtained with the engine running and hot, to provide a homogeneous and representative sample; they were stored in poly(propy1ene) bottles. Reference Method Validation of the method described herein was performed by comparing the results with those obtained by FAAS after sample ashing and acid dissolution of the residue.964 ANALYST, JUNE 1992, VOL.117 Procedure for Samples Shake the sample container vigorously in order to obtain a representative sample. Weigh approximately 2 g of used oil into a 50 cm3 Erlenmeyer flask. Add 1 cm3 of concentrated HCl-concentrated HN03 (6 + 1). Heat the mixture on a hot-plate for 20 min. Allow the mixture to cool to room temperature, then dilute it to 25 cm3 with IBMK-Tergitol (4 + 1). Procedure for Standards Prepare a 1000 pg 8-1 solution of Fe, Cu, Pb and Cr from iron reduced and the corresponding nitrates with doubly distilled water. Dilute the solution to 50 pg 8-1 with IPA-Tergitol(4 + 1).Weigh approximately 2 g of an un-used sample of the oil to be analysed and proceed as described under Procedure for Samples adding the appropriate volume of the 50 pg g-1 multi-element solution before diluting with IBMK-Tergitol. A blank sample is prepared by the same method but without metals in order to establish whether the un-used oil contains trace metals. Results and Discussion In order to establish the behaviour of the samples, attempts were made to apply the techniques described by other workers for different types of lubricating oil. Firstly, several oil samples (5 g) were prepared by simple dilution to 50 cm3 with IBMK and another set of samples were prepared by the same method but adding inorganic acids before diluting with IBMK. The standards were prepared in an identical manner but with the addition of metals from inorganic salts dissolved in aqueous ethanol and the same amount of un-used oil.The appearance of a fine precipitate was observed at the bottom of the container when a used oil sample was diluted with IBMK and when 1 cm3 of mineral acid was added to used or un-used oil followed by dilution. The liquid phase was not miscible and shortly after shaking manually or ultrasonically the mixture separated into two layers. The absorbance values were not reproducible, but it could be concluded that the absorbance values were higher with the addition of acid. The next step was to obtain a mixed solvent of IBMK and ethanol which was capable of dissolving the oil, a small amount of inorganic acid and the ethanolic solutions of the inorganic salts.Several IBMK-ethanol mixtures were tested but none was suitable. Other solvents were tried with the same aim, such as toluene, acetic acid and chlorobenzene but without success. Finally, it was found that a mixture of IBMK and IPA (4 + 1) was a suitable mixed solvent. A set of standards of 1, 2 and 5 pg g-1 was prepared as follows. A mixture of 5 g of un-used oil, 1.5 cm3 of HCI-HN03 (8 + 1) and the corresponding amount of the metallic salt dissolved in ethanol was diluted with IBMK-IPA (4 + 1). A precipitate appeared with the addition of acid as described above, which settled slowly to the bottom of the flask. The absorbance of this set of standards was measured by two methods: (a) vigorous shaking just before measuring; and ( b ) allowing the precipitate to settle to the bottom of the flask and analysing the clear supernatant liquid.The response for metals was higher with the shaken samples; however, the reproducibility was poor. It can, therefore, be concluded that the precipitate may contain metals and filtration is not advisable. The calibration graphs and reproducibility were acceptable when measuring without shaking for all the metals except Pb. In the following experiments samples were analysed with- out shaking and aspirating the supernatant liquid. Effects of Oil Basicity and Added Acid As it had been observed that the formation of a precipitate on addition of acid did not occur with all types of marine engine lubricating oil, the total basic number (TBN) of several oils was measured following the ASTM D-2896-80 (IP 276183) method22 in order to relate the formation of the precipitate to the TBN value of the oil.The following experiment was carried out with use of these values: 1.5 cm3 of concentrated HCI were added to several un-used oil samples with TBN values of about 6 and to some un-used oil samples with TBN values of about 30 followed by dilution with IBMK-IPA. The same experiment was repeated with 1.5 cm3 of concentrated HCl-concentrated HN03 (6 + 1). In both instances a precipitate appeared; suddenly, when the TBN of the oil was high, and very slowly, when it was low. From these results it appears that the solid product formed could be due to the oil additives. Several different mineral acids and mineral acid mixtures were tested.The acids tested included concentrated HCI, concentrated HN03, and the mixtures tested included concen- trated HCl-concentrated HN03 [(8 + l), (6 + 1) and (4 + l)]. The samples were heated with the acid for 15 min on a hot-plate and diluted to 50 cm3 with IBMK-IPA (4 + 1). The absorbance of the supernatant liquid was measured and it was found that acidifying with the 6 + 1 acid mixture gave the highest absorbance values. The heating time of the samples after addition of this mixture was investigated and the optimum heating time to attack the metallic particles was found to be 20 min. The behaviour of used oils of different TBN values was verified by measuring the metal concentration of the samples with two sets of standards, in one instance with the addition of concentrated HCl-concentrated HN03 (6 + 1) and in another instance without adding acid.Two of the samples (one with a high TBN and another with a low TBN) were prepared as follows: to 4 g of used oil, 1.5 cm3 of the 6 + 1 acid mixture were added and the mixture was heated on a hot-plate for 20 min; when the mixture had cooled to room temperature it was diluted to 50 cm3 with IBMK-IPA (4 + 1). The other two samples were prepared by the same method but without the addition of acid. Two sets of standards were prepared following the same procedure, with the same amount of un-used oil and the addition of the corresponding aliquot of a 50 pg g- 1 multi-element solution in IPA obtained by dilution from a 1000 pg g-1 stock solution in doubly distilled water.The supernatant liquid was measured leaving the precipitate at the bottom. The results are shown in Table 1. It can, therefore, be concluded that when the oil has a high TBN value the metal concentrations obtained are higher without adding acids but when the TBN value of the oil is low the metal concentrations with addition of acid are higher for metals other than Pb. Table 1 Concentration of the supernatant liquid. A mixture of IBMK-IPA (4 + 1) was used for the autozero Concentratiodpg g-1 Sample Cu Pb Fe Cr Comments 1 0.07 0.14 0.62 0.05 High TBN oil with acid attack 1 0.09 0.52 1.0 0.39 High TBN oil without acid 2 2.3 0.37 7.6 0.20 Low TBN oil with acid attack 2 2.0 0.87 6.3 0.13 LowTBNoilwithoutacidANALYST, JUNE 