Document aJ0zgvwnkM0r8VBEEOKomR4Y

29-10 JNL OF THE ASS0CIH10H OF OFFICIM. WtALFTlCM. CHEMISTS U. MOO JUL 1974 [awaiS] I KBL; COMPUTER APPROACH TO IDENTIFICATION OP PCB* 7BL INDUSTRIAL CHEMICALS Quantitative Determination of Polychlorinated Biphenyl*-- A Computer Approach MICHAEL C. R. ZOBEL Chawintry Divition, Dtpartment of Scientific and [ndxntnoi Reiearch, Private Bag, PtUma, Nrw Zealand A emniiBltf method U pnal*4 wIiMi *niHri polychlorinated biphenyl (PCB) hiMm ekrafliitgpami'io m NNit4 and qumitelively MtuurW in term* of i nm of different UadiH Aroclors. The Aroclor* 1242, 1254, and 1262 were u4 In this work, but il method is capable of using i*f number or com bination from 1 Aroelov upward. The propram MMndi teat chromatograms from standard peak heights, and, bp a minimising procedure, selects the teat chromatogram nearest to ilir anhaean Spuriously Urge or small peak heights, aaoeed bf interfering eompounda or metabo lism, are automatically sorted and rejected. The naallt are output aa both total and rouipo*t iIomI PCB content, and In the form of a con structed line chromatogram. The method waa applied to cardboard and pulp samples aj widely differing PCB content. A modified program has enabled analysis of a heavily metabolised pork fat sample by comparing only the later eluting psahsi Uu "pro-metabolism" PCB content waa estimated. Tbo elsae of compounds known as polychlor inated biphenyls (PCB, PCBe) have become well known as environments! pollutants (1-2). They am marketed aa mixtures of up to 70 components, ranging from mono- through docnehlorobiphcnyl. Oat-liquid chromatographic (CLC) nnalyaes of environmental samples containing traces of these eompounda yield complex chromatogram* conating of as many as 20 peaks, depending on the resolution of the equipment. Such chromato grams may be derived from a combination of many different PCB mixtures (e.g., 8 Arociont*. 1221-1268) that may have been 'subject to ehangee in composition by srlectiv* metabolism, diffusion, or chemical action. Quantitative measurement of such chromato grams is a formidable task. Even if all component ieomers in a sample could be resolved, identified, and measured accurately, the resulting man of data would be of limited use to the layman, sinco moat of the information required (or assenment of environmental parameters, eg., toxicological data and physical and chemical properties, has been evaluated in terms of the commercial mix tures. It would seem desirable, therefore, to t# able to describe the PCB content of an environ mental sample in terms of the quantity of each Aroclor present. The purpose of this paper ta to present a computer method capable of describing a residue chromatogram as s sum of-3 of the more com mon Aroclor mixtures. Present Methods of Quantitation The simplest and perhaps moat widely used method i* to compare the sample chromatogram with the most similar standard ItB chromato gram and calculate the ratio of I or more peak heights. While this is a simple and rapid proce dure, ii ia only itccuraie when the sample chro matogram corresponds very cloeely in slutiw and peak ratios with the selected standard chromato gram. Some authors use an averaging method. CoUina et of. (4) have experimentally deiermmrd that the mean electron capture rraponM to Arocion 1254-1262, taken on a basis of the total area un der all peak*, approximate* that to p.p'-DDE. Rote und Murphy <5) have used the total area under each chromatogram to estimate the detec tor rraiwiwc to mch chlorinated eum|>oncitt I lowever, this work is baaed on the assumption thui the delect or response is proportional to ihe num ber of chlorine atoms in the molecule, wherein it lias been shown that detector response is highly variable, eg., tetraehlorobipltcnyls differ by x 15 in response, and a hexachlurolnphenyl has a weaker response than some ictrachloro isomers (0). Berg rt af. (7) combine quantitative deter mination with confirmation by forming the bi phenyl and deenrhlorobtphenyl derivatives, thus describing the PCB content in terms of total biphenyl equivalent. MONS 084417 <T'--------- * ................ f 792 JOURNAL oj the aoac (Vol. 57, No. 4, 1074) A Computer AppiouH A computer program, written in the language BASIC, generates chromatograms by using vari ous combinations of the 3 Aroclors^ 1242, 1254, and 1262. The computed chromatogram* are made to converge to the actual GLC data by an iterative procedure which minimizes the sum of the absolute difference of the peak heights. In the first instance, a set of 27 chromato grams is generated by summing the individual peak height contributions from the 3 standard Aroclors, using the relative coefficients 0.50-5. The combination of coefficients which generates the chromatogram nearest, to the actual data, judged on the criterion of the minimum sum of absolute peak difference over all peaks, is used as the basis to generate a further 27 chromatograms, this time using an interval of 0.25. Six such iterations, using successively smaller coefficient intervals, arc sufficient to produce the best fit to the