Document 1wa6GNDBrB07RDDYkvmgOymX

C o m m e n t a r ie s .*** Reconciling Old and New Findings on Dioxin James J. Collins, John F. AcquaveUa, and Barry R. Friedlander W hether dioxin, or more specifically 2,3,7,8-tetrachlorodibenzo-p-dloxin (TCDD), has caused cancer in humans has been the subject of many epidemiology studies. A recent review concluded that the findings of these studies were not consistent, although the evidence was considered "weak to strong'* for nonHodgkin's lymphoma and too inconsistent for soft tissue sarcomas.' More recently, however, two studies have been published that report consistent findings indicative of dioxin carcinogenicity. These two studies are an update, by Zober et al,2 of a study of workers exposed as a result of a 1953 accidental release of dioxin in Germany, and the National, Institute for Occupational Safety and H ealth (NIOSH) Dioxin Registry study by Fingerhut et al.3 W e will consider how these new studies add to our knowledge of the potential dioxin health effects. In 1977, a case report appeared, followed by a related case-control study in 1979, that suggested a connection between herbicide exposure and soft tissue sarcoma.45 Dioxin can be a contam inant in some herbicides. This finding appeared to receive some support from three occupational studies** that reported a total of four soft tissue sarcomas, a relatively rare tum or, among small groups of workers exposed to relatively high levels of dioxin. These findings, as well as public concerns, motivated the formation of the N IO SH Dioxin Regis try- . After these early findings, however, the potential causal relation between dioxin and soft tissue sarcoma was questioned in two separate areas. First, the earlier reported association was not replicated in later studies,* and two o f the original four soft tissue sarcoma deaths were not confirmed on review o f the pathology re ports.10 Thus, the general view o f these findings shifted, and the evidence for a causal relation between From the Monsanto Company, St. Loula, MO. Address reprint requests to: Jam J. Colling Epidemiology Director, Monsanto Company A2SL, 800 North Lindbergh Boulevard, St. Louis, MO 63167- >1992 Epidemiology Resources Inc. dioxin and soft tissue sarcoma was considered "sugges tive b u t unproven."11 O ther cancer sites have occasionally been reported to be associated w ith phenoxy herbicides or dioxin exposure.12These sites include H odgkin's disease, nonH odgkin's lymphoma, and stomach cancer. As with soft tissue sarcoma, however, these associations were not consistent across studies, and so these sites were not recognised as related to exposure.1 Recently, however, two studies have been published . that show elevated mortality for total cancers in occu pational subcohorts with lung latency. Zober et a lj in updating an earlier study of workers exposed from ah accidental release, determined exposure from proxim ity to the accident O ne hundred fourteen cases of chloracnc, a condition related to significant exposure to dioxin as well as to other substances, and 13 cases of erythema (that is, suggestive of chloraene) w ere; observed among the 247 persons in the study. There were 67 deaths in this study; however, for 11 of these decedents, death certificates could n o t be located, and clinical reports were used to certain cause of death. It is unknow n w hat impact this case-finding procedure had on the final results, since fates foe die general population comparison group were ascertained from death certificate information only. In addition, the impact o f including nonchloracne (that is, erythema) With chloraene cases is notreported. In the Zober et <d study, mortality rates were evalu ated for four groups of potentially exposed persons: (1} 69 workers, all o f whom had chloracnc, identified from a plant list; (2) 84 workers involved in protective operations and demolition o f the building after the accident; (3) 94 persons from several sources known to be involved in the accident; and (4) 127 persons from the previous three groups w ith chloraene or erythema. N o m ortality excesses were observed among group 1 or group 3. Group 2 had a significant excess in the category o f "other and unspecified cancers." Group 4 had no significant excesses for any cause o f death, b u t persons w ith 20 or more years o f latency had a significant excess for all cancers combined [stand ardized m ortality ratio (SMR), = 2.0,95% confidence C O U IN S ET AL interval (Cl) * 1.1-3.4). From these data Zober et al conclude; "our results do not appear to support a strong association between cancer mortality and dioxin, but they do suggest that some hazard may have been produced."