Document GQBXpkBqBn2OyYbvkwGV6G2v

OCCUPATIONAL EXPOSURE TO PCBl AND VARIOUS CXMCiyj Report by the Xnduetriel Dleeeee Stenderde Peael IDSP Report No. 2 December, 1987 COMMITS by the NATIONAL KLSCTItICAL NAHUFACTUMRS ASSOCIATION WASHINGTON, D.C. USA Kerch, 1988 HONS 225001 0mi K. Ihitt V C* p'i'0"l P.O. C Dr. Robert G. Elgie Chairman Workers1 Compensation Board 2 Bloor Street Cast, 20th Floor Toronto, Ontario Canada M4W 3C3 March a, 1988 Dear Dr. Elgie: The National Electrical Manufacturers Association (NEMAI submits comments in this letter on the Report to the Workers' Compensation Board on Occupational Exposure to PCS. end v.ru,,. RCeanpcoertrsNpor.ep21areddatebdy DtheecemInbdeur,stIrfia87l . DisNeEaKsAe, wSthaincdha'rrdesprPeasnine^l .'(I2DMSP manufacturers of electrical equipment such as transfoJI2 .nd capacitors, has had a long and very close involvement'with Peis even pre-dating enactment of the Toxic Substances Control xJ?* i(MTSACuAfA) Cinturt*heolU,nCite,,d.MStatie.sgwrahdicuhasllpypehcaifaicina,llyoubtathnnaeidr*eontinuad NEKA also has a close association with the Electrical and Electronic Manufacturers Aasoclatlon of Canada (EEMAC) with several manufacturers having production facilities in both countries. in making these coranents, we have drawn upon our own extensive experience and knowledge of the health effects of pCBs. In addition we have asked Dr. Jack S. Mendel of the University of Minnesota to provide an objective scientific review of the XDSP report. The review of Dr. Handel and his associates is attached and forms the basis for our eosnents. we beliave that the large body of toxicologic and epidemiologie data does not justify findings of the Panel that there is a "probable connection between liver, biliary tract and gall bladder cancers and occupational exposure to PCBs." To the contrary, we believe the evidence does not support this finding. animal studies The review of the animal studies by the Panel is not complete and does not include a balanced perspective on the studies, studies showing an increase in liver carcinoma did not show an increase in total tumor incidence, in fact, some show a decrease in the non-liver carcinomas, in two of the studies mortality among the PCB exposed animals was actually decreased. MOMS 225002 lurch B, 19 SB Rsge 2 The studies do not "demonstrate ressonsble consistency" es stated in the Panel's report. The studies ere inconsistent, and pert of this inconsistency possibly stems from comparing the carcinogenicity of the different commercial PCB mixtures. Results show an association between fairly large doses of only the most highly chlorinated (e.g. Kanechlor 500 or AroclorTIeO) PCS compounds and site-specifie divert tumors. The effect appears to be species-spocific (primarily in rats). Extrapolation of the isolated findings in animals to humans is tenuous at best. Epidemiology Studies we do not believe the Bradford Hill criteria to assess the causal significance of an association are met. Contrary to claims in the Panel report, association is neither strong nor consistent. The strength of the association is substantially reduced if a two-tailed test is used, if an adjustment is made for multiple hypotheses testing, if only liver cancers are considered and the effect of confounding factors is removed. Consistency is not demonstrated since the individual studies do not all show a significant association. No dose-response relationship has been demonstrated. None of the individual human mortality studies presented by the Panel shows a significant association. The authors of these reports conclude that no association has been demonstrated. None of these studies accounted for confounding factors sueh as alcohol use, hepatitis infection, or use of oral contraceptives, known risk factor for liver cancer. All the studies lumped together cancers of the liver, biliary tract and gtll bladder even though the animal studies dealt primarily with liver cancer. Of the seven cancers of the liver, biliary tract or gall bladder considered to be signifi cant, four people had a duration of employment of one year or less, three were not confirmed and only two were classified on the death certificate as primary liver cancer. Three of the seven cancer eases occurred in one plant (Massachusetts) in one study (Brown and Jones). None of these was a primary cancer of the liver. Two of tbs three cases worked 1.5 years or less. The use of a meta-anelysis to sum the observed and expected results from the "five" studies is questionable because of the substantial dissimilarities among the studies. Such aggregated analyses have been used to date only with randomised controlled trials. HONS 225003 March 8, 198t Page 3 In summary, we do not agree with the finding of the Panel of a "probable connection batvaan livar, biliary tract and gall bladder cancers and occupational exposure to PCBs". Based upon our analysis of the data presented in the IDSP Report, w* are convinced that the evidence is not strong and the association cannot be considered causal. "Strong" and "causal" are used to describe associations such as those between cigarette smoking and lung cancer or asbestos and mesothelioma where the data ere consistent and persuasive. To use these words to characterize the association between liver cancer and PCB exposure is to distort and grossly exaggerate the weaX and inconsistent data that are available. we appreciate the opportunity to submit these comments to the Workers' Compensation Board of Ontario and trust that they will be given fair consideration. Please feel free to contact either me or Dr. Mendel directly should you need clarification of our remarks. We are also very willing to come to Toronto to discuss our review with the Board. Very truly yours DG3/bS CC-. MS. Linda Angove, Acting Secretary of the Board MOMS 225004 _ A Review of "Report to the Workeri Compensation Board oo Occupational Exposures to PCBs and Various Cancers* IOSP Report No. 2 Submitted by Jack S. Handel, Ph.O., H.P.H., Allen Williams, N.A., M.P.N. and Pat Coin, Ph.O, (cand.) March 1, 1988 HONS 225005 Summary Animal Studlet 1. The rwlow of th animal studies by th Panel It not complete and dott not Includo l balanced perspective on the ttudlet. 