Document e5JyD9YEQ4rx5e9ZZQLZmZ7Zm

Monsanto Company 8 0 0 N it V iftbergh B o u lev ard * Su Louis, Missouri 63167 Bulk tta lr U. N Postage P A ID Sr. LouU. M O V n iiir N o . 2935 \ il l muv L f i i ^R FROM THE PRESIDENT .. a^- h; . U.^' In 1989, Monsanto Company joined with other major chenw^.a>inp^wes:to- establish the Chemical Manufacturers AisociationVRespusibje Care Initiative. Responsible Care is the most ambitious environmental, safety and health c()mmitmcnt wer undertaken by a nlatiufiaurhigindustry in the United States. Monsanto extended this commitment worldwide, and it is one o f the leading companies encouraging die adoption o f this type o f initiative throughout the world. The 10. guiding principles I. To recognize and respond co community concerns about chemicals and our operation. 2. To develop and produce chemicals dint call be manufactured, transported, used and d isused o f safely. 3. To make health, safety and environmental considerations a priority in our planning for all existing and new products and processes. 4. To report prompt!}' -- to officials, employees, customers and die public -- information on cheinical-rehued health or environmental hazards and to recommend protective measures* 5. To counsel customers on die safe use, transpoiradon and disposal o f chemical products. To operate our planes and facilities in a manner that protects the environment and th e . healrh and safety o f our employees and the public, 7. To extend knowledge by conducting or supporting research oh health, safety and environ ment^ effects o f our products, processes and waste materials. , . 8, To work with others to resolve problems created by past handling and disposal o f hazardous substances. 1' 9. To panicipatc with govcrnhicnt and others in creating responsible laws, regulations and standards to safeguard die community, workplace and environment. 10. To -promote the principles and practices o f Responsible Care by sharing experiences and offering assistance to others who produce, handle, use, transport or dispose of chemicals. Wc believe that Responsible Care and the Monsanto Pledge are part o f one unified fabric. Responsible Care was created for the industry as a whole, and Monsanto fully supports and participates in this iniportant initiative. The Monsanto Pledge guides the employees o f our particular company. Furthermore, our Worldwide Environmental Guidelines express definiuve requirements embodied in the Pledge and in llesponstble Care. As we continue to weave specific Responsible Care elements into our own initiatives, we arc confident that Monsanto has the kind o f structure we need to ensure the welfare o f our workers, our customers, our communities and the environment. This annual review o f our commitments, our accomplishments, our statistical data, and areas'in which we need improvement is one way ip which we say to the world at targe: *Wc arc publicly accountable for our performance." President and Chief Operating Officer * .* Monsanto Company . fj r <.uIo.-.j.'-IxUp * it'-VaOltUrt'*'**Jivii .IV^IUUU p^tfofwhich lop* cent is derived fivro pou-ccntum^ 905171 T he M o n s a h t o P i i d g i he Monsanto Pledge is our vision o f environmental responsibility. It shapes' the on-the-job decisions o f every Monsanto employee worldwide and guides the company hi achieving its overall commitment to meaningful environmental, safety and health solutions. . It is ourpledge to: Reduce ad toxic and hazardous releases and emissions, tvorking toward anadtimategoalo fzero effect; ** Unsure no Monsanto operationposes any undue risk to our employees and our communities; Work to achieve sustaimhle agriadture through new technology andpractices; ' Ensuregroundwater safety; Keep ourplants open to our communities and intake the community inplant operations; tm- Manage all cotf>orale real estate, iwludingplant sites, to benefit nature; and to ' Search worldwidefor technology to reduce and eliminate wastefrom our operations, with the toppriority being not making it in thefirstplace. ** 905172 0 1-- 0 8 --'9-3 is: 07 M C C R E A A N D M C C R E A I D = 0 12 3 35 53 0 7 P .02 \m AmericanJournal d Epidemiology Copyright 0 1992 by The Johns 1tepns UnwtfSify School of h . jitine end Public Health Ait rights <etttrvft4 Mortality among Workers Exposed to Polychlorinated Biphenyls Vd. 136. No. 4 FM ed to U $.A Thom as Sinks,1Gregory Steele,2 Alexander B. Sm ith,'* Kathy W atkins,1 and Ruth A. S h o ts ' On the basis of evidence from anim al studies, polychlorinated biphenyls (PCBs) are considered potentially carcinogenic to humans. However, the results o f studies in human populations exposed to PCBs have been inconsistent. The authors conducted a retrospective cohort analysis (1 9 5 7 -1 9 8 6 ) comparing the mortality o f 3 ,5 8 8 electrical capacitor manufacturing w orkers w ith known exposure to PCBs w ith age-, sex-, and calendar time-specific m ortality rates for at! w hites In the United States. Proportional hazards modeling w as also perform ed to exam ine the association betw een cum ulative PCS exposure and site-specific cancer m ortality. Alt-cause m ortality (19 2 deaths ob served. 