* In the second recent study, Flngerhut it al5 included 5,172 workers from U U .S. manufacturing plants that produced chemicals potentially contaminated w ith dioxin. Dioxin exposure was assumed on the basts o f proximity to a process. Overall, none o f die a priori hypothesized cancers (for example, soft tissue sarcoma, stomach cancer, and non-Hodgkin's lymphoma) was significantly elevated. In a tubeohort o f 1,520 workers with 1 or inore years o f exposure and 20 years o f latency, there was a statistically significant SMR for unconfirmed soft tissue sarcoma (SMR * 9.2, 95% C l m 1.9-27.0), respiratory cancer (SMR " 1.4, 95% C l *1.0-1.?), and 1 cancers combined (SMR 1.5,95% C l-1 .2 -1 .8 ). Mortality for all cancers combined (SMR - 1.2, 95% C l = 1.0-1.3) was also elevated for the entire study population. From these data, Flngerhut et al conclude; "the Increased mortality, especially in the subcohort with one year or more o f exposure, is consistent with the status o f dioxin as a carcinogen."3 Because o f these recent studies, some have con' eluded that dioxin is more likely a carcinogen than not and that the weight o f evidence has contributed to the "weakening o f the position o f those who believe that low levels o f dioxin are entirety safo for humans."15 W e believe, however, that any such strengthening o f die conclusion regarding dioxin carcinogenicity is pre mature and not fully supported by the data. Four issues need to be resolved. The first issue is that the recent studies focus atten tion on a new group o f cancers not previously re ported. The only exception might be die finding in the Fingerhut et al study o f a soft tissue sarcoma excess in a subcohort o f workers, but (his reported excess Is difficult to interpret because; (1) upon pathologic re view, two o f the four deaths were found not to be soft tissue sarcoma10; (2) no new soft tissue sarcoma deaths were found, although 10 additional plants were studied (that is, all four deaths were known from prior pub lished studies4'o f two plants); and (3) at least one o f the reported deaths has been coded by three different nosologists to a non-soft tissue sarcoma a K iim t vwwoincy in nosologic decision making for this cause o f death. Furthermore, there were no soft tissue sarcoma deaths reported in die Zober et al study. T he substantia] variability In both diagnosis and n o sology coding indicates that diere may be no unbiased way to compare worker and general population death rates for soft tissue sarcoma. T he comparison o f ex posed and unexposed workers at the studied plana may be the only viable approach for assessing the possible relation between dioxin and toft tissue sar coma. The second issue Is that d ie findings are n ot con centrated in specific cancer sites and instead relate to all cancers combined. Such findings are not consistent with die patterns reported for known occupational carcinogens. Furthermore, the Fingerhut et rti study found no trend to risk w ith increasing exposure or duration o f exposure to dioxin (as shown in Table 4 o f the Flngerhut et al report5), and die pattern o f risk was not indicative o f an occupational carcinogen (that is, the largest risk was not found in d ie highest expo- sure/longest latency strata), although the risk, as meas ured with die SMR* did increase with latency alone. Additionally, die primary findings were unexpected at the beginning o f the study. Therefore, these new findings should be viewed cautiously as warranting further analysis and awaiting replication from other epidemiologic investigations, but not as direct evidence o f causality. The third issue