2. Th# ratultt fro* th# animal ttudlct thow an at toe 1it ion betw#*n fairly larga dot#! of only th# most highly chlorinated (e.g. Kancchlor 500 or Aroclor 12(0) PCS compound! and f1te*tpec1f1c (11vr) tumort. Th# effect appears to b# Ip#d#t*tp#c1f 1c (primarily In ratt). 3. Extrapolation of th# Itolatcd findings In anfauls to humans It tanuout at bett. Human Mortality Studies 4. Non# of th# Individual human mortality ttudlet pr#t#flt#d by th# Panel shows a tlgniflcant association. Th# authort of th#t# rportt conclude that no association hat bn d#monttratd. 5. Non# of that# ttudlet accounted for confounding fee tort tuch at alco* hoi use, hepatitis Infection, or use of oral contracptlv#t, known rltk factors for liver canc#r. (. All the ttudlet lumped together cancers of the liver, biliary tract and gallbladder even though the animal studies dealt primarily with 1iver cancer. 7. The ute of a mete-analysis to turn the observed and expected results from the *flve" ttudlet It guettionnable because of the substantial dissimilar 111 at among the ttudlet. 1 HONS 225006 8. The ut of a one-tailed statlitical iigniflcance test 1 not approprlate. 9. No account was made for tho fact that 21 tests were conducted to determine significance. Chine# lion# would result in th# significance of on# SMN. An adjustment to th# p v#1u# ihould 6# made for aultlpl# hypothec#! testing. 10. Of th# 7 cancers of th# liver, biliary tract or gallbladder considered to be significant, 4 had a duration of employment of one year or less, 3 wort not conflrned and only 2 were classified on the death certiflcate as primary liver cancer. 11. No dose*r#sponse effect was demonstrated. 12. Three of the 7 cancer cases occurred in one plant (Massachusetts) in one study (Irotm and Jones). None of these was a primary cancer of the liver. Two of the 3 cases eorked 1.8 years or less. 13. The Iradford-Hlll criteria to assess the causal significance of an association are not met. The strength of the association Is substan tially reduced If e two-tailed test Is used. If an adjustment is made for multiple hypotheses testing. If only liver cancers are considered and the effect of confounding factors is removed. Consistency Is not demonstrated since the Individual studies do not all show a signifi cant association. No dose-response relationship has been demonstrated. 2 MOMS 225007 Review pf "Report to the Horkprs Compensation Board on Occupational Fiooiurt to PCBs nd Various Cancers" IOSP Report No. 2 Introduction In Oeceeber 1987, ths Industrial Olsease Standards Panel of the Ontario Ministry of Labor subolt tad to the Mortars' Compensation Board a raport on occupational exposure to polychlorinated biphenyls (PCBs) and cancart. This raport was in rasponsa to a raduast from tha Workers' Compeniatlon Board to considar tha issua of tha human carcinogenicity of PCBs for pur. posas of determining entitlement for workars occupationally tipostd to PCBs ho hava developed cancar. Tha basis of tha Panal's rec (emendations is th* Raport of tha Special Panal on Occupational PCB Exposure and Various Cancars: Hunan Health Effects and Carcinogenic Risk Potential of PCBs*. This Special Panal. chaired by Dr. Mil Han Nicholson of the Mount Sinai School of Nadtclna, New York, raw lowed tha experimental an Inal studies and observational human studies and conducted a neta-inalytls of tha epldemlologic aortality studies to reach tha conclusion that there Is, *a probable connection between IWer, biliary tract and gallbladder cancars and occupatlonal exposure to PCBs". following is a critical review of the Special Panal's raport. Experimental Animal Studies Although the Special Panel does not explicitly define tha scope of. or approach to. tha review of tha exparlaantal PCB literature, it appears that the Panel sought to provide a soaawhat broad and general overview of this literature, as opposed to a coaprehenslve and critical review. The Panel 3 MOMS 225008 cites soft, but not *11, of th major animal studies cltad by other reviewers. Furthermore, th*y do not dlicuti all the Important finding! m tha itudlai that they review. Tha first study cltad on PCI*Induced liver cancar was that of Ito at al., 1973, who fad mala mica a dfat of 100, 250, or SOO ppm of Kanachlor 300, Kanachlor 400, or Kanachlor 500. Neoplastic nodules and hepatocellular carcinomas were seen after 32 weeks only In tha proop exposed to 500 ppm of Kanachlor 500. Sevan of 12 animals ware found to have neoplastic nodules and S of 12 hepatocellular carcinomas. It should be noted that tha affect was observed only In tha mica fad tha diet with tha greatest (500 ppm) dote and most chlorinated (54*) PCI compound. Tha second study of mica (Kimbrough and Linder, 1974) found evidence of nonmallgnant lesions in 47* of the mice after an 11 month diet of 300 ppm of Aroclor 1254. Six studies on rats were reviewed In the Special Panel's report. The first (Klmura and laba, 1973) found neoplastic nodules In only female rats fed over 1200 mg of PCI In their diet. No mention Is made of the second study by Klmura at *1. In 197* which did not find an association between hepatocellular carcinoma and PCI*. The Panel concluded that the study by Ito at al. in 1974 demonstrated that, "all three Kanechlors produced liver changes related to the concentration In the diet.