1992, VOL. 117 965 The results obtained with the high TBN oils could be explained by considering the two opposite effects caused by the acid attack: firstly, the solid metallic particles are dissolved, which should increase the metal concentration; however, on the other hand the precipitate formed on addition of acid carries down a portion of the metals which are, therefore, not determined and a decrease in the metal concentration found in the supernatant liquid results.Oils that have low TBN values show an increased metal concentration with added acid; therefore, in this instance the predominant effect seems to be the dissolution of the metallic particles. The behaviour of Pb is not clear, as has been found in earlier experiments and an attempt to explain it will be made later. As can be observed, the behaviour of oils with high and low TBN values is different when mineral acids are added.Metal Content in the Precipitates In order to prove that the precipitate formed contains metals, 4 g of a high TBN used oil sample were mixed with concentrated HCI-concentrated HN03 (6 + 1) and heated; when the mixture was cool, it was diluted to 50 cm3 with IBMK-IPA (4 + 1). The precipitate that formed was separated from the liquid phase, dried in a crucible on a hot-plate and ashed in an oven at 500 "C. The insoluble residue that remained was dissolved in acid and diluted with doubly distilled water. A set of standards was prepared from a 1000 pg g-1 multi-element stock solution in doubly distilled water and the metal concentrations were measured by FAAS. The same experiment was undertaken with use of two standard samples containing un-used oil and with the addition of metals.The precipitate that formed from the used oil samples contained 3.2 pg of Pb, 80 pg of Fe and 4 pg of Cr. To one of the standards 125 pg of metals were added and to the other 250 pg of metals were added. The metal content of the precipitates formed from the standards was 1.5 and 2 pg of Pb, 45 and 88 pg of Fe and 72 and 73 pg of Cr, respectively. From these results it can be concluded that the precipitate from the samples and standards cannot be discarded and that analysis of the supernatant liquid alone is not satisfactory. Therefore, attempts were made to dissolve the precipitate, or to clean it from metals after separating it from the liquid with the aim of mixing the metals with the liquid phase and determining the total metal content.The precipitate was treated with H202, with KOH in ethanol and with KOH and citrate in ethanol; the mixture was heated but the results were not satisfactory. Effect of Surfactants The next step was to attempt to homogenize the sample using a non-ionic surfactant which should be able to maintain the precipitate in suspension; this would allow the liquid and the precipitate to be determined together. Other workers have used surfactants previously but with another purpose. Kauff- man et al.11 used Neodol 91-6 to stabilize the organometallic standards; they also noted that this agent widens the applica- bility of the method to synthetic ester oils and to hydraulic fluids. On the other hand, Salvador et a1.15 used Nemol K 39 to obtain an emulsion, which allowed inorganic salts to be used as standards.Several non-ionic surfactants were tried and it was found that Tergitol Type 15-S-3 was satisfactory for maintaining the precipitate in suspension. It was intended, initially, to add a small amount of Tergitol to the mixed solvent, to dilute the sample after adding the acid mixture; however, after shaking the sample the precipitate started to settle slowly to the bottom of the vessel. It was, therefore, decided to substitute the IPA with Tergitol. Several samples were prepared, some with used oil and others with un-used oil as described under Experimental. The samples were shaken manually and it was observed that the precipitate remained in suspension.The absorbance measurements displayed good reproducibility and stability and it was, therefore, shown that Tergitol-IBMK (4 + 1) was a suitable homogenizing solvent for marine engine lubricating oil. Instrument Parameters Several samples were prepared by the procedure described above and the following parameters were optimized for each element: burner height, acetylene flow rate, wavelength and impact bead position. The results are shown in Table 2. In all instances the wavelength used was the principal wavelength except for Pb, the measurement of which was made at the 283.3 nm line, as the 217 nm principal line showed a high noise, which, in many instances, exceeded the signal corresponding to the Pb content. This noise can be attributed to scattering caused by the presence of small solid particles in the flame arising from the complicated matrix of the sample, as the scattering affects measurements made at lower wavelengths.23 When the absorbance was measured at 283.4 nm the signal-to-noise ratio was much improved.Therefore, the anomalous behaviour of Pb found previously can defin- itively be ascribed to the high noise level at 217 nm. Stability and Reproducibility The stability of several samples with a TBN value of 23 and of several standards prepared with un-used oil with a TBN value of 30 was assessed by preparing a set of multi-element standards of 1, 2 and 5 pg g-1 and three samples as described under Experimental. The absorbance of the standards and the concentration of samples were determined periodically, with use of IBMK-Tergitol (4 + 1) for the autozero.It was found that the samples were stable for at least 24 h and the standards for 3 d. The reproducibility of standards and samples was evaluated with 11 standards of 2 pg g-1 prepared with un-used oil (TBN=30) and 11 samples with TBN values of 18 prepared as described under Experimental. The results are shown in Table 3. Table 2 Instrument parameters Cu Pb Fe Cr ?Jnm 324.8 283.3 248.3 357.9 Acetylene flow rateldm3 min-1 1.0 1.0 0.9 1.2 Impact bead On On On On Burner height/mm 8 a 8 10 Table 3 Reproducibility of standards and samples. Concentration of standards, 2 yg g-1 c u Pb Fe Cr Standards- Standarddeviation/ygg-l 4.9 x 1.5 x 5.5 x 5.2 x 10-3 10-3 10-3 10-3 Relative error (%) 1.2 1.7 1.9 3.3 Standard deviatiodyg g-1 2.9 x 1.0 x 2.3 x 1.2 x 10-3 10-3 10-3 10-3 Relative error (YO) 2.6 4.5 0.8 2.5 Sampies- Table 4 Linear working range, detection limit and sensitivity. A11 values in yg g-1 c u Pb Fe Cr Linear working range 0.05-5 0.2-5 0.2-10 0.5-5 Detection limit 0.01 0.01 0.05 0.5 Sensitivity 0.02 0.16 0.08 0.06966 ANALYST, JUNE 1992, VOL.117 Table 5 Concentration of metals in used marine engine oils determined by the proposed method (A), flame atomic absorption spectrometry after sample ashing and acid dissolution (B), and flame atomic absorption spectrometry by simple dilution with IBMK (C). The results are the mean of two determinations c u Pb Fe Cr Sample Method A Method B Method C Method A Method B Method C Method A Method B Method C Method A Method B Method C 1 3.0 3.1 1.9 6.7 - 2.0 18 11 11 2.1 4.1 1.0* 2 