is that d ie findings o f Zober et al and Fingerhut ct ol are each internally inconsistent. That is, each study is a mixture o f previously studied workers and new data, and the old and new results differ. Characterized in this way, each study can be evaluated for internal consistency. T he Zober et al study is in forge part an update o f a previous study14 In which three stomach cancers were reported (expected 0.5, SM R 5.8,95% C l 1.2- 16.9). Although die populations were slightly different between studies, Zober tt al found no additional stom ach cancers during 12 additional years o f follow-up (expected - 0.5, SMR - 0 A 95% C l - 0.0-7.4) but only reported combined results for d ie total follow-up period (SMR * 3.0,95% C l 0.6-8.?). T w o conclusions are possible from contrasting the old and new data; d ie stomach cancer findings from the original study were not confirmed by further follow-up, or d ie max imum Induction period for dioxin-related stomach can cer passed prior to the additional follow-up period. Assessm ent o f result* from other dioxin populations with similar time since exposure does not show ete- ,.^ .4 ...------ v. .,--- upportm i ttve former conclusion. T he Fingerhut ct at study la a com bination o f pre viously published data*'*'1*'17 from two plan a and data from 10 additional plants that were not studied previ ously- The former plant* Include 49% (2.JH4 worker) of the total study y ^ o o n , and Aw dioxift-relatid 66 Epidemiology January 1992, Volume 3 Number l RECONCILING OLD AND NEW FIND IN GS O N DIOXIN manufacturing periods and th e subsequent follow-up periods overlap substantially, with die 10 additional planes providing a comparable basts lor contrasting the findings.u Although the data were not reported sepa rately for the 10 newly studied plants, this separation can be approximated (Table 1). The tw o plants for merly studied show no remarkable cancer patterns w ith the exception o f the previously reported findings for oft tissue sarcoma. This finding was, In part, the basis for the NIO SH Dioxin Registry. Findings from th e 10 newly studied plants show no cases o f soft tissue sarcoma (SMR 0), although ft expected number o f deaths is almost the same as for the previously studied plants (0.6 vj 0.7). Thus, the previously reported find ing for soft tissue sarcoma was not confirmed in a very similar study among a second exposed worker popu lation. A second possible explanation is that the period since exposure for workers differs between studies despite the overlap o f manufacturing activities and follow -up periods between studies: but that seems unlikely. It la also noteworthy that stomach cancer m ortality was n ot elevated In either subgroup-- find ings inconsistent with those in the original study o f the Zober et al cohort. O ther findings from the 10 additional plants indicate an excess o f all cancers combined snd a borderline excess o f lung cancer--findings not evident in the two plants previously studied. These "new1' findings should be view ed as Internally inconsistent and provide a rationale for evaluating potential confounding factors as n oted in the remainder o f this paper. T h e fourth issue is that the interpretation o f the findings is obscured by the potential Influence o f im portant confounding factors such as: (1) other oc cupational exposures, (2) smoking, and (3) failure to control for regional variation in the general population cancer m ortality. A s Fingerhut et al point out, duration o f dioxin exposure was correlated With duration o f tim e em ployed in these plants. The two Monsanto plants included in the study have been studied with respect to various substances other than dioxin; other expo sures,in these plane could be important confbunders In subgroups o f workers. The same may also be true for th e other plants. T he relative nonspecificity o f the reported cancer findings in this study and in the Zober a( study are consistent with exposures to a variety o f substances, but this issue has not been addressed in detail by th e authors. W hile both the Fingerhut et al and th e Zober et al studies assumed that non-dioxin exposures were