* Table 3, presented in support of this statement, contains data on the hlstopathological findings in the liver of rats given Pels. At the highest dose (1000 ppm) of the most chlorinated compound (Kanechlor 500), 30.11 developed nodular hyperplasia. Thirty percent and 31.3* of the rats developed nodular 4 MONS 225009 hyperplasia when fed 1000 ppm of Kanechlor 400 and 500 ppm of Kanechlor $oo respectively. Lower doses and lets chlorinated PCIs thowed virtually no effect. Noteworthy it the Panel's failure to ttate that no hepatocellular carcinoma! were found in any of the rati Including those fed the highest dose-chlorination combination. In describing (p. 20) the results from Kimbrough et al. (1975), the Panel approprlately acknowledges the substantially increased incidence of hepato cellular carcinomas and neoplastic nodules in experimental animals In this study. They go on to mention that uterine cancers were "also elevated, but not to a statistically significant level*. (lamination of Table 4 in the Panel's report, which presents Kimbrough's data suggests a more complex picture than the Panel's description. It can be seen that for a number of tumors Incidence appears to be reduced, perhaps slgnlftcantly, in the exposed animals compared to the control animals. In fact, for all tumors other than those of the liver, the Incidence was somewhat greater among the control animals than In the animals fed PCIs. PCIs were also not life shortening as almost twice as many control animals died before the experl* ment was terminated. In the study of Norbeck and Weltman (191$), some significant findings described by the authors are net noted by the Panel. Although the striking sex-related difference In tumor rates Is mentioned, the Panel did not men* tlon that the mortality rate was not Increased In the PCI exposed group. An example of a major animal study that is not cited by the authors is that 5 HOMS 225010 by Schaeffer and bis colleagues published In 1984. This study reported significantly increased incidence of neoplastic nodules and hepatocellular carcinomas In Ulster rats receiving CTophen A60 (but not Clophen A30). As noted in a subsequent letter by Young (1985), this study also showed a significantly reduced risk of other tumor types and reduced mortality in both sets of PC8 treated animals. Schaeffer et al. considered the protec* tlon from neoplasms of the thymus and from nephritis to be the result of PC8*1nduced alterations In the lumune system. It Is not clear why the Panel did not include this study; such an omission reduces the confidence in the completeness and balance of the review. Another study not acknowledged by the Panel was by Kerkllvet and Mmeldorf (1977) In which they reported that PCS exposure In rats, *had a significant protective effect on the host in response to the (plantation and growth of Walker 256 tumor cells.* The one study of dogs was essentially negative at neither hepatocellular carcinoma nor neoplastic nodules were found In the dogs fed up to 100 ppm of Aroclor 1242, 1254 or 1210. five animal studies were reviewed that deal with the cancer*promotlng effects of PCIs. As with the studies on carcinogenesis, the effect appears to be limited to the highest dose-chlorinated PCI compounds. Summary of Experimental Animal Studies 1. The review of the animal studies provided by the Panel is not complete and does not include a balanced perspective on the studies. 6 HONS 225011 Z. In addition to tho incomplete presentation of the major findings, the authors alto do not provide adequate doscrlptlons or critiques of the conduct of tho studios - Information that portalni to tho overall intorprotatlon and validity of tho studios. For example, they do not consistently indicate overall mortality or lost of body weight. Thus it is much mort difficult for tho reader to evaluate these finding! without searching out the original literature. 3. The studies showing an increase in hepatocellular carcinoma incidence show these increases at fairly substantial doses with the most highly chlorinated compounds. 4. Studies shoeing an Increase In liver carcinoma did not show an increase in total tumor Incidence, In fact, some show a decrease In the non11vor carcinomas. 5. In two of the studies mortality among the PCI exposed animals was actually decreased. 6. The studies do not, 'demonstrate reasonable consistency* as stated in the Panel's report. The studies ere Inconsistent, and part of this Inconsistency possibly stems from comparing the carcinogenicity of the different commercial PCI mixtures. The more highly chlorinated PCI isomers are most carcinogenic, as the Panel's report states. Kimbrough (1917) goes further, stating that It is not certain that the less chlorinated fractions (pentachloroblphenyls and less chlorinated) cause hepatocellular carcinomas In rodents. This has implications for the 7 HONS 225012 possible risk of occupational exposures, iinca the lata chlorinated comarcUl mixtures (0.9. Aroclor 1242, 1254 and 1016) vara more com monly mad than Aroclor 1250 (Kimbrough, 1987). of course tha lata chlorinated commercial mixtures ara not davotd of tha highly chlorlnatad congeners, but tha concentrations of these congeners ara lower in tha less-chlorinated rnUes (a.9. see Tabla 2 of tht Panel's report). 7. Aroclor 1254 (as opposed to Aroclor 1260) has not definitely bean shown to be hepatocarcinogentc In anlsals. Tha study most frequently reviewed is that of the National Cancer Institute (1978). Thera ware statis tically non-sipnlfleant Increases of hepatocellular carcinomas and hematopoietic cancers In exposed rats. Therefore, the study was equiv ocal at best and Is negative by the criteria set for the NCI bioassays. The fanel criticises this study on the basis of an Increased eortal1ty in the exposed groups, which they state leads to a decreased number of animals at risk for cancer In the dosed groups. While this Is true, an alternative argument may also be valid. The excessive mortality In the dosed groups Indicates that the doses exceeded the maximum tolerated dose and that therefore the assay Is Invalid. The NCI guidelines require that the maximum dose of test agent used in a bioassay be one that does not alter the animal's normal longevity unless death is due to tumors (National Cancer Institute 1976). In this NCI report it sta tes, *A Maxi** Tolerated Dose OTTO) should be selected for each sex of each strain to be used In the chronic study. The NTD Is defined as the highest dose of the test agent given during the chronic study that can 8 HONS 225013 be predicted not to titer the tnimtls' normtl longevity from effects other then ctrctnogenedty. 