3.6 3.5 2.1 7.9 11 2.4 23 28 12 5.5 5.5 2.0" 3 3.5 3.6 2.6 7.9 10 1.8 9.3 11 6.3 3.0 4.2 0.8* 4 6.7 7.3 5.7 9.5 12 3.1 24 23 15 2.6 5.7 1 .o* 5 2.1 1.4 3.6 8.7 11 0.3 5.8 6.4 1.2 2.3 5.7 0.8* * Inductively coupled plasma atomic emission spectrometry by simple dilution with kerosene.Calibration Graph In order to establish the calibration graph a set of standards from 0.05 to 10 pg g-1 was prepared as described under Experimental; the linear working range is shown in Table 4. The detection limit was taken as the metal concentration in the un-used oil giving an absorbance equal to that of the blank plus three times its standard deviation. The detection limit and sensitivity are shown in Table 4. The total content of Fe, Cu, Pb and Cr of five types of used marine engine lubricating oil was determined as described under Experimental.The results were compared with those obtained by FAAS after sample ashing and acid dissolution of the residue and with the simple dilution method with IBMK using Conostan organometallic standards. The results are shown in Table 5. Conclusions The data reported in Table 5 show good agreement between the results obtained with the proposed method and those from ashing of the samples; data obtained by the simple dilution method are lower. It can, therefore, be concluded that the proposed method allows the detection of the total metal content in used marine engine lubricating oil, which, as expected, is not possible with the simple dilution method.References Burrows, J. A., Heerdt. J. C., and Willis, J. B.. Anal. Chem., 1965.37, 579. Kriss, R. H., and Bartels, T. T., At. Absorpt. Newsl.. 1972, 11. 110. Hon, P. K., Lau, 0. W.. and Mok. C. S.. Analyst, 1980, 105, 919. 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Holding, S. T., and Rowson, J. J., Analyst, 1975, 100, 465. Wittmann, Z., Analyst. 1979, 104, 156. Wittmann, Z., Acta Chim. Acad. Sci. Hung., 1982, 109, 295. Brown, J. R., Saba, C. S.. Rhine, W. E., and Eisentraut, K. J.. Anal. Chem., 1980, 52,2369. Kriss, R. H., and Bartels, T., At. Absorpt. Newsl., 1970,9, 78. Saba, C. S., and Eisentraut, K. J., Anal. Chem., 1977,49,454. Saba, C. S., and Eisentraut, K. J., Anal. Chem., 1979,51,1927. Kauffman, R. E., Saba, C. S., Rhine, W. E., and Eisentraut. K. J., Anal. Chem., 1982, 54, 975. Berenguer-Navarro, V., and Hernandez-Mendez, J . , Quim. Anal.. 1977, 31, 81. Hernandez-Mendez, J . , Polo-Diez, L. M., and Bernal-Melchor, A.. Anal. Chim. Acta, 1979, 108, 39. de la Guardia, M., Salvador, A.. and Berenguer, V., An. Quim., 1982,78,321. Salvador, A., de la Guardia, M.. and Berenguer, V., Talanta, 1983, 30,986. de la Guardia, M., and Salvador, S., At. Spectrosc., 1984, 5 , 150. Arfelli, W., J. Test. Eval., 1984, 12, 152. Beferull-Blasco. J. B., de la Guardia-Cirugeda, M., and Salvador-Carreiio, A., Anal. Chim. Acta, 1985, 174, 353. Cardarelli. E., Cifani, M., Mecozzi, M., and Sechi, G., Talanta, 1986, 33, 279. Taylor, C. F.. The Internal Combustion Engines in Theory and Practice, Massachussetts Institute of Technology Press. MA, 2nd edn., 1986. Frederick, S. H., and Capper, H., Materials for Marine Machinery, Institute of Marine Engineers, London, 1980. American Society for Testing and Materials, ASTM D-2896-80 (IP 276/83), Philadelphia, PA, 1980. Larson. G. F., Fassel. V. A., Winge, R. K., and Kniseley, R. N., Appl. Spectrosc., 1976,30, 384. Paper I I051 92J Received October 14, I991 Accepted January 7, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700963

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 967 Direct Determination of Trace Metals in Graphite Powders by Electrothermal Atomic Absorption Spectrometry Yukihiro Koshino and Akira Narukawa Analysis & Properties Research Laboratory, Corporate Research and Development Group, NGK Insulators, Ltd., 2-56 Suda-cho, Mizuho-ku, Nagoya-shi, Aichi 467, Japan Trace concentrations of Li, Na, K, Mg, Al, Fe, Ni and Cr in graphite powders were determined directly by electrothermal atomic absorption spectrometry (ETAAS). As matrix interferences from graphite powder were easily removed by pre-heating at 800-850 "C for 180 s in an air stream, good peak profiles were obtained. With a pyrolytic graphite coated graphite tube and L'vov platform, simple metal standard solutions were used as calibration standards.The results obtained for the graphite powders JAERI G3, G5 and G6 agree with those reported previously by Watanabe and Takashima. The relative standard deviations are within +lo% for each metal determined. The time required for the determination of the eight trace metals studied is 10 h. Keywords: Electrothermal atomic absorption spectrometry; L'vov platform; graphite powder; direct determination; air stream Graphite is widely used in nuclear reactor material, as a crucible for various instruments, in electrodes, and also has many other uses. Therefore, it is very important to determine trace amounts of impurities in graphite which may affect its characteristics. Emission spectrometry is usually used for the determination of impurities in graphite powders,l but this method requires reliable standard samples which are very difficult to obtain.The ashing method is also used;* however, with this method there is the possibility of loss of low boiling-point elements such as Na and K, or the adsorption of transition metals such as Fe and Ni when platinum vessels are used. There are several reports on the decomposition and determination of impurities in graphite powders using potas- sium dichromate-phosphoric acid,3 a mixture of perchloric acid and periodic acid,4 a mixture of nitric acid and sulfuric acid in an open system,5 and by using a microwave oven.6 However, in these methods, Na and K cannot be determined because of the use of glass vessels, or because the graphite powders cannot be decomposed completely. This paper describes a method for the direct determination of trace metals in graphite powders by electrothermal atomic absorption spectrometry (ETA AS) without con taminat ion from reagents or the environment.Matrix interferences are easily removed from graphite powder and Li, Na, Mg, Ca, Al, Fe, Ni and Cr can be determined accurately in only 0.20-1.00 mg amounts of sample. This method has been used success- fully to determine metal impurities in graphite powders produced by the Japan Atomic Energy Research Institute (JAERI). Experimental Apparatus A Perkin-Elmer Model 25100 atomic absorption spec- trometer equipped with a Zeeman-effect background correc- tion system was used. The apparatus was attached to an AS-60 autosampler and a ZHGA-600 atomizer. A Type 7700 computer was used to control the furnace temperature and the amount of solution injected.Coded hollow cathode lamps were used as the radiation source. A pyrolytic graphite coated graphite tube and L'vov platform were used throughout. A Mettler AE-163 balance was used to weigh the samples. Reagents Metal stock standard solutions (1 mg cm-3) were prepared by dissolving 0.25 g of high-purity metals (Al, Fe, Ni and Cr), 0.25 g metal