not important confbunders, our ex perience w ith tw o plants in the Fingerhut et a l study TABLE l . Comparison o f O bserved Deaths (Ob), Expected Deaths (Exp), and Standardised Mortality Ratio (SMR) w ith Confidence Intervall (C l) o f Old Results (Two Plant Previously Studied) w ith N ew Results (Ten Planes not Previously Studied) from the Fingerhut *t a l Study11 SMR Obi/Exp* (95% Cfif Site of Cancer (ICD9 Code) Old Result? NewResuItsf All cancer (140-208) 1-06 1.28 140/132.6 125/97.3 (0.89-1.25) (1.07-133) Non-Hodgkin's lymphoma (200,202) Soft tissue sarcoma (171) 1.67 7/4.2 (0.67-3.45) 0.96 3/3.1 (p.20-2.81) 0.00 0/0.6 (p.O-6.71) Hodgkin's diseue (201) 1.33 2/1.5 (0.16-4.82) 1.00 1/1.0 (0.03-537) Stomach (151) U4 7/5.7 (0.50-2.55) 0.68 3/4.0 (0.15-2.1 Liver (155,156) 0.68 2/2.9 (008-2.47) 1.76 4 /U (0.48-4.51) Lung (162) <193 43/46.3 (0.67-1.25) 1.36 46/33.8 (1.00-1.62) `Eitlmattd (tomTible 6 ofFingerhut <t t Confidence totervali using Byer approximation for 5 or more deathand the fisher m et method for fewer than 5 d*thi,n f The Monsantoplantappears Instudies6 and7. and thaDowplant appears Instudies 15- 1?, Only two ofthese cue* were confirmedas toft tissue sarcoma do pathologic review. suggests otherwise. For example, in one o f our plants, a subgroup o f the dioxin cohort had exposure to 4- aminobiphenyl, a potent bladder carcinogen,'* which was discontinued in 1955 after it was discovered to be a carcinogen. The excess o f bladder cancer reported by Zack end Gaffey* in a subset o f th e N IO SH cohort illustrates the effect o f 4-am lnobiphenyl. Aromatic amines were also produced In th e plant studied by Zober et dl, and several bladder cancers were attributed to such exposure.1 The bladder cancer attributable to 4-amlnobiphenyl contributes to th e total cancer excess observed in the Fingerhut et o l study. - A second example o f a potential confounding ex posure is indicated by the m esotheliom a deaths present in both the Fingerhut et a l study (tw o deaths) and the Zober et al study (one death). T hese deaths indicate Epidem iology January 1992, Volume 3 Num ber 1 7 COLUNS ET AL TABLE 2. Observed Deaths and Standardized Mortality Ratios (SMR) by Plant for Lung Cancers and All Cancers Combined* Plant Code 1 2 3 4 5 6 1 8 9 10 11 12 Lung Cancer Observed SMR 7 0.72 1 1.55 4 1.07 3 1.01 3 1.66 2 1.04 3 2.38 15 1.44 28 0.78 13 2.14 5 2.39 $ 1-25 All Cancers Com bined Observed 32 2 10 7 7 7 3 35 105 30 11 16 SMR 1.15 U1 0.87 0.75 1.41 1.29 0.86 1.18 1.02 1.81 1.93 1.41 Total 89 111 265 1.15 #Source; Fingerhut et a l1* the potential for asbestos exposure among one or more o f the workforces in the study and could be contrib uting to the lung cancer findings in these studies.19 T he possibility that these and other occupational factors may be affecting the reported findings is suggested by the wide variation in the SMRs for lung cancer and all cancers combined for the 12 plants in the Fingerhut it al studyu (Table 2). For lung cancer and all cancers, the SMRs range from 0.7 to 2,4 and from 0.8 to 1.9, respectively. Although sampling variation is a possible explanation for the wide range of SMRs, it would be useful to assess the potential confounding at the plants w ith high SMRs. Smoking is another confounder whose impact has likely been underestimated, at least in the Fingerhut it al study. This underestimation takes on added im portance because smoking is a strong risk factor for lung cancer that has a synergistic effect with asbestos exposure.19To assess the potential confounding effect o f smoking, Fingerhut et at used estimates of smoking prevalence for the cohort in 1965 based on a survey of workers alive at two of the 12 plants in 1987. Smokers w ould be underreported in this cross-sectional survey since nonsmokers would be overrepresented among surviving workers. Additionally, smoking prevalence in 1965 may un derreport the smoking prevalence over the subsequent study period (that is, 1965-1987), compared with the general U.S. population, since the cohort is comprised predominantly o f blue collar workers who may take longer than the general population to quit smoking.20 These two factors