8. Still controversy is the relttlonship of hyperpltstic liver nodules to the development of ctrdnone. The Ptnel recognises thtt the studies suggesting thtt the hyperpltstic nodules ere precursors of mtllgntnt tumors ere. "not completely definitive". Pitt (19M) cites this Interpretttlon ts one of two theories of ptthologlc progression of liver etneer, Stfe (1984) stttes thtt these hyperpltstic nodules only tppetr In experiments tnlmtls tfter long exposure to high doses. Hurntn Morbidity Studies t. Yusho Oisetse Kimbrough (1917) egrets with the Panel's report thtt the Yusho end Yucheng poisonings ctn atlnly be escribed to the polychlorinated dibentofurtns (PCFOs) present In the conttmlntted rice oil. Kimbrough tlso mentions t highly tonic PCI Isomer (3*4,3',4' tetrtchlorobiphenyl) ts being present In the conttmlntted oil. Kimbrough notes thtt this isomer hts not been Identified In cooMrclel mixtures used in the U.S. The Yusho end Yucheng episodes do not seem very relevtnt to the possible ctrclnogenlc properties of PCIs. Kimbrough (1917) stttes thtt there Is no definitive evidence ebout cencer rotes In the Yusho cohort. Tho number of ctses tppetring so ftr hts been smell end t sufficient lttent period mty not htve yet elapsed. Furthermore, although PCIs htve toxicologic properties eutlltttively similtr to 9 MOMS 225014 those of polychlorinated dlbentodloxlns (PCOOs) end PCOFs, the responses *rt not Identical. In particular 2,3, 7,1 dlbeniodloxln (TCDD here) con Induct tumors tt variety of sites In rodents, while pels Ion# hv# only Induced 11v#r tnd ttomtcb tumors (Kimbrough, IBIS). Furthermore, the exact petterns of eniyne induction of TCOO r# mimicked only by a few PCI congeners heving specific chlorination pet* terns (Sift tt 1., 1915). b. Skin end Othtr Cutmtous Tlssuts Skin lesions (cneform eruptions, folliculitis and possible skin thickening) appear to be associated with high exposure to PCIs, par ticularly the higher chlorinated isomers. c. Hood Chemistry Several studies have been conducted of workers exposed to various con centrations of PCIs in which serum blood analyses were obtained. The results from these studies are summer lied in Table 12 o' the Panel's report. The Panel concludes *in considering the overall results, SWT and fiSTP particularly are increased with exposure to PCIs*. The results from the SWT studies given in Table 12 show that only 3 of the 10 studies had significantly positive associations and the studies of MTP show that I of 10 studies had significantly positive association. The Panel concludes that, amost of the results are negative and when positive, the effect Is relatively minor.* A number of studies have looked at cholesterol and triglyceride con centration In relation to PCIs. Mar# the Panel over Interprets the 10 HOMS 225015 results tumnarlzed In Table 13. They conelude that, "statistical1/ significant Increases In trfglycerldes nd total cholesterol were asso ciated In tone studies with incraitad serum PCI concentrations11. However, only 8 of the 20 r*|0lts in Table 13 are significantly ano. dated. Furthermore, the Panel notes that "some studies" did not fully control for body mass and It Is not clear whether the studies showing a posi tive association were among those that controlled for body mass. The Panel also does not make It clear that the association can probably be explained by the increased solubility of PCBs In serum with a higher lipid content as suggested by Kimbrough (1987). Liver Abnormalities The report states that the increased serum liver enzyme levels are of uncertain clinical significance. They find more alarming a report of hepatomegaly In capacitor workers (Moroni et al, 1911), but state that this Is an Isolated finding. They point out (p.2l) that, "Other clini cal studies have not reported liver disease of such severity, if at all." It Is also 1shorten t to note that the study of Maron1 et al. did not have a control group( nor were the examining physicians blind to the PCI exposure status of the workers. They did, however, find an association between serum PCI levels and the frequency of "abnormal" liver findings as "deduced from symptoms, clinical examination of the liver and laboratory tests." 11 MOMS 225016  . Cardiovascular Effects Generally, th associations between ret exposure and blood prassure have disappeared when the confounding factors of ape, sex, imotlng and bod/ nasi were considered in the anal/sls. One study (Krelss et el, 1981) did find an association with blood pressure after eontolllng for age, sex and body ness. Apparently smoking was not controlled and given the low PCI serum levels (average 17.2 ppb, range 3.2 - 1S8). It Is unlikely that the association was real. f. Respiratory Effacts The Panel properly concludes that, "the single study Indicating decreased forced vital capacity among PCI exposed workers must be con* firmed In other populations*. g. Reproductive Effects The finding from the studies on reproductive outcomes are weak and unlmprasslve. The Panel refers to the results as 'uncertain* and 'suggestive*. h. Cancer Morbidity One study of cancer among workers Is reviewed. The study has not been published in a peer-reviewed journal and apparantly no control group was included. However, no case of liver cencer was reported among the exposed workers. The Panel appropriately points out (p. 31) some of the limitations of the existing morbidity data from occupational exposures and presents 12 HONS 225017 toff* cl In 1c*1/p1d<o1o1ogIc considerations for future studies. They mention thet in eppropriite group for the study of pregnency outconws would be difficult to obtein. Perhaps in this regard, the potential for exposure