2 8 3 .3 expected) and total cancer m ortality (54 dean is observed, 63 .7 expected) w ere low er than expected. More deaths w ere observed than expected for m alignant melanoma (8 observed, < 2 .0 expected) and cancer of the brain and nervous system (5 observed, 2 .8 expected). The average estim ated cum ulative dose for the cases of brain cancer (22.9 units) w as greater than for other w orkers (12 .9 units), but the 95% confidence intervals around tin's difference w ere broad. The risk o f malignant melanom a w as not related to cum ulative PCB exposure. These results provide som e evidence of an association between employment a t this plant aruj malignant melanoma and cancer of tho brain. The possibility that the results are due to chance, bias, or confounding cannot be excluded. Am j Epidem iol 1 9 9 2:1 3 6 :3 8 9 -9 8 . brain neoplasms; electricity; melanoma; m ortality; occupational diseases; polychlori nated biphenyls Although banned from production and distribution in the United States, polychlo rinated biphenyls (PCBs) remain in the en vironm ent Exposed workers include those involved in the maintenance and replace ment of electrical transformers and capaci tors and those working in the disposal of materials containing PCBs (1). In 1985, the Environmental Protection Agency esti mated that 1.6 million substation capacitors and 21,000 transformers containing PCBs remained in use (2). Another two million mineral oil transformers Were contaminated with PCBs at levels of >50 ppm. With ap proximately 2.5 percent of transformers re moved from service annually, 1.4 million Received for publication August 14. 1091 and in final Reprint requests to Dr. Thomas Sinks at the Division of form March 16.1992. > . Environmental Hazards and Health Effects, Center for Abbreviations: a . confidence interval: PCB3, polychlo Environmental Health and Injury Control, Centers for Dis rinated biphenyls; SMH, standardized mortality*rat. ease Cont/ot, 1600 Clifton Road NE, Maitstop F46, Atlanta, 1Division of Surveillance, hazard Evaluations end Field GA 30333. Studies, National Institute for Occupational Safety and The authors thank Dm. Kyle Steenland. Lawrence Fine, Health, Cincinnati, OH. William Matperin, Robert Herrick, and Theodore Meinttardl * Environmental Epidemiology Section. Epidemiology for thrir comments on preliminary drafts of the manuscript; Resource Center, Indiana State Department of Health. Dr. Richard Homung tor assistance with the statistical i Indianapolis, <N, analyst; and Patricia Dill foe data management and com a Current address: Department of Fomfy and Commu puter programming. nity Medicine, Eastern Virginia Medical School, Norfolk. VA. 359 905173 O 1 -08-93 is:07 MCCREA AND MCCREA 10=812 33B 5307 P .0 3 * 4 i 390 Sinks et al. contaminated transformers will remain in service in the year 2000. On the basis of evidence from animal studies, PCRs are considered potentially car cinogenic in humans (3-5), but no consis tent results have emerged from studies of humans exposed to PCBs. Some studies have found an excess of malignant mela noma, liver and biliary tract cancer, cancer of the rectum, hematopoietic malignancies, and lung cancer (6-9). To further evaluate the carcinogenicity of PCBs, we conducted a retrospective cohort study of workers with known exposure to PCBs because they had worked in a plant that manufactured electrical capacitors con taining PCBs (10). MATERIALS AND METHODS The study cohort consisted of individuals who worked at a plant in which electrical capacitors containing PCBs were manufac tured from January 1957 through March 1977, at which time the PCBs were replaced with isopropyl biphenyl. Aroclor 1242 (Monsanto Chemical Company, St. Louis, Missouri) was used through 1970, and Aro clor 1026 was used from 1971 onward. The plant was contained under a single roof with administrative offices and certain processing areas isolated by walls (figure 1). The details of the manufacturing process have been de scribed elsewhere (10). The primary area in which exposure to PCBs occurred was the area around the impregnation ovens in which the capacitors were heated. When the doors of these ovens were opened, fumes were released, spreading PCBs throughout the facility. Approximately 10 percent of the work force was directly involved in capacitor pro duction. In 1977, the concentration of PCBs in the serum of workers in the plant was sevenfold greater for salaried workers and 50-fold greater for capacitor-processing workers than for individuals in the com munity (11). The solvents used at the plant included toluene, xylene, methyl ethyl ke tone, trichlorethylene, and 1,1,1-trichioroethane. Personal and area environmental sampling for several metals indicated that metal exposures were well below the rec ommended standards. Retrospective cohort study The study cohort consisted of 3,588 men @ MAINTENANCE -- ------r -------------- (D -- .