suggest an underestimate of the true confounding effect of smoking on lung cancer in the study. The general pattern o f elevated SMR for smok ing-related cancers in both studies supports a potential confounding effect due to smoking. An additional im portant factor could be the failure in both studies to account adequately for geographical variation in m ortality rates by comparing w orker m or tality with national rates. Because of known regional variations, it is more appropriate to use regional com parison populations.21*22Cancer SMRs in the range o f the present study can dim inish substantially w hen regional comparisons over the total study time frame are used.21*22 For the two M onsanto plants, using re gional comparison rates reduces the SMR 24% and 28% for lung cancer and 11% and 20% for all cancers combined. H ence, geographical variation is often sig nificant and may obscure the estimation of cancer risk. In summary, the recent dioxin studies report impor tant new inform ation relevant to potential effects o f dioxin, but the inconsistency of results with previous findings makes causal inference difficult. The recent findings that have been emphasized were not those expected on th e basis of earlier studies, were n o t related to dioxin exposure, and were of such small magnitude th at bias should be a critical concern. Fur ther studies o f these populations could evaluate die potential confounding factors mentioned above to clar ify the potential cancer risk from dioxin or other exposures. R eferences 1. Lilienfeld DE, Gallo MA 2,4~D, 2,4,5-T and 2A7.B-TCDD: an overview. Epidemiol Rev 11:2S-5S. 2. Zober A, M sserer ? , Huber P. Thirty-four-year mortality fol low-up of BASF employees exposed to 2,3,7,8-TCDD after the 1953 accident. 1m Arch Oceup Environ Health 1990#2;139- 157. 3. Fingerhut M* Halperin W, Piacitelli L, Ilonchar P, Sweeney M, Greife A, D1 P, SteenUnd K, Suruda A. Cancer mortality in workers exposed to 2v3,7l^tetrachlorodlbenaO`P^cboxln. N Engl JMed 1991s324;2l2-2i8. 4. Hardell L Soft tissue sarcomas and exposure to phcnaryacetic acids and eaaeer (in Swedish). Lak*rtidningen J977742753- 2754. 5. Hardell L, Sandscrorn A. Case-control study: soft tissue sarcomas and exposure to phenoxyacetic adds or ddorophenok Br J Cancer 1979;39:711-717. 6. Zack JA, Suskind R. The mortality experience of workers exposed to tecrachlorodibetizodioxin in a trichSorophenol proc ess accident. J Occup Med 1980.22j11-14, 7. Zack JAt Gaffey WR. A mortality study of workers employed at the Monsanto Company plant in Nitro, West Virginia. Environ SlRes 198306;575-59i. 8. Cook RR. Dioxin Chloracne, and soft tissue sarcoma. Lancet 1981;1:618-619. , 9. Bond GG, Bodner KM, Cook RR. Phcnoxy herbicides and cancer; insufficient epidemiologic evidence for a causal relation ship. Fundaro App! Toxicol 1989;12:172-188. 10. Fingerhut MA, Halperin WE, Honchar PA, Smith AB, Groch 6 8 . Epidemiology January 1992, Volume 3 Number 1 OLD AND NEW REFLECTIONS O N DIOXIN DH, Russell WO. An evaluation o f reports of dioxin exposure and soft tissue sarcoma pathology in U.S. chemical workers. ScandJ Work Environ Health 1984410s299-303. 11. Council on Scientific Affairs. Cancer risk of pesticide* in agriCultural worker*. JAMA 1988$260i959-966. 12. Fuigerhut M, Halperin W, Marlow 0 , Placitelli L, Honchar P, Sweeney M, Grdfe A, Dill P, Steenland K, Suruda A. Mortality among U-S- workers employed in the production of chemicals contaminated with 2t3t7(S^tetrach1orodtbenso*p-dioxin (TCDD), NIOSH Final Report 91125971. Springfield, V A National Technical Information Service, 1990. 13. Bailar jC . How dangerous b dioxin? N Engl ] Med 1991^24:260-* 262. 14- Theis* AM, Frenttel-Beyme R, Link R, Mortality study of persons exposed to dioxin in a trtehlofophetiol-proctM accident that occurred in the BASF AG on November 17 1953. Am ] Ind Med 1982;3:179-189. 15. Cook RR, Bond GG, Olson RA, O tt MG, Gonder MR. Eval uation of the mortality experience of workers exposed to the chlorinated dioxins. Chemosphere 1986;15:1769-1776. 16. O tt MG, Olson RA, Cook RR, Bond GG. Cohort mortality Study of chemical workers with potential exposure to the higher chlorinated dioxins. J Qccup Med 1987;29:422-429. 