to spouses or other household contacts of PCI exposed workers should be considered (Fischbeln and Wolff, 19|7). Human Mortality Studies The mortality studies of wrkers presumably exposed to PCIs are used as the foundation for the recommendations by the Panel. Four epidemiologic mor tality studies have been done; one drown) was conducted at plants in two different states and hence, the results are presented as if S studies have been done. Table IS provides the characteristics of the populations from the four major studies. Particularly noteworthy from this table is the fact that all the studies were of relatively few people and given the rela tive rarity of liver cancer In the population, none of the studies would have sufficient statistical power unless the magnitude of the 941 was extremely large, which apparently it Is not. Somewhat troubling is the manner In which the data were pooled (ate*enalysis), particularly since the panel had to rely on unpublished data provided by the authors to con duct the combined analyses. Tables 1$ and 17 of the report provide the observed and expected number of deaths by cause of death for males and famales from the 'five* studies. The cohort described In Table 15 for the Mcholson et al. study is not the same as the one used to obtain observed numbers of deaths in Tables 16 and 17. This Is a bit misleading and confusing as It suggests that the study 13 HOMS 225018 ttic was larger then it actually was. Tablet 16 and 17 include deaths among study subjects not included In the Broun study. In Table 16, the number of deaths from all causes (43) Is less than the number of deaths from cardiovascular disease ($3). It is also somewhat surprising that the expected and observed numbers of deaths for all causes are identical for males and females (43 observed and 4S.49 expected). Individually, the results from the studies would be Interpreted as showing no association between any cause of death and employment In that particular company. However, when the observed and expected number of deaths are tunned over all the studies, a significant (p .016) SMI is found (see Table 18) for death from cancers of the liver, biliary passages and gallbladder based on 7 observed deaths and 8.64 expected. The Panel used a one-sided statistical test when a two-sided test would have been more appropriate. Also, since 21 comparisons were made, the conventional p value of 0.06 Is not appropriate because of multiple hypothesis testing. The Panel attempts to summarise the results within the framework of the Bradford-HIll criteria for assessing the causal significance of obser vational data. First, with respect to strength of the association, the Panel accepts the feet that since the SMI Is statistically significant, this criterion is fulfilled. However, none of the individual stud las is significant end, using the meta-analysis, the result Is not sufficiently strong to satisfy this criterion (particularly If a more conservative p value and 2 tailed test are used at wuld be appropriate in this situation), under consistency, no mention is made of the fact that none of 14 HONS 225019 the five studies reviewed show a significant association, in fact, th studies are remarkably consistent In that they all fall to demonstrate an association between PCI exposure and liver cancer, with respect to bio logical plausibility In reference to th supporting evidence from animal studies, the Panel concludes that such support is apparent from the studies reviewed in the document. However, although some of the animal studies do suggest an association, they are far from conclusive as sumaerized in the earlier sections of this report. Furthermore, no discussion Is provided on extrapolating from the doses used in the animal studies to the doses experienced by the workers in the four studies. Hor Is any consideration given to the type of PCB compound. This is important because only the imre highly chlorinated ones showed a carcinogenic effect with animals. Also, the general issue of extrapolating to humans from animal data Is not addressed In the paper. Kimbrough (IBB?) states, "From empirical obser vations, humans also appear to be less sensitive to the toxic effects of PCBs. The ability to store these chemicals In adipose tissue may be protective." With respect to time relationships, the Panel overlooks the fact that 4 of the 7 cases had a duration of exposure of 1 year or less (see Table 22). Furthermore, only two of the cases were actually liver cancer, the others were biliary system or gallbladder cancers and of the 7 cases, 3 were not confirmed. In their analysis, the Panel combines tumors of the liver, biliary duct and gallbladder. The Justification given for this grouping Is the finding of "preneoplastic lesions" In the biliary tract in a rat 15 MONS 225020 study. However, there ere many differences awing these sites in humans with respect to geographic variations in incidence and mortality rates, secular trends, sex ratios and known or suspected risk factors (Fraumonl and Kanton, 1982; Falk, 1982). Thus grouping them is not appropriate. Absent from the document Is any discussion of confounding factors. This Is particularly Important given the small numbers Involved In arriving at the conclusion that an association exists between PCIs and liver, biliary tract and gallbladder cancers. Seven cases wore observed, 2.54 expected. It Is appropriate to evaluate the potential effect on this association of a con founding factor such as alcohol use or hepatitis I Infection (Falk, 1912). A smell difference In the prevalence of risk factors between exposed and control groups could be of importance. Of significance is the fact that no doso>response relationship was found. This would be another issue in evaluating the causal significance of the association. As stated In the document, "as can be seen from Table 24 [should read 23], there Is no significant trend with Increasing estimated risk when the durations of exposure are periods of employment in the high PCI exposed Jobs*. The report goes on to state, "however when the deaths In the cohort of Irown are categorized according to total plant employment as other deaths are, the distribution corresponds to an increasing risk of cancer with Increasing estimated risk*. This Is a clear over interpretation of the results since, as noted by the Panel, the trend Is not apparent and the data are not statistically significant. 