experimental data. Theoretically, any FOB chromatogram, de rived originally from Aroclors and not subject to changes by metabolism, spurious peaks, etc., can be represented exactly by n linear sum of the 6 marketed Aroclor standards. However, a high degree of correlation exists between certain Aroclors (Table 1), so that approximate fits may be obtained by using 3 or 4 standards. In this work Aroclors 1221, 1242, 1254, and 1262 were chosen: 1221 was later rejected because of a lack of adequate major peaks. - Fifteen well defined peaks were selected from the approximately 30 resolved by normal residue analysis GLC (Table 2). Heptachlor epoxide was used as a reference standard to provide correla tion between retention lime and relative peak heights, thus obvinting the necessity of multiple Aroclor Btandnrd injections between each sample. The quality of the computer fit is evaluated as the average peak difference (APD), i.e., the ratio of the sum of absolute differences to the sum of Tati* 1. Corratatian coefficient* ter Aractar *aak hoieht*--IS paafca compared Aroclor 1221 1232 1242 1241 1254 1262 1221 1 1232 on 1 1242 0.14 ON 1 1241 0.63 S3 0.11 1254 0 01 o \* o n 0.5 1 1212 0.63 0.54 0.54 0.41 0.20 TaAla 2. Aroclor * utad by h* program ta fit PCS chromatogram*--15% DC-200 column maintain** at lf5*C Peak No. 1 2 3 4 5 i 7 1 10 11 12 13 14 15 Hoptochlor opoaido Retention Umt mm roiotivo 9 0.47 11.5 0.55 IS 0.69 17 Oil 21.5 1.00 25 1.17 32 1.50 >1 1.01 46 2.1 53 2.S 63 2.9 73 3.4 *7 4.0 103 4.0 113 5.4 21.6 1.00 P**h height*,* mm 1242 2254 1202 52 131 _ _ 35 31 49 13 113 167 31 241 26 154 16 197 2 151 _ 121 _ 27 _ 20 _ 11 ---- _ __ 21 33 25 130 171 7* 145 92 140 40 40 * Sensitivity - 2.00 mm/pj hopUchlor *po**do. peak heights. GLC baaeline and nonlinearity errors contribute up to 5% APD (Fig. 1). By experience, u fit giving less than about 20% APD produces acceptable and stable data, i.e., the program will always converge to essentially the same fit with small variations in peak heights Poorer fits arc characterized by large fluctuations in fit parameters. One of the problems associated with quanti tative measurement by total peak area methods is the falsely high values generated by the oc currence of extraneous peaks with similar reten tion times. Such errors are overcome by using the sum of the absolute difference rather than the sum of squares ns the minimizing parameter. Any spuriously high (or low) peaks are relatively poorly fitted, and can be sorted out and rejected by the computer. Refitting then produces more accurate data. Rejection criteria have been arbitrarily set as follows: (f) The fitting error is greater than half the peak height, and (t) the peak or fit must be significant, i.e., greater than 10% full scale deflection. Once determined experimentally, the relative peak heights for each standard arc written into the program. To analyze an unknown chromato gram, the only data required are the 15 corre sponding jx'ak heights measured from the re corder churl, the GLC sensitivity in terms of the reference standard, and I rnntrol variable The MONS 084418 COREL: COMPUTER APPROACH Tb IDENTIFICATION OF PCBa 793 program outputs the actual PCB content in nanogrsms, the fitted jicak heights, and the aver age peak difference; specifies which peaks (if any) hove been rejected; and, if desired, pro duces & line chromatogram similar to Fig. 1. Running time is less than 2 min. Variation in conditions of manufacture mean that the composition of Aroclor standards will be slightly variable. This fact, coupled with other errors inherent in the gns chromatography, means that any 1 Aroclor comprising less than about 5% total PCBa is probably not significant and can be regarded as an artifact of the fitting pro cedure. It must be emphasised that the computer re sult does not necessarily determine which com bination of PCBs are or were actually present in the sample. Il merely describes the chromato grams in terms of a linear sum of 3 or more Aroclora. However, if one assumes that the iso mer composition of each corresponding peak is simitar in nil Aroclora, then the total weight of PCBs as determined by the computer program will be a good estimate of the true total PCB oontent. Metabolism and Weathering Residues in many environmental samples are subject to chemical change by metabolism, sun light, etc. Since PCBs are mixtures of many iso mers, the relative composition may change as well as the absolute concentration. Further changrs in compound ratio may be elicited by difference* in physical characteristics, for ex ample, the rate of diffusion. Published work indicates that the lower chlor inated isomers of PCBs are more easily metabo lized, whereas the Itcnvicr molecules are relatively resistant to change and tend to maintain com positional integrity (8, 9). If a considerable pro portion of peaks has been reduced in this way, then the computer fit will tend to bn poor and unstable. Such chromatograms may be analyzed by a modification to the program, which enables the fit to be carried out over a senes of selected