equivalent chlorides (Na and K), and carbonates (Li) in 5 cm3 of nitric acid (1 + 1) except for Al, which was treated with a mixture of 5 cm3 of nitric acid, hydrochloric acid and water (1 + 1 + 2), and diluted to 250 cm3 with water. The resulting solutions were stored in PFA bottles. Calibration standards were prepared by diluting these stock standard solutions with 0.2% nitric acid.Tama Chemical Tamapure AA-100 nitric acid, and Cica- Merck Ultra Pure hydrochloric acid were used. Distilled water was obtained using a Milli-Q purification system. Graphite powders (JAERI G3, G5 and G6) were provided by JAERI. Recommended Analytical Procedure Weigh 10 mg of graphite powder into a 15 mm diameter weighing bottle, then place the bottle and a micropipette tip on the balance and record the mass. Insert the tip into the furnace, then introduce 0.20-1.00 mg of graphite powder using a microspatula. Pull out the tip, then place both the tip and the sample bottle on the balance in order to determine the sample mass introduced. Inject 20 mm3 of 0.2% nitric acid into the furnace with the autosampler and proceed using the furnace conditions given in Table 1.Prepare the calibration graphs with a series of calibration standards using the same furnace conditions as those used for the samples. Results and Discussion The optimum operating conditions for determining metallic elements in graphite powders by ETAAS are shown in Table 1. Although graphite powder was easily removed by pre- heating at 8004350°C for 180 s in an air stream, the metals remained in the furnace. Consequently, good peak profiles, without matrix interferences, were obtained. Profiles of Na and K are shown in Figs. 1 and 2, respectively. In Fig. 1, the profile of Na in graphite powder was the same shape as that of the standard solution. In Fig. 2, the peak appearance time of K in graphite powder was longer than that for the standard solution. However, integrated absorbances were measured in all experiments, rather than peak height, and zero absorbance was measured after the second atomization. This suggests that the K in the graphite powder is atomized completely.With other elements, zero absorbances were also obtained after the second atomization. When Ar + 20% oxygen was used instead of air for the conditions given for pre-heating periods 2-4 in968 ANALYST, JUNE 1992, VOL. 117 Table 1 Optimum operating conditions for ETAAS. Ar flow rate, 300 cm3 min-' Element Parameter Wavelengt h/nm Slit-widthhm Lamp current/mA Pre-heating period 1PC (Ramp/s-Holds) Pre-heating period 2PCt ( Ramp/s-Holds) Pre-heating period 3PCt ( Ramp/s-Hold/s) Pre-heating period 4/"Ct (Ramp/s-Holds) AshingPC (Ramp/s-Holds) AtornizingPCI (Ramp/s-Holds) Clean up/"C (Ramp/s-Hold/s) Li Na K 670.8 589.0 766,5 0.4 15 120 300 800 (10-90) 850 (1-90) 900 2600 2800 (1-40) (20-20) (1-20) (0-5) (1-3) 1.4 8 120 (1-40) 300 (20-20) 800 (10-90) 850 (1-90) 900 1500 2500 (1-20) (0-5) (1-3) 1.4 12 120 300 800 850 950 1500 2500 (1-110) (20-20) (10-90) (1-90) (1-20) (0-5) (1-3) * Fe and Ni in JAERI G5 were determined with these conditions. 7 Air was used instead of Ar.I Ar flow was stopped in this step. Mg 285.2 0.7 6 120 300 800 (10-90) 850 900 1700 2500 (1-110) (20-20) (1-90) (1-20) (0-5) (1-3) A1 309.3 0.7 25 120 300 800 ( 10-90) 850 1700 2500 2800 (1-40) (20-20) (1-90) (1-20) (0-5) (1-3) Fe 296.7 (248.3)* 0.2 20( 30)* 120 300 800 (10-90) 850 1400 2500 2800 (1-40) (20-20) (1-90) (1-20) (0-5) (1-3) Ni 300.2 (232.0) * 0.2 25 120 300 800 (10-90) 850 1 400 2500 2800 (140) (20-20) (1-90) (1-20) (0-5) (1-3) Cr 357.9 0.7 25 120 300 800 ( 10-90) 850 1650 2500 2800 (1-40) (20-20) (1-90) (1-20) (0-5) (1-3) 0 2.5 Ti m e/s 5.0 Fig.1 Peak profiles of Na standard solution and Na in graphite powder. A , Standard Na solution, 400 pg; and B. 0.48 mg of graphite powder G6 0 2.5 Time/s 5.0 Fig. 2 Peak profiles of K standard solution and K in graphite powder. A , Standard K solution, 200 pg; and B, 0.36 ng of graphite powder G3 0 100 200 Na added/pg Fig. 3 and B, standard solution + 0.20 mg of G3 Investigation of calibration standards. A , Standard solution; Table 1, the same profiles were obtained for the analyte elements. Good profiles were obtained even when pure oxygen was used, but the graphite tube was damaged by oxidation and it could only be used for five runs. When argon gas was used, graphite powder remained in the graphite tube.Accordingly, matrix interference occurred. From these results, it is assumed that air is useful for the removal of graphite, converting it into carbon dioxide, and also for oxidizing analyte elements in the furnace in order to obtain ideal atomization. A series of diluted standard solutions of 0.2% nitric acidity were used as the calibration standards. Atomization was carried out using the conditions given in Table 1. In both examples 20 mm3 of 0-10 ng cm-3 diluted Na standard solutions plus 0.20 mg of graphite powder sample were injected into the furnace and atomized.The results obtained are shown in Fig. 3. A series of diluted standard solutions could be used to construct the calibration graphs because the slopes of the two lines agreed. This relationship was also achieved with other elements. Blank values were measured by injecting 20 mm3 of 0.2% nitric acid into the furnace. The results for Li, Na, Mg, Ca, Al, Fe, Ni and Cr in JAERI graphite standard reference powders (JAERI G3, G5 and G6) are shown in Tables 2 4 . The results agreed with those reported previously by Watanabe and Takashima.6 The relative standard deviation of five replicate analyses for each element was within +lo%. Conclusion A direct determination method for Li, Na, Mg, Ca, Al, Fe, Ni and Cr in graphite powder has been developed using ETAAS. Graphite is easily removed by pre-heating the furnace at 800-850 "C for 180 s in an air stream.As the proposed method needs no sample pre-treatment, contamination from reagents and vessels cannot occur. Sub-ppm amounts of metals were determined accurately in a 0.20-1.00 mg sample. This method can also be used to determine metal elements in poly- (acrylonitrile) (PAN) or pitch-based carbon fibre.ANALYST, JUNE 1992, VOL. 117 969 Table 2 Results obtained for JAERI G3 graphite powder. The relative standard deviation (n = 5) is given in parentheses Element/pg g-' Method Proposed Literature6 Li Na K Mg A1 Fe Ni Cr 0.03 0.48 0.28 0.09 0.70 37.0 25.1 1.25 0.03 0.51 (9.1) 1.0 (2.8) (7.2) 39.7 26.6 (5.7) 0.6 (6.5) 0.08 (9 * 4) (10) 0.20 (8.3) Table 3 Results obtained for JAERI G5 graphite powder.The relative standard deviation (n = 5) is given in parentheses Element/pg g-1 Method Proposed Literature6 Li Na K Mg A1 Fe Ni Cr <0.01 0.03 0.08 <0.01 0.04 0.33 0.07 0.04 0.02 <0.2 0.2 <o. 1 <o. 1 <0.01 <0.01 <0.01 (-1 (5.7) (10) (9.5) (8.3) (-1 (7.6) (10) Table 4 Results obtained for JAERI G6 graphite powder. The relative standard deviation (n = 5) is given in parentheses Element/pg g-l Method Proposed Literature6 Li Na K Mg A1 Fe Ni Cr 0.16 0.51 1.10 0.07 1.99 11.9 0.24 0.45 0.14 (4.3) (2.3) 0.3 0.5 (6.8) (8.9) (6.5) (6.7) (10) 0.44 1 .oo 0.16 1.6 10.3 (10) The authors are grateful to Dr. K. Takashima of the JAERI 3 4 Analytical Center for providing JAERI graphite powders. 5 6 References American Society for Testing and Materials, Standard Method For Chemical Analysis of Graphite, ASTM C 560-77, ASTM, Philadelphia, PA, 1982. Japanese Industrial Standard, Chemical Analysis of High Purity Graphite Material , JIS R 7223, Japanese Standard Association, 1979. Tagawa, H., and Nakajima, S., Kougyou Kagaku Zasshi, 1960, 63. 20. Hashitani. H., Yoshida, H . , Adachi, T., and Izawa, K., Bunseki Kagaku, 1986,35, 911. Kawakami, O., Takeya, M., and Sayama, Y., Abstracts of Slst Analytical Chemistry Debate, The Japan Society for Analytical Chemistry, 1990, p. 135. Watanabe, K., and Takashima, K., Abstracts of Slst Analytical Chemistry Debate, The Japan Society for Analytical Chemistry, 1990, pp. 431, 523. Paper 1105707C Received November 11, 1991 Accepted January 20, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700967