18 HONS 225021 Son* specific consents regarding the Individual mortality studies are given below. Brown and Jones {1911) and Brown. (1986) As of the end of February 19SI, the Brown study update had not yet appeared In the published literature. The 1981 study was published and available for review. 1. The original study reports that 163 Individuals died from 1940-1976, not 160 at stated in the Panel's report. 2. In addition to not observing an increased risk with greater latency. Brown and Jones also did not observe an association with duration of employment. 3. The Panel addresses the possible impact of using u.S. mortality rates, rather than local (county) rates where plants are located, in calcu lating expected numbers of deaths. With respect to liver and biliary tract cancers, the Penal states that local rates were higher near one plant (Massachusetts), but lower near the other plant (New York). Three out of the 4 liver cancer or biliary tract deaths occurred in the Massachusetts cohort. This cohort also accounted for two-thirds of the total person-years (1911 study). Contrary to the assertion of the Pane), these findings do suggest a possible bias. If the higher local rate of liver cancer where the Massachusetts plant is located were used to calculate the 'expected* value rather then the lower u.S. rates, a higher expected number would probably be obtained, which In turn would lower the SMR for these sites. 17 MOMS 225022 4. The originl 1981 Study found no trond In liver cmcir deaths 1th Increasing latency. 5. Brown and Jones (1911) found 4 trond In mortality from cirrhosis of tho liver with Increased length of employment which they considered to be consistent with the notion that PCIs have a toxic effect on the liver.* The 9* Increased from 74 for those employed between 3 months and S years to 43$ for those employed between 15 and 19 years. They state that alcohol consumption is a possible confounder of the asso ciation between PCI exposure and liver cancer. Custavasson et al. (1918) Several important aspects of this study were not noted by the Panel. 1. There appears to be an inconsistency In the mortality data presented In Table III in the original article. Total deaths are listed as 21. Under this Total are five major categories of deaths which sum to 27. 2. Neither the study authors nor the Panel provide the person-years of observation for this study, although study power was briefly discussed by the authors. 3. The Panel notes that this study utilised Swedish national rates to obtain expected deaths and that this may have underestimated the deaths expected in this cohort. The Panel then states that use of local rates would have, *only a trivial effect on the overall results of this review*. While that may be true, the effects may not be trivial II NONS 225023 regarding tht findings of this study. Tht offsets could ba quits signlflcsnt if thsrs art largo dlffsrsncss botwaan local and national rstss for spsclfic tumor sltas. 4. Ths on* llvsr canctr (actually cancsr of tht ampulla of Vatsr) rtportsd from this cohort was not rtporttd as such In tht original study which did not havt a dttalltd brsakdown by site. Tht Inclusion of this cast is bastd on a ptrsonal communication from Sustavasson to th* Pans). Bartani tt al. (1912) and lartani at al. (1917) 1. According to tht Panal and tht Investigators, tht tarlltr study (1902) Included only production worktrs who had boon tmploytd for at least six months bttwttn 1940 and 1970. In tht updattd study (1917), savora1 major changts wtrt madt In tht dtflnltlon of tht study cohort, first, tht tllglblllty ptrlod was txttndtd to 1970. Sacond, tht minimum ptrlod of employment was raduetd from six months to ona waak. Tht authors gavt two rtasons for this changt; ont wttk was appartntly tht minimum ptrlod ntetsstry to obtain reliable data on an employee, and loss than six months of employment could product significant txposurt to PCBs. Third, til company tmploytts, not Just production amploytas, war# includtd In tht lattr study. Thus any company amploytt with at Itast ont wttk of amploymant bttwttn 1940 and 1970 was Includtd In tht study* 2. Tht study authors did not makt a particularly strong cast for tht changts thay madt In dtflnlng tht study cohort. I.t., tht Inclusion of 19 HONS 225024 nonproduction employees nd thou with work histories at short at 0na week. Tht avldtnca dud for significant short-term xposures at a 1954 study of thru casts of chloracn# in tht company among autoclave worktrs who had worktd 4, $ and 7 months, respectively. Although four othtr caus of chloraent wort dlagnostd In 1977 among worktrs Imprtgnatlng capacitors with KBs, tht work duration of thtst worktrs Is not nottd and ntlthtr thalr own data nor thost of othtr publlshtd studlts trt prtstnttd to support tht coattntlom that short-term workars can gtntrally have significant exposures. Tht Inclusion of nonproduc tion worktrs was also not will justified. In addition to several air samp Its of KBs, a numbtr of worktrs wort assayod for KBs on thtlr hands and In thtlr blood. Unfortunattly, tht authors do not adequately sptclfy tht locations of tht tnvlronmtntal supIts, or provide Infor mation on how tht worktrs wort stltcttd, thtlr occupations or work histories. It was only stated that thay represent "typical* exposure conditions. Thus, It Is aot at all clear that nonproduction workers or very short-term worktrs had significant exposures to KBs or that tht exposure potential would havt been constant over different time periods. The authors do not Indicate tbe proportion of the cohort that fell into these categories. 