--------------- A-J6 PAINT ROOM FIGURE 1. Floor plan of capacitor manufacturing plant allowing impregnation ovens and polychlorinated biphenyl (PCD) exposure zones (800 fact = 274 m). lc tl. tfc sc P1 St: oi of cc kT of ca th be m: nu at S' in Sc 19 of COAt v ie int W< we em CO! dci kxv e it). the foil \ De a li'. t 19f The the. con wo ly il: ing sifu the P mu1 905174 0 1 -0 8 -3 3 is :a a mccrea and mccrea ID =312 336 5307 ivifh 4ing tails i de a in the k li ftthe mz$ lout vork proI CBa " Wtl$ and jsing om4ant keoromuti that rce- men Jh$ny< Polychlorinated Biphenylsand Mortality 391 and women who worked at the plant for at the first day of employment at the plant, least 1 day between January l, 1957, when whichever occurred later, and continued un the plant opened, and March 3 1 ,1977, when til the date of death or the study end date the use of PCBs was discontinued. All'per (June 30, 1986), whichever occurred first. f sonnel records and death records in the com March 31, 1977, was considered the last pany's possession were microfilmed and ab possible day o f employment. The Life Table stracted. Race was indicated in 12 percent Analysis System of the National Institute for of the personnel records, and the majority Occupational Safety and Health was used ofthese workers (96 percent) were white. We to distribute person-years at risk over sex- considered workers whose race was un specific 5-year calendar time periods and 5- known to have been white, since 98 percent ycar age groups (13). The expected numbers of the local population was white (12). Be of cause-specific deaths were calculated by cause of the small number of nonwhites in multiplying the age-, sex-, and calendar this cohort, we excluded persons known to time-specific US mortality rates for all be nonwhitc from the analysis. Workerswith whites by the corresponding number of missing dates of birth or Social Security person-years at risk. The number o f ob numbers and those who had not worked for served cause-specific deaths was divided at least 1day at the plant were also excluded. by the number of expected cause-specific Vital status was determined through the deaths to yield a standardized mortality ra Social Security Administration. We received tio. Ninety-five percent confidence intervals information on deaths reported to the Social around the standardized mortality ratio were Security Administration as of the end of calculated using an approximation based on 1987 and on persons known to be alive as the Poisson distribution (14). of December 31, 1984. For persons who could not be traced by the Social Security Proportional hazards analysis Administration, the internai Revenue Ser The primary purpose of the proportional vice provided both vital status and the mail hazards modeling (15) was to determine ing address of the most recent tax return. whether a dose-response relation existed be We considered these individuals alive if they tween cumulative exposure to PCBs (dura were indicated to be so by the Internal Rev tion of employment multiplied by an expo enue Service and if the local post office sure intensity rating) and mortality from confirmed that their mail was still being either malignant melanoma or brain cancer. delivered. Vital status was considered un Pertinent exposure information included known if it could not be determined from our knowledge of the manufacturing pro either the Social Security Administration or cess, environmental data collected in 1977 the Internal Revenue Service with post office (10), and data on scrum PCB levels collected follow-up. during a cross-sectional study conducted a Workers known to have been alive as of few months later (I l) (table 1). December 31, 1984, were assumed to be Our principal cumulative dose estimate alive as of the study end date (June 30, (cumulative PCB l) was based on the avail 1986), unless they were known to have died. able environmental sampling results and the Those whose vital status was unknown and assumption that airborne and dermal PCB those who died after June 30, 1986, were exposures decreased with distance from the considered alive in the life table analysis. For impregnation ovens. We assigned the lowest workers known to be deceased, the under exposure weight to the office area, giving it lying cause o f death was determined accord a weight of 1 and designating it as zone 1 ing to the revision of the International Clas (see figure l). Two production departments, sification of Diseases that