出版商:RSC    

年代:1992

数据来源: RSC

摘要:

ANALYST, JUNE 1992, VOL. 117 971 Arsenic Speciation by Ion Chromatography With Inductively Coupled Plasma Mass Spectrometric Detection Brenda S. Sheppard" and Joseph A. Carusot Department of Chemistry, University of Cincinnati, Cincinnati, OH 4522 I , USA Douglas T. Heitkernper and Karen A. Wolnik National Forensic Chemistry Center, US Food and Drug Administration, Cincinnati, OH 45202, USA Four As compounds were successfully separated and detected by single-column ion chromatography with inductively coupled plasma (ICP) mass spectrometric detection. The mass spectral interferent ArCI+ was reduced by chromatographically resolving chloride from the negatively charged arsenic species. Determina- tion of four As species was investigated in urine, club soda and wine. Detection limits of 0.16 ng of As"', 0.26 ng of AsV, 0.073 ng of dimethylarsinic acid (DMA) and 0.18 ng of methylarsonic acid (MMA) in wine were obtained.Sensitivity was further improved by using an He-Ar mixed gas ICP as the ionization source. However, the intensity of the ArCI+ interference was also increased using this plasma. Detection limits of 0.063 ng of Aslll, 0.037 ng of AsV, 0.032 ng of DMA and 0.080 ng of MMA in club soda were achieved using the He-Ar plasma source. Similar limits of detection were found in urine and wine. Keywords: Arsenic speciation; ion chromatography; inductively coupled plasma mass spectrometry; argon chloride polyatomic interference The toxicity of As is known to depend on its chemical form. Arsenite (As"') is the most toxic of the water-soluble species commonly found in clinical and environmental samples.Arsenate (As") is also relatively toxic, whereas the methylated forms, methylarsonic acid (MMA) and dimethylarsinic acid (DMA), are much less toxic.' Human exposure to As compounds can be monitored by sampling urine, blood and hair. Arsenic levels found in these samples are good biological indicators of poisoning.2 Methods which determine the total amount of As present in a sample do not adequately assess the danger of exposure. Chemical speciation is required to determine the types and amounts of As compounds present in a given material. The speciation of As in biological fluids is important in assessing a subject's exposure to toxic As compounds. It is also important to determine the species present in the source of exposure such as environmental samples or, as with poisonings, in foods and beverages.Several workers have reported As speciation by high-perfor- mance liquid chromatography-inductively coupled plasma mass spectrometry (HPLC-ICP-MS) .3-7 Plasma MS is a widely recognized technique for trace element analysis and is being increasingly used as a detector for chromatographic determinations.3-12 The benefits of coupling plasma mass spectrometric detection with chromato- graphic separation include element specificity, real-time chromatograms, the ability to separate interferences from peaks of interest, multi-element capability and low levels of detection (sub-nanogram for most elements). These features of ICP-MS are valuable assets in the speciation of trace elements, where sample pre-treatment should be minimized.Sample pre-treatment and preconcentration can lead to changes in the relative concentration of individual species. Ion chromatography (IC) is an attractive analytical tech- nique for element speciation because it can separate both inorganic and organic charged species in addition to free ions. Unlike ion-pairing or strong ion-exchange chromatography, IC does not generally require high concentrations of organic * Present address: National Forensic Chemistry Center, US Food t To whom correspondence should be addressed. and Drug Administration, Cincinnati, OH, USA. solvents such as methanol and acetonitrile. When used, organic solvents are usually less than 5% of the mobile phase. The introduction of organic solvents can destabilize the ICP and lead to elevated background levels.'3 In addition, single-column IC utilizes low buffer concentrations which help to reduce matrix-induced effects on the analytical signal and problems associated with salt deposition at the nebulizer tip and MS sampling cone.Therefore, the combination of IC with ICP-MS (IC-ICP-MS) should provide high potential for trace element speciation. Additionally, alternative ionization sources, such as the He-Ar mixed gas ICP, may be used to enhance the sensitivity of plasma MS.14 The He-Ar mixed gas plasma combines the familiar features of Ar ICP with the benefits of He plasmas. Plasmas containing He are generally thought to be more energetic than their Ar counterparts because of the higher ionization energy and metastable state energies of He.Higher plasma energies are required to achieve better sensitivity with higher ionization potential elements such as the halogens, As and Se. The addition of He to Ar plasmas has been shown to increase sensitivity for elements with higher ionization ener- gies while maintaining high sensitivity for other elements. 