3. A mere significant U^adlment to Interpretation of this study Is that the authors did not havt available to them detailed employee work histories or Job descriptions, and could not classify workers according to work history. 20 HONS 225025 4. The authors did not find an/ relationship between mortality nd length of employment, latanc/. or /car startad. However, as the authors and Panel note, the snail number of daaths Unit an/ conclusions that can bt drawn. 5. It Is sonowhat curious that tha authors prasantad characteristics of `selected" cases of cancar daaths in Tablas VI and VII for nalas and females, respectively. In Tabla VI, tha/ actuall/ prasantad all 12 mala daaths, whila in Tabla VII, tha/ prasantad onl/ 4 of tha 14 female daaths. hicholson at al. (1987) This is an unpub11 shad stud/, howavar a draft of tha papar was made aval labia for ravlaw. 1. Tha Panel notes that this study was conductad on ona of tha sane plants (Plant 1, haw York) that was includad In tha drown and Jonas stud/. Howavar, tha invastlfators daflnad thair cohorts quite dlffarantly. drown and Jonas defined thair stud/ cohort as an/ workar with at laast ] months of sarvlca at an/ time In an eraa with high exposure potential - as daflnad by tha company and corroborated by tha union and a NIOSH industrial hyflana survey. Tha Nicholson study cohort consisted of all plant workers with at least five /ears of sarvlca whose employment bapan before 1*54. furthermore, Nicholson allowed for a 10-yaer laten cy; follow-up did not occur until after 10 /ears from first employment, because of tha different requirements for length of 21 HONS 225026 MplojMfit and tlaa parlod of anploynant, It 1i not surprising that tha Nicholson cohort was smallar 7*9 workars. Tho Panal notas that ovar two-thirds of thasa wrkars (521/769) war a not includad in tha Brown and Jonas cohort (and It is only data for thasa workars that is apparantly prasantad In Tablas 16 and 17). This finding would suggast that aost of thn Nicholson cohort did not hava tha high-axposura Jobs as idantlflad by Brown and Jonas. In any avant, In tha absanca of mra datallad axposura Inforaatlon. It Is unclaar whathar tha Nicholson cohort or tha Brown and Jonas cohort providat a battar or adaquata study group* Usa of a Mata-Analysls Tha application of ata-analytls to ratrospactlva cohort studios is ralatlvaly now. It hat baan nora oftan ustd for axparlaantal studios whara rosults froa a nuiabar of randoalxad trials hava boon poolad. In thasa situations a nuabar of crltarla had to ba not bafora combining tha data froa aany studlas. Tha Panal's raport doas not addrass tha approprlatonass of conducting tha aata-analytlt in this situation. Thara ara no foraal spaclflcatlont In tha Banal** raport for Judging that tha ttudlat wara sufflclantly slallar to alloa pooling. Thara nri tufflclant dlffarancas among tha studlat that caald Invaildata a aata-analysls, Tha studlas wara dona in dtffarant countrlat (U.S.A., Italy and Swadan) and Involvad dlffarant avaraga durations of aaployaant, dlffarant Intansltias of axposura and dlffarant daflnltlont of allglblllty for Inclusion Into tha study cohort. Bartaisl at al. did not hava work hlstorlas and includad all plant 22 MOMS 225027 employees Mho worked one week or nore, whereas Irown and Jon** Included only ork*r* with likely exposure to PCI* with *t least three months of work experience. Nicholson et *1. Included only employees with five year* of service and at least ten years of followup (latency). Inclusion of all relevant articles is important In n*t*-*n*1ys1s. The Panel does not state what Methods ware used to ensure inclusion of all studies. Studies with negative results are less likely to he published than studies with positive results (Sacks et al, 1917). Thus, it is iwportant to provide a well defined framework to survey the research field and identify all studies which address the Issue. The Panel presumes to have identified all studies, however, they do not describe a process for con ducting the search. Sunwary The panel concludes (p. 49), In suMaary, while only seven deaths from liver and biliary tract cancer occurred, the criteria of Hill for an association with exposure are reason ably well net." The Hill criteria are not aet. The only criterion that has soae Merit is biological plausibility. The association Is Strong (991 27$) and achieves statistical significance." It is euestionable If significance would be attained If a nor* conservative p value was used (because of Multiple hypotheses testing) along with a 2-talled test. Furthermore, 3 of the 7 cases were not confined. 23 MOMS 225028 *A finding of excess cancer It present In the workforces of four 0f the five plants studied; tho overall excess did not come from on# sporadically high finding." 7 Of tho S pUntt in idilch Mloi were studied, 2 had no deaths from cancan of tho liver, biliary passages and gallbladder, 2 had 1 death and 1 had 2 deaths. Of the 4 plants chi eh Included women, 3 had no deaths from these causes and 1 plant had 3 deaths. (The Utter is from an unpublished study.) Of the four published studies, Irown and Jones state, "There is no relationship between Increasing duration of amplo>ment In jobs Involving PCS exposure and the risk of mortality due to cancer or cirrhosis of the liver*, lertaxxl et al. state, "...both workers who died from cancer of the liver and biliary tract had been employed in the production area, albeit for rather short periods". They later state, "The limitations discussed did not permit a causal assocatlon to be either proved or dismissed." lustavsson et al. does not even mention the one liver cancer cited in the Panel's report but do state, "A subgroup of It individuals with higher exposure (capacitor fillers and repairers) was analysed separ ately but there was no tendency towards an increased mortality or cancer incidence in this subgroup." And finally hlcholson, the chair of the Special Panel concludes Ms own study with the statement, "The overall results do not Indicate any association of PCI exposure with excess mor tality for any cause." A final point, the excess did Indeed come mainly from one sporadically high finding. Three of the 7 deaths (431) occurred among females In one of tho plants in the Irown and Jones study. "The concordance of the animal data and the finding of liver tumors in humans exposed to dlbeniofurans, structurally similar compounds, adds weight to the human evidence." 