peaks (Table 3). Thus if the early eluting com pounds have been heavily metabolized, the chro matogram can be fitted by using the longer re tention time peaks. Care must be exercised that enough peaks are used to produce a meaningful result, and there are sufficient major peaks for each Aroclor; other wise instability may result. If n sufficiently good fit is obtained thus, the original pre-metabolism chromatogram can be constructed by the com puter, and the degree of metabolism may be determined. Application to Cardboard The computer method is particularly suited to analyzing PCB contamination in cardboard and waste pulp samples, where metabolism is not ex pected and the integrity of die PCB patterns is largely maintained. Samples of virgin and waste pulps and cardboard were prepared for GLC by Soxhlct extraction, chromic acid oxidation (10), and silica gel chromatography (11) suitably modified to ensure that all PCB peaks were col lected (12). Good computer fit* were obtained with all samples: about 10% average peak differ ence for waste pulps and lionrda, and about 18% for the virgin pulp samples. Typical fits Are shown in Figs. 2-4. The computer analyses indicated a distinct mi WICWTI0H 1,Mt no. I--Compular fit of standard mhrhirs. taild Oar = campular fit hatehad bar a SIC data. Table I. Computer fit af matabaHraO mala, using various pa*** far fitting Peaks u**d Rl. wt Ral. wt Av. % for fitting Aroclor 1254 Aroclor 1262 peak cliff. 1-15 4-15 5-15 6-15 7-15 1-15 6-15 10-15 11-15 0.25 0.25 0.25 0.21 0.44 0.72 0.72 0.75 0.(6 0.21 0.21 0.20 0.22 0.25 0.16 0.16 0.19 0.16 54 12 21 21 25 26 10 15 21 r MQNS 084419 7M Mi. MftnriM rnN F1<L Z-Campulir fit f virgin groxindvrasd puis- JOURNAL OF TH* aoac (Vol. 57, No. 4, 1974) qualitative difference between all 3 classes of sample, pinpointing the additional conlumination as Aroclor 1242. This result concura wit h the use of 1242 in carbon-free copying paper, a well known source of PCBs in cardboard. Application to Pork Fat A sample of pork fat was analyzed in a similar fashion, except that a sulfuric acid shAke-out was added after the extraction to remove the bulk of the fat. A chromatogram resembling a heavily metabolized PCB pattern resulted. A modified program was used to fit these peaks, using the later eluting peaks only. The best fit, with 10% APD, was obtained by using peaks 9 through 15 (Table 3). The fitted chromatogram is shown in Fig. 5, demonstrating how a pattern of pre metabolism contamination can be built up In this case only Aroclors 1254 and 1262 were used because of the lack of adequate major peaks for 1242. Caution is required when labeling peaks as metabolized. The improvement obtained by fit ting only the later eluting peaks must be highly significant to justify acceptance of the fit. Discussion A new method of measuring PCB residues by computer analysis is presented. The total PCB content is calculated as well as the composition of the residue in terms of 3 (or any number of) standard Aroclors. While the computer program is primarily designed to quantify extracts which have maintained their compositional integrity, a HI IM < I' It It mi Half no. 4- CsmpUtsr fit W MmcM csrtvn *Mr4. IhFI|.1. SONS 084420 ZOBEL: COMPUTER APPROACH TO IDENTIFICATION OF PCBft 7W simple modification of the technique affords a description of heavily metabolized residues. It may prove possible to use this method to trace a source of contamination by resolving a residue into PCB types. The technique provides a means whereby complex chromatograms may be de scribed as simple mixtures of Aroclors, and thus enables the analyst to provide the layman with a more accurate and informative picture of PCB contamination than previous quantitation meth ods allow. Listings of the programs may be ob tained from the author on request. Acknowledgments The gift of Arocior standards from Monsanto (Australia) Ltd. is gratefully acknowledged. I thank L. P. Aldridge and J. P. M. Bailey for computing advice and for testing the feasibility of the method, and L. J. Porter for discussions on the cardboard analyses. RamusNcu (1) Gustavson, C. G. (1970) Environ. Sei. Ttch- nol. 4, 814-819 (2) Interdepartmental Task Force on PCB (1973) Rapt. Na. ITF.PCB.72-l, Nktionnl Trrhnical Information Service, US. Deportment of Commerce. Springfield, Va. (3) Hammond, A. L. (1972) Sctenre 173, 158-158 (4) Collins, G. B , Holme, D C 4 Jackeoa, F. J, (1972) /. Chromatogr. 71, 443-449 (5> Rote, J. W.. 4 Murphy, P G, (1971) BuU. Environ. Conlam, Toxicol. 6, 377-384 (6) Zitko, V, Hutringer, 0.. A Safe, S. (1971) BuU. Bnihron. Conlam. Toxicol. 8, 180-183 (7) Berg, 0. W., Diossdy, P. L., 4 Rea, G. A. V, (1973) BuU. Environ. Conlam. Toxicol. 7, 338-347 (8) Ramsey, L. I. (1973) Am. Food Drug Ofic. Quart. BuU. 37 , 43-57 (9) Bourne, W. R. P,, 4 Bogan, J. A. (1972) Mar. PoUut. Bull. 3. 171-174 (10) Westoo, G.. 4 Norm, K (1670) Acta Cham. Scand. 24, 1639-1844 (11) Snyder, D., 4 Remcrt, R. (1971) Bull. En viron. Conlam. Toxicol. 8, 385-390 (13) Moiumoto, H. T. (1972) JA0AC 55, 1092 1100 Rwvd November IS, IfTJ. MQNS 084421