was in effect at isolated by walls and separate ventilation the time o f death. systems, were also assigned to zone 1. The Person-years at risk of dying were accu production area was then divided by three mulated starting on January 1, 1957, or on equidistant and concentric semicircles cen- 905175 ^ -- {7tQ-- 9 3 , = .n a M C C R E A A N D M C C R E A ID -Q 12 33B S307 p .0 5 392 Sinkset ai. TABLE 1. Mean level* of polychlorinated biphenyls In a US electrical capacitor manufacturing plant and in serum samples from workers In the plant, 1977 Zone Unknown* 2 3 5 No.f Uwrf No. Uvei No. Level No. Level No. Level 1 Environmentalrrmsuremente (pgjm*) Area air sampling Personal breathing zone sampling % 2 16 15$ 4 48 13 8 59 1 9 4 76 52 0 N5 0 NS 0 NS 38 94 68 Sdrum samptos8 (ngfmi) Current job Salaried workers 66 126 101 Houriy workers only Only worked inone areal Salaried workers 36 119 2 0 Hourly workers only $1 199 3 7 7 23 96 4 5 71 305 479 7 121 6 1 5 100 2 7 8 7 6 3 1,117 * Zona was unknown for eatoried workers, t No. of observation*, t Mean standard deviation. NS, not sampled. 1 Serum teveis of polychlorinated Diphenyls m to measured as ! wor chtorinatad biphenyl moiecuiee with no more than four chlorine atons per motooife <6). 1 Persona who only worked mthis area during the* emptayment. tered upon the impregnation ovens. A value of 5 (based on the environmental sampling data shown in tabic 1) was given to the highest exposure area immediately sur rounding the ovens (zone 5), The process area furthest from the ovens was assigned an exposure weight of 2 (/.one 2), and the area adjacent to the ovens was assigned a value o f 3 (zone 3). Maintenance workers (rt ~ 34) were assigned an exposure score of 4 (zone 4) if their primary work area was in zone 3 but they were called upon to work in zone 5. Hourly workers (n = 125) who could not be located by department were assigned to zone 2. Cumulative PCB 1 was calculated by multiplying the number of days worked in each department by its exposure weight, summing across departments, and dividing by the number of days in a year. Thus, cumulative PCB exposure from working in zone 5 for 1 year was considered equivalent to that from working in'zone 1 for 5 years. Since the accuracy of our estimate of cu mulative PCB could not be verified, wc es timated it using two additional weighting schemes. Our second estimate (cumulative PCB 2) was based on the results of the serum PCB values, assigning a weight of I to zones 1-4 and a weight of 5 to zone 5. The third clim ate (cumulative PCB 3) assumed no exposure difference in zones 2-4, which were weighted by a factor of 2.5. Zone l and zr.ne 5 retained their original weights. We examined several exposure variables besides cumulative PCB dose. Workers were categorized as exposed (or not exposed) to 1,1,1-trichloroethane, trichlorethylcne, tol uene, methyl ethyl ketone, and xylene if they had worked in a department where these solvents were detected during environmen tal campling. Work in each PCB exposure zone (dichotomous) and ever having worked outside of 2one 1 were also analyzed. We also examined the duration of employment and years since first employment. We did not consider exposure to the various metais, because the environmental measurements indicated that these exposures were mini mal. Cases for the proportional hazards analy sis included workers from the population at risk with'a "primary cancer of the brain or a T 1\ j 905176 0 1 - 0 8 - 0 3 I S : 10 M C C R E A A N D M C C R E A ID =Q12 3 36 5307 F .0 B wdfn I A 2 8 '9 U7 fi four rum mes bird no rich md )Ies ere 1 *to I ;oiiey esc inire ed Vc ;nt lid .Is, its li- y- at a % + Polychlorinated Biphenyls and Mortality 393 malignant melanoma listed as an underlying or contributory cause of death. Age was used as the time variable. Cases whose illness was diagnosed before their employment began were excluded. All workers who were born within 5 years of a case and were the same sex as the case were eligible for inclusion in a comparison group (risk set) for that case. The risk sets were further limited to workers who survived to the age at which the case died and were employed at the facility prior to that age. The work history of each mem ber of the risk set was truncated at the age at which the index case had died. An association between cumulative dose of PCB exposure and cancer outcome was considered to exist if 95 percent confidence intervals surrounding the risk estimate for cumulative dose did not include 1.0. Ninetyfive percent confidence intervals were cal culated for the estimated rate ratios using a test-based method proposed by Miettinen (16). Variables that, when included in the model, altered the coefficient representing PCB exposure were maintained for multi variate analysis. All two-way interactions were examined. RESULTS Retrospective