17. Bond GG, Melateti EA, Upps TE Cook RR. Update of ao r- caliry among chemical workers with potential exposure to the higher chlorinated dioxin*. ] Occup Med 1989l:12t-123. 18. Mclick WF, Narykafl, Kelly R E Bladder cancer due to exposure to parwaminabipheny: a 17-year foOowup,J Urol 1971,106:220226. 19. $e!ikoff QLee DH. Asbestos and Disease. New York: Academic Press, 1978. 20. PeD S, Fayerweather W . Trends in the Incidence o f myocardial infarction and in assodateci mortality and morbidity in a large employed population, 1957-1983. N EnglJ Med 1985^312:100$- 1011. : ' 21. Ehrerline PE. Pitfall* in epidemiologic research: an examination of the asbestos literature. J Occup Med 1976:18:150-156. 22. Veys CA. Towards causal inference In occupational cancer epidemiology. L A n example of the interpretive value of using local rates as the reference statistic. Ann Occup Hyg 1990,34:349-359. 23. Rothman K} Bence JD. Epidemiologic analysis with a program mable calculator. N1H Pub. No. 79*1649. Washington DC: National Institutes of Health, 1979. Old and New Reflections on Dioxin Marilyn A. Fingerhut, Kyle Steenland, Marie Haring Sweeney, William E. Halperin, Laurie A. Piacitelli, and David A. Marlow W hat is "old* and what is "new"? As w ith most things epidemiologic, it depends on your definition. The paper by Collins et a l1 claims th at a recent study2 of dioxin-exposed workers by the National In stitute for Occupational Safety and H ealth (NIOSH) is a mixture o f "old" (that is, previously studied) and "new" findings which are internally inconsistent We present the historical context in which the NIOSH study was conducted to point out the erroneous assu m p tio n s of Collins et al regarding "old" and "new." W e also explore the question o f how to evaluate "consistency" in an epidemiologic study. From the U.S. Department of Health and Human Service*, Public Health Service, Centers for Disease Comrol. NationaJ Institute for Occupational Safety and Health, Industrywide Studies Branch, Di vision of Surveillance, Hazard Evaluations and Field Studies, Cincin nati, OH. Address correspondence to: Marilyn A. Fingerhut, U.S. Department of Health and Human Services, National Institute for Occupational Safety and Health, 4676 Columbia Parkway, Cincinnati, OH 43226. <1992 Epidemiology Resources Inc. T h is study was initiated in 1978 by NIOSH to include all (5,172) chemical workers in the United States who had work records verifying assignment to the production of 2,4,5'trichJorophenoxyacetic acid or its precursor 2,4,5'trichlorophenol, both contaminated with 2,3,7,&'tctrachlorodibenzo*p-dioxin (TCDD), commonly referred to as dioxin.2By 1976, toxicity and teratogenic and carcinogenic effects had been seen in animals, but only lim ited epidemiologic investigations had been conducted.* T he NIOSH study was initiated because o f the animal data and out of concern about potential effects on the Vietnam veterans exposed to Agent Orange, which contained dioxin, and the work, ers who produced the products contaminated with dioxin. All o f the M onsanto and Dow studies cited by Collins et a t1as "previously published" were actually published after the 1976 initiation of the NIOSH study. Table 1 lists key aspects of four studies of M onsanto and Dow w orkers published in the early 1980s.4"7 Since most o f the workers fit the NIOSH protocol definitions and were included in the NIOSH' FINGERHUT ET AL study, these four studies should he viewed as small subsets of A e large NIOSH study of all U S. workers, With the vital status of the workers updated through 1987. Table l shows that the total of 444 workers identified in the four early M onsanto and Dow studies constitute 17% of the 2,544 dioxin-exposed workers identified by NIOSH at the M onsanto and Dow plants and less than 9% of the 5,172 workers in die full NIOSH study of 12 U.S. plants. This number is well below the 49% cited by Collins et a l,1who erroneously suggest that the large Dow cohort studied in the late 1980s8"10 is independent of th e NIOSH study. That cohort was actually the N IOSH cohort of Dow work ers identified