14 One problem associated with the determination of As via ICP-MS is the polyatomic interference at mlz 75. Chloride present in samples combines with the plasma gas and forms 40Ar35CI+ (mlz 75), which is detected together with any As species and thus yields inaccurate results. In a previous study7 it was shown that chloride could be separated chromatograph- ically from As species present in samples.Unfortunately, under the conditions used the methylated As species were not resolved. This paper will discuss the use of single-column IC to determine four forms of As and to resolve chloride from As species chromatographically, thereby eliminating the interfer- ence from polyatomic ArCI+ . Three chloride-containing matrices (urine, club soda and wine) were chosen to illustrate the effectiveness of the technique. Urine contains a high level of chloride and is often used as an indicator of exposure to As. Wine and club soda were chosen as possible vehicles for poisons. Analytical figures of merit for As in urine, and in the other matrices, are reported for Ar ICP-MS. Figures of merit for As in wine are reported for He-Ar ICP-MS. Several reference urine samples were analysed for comparison of total As levels.972 ANALYST, JUNE 1992, VOL.117 Experimental Instrumentation A VG PlasmaQuad (VG Elemental, Winsford, Cheshire, UK) mass spectrometer was used. The operating conditions are shown in Table 1. Sensitivity for As was investigated with both the Ar and He-Ar mixed gas ICP as ionization sources. Instrumental modification required for the operation of the mass spectrometer with He-Ar ICP has been discussed previously. l4 The sample introduction system was equipped with a Type C-1 concentric nebulizer (Precision Glassblowing of Colorado, Parker, CO, USA) and a double-pass Scott-type spray chamber cooled to 5 "C with a Neslab Endocal refrigerated chiller (Neslab Instrument, Portsmouth, NH, USA). A nickel sampler and skimmer, each with a 1.0 mm diameter orifice, were used.An Isco Model 2350 (Lincoln, NE, USA) HPLC pump and a Rheodyne Model 7010 injector (Cotati, CA, USA) were employed. Samples were injected using a 100 mm3 loop. The analytical column was a Wescan Anion/R IC column (250 X 4.1 mm i.d.) (Wescan Instrument, Deerfield, IL, USA). A guard column of the same packing type was used in the analysis of urine and wine samples. An aqueous solution of carbonate buffer, equal amounts of carbonate and hydrogen carbonate, served as the mobile phase. Separations were performed using a gradient programme, which is described under Results and Discussion. A 71 cm length of Polyplex tubing (A in 0.d. x 0.020 in i.d., Alltech Associates, Deerfield, IL, USA) was used to connect the analytical column directly to the nebulizer.Data Acquisition Data were collected by setting the quadrupole to mlz 75 and then adjusting the sampling position and ion-lens voltages for the optimum As signal using a 100 pg dm-3 solution of As"'. The multi-channel analyser was set to monitor a single mass with an integration time of 1.3 s . Background was determined using at least 25 points prior to the elution of the first peak. Peak heights were determined by averaging integrated counts for three consecutive channels across the peak maximum and subtracting the average background. All data presented have undergone seven-point Savitzky-Golay smoothing. 15 Table 1 Instrumental operating conditions Ar ICP-MS- Forward r.f. power Reflected power Outer gas flow rate Intermediate gas flow rate Injector gas flow rate He-Ar ICP-MS- Forward r.f.power Reflected power Outer gas flow rate Intermediate gas flow rate Injector gas flow rate Ion chromatography- Column Flow rate Sample loop Eluent 1 Eluent 2 Gradient programme- Step 1 Step 2 Flow rate 1.35 k W <5 w 16 dm3 min-l 1 dm3 min-1 0.660 dm3 min-' 1.55 kW <5 w 16 dm3 min-'(20% He) 3 dm3 min-'(20% He) 0.666 dm3 min-* WescanAnion/R-IC,250~4.1 mmi.d. 1 .O cm3 min- * 100 mm3 2% propan-1-01 50 mmol dm-3 carbonate buffer, pH 7.5 3 min: 70% eluent 1,30% eluent 2 Step to 100% eluent 2 1 cm3 min-1 for 6.5 min 2 cm3 min-1 to end Standards and Reagents The mobile phases used were prepared from ammonium carbonate and ammonium hydrogen carbonate (Fisher Scien- tific, Fair Lawn, NJ, USA) and distilled, de-ionized (DDI) water with a metered resistance of 18 MSZ (Barnstead, Boston, MA, USA).Equal amounts of carbonate and hydrogen carbonate were used. The pH of the mobile phase was adjusted using ammonia solution or nitric acid. A 2% propan-1-01 solution was also prepared for use in mobile phase optimization studies. Two freeze-dried urine standards, UriChem, Urine Chemistry Control Level I (Fisher Scientific, Orangeburg, NY, USA), and National Institute of Standards and Technol- ogy (NIST), Standard Reference Material (SRM) 2670 Low (Toxic Metals in Freeze-Dried Urine) were analysed. Ad- ditionally, urine for chromatographic development was col- lected from laboratory personnel. All samples were refriger- ated and stored in poly(tetrafluoroethy1ene) (PTFE) bottles.Urine samples were filtered through a nylon Acrodisc 0.45 pm filter (Gelman Sciences, Ann Arbor, MI, USA) after dilution with the chromatographic mobile phase. In addition to urine samples, several beverages were used as matrix solutions. Club soda was chosen as one of the beverage solutions, because it is known to contain a significant amount of chloride. Portions of club soda were spiked to contain As species at concentrations of 10, 25, 50, 100, 500 or 1000 pg dm-3. The second beverage studied was blush wine which was also spiked to contain As concentrations of 10,25,50,100 and lo00 pg dm-3. The arsenate, arsenite and DMA were obtained from Fisher Scientific (Fair Lawn, NJ, USA). The sodium salt of MMA was obtained from J. S. Thayer (University of Cincinnati, OH, USA).Stock solutions (lo00 ppm of As) of sodium arsenite, sodium arsenate, DMA and MMA were prepared from the pure compound in 1% nitric acid and diluted with DDI water. Working standards were prepared daily in the chromato- graphic mobile phase, from the stock solutions. Standards (10 pg dm-3) in DDI water were prepared for standard additions to urine. All solutions and mobile phases were filtered through 0.45 pm nylon filters. Results and Discussion IC-ICP-MS Arsenic Speciation in Urine Separation in ion exchange and ion chromatography is achieved by competition for stationary phase sites between the analytes and ionic components of the mobile phase.17 In a previous study,' a phthalic acid mobile phase was used to separate the inorganic As species and chloride.Phthalic acid has pKa values of 3.10 and5.40, whereas HAs02, H3As02 and (CH3)2A~02H have pK, values of 9.2, 2.2 and 6.3, respec- tively. Under these mobile phase conditions the elution order was AS'", DMA, AsV and C1-. All species were separated in 8 min and the two inorganic species were totally resolved. Arsenic(rrr), as HAs02, was not ionized and eluted with the void volume, whereas DMA and MMA co-eluted. A mobile phase with a higher pKa value and a higher pH range would be more desirable for increased ionization of DMA and, conse- quently, better separation of the methylated species. A common mobile phase used in suppressed IC is the carbonate- hydrogen carbonate buffer. 