cohort study The study cohort included 3,588 work ers-- 2,742 men and 846 women (table 2). Overall mortality was lower than expected (192 deaths .observed; standardized mortal ity ratio (SMR) = 0.7,95 percent confidence interval (Cl) 0.6-0.8), as was mortality from diseases of the heart (60 deaths observed; SMR - 0.7, 95 percent Cl 0.5-0.9) and accidents (28 deaths observed; SMR 0.7, 95 percent Cl 0.5-1,0) (table 3). The stan dardized mortality ratio for all cancers was also below that expected (54 deaths ob served; SMR - 0.8, 95 percent C l 0.6-1.1). The standardized mortality ratio for deaths due to skin cancer was elevated (8 deaths observed; SMR 4.1, 95 percent Cl 1.88.0). All eight skin cancer deaths were due to malignant melanoma. (The Life Table of the National Institute For Occupational Safety and Health does not contain a calcu- 7ABLE 2. $4* and vital statua of worker* exposed to polychlorinated biphenyl in US electrical capacitor manufacturing plant, 1957-1977 Factor Total no. No. rejected No. in foal cohort Sex* Male Female Total 2,785 43 2,742 858 12 846 3,643 5 5 t 3,588 Vital status Afive Deceased Unknown Total 3 ,2 8 8 216 139 3 ,6 4 3 47 7 1 55 3,241 209$ 138 3 ,5 8 8 ' Persons of unknown race (68 percent) were mekKtod as white. t Rejected from the analysis because their work histories ddnot meet etudy inclusion effleria (n 40) or their race was Known tobenonwntte(ns 15). t Seventeen workers died after June SO. 1986, and were considered etveaa of the study end date (June 30.1986). Workers wftn art unknown vital status (r* = 138; 3.8 percent) were considered aMve as of the study end dale (June 30.1986). lalion for the expected number of deaths from malignant melanoma.) An increase was also noted for death from cancer of the brain and nervous system (5 deaths ob served; SMR 1.8, 95 percent Cl 0.6-4.2). The excess mortality from melanoma and brain cancer affected both men and women. All eight melanoma deaths included in the life tabic analysis occurred 5 or more years after initial employment, and three deaths occurred in individuals who had worked at the plant for more than 10 years (table 4, top). One person had been diag nosed with malignant melanoma approxi mately 2 months prior to working at the facility. The observed excess skin cancer mortality remained after this case was re moved from the life table analysis (SMR ~ 3.5,95 percent Cl 1.4-7,3). A ninth worker died in 1987 with malignant melanoma listed as a contributory cause of death; this individual had worked at the plant for 1 month and died 20 years after exposure. Pathology reports or medical records con firmed the diagnosis of malignant mela noma for ail cases. All five brain cancer deaths included in the life table analysis occurred 5 or more years after the date of hire, and were more * 905177 01-08-93 1BS10 MCCREA AND MCCREA ID=012 336 S307 P **%. 394 Sinks et ai. TABLE 3. Observed and expected numbers of deaths and standardised mortality ratio* for 3,586 electrical capacitor workers axposed to polychlorinated biphenyls 1957-1977 Underlying cause of death Obst Expf SMRf 95% C l| AN causes 192 283.3 0 .7 " O.6-0.8 AN cancers 54 83.7 0.8 0.6-1.1 Site* specific cancers Buccal cavity and pharynx 0 1,7 Digestive organa 8 13.9 0.6 0.2-1.1 Biliary passages, liver, and gallbladder 1 0.8 1.1 0 .0 -8 .4 Pancreas 2 2.8 0.7 0 .1 -2 .5 Rectum 1 1.2 0.8 0 .0 -4 .5 Respiratory system 15 20.2 0.7 0 .4 -1 .2 Kidney 2 1.5 1.3 0-2-4,8 Lymphatic and hematopoietic tissue 7 1L 1.0 0 .4 -2 .0 Siting 8 2.0 4 .1 " 1.8-8.0 Brain and nervous system^ 5 2.8 1 4 0 .6 -4 .2 AN other sites combined 5 8.5 0.6 0 .2 -1 .4 Diseases of the heart 60 85.4 0 .7 " 0 .5 -0 .9 Diseases of the respiratory system 10 12.3 0.8 0.4-1.5 Accidents 28 41.1 0.7* 0 .5 -1 .0 Violence 14 21.5 0.6 0 5 -1 .1 p o .o s :**p < o .a i. t OtJS. observed; Exp. expected; SMR, standaidfced mortality ratio:-CL conikfcnca interval. f The expected number of deaths was calculated using mortaiiiy rates lor basal ceH caroncma. squamous cefi carcinoma, and mafipnant melanoma combined. Alt observed skin cancer deaths were due to malignant melanoma. Cancer of the orain and central nervous system included the followinginternationalClassification of p&eesee codes: code 193. Sixth and Seventh Revisions; and C0dBS 191 and 192, Eighth end Ninth Revisions. TABLE 4. Mortality from akin cancer and cancer of ths brain and central nervous system In 3,568 electrical capacitor worker* exposed to polychlorinated blphenyta, by duration of employment and number of year* ' since first employment, 1957-1977 Years since first employment Duration of employment (years) <10 10 TcdJ <10 10 Total O /E f SMRt O/E SMR O/E SMR Skin cancer* 2 /0 46 3.59 3/1.01 2.97 5/1.57 3.19 3/0.43 6.98 3/0.43 6.98 2/0 56 3.59 6/1.44 4.16 8/1.99 4.02 Cancer of tno brain and C$;tirai nervous system^ <10 2:10 Total o /e SMR O/E SMR O/E SMR 2/0.88 2.28 0/1.29 0.00 2/2.18 0.92 3/0.62 4.64 3/0.62 4.84 2/0.88 2.28 3/1.91 1.57 5/2.79 1.79 * Expected numbers of skin cancer deaths were calculated for basal con carcinoma, squamous cefl carcinoma, and malignant melanoma combined. ANobserved skin cancer death* were due to malignant melanoma, t O/E, otaerved/expected; SMft. standardized mortality ratio. t Expected numbers of cancers of the brain and central nervous system included the following knernetiontf C/as&ficarian of Diseases oodos: code 193, Sixth and Seventh Revisions; and codes 191 and 192, Eighth and Ninth Revisions. 