for the NIOSH study by NIOSH re searchers with the active collaboration of Dow epide miologists (Table 1). The Dow researchers, however, published several mortality analyses of this Cohort8"10 before publication of die full NIOSH study.1 Consequently, since M onsanto and Dow published studies o f fewer than 20% of the w orkers exposed to dioxin at the two plants, we cannot agree with the premise of Collins et al that *the Fingerhut et a1(NIOSH] study is a combination of previously published data... from two plants and data from 10 additional plants that were not studied previously." A more substantive issue regarding the re-analysis o f published studies is raised by consideration of the paper by Collins et a l.1 W hat is an appropriate epide miologic rationale for splitting a m ultiplant study con ducted from a single protocol to look for internal consistency? W hat is the definition of "internal con sistency''? Collins et al appear to be seeking similar cause-specific standardized m ortality ratios (SMRs) in the various plants. W e suggest that this standard is not appropriate in th e absence of thoughtful consideration about age distributions, levels (or duration) of expo sure, latency (time since first exposure), and power. In our study,2 as in other m ultiplant studies, there are a number of reasons why one might n ot expect CO find similar SMRs across plants. For example, aDwork ers at two of the 12 plants in the NIOSH study had been first exposed less than 20 years earlier, so cancers may not have had time to develop. Several plants had only about 100 exposed workers, leading to few deaths, limited statistical power, and unstable numbers. We found that the excess cancer was concentrated in workers with more than 1 year o f exposure. Thus, at two plants, where most workers were exposed for less than 1 year, the overall risk would be lower d u n at the plants where workers were exposed for many years. The intensity of exposure also differed at the plants. T he NIOSH study utilized years o f assignment to 70 tpidemioiogy January 1992, Volume 3 Number 1 OLD AND NEW REFLECTIONS ON DIOXIN dioxin-contaminated processes as a measure o f cum u W hat is truly 'old* in the paper by C otilos t o is lative exposure for all workers. We are currently de the listing o f possible confounders in the NIOSH veloping an industrial hygiene-based exposure m atrix study. All of these were discussed in some detail by to estimate relative exposures of the workers in the Fingerhut et oL1,w Collins et aV have om itted several cohort The planned re-analysis of the cohort will balancing considerations presented in these papers. For further examine the question of exposure-response, example, they point out that we used a smoking survey using an internal comparison. conducted at two plants to estmate the possible influ Collins et a l1 expressed concern that the "recent ence of smoking on the observed lung cancer excess {NIOSH2and Zober et a in ] studies focus attention on in the entire cohort of 12 planta. This survey indicated a new group of cancers not previously reported " W e little difference in smoking between these two plants agree that our finding o f an excess o f total cancers was and the U.S. referent population. In drawing our . unexpected, since it is unusual in studies o f chemical conclusions, we also considered other factors, includ workers. Earlier studies o f dioxin-exposed animals, ing an evaluation of m ortality from other diseases however, are consistent w ith our finding in that they know n to be associated with smoking. W e found that have found tumors in a number of organs.12-11 The mortality from nonmalignant respiratory disease was Zober study of German workers, although small and lower than expected (15 deaths, SMR 96, 95% C l 54- limited in its assessment of exposure, reported a sim ilar 158) in the subcohort with over 1 year of exposure finding of excess total cancers in a subgroup with and more than 20 years of latency, litis is the subco- chloracne who had been exposed 20 or m ore years hort in which the excess respiratory cancer was found earlier.11 (43 deaths, SMR 142,95% C l 103-192). Additionally, The