24 MOMS 225029 The animal data and their limitations were in thii report. Begirding the dlbenzofurans (PCOFs) Kimbrough (1987a) states, `The toxicity of Individual isomers of the PCBs, PCOOs (dibentodloxlns) and PCOFs varies graatly. The type of effects produced by the same compound and the target organs nay differ in different species and may differ for various honologs of this mixture of chemicals. Species variations in response to PCOOs and PCOFs have some similarities, but also some differences. This makes risk esti* mates extremely difficult.* The exposure and time relationship of the cancer Incidence are In accord with expectations.* no dote retponto relationship was demonstrated In any of the Individual studies nor in the combined data, four of the 7 cases were employed for one year or lets. Finally, no other exposure or circumstance could be Identified that would provide an alternative explanation for the excess.* It must be pointed out that none of the studies considered other exposures. Important confoundlnf factors such as alcohol uso, hepatitis I virus infectlon or use of oral contraceptives were not considered although trowi and Jones recopnlted the need to consider alcohol use, particularly since they found a larpe Increase In clrrhosla of the liver. And finally the Panel concludes, 'Considering all the data, the evidence Is strong for a causal association between the excess of these cancers and exposure to PCIs.* 2S MOMS 225030 The evidence It net strong and the association cannot be considered causal. 'Strong* and 'causal* are used to describe associations such at those between cigarette saoklng and lung cancer or asbestos and aesothelloea where the data are consistent and persuasive. To use these words to characterise the association between liver cancer and PCI exposure Is to distort and grossly exaggerate the weak and Inconsistent data that are available. The an leal data are suggestive, however, the huaan data are weak and unconvincing. 21 HONS 225031 EFCNCES Bartazzl PA, klboldl l, PosatoH A, kadlce L, Zocchettl C. Cancer nortelIt/ of captcltor manufacturing workers. Am J Ind Mod 1917;11:1(5*76. Brown OP, Jon#* M. Mortality end IndottrUI hygiene study of workers exposed to polychlorinated biphenyls. Arch Environ hat 1th 19tl;3G:120-9. Falk H. liver. In: Schottanfald 0, Fraunanl JF, ad*. Cancar Eoldaaioloov and Prevention. Phlladalphla; Ml Saundars, 19I2:*-I2.*"--"------------ " Flschbeln A and Wolff, it. Conjugal exposure to polychlorlnatad biphenyls (PCI*). Brit Indust Itad 1917;44:284-6. Fraunanl JF, Kantor AF. Hilary tract. In: Schottanfald 0, Fraunanl JF, ads. Cancar Coldanloloav and Fravantlon. Phlladalphla; Wl Saundars, 1992: GH-M. Gustavston P, Hogstedt C, kappa C. Short-tarn Mortality and cancar Ind* danca In capacitor manufacturing workers exposed to polychlorlnatad bl* phanyls (PCI*). Am J Ind Mod 1917;10:341-4. Karkllvat HI and Klaaldorf DJ. Antltunor activity of a polychlorlnatad blphanyl nlxtura, Aroclor 12S4, In rats Innoculatad with Walker 2SG car* clnosarcoMa calls. JNCI 1977; S9:951-SS. Karkllvat HI and Klaaldorf DJ. Inhibition of tuaor growth In rats by faadlng a polychlorlnatad blphanyl, Aroclor 1294. lull Environ Contaa Toxic 1977;11:243. Klnbrough 10. laboratory and human studios on polychlorlnatad blphanyl* (PCIs) and ralatad coMounds. Environ Koalth Porsp IMS;$9:99* 101. Kimbrough RD. Human haalth effects of polychlorlnatad blphanyl* (PCIs) and polybroalnatad blphanyl* (PMs), Am Rav Pharmacol Toxicol 1M7;27:I7-111. Kimbrough R0. Toxicology of haloganatad blphanyl*, dlbanzodloxlns and dlbenzofurans* I SI Atlas of Science:Pharmacology lM7a; 1:139-42, Klauro RT at al, Polychlorlnatad blphanyl* as a proaotar In axparlaantal hap atocarcInogeno*1s 1971. in: Karblson R0 at al. Biological data rele vant to tha evaluation of carcinogenic risk to humans. August, 1M7. Msronl M, CoTombl A, Canton 1 S, at al. Occupational exposure to poly chlorlnatad biphenyls In electrical workers. II Health affects, lr J Ind Mad 1991;3t:SS-IO. National Cancar Institute. Guidelines for carcinogen bioassay In mall rodents. National Cancer Institute Carcinogenesis Technical Report Sorias 1974;1. (OHCM Publication no (NIH) 74*101). 27 HONS 22S032 National Cancer Institute. Hoessay of Aroelor 1254 for possible car. dnogentclty. National Cantor Inatltuto Carcinogenesis Technical Report Series 197i3. (Wild Publication No (NIH) 70-030). Nor back ON, Neltaen RK. Polychlorinated biphenyl Induction of hepetocellular cerctnoaa In the Sprejue-Dewley rat. Environ Health Perspect 1905; 50: *7-105. Ptaa OL. Toxic responses of the liver. In; Klaassen CO, Aadur MO, Ooull j. eds. Casarett and Ooull*s Toxicology. Now VorktNacalllan, 1900:201-309. Socks HS, Oerrler 3, Reltnan 0, Ancona-Oerk VA, Chalaert TC. Mote-analyses of random led controlled trials. N Cn1 3 Med 1907;310:450-5. Safe S. Polychlorinated biphenyls (PCOs) and polybroalneted biphenyls (POIs); Hochenlstry, toxlcolofy, and aechan Isa of action, CRC Crlt Rev Toxicol 1904;13:319*95. Safe S, Oandlere S, Sawyer T, at el. PCOs: Structure-function rela tionships and aechenlsa of action. Environ Health Perspect 1905;50:47-55. Schaeffer , Orela H, Soossner w. Pathology of chronic polychlorinated biphenyl (PCI) feeding In rats. Toxicol Appl Pharaacol 1904;75:270-00. Young S. Letter to the editor. Toxicol Appl Pharaacol 1905;70:321-2. 20 HONS 225033