905178 0 1 - 0 8 - 3 3 i s : 11 M C C R E A A N D M C C R E A 1D = Q 12 3 3 6 5 3 0 7 P .0 0 iffiactrical at IB 1.1 1.1 3.4 ?.5 15 L2 16 *.0 3.0 12 ! U! 3.9 15 10 11 mjEtoma, and ft etoctiiccf of yosra k and mtfigitciU bt&iitC&Uvn a f Polychlorinated Biphenyls and M ortality 3 9 5 common among those with a longer dura tion of employment (table 4, bottom). Two additional brain cancer deaths were ob served that were not included in the life tabic analysis. One worker died of a glioblastoma shortly after the study end date. He had worked at the plant for 6 months and died 24 years after his first exposure. A black female who had worked at the plant for 11 years died of a glioblastoma 2 years alter leaving work. Pathology records or medical records confirmed the diagnosis of primary brain cancer in the five workers who had died of brain cancer. Medical records could not be obtained for two of the cases, but death certificates indicated that the under lying cause of death was carcinoma of the brain. Proportional hazards analysis 0 35 i SJ EQC Jd25W & *g l <n n K *o O) cv o 0.54 2.18 S CO CO Nr V-- T" V** a *0 r- o i to CM v tO S to to <o ad 22.9 14.2 21.8 22.0 16.4 20.4 15.8 12.7 16.1 0 .B 8 -1 .88 0.75-1 .84 0.84-1.75 a 0.47-1.44 0.50-1.88 0.50-1.46 The proportional hazards analysis for ma lignant melanoma included eight cases with risk sets that varied in size from i l l to 1,112 workers. The distributions of the cases and their comparison groups by exposure zone were similar, with 25 percent of the cases and 33 percent ofthe comparison group ever having worked in zone 5. Only one cr.se worked in a department that had been mon itored for solvents. The 95 percent confi dence intervals surrounding the odds ratios for years since fust employment, duration of employment, and cumulative PCB expo sures ail included 1.0 (table 5). The proportional hazards models for brain cancer included seven cases, and the risk sets varied in size from 42 workers to j 489 workers. Cases were more likely than i the comparison group to have worked in zone 3 (rate ratio 3.4, 95 percent Cl 0.7- 16.8). Only one case was considered poten tially exposed to any of the solvents. The number of years since first employment was similar between the cases and their compar ison group. Brain cancer cases had a longer average duration of employment than the comparison group, but the 95 percent con fidence interval for this association was broad (table 5). O f the three measures of cumulative exposure, the two estimates that 0 iO o to cd <4. o!i CM CM bb 0.83 0.97 0.85 1.16 0.82 CO (O CO to O) q r- d*- r r in r to <o p) *r q tq b d1 to tq v- CO CM O) br- rs! bv- i id tO rr Or IrN- O* CO <0 i 8a 8cl8K CO rsi <o * 905179 0 X-08-93 15:12 MCCREA AND MCCREA ID = Q 12 3 35 5307 P .0 3 `H 3 9 6 Sinks et al. weighted departments by proximity to the impregnation ovens (cumulative PCB 1 and cumulative PCB 3) were stronger predictors of brain cancer. On average, brain cancer cases had more than twice the estimated cumulative PCB dose (cumulative PCB 1) of the comparison group, but again, the 95 percent confidence interval for this associa tion was broad. DISCUSSION 'l*his group of workers had an overall sur vival rate that was better than expected when they were compared with all white men and women in the United States. At the same time, they experienced greater than fourfold excess mortality from malignant melanoma. The risk for malignant melanoma did not vary by duration of employment, lime since fust employment, or estimated cumulative PCB exposure. There was also an excess of brain cancer mortality which increased v/ith duration of employment. The workers who died of brain cancer had a higher cumulative PCB exposure than the other workers at the plant However, the 95 percent confidence intervals for these measures of brain cancer risk were broad, and the results should be interpreted with caution. Predisposing risk factors (17) or environ mental exposures probably do not account for the excess deaths reported in this study. Interviews were conducted with the next of kin of the eight workers who had died of malignant melanoma. None of the cases were related to each other, none had a con genital mole, and none had had an earlier primary