17 This buffer system has pK, values of 6.3 and 10.3 and can be used at higher pH values. Chromatographic development was accomplished using both standards in buffer solutions and spiked urine samples which were collected from laboratory personnel. Optimiza- tion of mobile phase conditions included varying both concentration and pH of the buffer system.With a 25 mmol dm-3 carbonate buffer mobile phase concentration and a pH of 9, DMA co-eluted with As"'; however, MMA and AsV were resolved (see Fig. 1). Therefore, increased buffer concentration was investigated. The resolution, R , betweenANALYST, JUNE 1992, VOL. 117 973 each of the species was calculated using the following equation: R = 1.176(tR2 - tRl)/(WI + Wz) where t R is the retention time and W is the width at half the peak height. With a buffer concentration of 50 mmol dm-3 and a pH of 8 the resolution between the species was as follows: As"' and DMA, R = 0.96; DMA and MMA, R = 1.26; MMA and AsV, R = 3.5.Two peaks are generally considered baseline resolved when R = 1.5; however, the separation is often considered acceptable when R = 1. Arsenic(ii1) and DMA were further resolved by the addition of a small amount of propan-1-01 to the mobile phase. Alcohols can be used as column wetting agents to solvate the stationary phase functional groups. The pH of the mobile phase was varied from 7.0 to 8.6 at a mobile phase concentration of 50 mmol dm-3 and propan-1-01 concentra- tion of 2%. It was determined that the best separation was achieved when the mobile phase pH was 7.5. Additionally, it was necessary to use gradient programming to achieve complete separation between the As"' species and the two organic species.The gradient programme used is shown in Table 1. During the first 3 min of the programme, 70% eluent 1 (2% propan-1-01), and 30% eluent 2 (50 mmol dm-3 carbonate buffer) was used. At 3 min the mobile phase was stepped to 100% eluent 2. The flow rate was initially 1 cm3 min-1 but increased to 2 cm3 min-1 after elution of MMA (6.5 min) in order to decrease the elution time of As". Under these conditions all four of the As species of interest were not only resolved from each other but they were also separated from the eluting chloride and subsequent interference from ArCI+. The order of the elution was As''', DMA, MMA, Cl- and AsV (see Fig. 2). ArCI+ Interference Human urine is about 0.15 mol dm-3 in sodium chloride (or approximately 0.9% by mass).16 Assuming normal levels of As in urine of 100 pg dm-3 or lower, chloride concentrations would be approximately 105 times more concentrated than any of the As species present.16 The interference resulting from the elution of chloride and subsequent formation and detec- tion of ArCI+ at mlz 75 can, therefore, be substantial.In a previous study,' where a phthalic acid mobile phase was employed, a 1 + 19 dilution of urine samples was required to MMA 4 8 12 Time/min Fig. 1 Separation of three As species by IC with Ar ICP-MS detection. All species were at a concentration of 500 pg dm-3 As. Mobile phase, 25 mmol dm-3 carbonate-hydrogen carbonate buffer at pH 9; flow rate. 1 cm3 min-I; sample loop size, 100 mm3; monitoring mlz 75 reduce column overloading due to chloride and reduce the ArCI+ interference.With the carbonate buffer system the required dilution was reduced to 1 + 4. The separation of the four As species, at 100 pg dm-3, and chloride in 1 + 4 urine, is shown in Fig. 2. The identity of a chromatographic peak at m/z 75 due to ArCI+ was confirmed by monitoring injections at m/z 51 (35C1160+), 75 (4*Ar35CI+) and 77 (aAr37Cl+). The isotope ratios for the two peaks at mlz 75 and 77 compared well with the 3 : 1 chloride ratio at mlz 35/37. Figures of merit including detection limits, linearity and reproducibility were determined in urine by spiking As into diluted (1 + 4) urine and are listed in Table 2. This sample contained no detectable natural As.Detection limits of 4.9, 6.0, 1.2 and 3.6 pg dm-3 of As were obtained for AsV, DMA and MMA, respectively, in 100% urine using Ar ICP-MS. Linearity was investigated over three concentration decades, 10-1000 pg dm-3. Reproducibility was based on repetitive injections of a 10 pg dm-3 standard addition and calculated for peak height. For five injections, relative standard deviations (RSDs) of approximately 6% were obtained. The accuracy of the method was tested by comparing the experimental sum total of all As species measured with the total As concentration reported for freeze-dried urine stan- dards. The results are shown in Table 3. Samples were analysed using a single standard addition. Agreement of total As was within the 95% confidence limit reported for the reference standard and was obtained for both ionization sources.Urichem, Urine Chemistry Control, is a commer- cially available freeze-dried urine standard with a reported As concentration of 20 k 15 pg dm-3. In this work, values of 30 k 5 pg dm-3 (Ar ICP) and 26 k 6 pg dm-3 (He-Ar ICP) for As"' compare well with the reported value. In NIST SRM 2670, As"' and DMA were detected. Concentrations of 29 k 4 and I 0 4 8 12 Time/m i n Fig. 2 Separation of all four As species and chloride by IC with Ar ICP-MS detection. All As species were at concentrations of 100 pg dm-3 As in 1 + 4 urine. Mobile phase, 50 mmol dm-3 carbonate-hydrogen carbonate buffer at pH 7.5 (see Table 1 for gradient programme used); sample loop size, 100 mm3; and monitor- ing mlz 75 Table 2 Figures of merit in urine Detection Absolute limit*/ detection pg dm-3 limit/ng Ar ICP-MS- As"' 4.9 0.49 AsV 6.0 0.60 DMA 1.2 0.12 MMA 3.6 0.36 * 100 mm3 injection.Repro- ducibility, RSD (Yo) Linearity ( n = 5) 3 orders 8 3 orders 6 3 orders 2 3 orders 5974 ANALYST, JUNE 1992, VOL. 117 Table 3 Determination of As in urine by IC-ICP-MS Accepted As"'/ D M N Total As/ total As/ Sample pg dm-3* pg dm-3 pg dm-3 pg dm-3 Urichem N 30 f S - 3O-t-5 20k 15 SRM2670N 29+4 18-+10 4 7 f 1 1 607 Urichem N 26 + 6 - 2 6 f 6 20+ 15 SRM2670N 2 6 f S 14+12 40+13 607 * (pg dm-3 k 1 SD). t Provisional. Ar ICP-MS- He-A r IC P-MS- v) c ; 2.0 m z \ 21.6 - c 1.2 .- v) Q) 4- 0.8 0.2 8 12 Time/min Fig. 3 Separation of the four As species and chloride in club soda. Each As species present at SO pg dm-3 As.Chromatographic conditions as described in Fig. 2. Ar ICP-MS detection, monitoring mlz 75 18 2 10 pg dm-3 with the Ar ICP and 26 f 5 and 14 k 12 pg dm-3 with the He-Ar ICP were obtained for As"' and DMA, respectively. These values are comparable to the provisional value of 60 pg dm-3 for total As reported by NIST. These results indicate that the interference from ArCI+ has been sufficiently minimized to yield accurate results for As speciation in urine. Additionally, there is good correlation between the results obtained from the two ionization sources. Arsenic Speciation in Beverages Foods and/or beverages may be purposely or inadvertently used as vehicles for poisons. Therefore, the determination and quantification of contaminants need to be applicable to a variety of matrices.Club soda and wine are used here to illustrate typical beverage matrices which could be used as a means to deliver As poison. Club soda is known to contain chloride and does not normally contain As compounds. The salt content of this matrix is considerably lower than that of urine. The analysis of club soda showed a single chromatographic peak at mass 75; however, the same peak was present at m/z 51 (35C1160+) and mlz 77 (NAr37Cl+), thus demonstrating that no As was present in the original sample. The peak was due to chloride in the sample. Successful speciation of the four As compounds of interest, spiked into club soda, was accomplished under the same chromatographic conditions as those used with the spiked urine matrix.The separation of the As compounds and chloride in club soda with Ar ICP-MS detection is shown in Fig. 3. In this chromatogram As"' and DMA are no longer 2.0 1.6 1.2 0.8 c I? 0.4 v) 4- g o 2, 5.0 > v) 0 c .