malignant melanoma. One case may have had a dysplastic mole; two were considered sun-sensitive; and one was of Celtic ancestry. An environmental cause is also unlikely. Age-adjusted and calendar time-specific mortality rates (18) for malig nant melanoma and brain cancer for the county where the facility was located were similar to both state and national rates. The skin is a recognized target organ for several effects of PCBs, and it is a major route of exposure in the workplace. Chloracne (19-21) and hyperpigmentation (22) have been reported among FCB-exposed workers. While PCBs appear to affect melan ocytes, it is not known whether they pro mote or initiate neoplastic transformation in these cells, and the mechanisms involved in hyperpigmentation and carcinogenesis probably differ. Studies of human populations exposed to PCBs have yielded a variety o f results. No excess cases of brain cancer or malignant melanoma were reported in three cohort studies (7, 9, 23). An excess incidence of malignant melanoma was reported in a co hort of 72 workers exposed to PCBs (6), but that study was considered inconclusive (24). Excess numbers of brain cancer cases have been found in two other PCB-exposed co horts (8, 25). Unrecognized differences be tween study populations, different exposures from different manufacturing processes, ex posures to other carcinogens, or chance may explain the discrepancies between the var ious studies. At the same time, the statistical power (26) of all of these studies has been limited by the relatively small numbers of deaths observed. Our study had several limitations. Fewer than 10 percent of the person-years at risk were calculated with more than 19 years since the date of hire, and there have been relatively few deaths in this cohort. Thus, we could not assess the risk of cancers with long latency periods. In addition, the small num ber of observed deaths resulted in risk esti mates with broad confidence intervals. The manner in which the environmental and biologic exposure data were collected is perhaps the greatest limitation of this study. The small number of environmental sam ples collected outside the capacitor process ing area and the lack of any data from before 1977 prevented the construction of a vali dated job-exposure matrix. The data from serum PCB measurements, also collected at a single point in time, only included infor mation on current employees. Furthermore, the concentration of PCBs in scrum is af fected by their long half-lives (27) and by individual factors such as body weight, age, and sex (11), which further limits the use fulness of these data. Because of these limi- 905180 0 1 - 0 8 - 3 3 ISS 13 M C C R E A A N D M C C R E A ID = 812 33B 5307 P. 10 sxposcd melaney promation ivolved genesis oscd to Ji$. No Jignant cohort mcs cf naco(6), but ve (2 4 ). have sed co, ccii be- .Dosurcs 5CS, ccraay he varitisticii been bers of Fewer at risk y yCuis e been ius5%ve th long 1 nxmisk esti- mental sted is study. J $am* rocessbefore a valii from ctcd at inforrmorc, i is afuid by it, age, e usce limi- Polychlorinated Biphenyls and M ortality 3 9 7 tations, the weighting scales used must be considered crude approxim ations. While other weighting scales can be envisioned, all incorporate assumptions that cannot be ad equately tested. Thus, the limitations of the exposure data may have led to substantial misclassification, obscuring a possible doseresponse relation (28). Various forms of bias or confounding should also be considered. The healthy worker effect (29), together with the inclu sion of persons lost to follow-up, could partly explain the low overall mortality of this cohort, A small selection bias came about from our decision to exclude from the life table analysis four nonwhites whose race was determined from death certificates. At the same time, a small amount of persontime for unidentified, surviving nonwhites slightly increased the expected numbers of deaths. Finally, any association of excess mortality with PCBs may have been con founded by simultaneous exposure to PCB contaminants, such as polychlorinated dibenzofurans, or other unidentified sub stances. Despite the conflicting results from the various epidemiologic studies, PCBs are considered potentially carcinogenic to hu mans by the National Institute for Occupa tional Safety and Health (3), the Interna tional Agency for Research on Cancer (4), and the Environmental Protection Agency (5). This study provides some evidence of an association between PCB exposure in an occupational environment and mortality from malignant melanoma. The brain can cer finding suggests that this outcome should be carefully studied in further follow-up of this cohort. The continued follow-up of this cohort and several other large cohorts of PCB-cxposed populations will be essential for the final determination of the carcino genicity of PCBs to humans. REFERENCES 1. Smith AB, Brown DP. Polychlorinated biphenyls in the work-place. In: Waid J$, cd. PCBs and the environment. Vol 3. Boca Raton. FL: CRC Press, 1986:1-26. 2. environmental Protection Agency. PCBs in electri cal transformers: final rule. Code of Federal Regu lations, section 40, part 761. Washington, DC; En vironmental Protection Agency, 1985. 3. National Institute for Occupational Safety and Health, Current intelligence bulletin 45. Cincin nati, OH: National Institute for Occupational Safety and Health, 1986. (DHHS (NIOS1I) publi cation no. 86-111). 4. International Agency for Research on Cancer, Overall evaluations of carcinogenicity: an updating of IARC Monographs volumes l to 42. lARO Monogr Eval Carcinog Risks Hum Suppl 1987;7: 1-440. 5. Environmental Protection Agency. Ambient water quality criteria for polychlorinated biphenyls (PCBs), Washington, DC: Environmental Protec tion Agency, 1980. (ItPA publication no. 440-580-068), 6. Bahn AK. Rosenwaikc I. Herrmann N, et al. Mel anoma after exposure to PCB V ( Letter). N Engl J Med 1976:295:450. 7. Brown DP. Mortality of workers exposed to poly chlorinated biphenyls: an update. Arch Environ Health 1987;42:333-9. 8. Bcrtam PA, Riboldi L, Pcsatori A, et al. Cancer mortality of electrochemical workers exposed to PCBs. Am J Ind Med 1987:11:165-76. 9. Kuiatsunc M, Ikcda M, Nakamura Y, ct al. A cohort mortality study on mortality of "Yusho" patients: a preliminary report. In: Miller RW, et al, eds. Unusual occurrences as dues to cancer etiol ogy. Tokyo: Japan Scientific Societies Press Ltd, 1988:6!-6. 10. Blade L, Marlow p , Jones M. Industrial hygiene survey of Wcstinghouse Electric Corporation, Bloomington, Indiana, April 19-22, 1977. Cincin** nati, OH: Industrywide Studies Branch. National Institute of Occupational Safety and Health, 1977. 11. Smith AB,$chlocmerJ, Lowry LK ,etal, Metabolic and health consequences of occupational exposure to polychlorinated biphenyls. Br J lnd Med 1982: 39:361-9. 12. Bureau o f the Census. 1980 Census o f Population. General population characteristics; Indiana. Wash* ington, DC: Bureau of the Census, US Department of Commerce, 1982. 13. Waxweilcr RJ, Beaumont JJ, Henry JA, et ai. A modified life-table analysis system for cohort stud ies. J Occup Med 1983;25:115-24. 14. Rothman KJ, BoiceJD Jr. Epidemiologic, analysis with a programmable calculator. Washington, DC: USGPO, 1979. (DHHW publication no, 79-1649). 15. Cox DR. Regression models and life tables. J R Stat Soc 1972;34:187-202. 16. Mtettinen OS. Estimability and estimation in case-referent studies. Am J Epidemiol 1976:103: 226-35. 17. Rhodes AR, Wcinstock MA, Fitzpatrick TB, et al. Risk factors for cutaneous melanoma: a practical method of recognizing predisposed individuals. JAMA 1987;258:3146-54. 18. Riggan WB, Van Bniggcn J, Acquavelia JK, ct al. US cancer mortality rates and trends, 1950-1979. In: NCl/EPA Interagency Agreement on Environ* 905181 0 1 - 0 0 - 3 3 IS M 3 M C C R E A AND M C C R E A ID=812 336 5307 , 398 SinksetaL I1 menial Carcinogenesis. Bcthesda, MO: National Cancer institute. 983. (Publication no. EPA-60Q/ 1 -8 3-015 A). 19. Meigs JW, Albom JJ. Chlomcne from an unusual exposure lo Arod or. JAMA 1954;154:1417-18, 20. Ouw HK, Simpson GRr Siyali DS. IJsc and health effect* of Arodor 1242, a polychlorinated biphenyl, in an electrical industry. Arch Environ Health 1976;31:189-94. 21. Daker EL Jr, Landrigan PJ, Glueck CJ, ct aL Met abolic consequences of exposure to polychlorinated biphenyls (PCB) in sewage sludge. Am J Epidemiol 1980;112:553-63. 22. Fisdibdn A, Wolff MS, Ulis R, ct al. Clinical findings among PCB-cxposcd capacitor manufac turing workers. Ann N Y Acad Sei 1979;320: 703-15. 23. Gustavsson P, Hogstedt C, Rappe C. Short-term mortality and cancer incidence in capacitor man ufacturing workers exposed to polychlorinated bi phenyls (PCBs). Ara J Ind Med 1986;10:341-4. 24. Agency for Toxic Substances and Disease Registry. Toxicological profile for selected PCBs. Atlanta, GA: Agency for Toxic Substances and Disease Reg istry, 1989. (ATSDR publication no. TP-88-21). 25. Liss GM. Mortality and cancer morbidity among transformer manufacturing workers, FerrantiPacfcord Transformers Ltd-, St, Catherines. T o ronto; Ontario Ministry of Labor, Health Studies Service. 1989. 26. Beaumont JJ Breslow NE. Power considerations in epidemiologic studies of vinyl chloride worker*. Am J Epidemiol 1981;114:725-34, 27. Phillips DL, Smith AB, Burs VW, et aL Half-life of polychlorinated biphenyls in occupationally exposed worker*. Arch Environ Health 1989;44; 35 M . 28. Checkoway H, Pearce NE, Crawford-Brown DJ. Research methods in occupational epidemiology. New York: Oxford University Press, 1989:78. 29. Monson RR Occupational epidemiology. 2nd cd. Boca Raton, FL: CRC Press, 1990:100. 905182