- 4- - C 4.2 3.4 2.6 1.8 0 4 8 12 Time/min Fig. 4 Chromatograms showing the speciation of As in 1 + 4 wine. Each As species is present at SO pg dm-3 As. (a) Ar ICP-MS detection and (b) He-Ar ICP-MS detection. Chromatographic conditions as described in Fig. 2. Monitoring mlz 75 fully resolved. The difference in ionic strength between club soda and urine is believed to be the reason for the difference in resolution. Figures of merit for club soda are comparable to those in urine. All four compounds gave linear ranges of over two orders of magnitude (the concentration range investigated) with log-log slopes of 0.992, 0.996, 0.999 and 0.987 for As"', AsV, DMA and MMA, respectively, using the Ar plasma.Absolute detection limits for the As compounds were 0.25, 0.33, 0.51 and 0.81 ng for As"', AsV, DMA and MMA, respectively. Improvements in detection limits are obtained in the club soda matrix by using the mixed-gas plasma. Absolute detection limits of 0.072, 0.02, 0.034 and 0.044 ng were obtained for As"', As", DMA and MMA, respectively. Similar results for linearity were achieved with the mixed-gas plasma. One beverage that has historically been subject to poisoning is wine. Hence the speciation of As in wine was investigated. The wine was filtered before injection onto the column. A sample of blush wine was analysed to ensure that no natural As was present.Two peaks were obtained at mlz 75, one which eluted at the retentidime expected for chloride and a second which eluted at 6 retention time expected for AS"'. 75 and 51, leading to speculation that the first peak is also due to the elution of some chlorine-containing species and subsequent formation of 4oAr35CI+. The formation of ArCI+ in the plasma is not necessarily linear with the solution concentration of chloride and, therefore, a 1 + 4 dilution of the wine with buffer alleviated the background interference problem. Separation of the As compounds in this matrix resembled the separation in urine matrix. The separation of the four As compounds and chloride for both Ar and He-Ar ICP-MS is shown in Fig. 4. The sensitivity for As is increased with the more energetic He-Ar plasma.The chloride peak is noticeably absent when the ionization source is an Ar plasma and present when the source is the more energetic He-Ar plasma. (The mixed-gas plasma has more energy available for the formation of ArCI+ .) The ionization source must there- Both chromatographi it peaks, however, are present at massesANALYST, JUNE 1992, VOL. 117 975 Table 4 Figures of merit in 1 + 4 wine Detection Absolute limit/ detection pg dm-3 limithg Ar ICP-MS- As"' 1.6 0.16 AsV 2.6 0.26 DMA 0.73 0.073 MMA 1.8 0.18 As"' 0.63 0.063 AsV 0.37 0.037 DMA 0.32 0.032 MMA 0.80 0.080 * n = 5 . He-A r ICP- MS- Linearity 3 orders 3 orders 3 orders >2 orders 3 orders 3 orders 3 orders 3 orders Repro- ducibility. RSD(%)* 4.5 5 4 15 7 4 14 12 fore be taken into account when determining the extent of dilution necessary for a particular sample. Analytical figures of merit for the speciation of As in 1 + 4 wine with plasma mass spectrometric detection are listed in Table 4.Linearity was investigated for the concentration range 10-1000 pg dm-3; the compounds showed linearity in this region. Absolute detection limits of 0.16,0.26,0.073 and 0.18 ng were found for As"', As", DMA and MMA, respectively, using Ar ICP detection and 0.063, 0.037, 0.032 and 0.080 ng using He-Ar ICP detection. Conclusions Plasma MS has high potential in the detection and speciation of trace amounts of toxic elements such as As. Absolute detection limits for As species of less than 50 pg demonstrate the power of ICP-MS as a chromatographic detector.Single- column IC can separate free ions and charged species. Furthermore, the use of aqueous mobile phases with low buffer salt concentrations minimizes the problems associated with the coupling of ICP-MS to other forms of liquid chromatography. Ion chromatography has been used to separate four As species in urine, wine and club soda matrices. Additionally, the interference from ArC1+ has been minimized by chroma- tographically separating chloride from As species and when necessary using an appropriate sample dilution factor to keep chloride from overloading the column. The sensitivity of this method is limited by the need to dilute samples of high chloride concentration. However, the use of a carbonate buffer mobile phase reduces the dilution needed to prevent column overloading. Additionally, the He-Ar mixed-gas ICP can be used as an alternative ionization source to improve detection limits for As; however, the ArClf signal is intensi- fied with this plasma.The authors are grateful to the National Institute for Environmental Health Sciences for support of this study under grant number ES 03221 and ES 04908 and to the NIH-BRS Shared Instruments Grant Program for providing the VG PlasmaQuad through grant number SIORR02714. We also thank John Dorsey for his helpful suggestions. B. S. S. is grateful to the US Food and Drug Administration for support during this work. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 References Hodgson. E., Mailman, R. B., and Chambers, J. E., Dictionary of Toxicology, Macmillan, London, 1988, pp. 40-41. Buchet, J. P., and Lauwerys, R., in Analytical Techniques for Heavy Metals in Biological Fluids, ed., Facchetti, S., Elsevier, Amsterdam, 1981, pp. 75-90. Thompson, J. J., and Houk, R. S., Anal. Chem., 1986,58,2541. Beauchemin, D., Bednas. M. E., Berman, S. S.. McLaren, J. W., Siu, K. W. M., and Sturgeon, R. E.. Anal. Chem., 1988, 60, 2209. Beauchemin, D., Siu, K. W. M., McLaren, J. W., and Berman, S. S.. J. Anal. At. Spectrom., 1989, 4, 285. Heitkemper, D.. Creed, J., Caruso, J., and Fricke, F. L., J. Anal. At. Spectrom., 1989, 4 , 279. Sheppard, B. S., Shen, W. L., Caruso, J. A., Heitkemper, D. T., and Fricke, F. L., J. Anal. At. Spectrom., 1990, 5,431. Dean, J. R., Munro. S., Ebdon, L., Crews, H. M., and Massey. R. C., J. Anal. At. Spectrom., 1987, 2, 607. Suyani, H., Creed, J., Davidson, T., and Caruso, J. A., J. Chromatogr. Sci., 1987, 27, 139. Suyani, H., Heitkemper, D. T., Creed, J., and Caruso, J. A., Appl. Spectrosc., 1989.43, 962. Bushee, D. S., Analyst, 1988, 113, 1167. Jiang, S. J., and Houk, R. S., Spectrochim. Acta, Part B , 1988, 43,405. Boorn. A. W., and Browner, R. F., Anal. Chem.. 1982, 54, 1402. Sheppard, B. S., Shen, W. L., Davidson. T. M., and Caruso, J. A., J. Anal. At. Spectrom., 1990, 5 , 697. Savitzky, A., and Golay, M. J. E., Anal. Chem., 1964,36,1627. Low, G. K.-C., Bately, G. E., and Buchanan, S. J., Chromato- graphia, 1986.22, 292. Chromatographic Theory and Basic Principles, ed., Jonsson, J. A., Marcel Dekker, New York, 1987, p. 318. Paper I /(I4711 F Received September 10, 1991 Accepted February 12, 1992

ISSN:0003-2654    

DOI:10.1039/AN9921700971

出版商:RSC    

年代:1992

数据来源: RSC