history of soft tissue sarcoma studies described the subcohort with over 20 years of latency, b u t with by Collins et a! requires clarification. A lthough a case less than 1 year of exposure, did not have an excess of report about soft tissue sarcoma14was published before respiratory cancer (19 deaths, SMR 103, 95% C l 62- the NIOSH study was initiated in 1978, all of the other 161)- Consequently, it did not seem to us that smoking epidemiologic studies of soft tissue sarcoma in various would have fully accounted for the respiratory cancer countries were conducted during the period w hen the excess in the subcohort with long exposure. NIOSH study of dioxin-exposed U.S. ,,production ' _ Collins HS1suggestihat regional rates should have workers was-underway; Thus, the claim o f Collins et been used in the NIOSH study. The choice of com al1is incorrect that the 1978 initiation o f the N IO SH parison rntes, always a judgment call, has to address study was motivated by the soft tissue sarcomas re the limitations in the use o f local rates, such as the ported in the Swedish case-control studies and the instability of small numbers, the question of w hether U.S. occupational cohorts of the 1980s. the workers actually live within the region, the prob The difficulties of studying soft tissue sarcoma from lem that some cohorts may Influence the local rates, death certificates have been described, including the and the locations of the 12 plants in 10 states across problem o f m isdassifkation of cause of death2,15,14 The the country. Because of the lim itations o f local rates, NIOSH mortality analysis2 included four deaths iden we chose to use national rates, as did the authors o f tified from death certificates as 6oft tissue sarcoma the Dow and Monsanto studies listed in Table l.4' 10 deaths. Based on tissue review, two of these were found O ne new suggestion can be drawn from the Collins not to have died o f soft tissue sarcoma. Misclassification et al paper1, namely, that future epidemiologic m ortal in the opposite direction also occurred. Tw o other ity studies compare workers exposed to dioxin-contam men in the NIOSH cohort o f Monsanto workers died inated products with unexposed workers at the same of soft tissue sarcoma, based upon medical records and plant. Studies with internal comparisons would indeed tissue review. They were not counted as soft tissue be helpful, particularly If exposure to dioxin-contami sarcoma deaths in our study, however, because soft nated chemicals is carefully characterized. Internal tissue Sarcoma was not listed on their death certifi comparisons at the plants could address several poten cates.2,14Misclassification is also known to occur in the tial confounders not folly addressed in the NIOSH Coding of death certificates for the U.S. population.17 m ultiplant study of mortality. Thus, the usual practice is to conduct the cohort In summary, we differ with the views o f Collins et analysis using the data on the death certificate, w ithout a l.1 W e consider the NIOSH stpdy to be a "new" attem pting any correction for misclassification, and to study, which followed a single protocol to identify all add a discussion if additional information is know n workers in the United States assigned to the manufac about the deaths in the cohort. ture of products contaminated with dioxin. The study Epidemiology January 1992, Volume 3 Number 1 71 FINGERHUT ET AL utilized standard epidemiologic methods and found an overall 15% excess of all cancers, which was more pronounced (46% excess) in a subcohort with more than 1 year of exposure and over 20 years o f latency. In this subcohort, a 42% excess of respiratory cancer and a ninefold excess of soft tissue sarcoma were also found. Although the NIOSH study could not com- pletely exclude the possible contribution o f other occupatlonal carcinogens or smoking, we conclude that the results are consistent with TCDD being a carcin ogen. R eferences 1. Collin* JJ, Acquavella )F, Friediander BR, Reconciling old and new finding* cm dioxin, Epidemiology 1992;3:65-69- 2. Fingerhut M, Halperin W, Marlow D, Piacitelli L, Honehar P, Sweeney M, Grcife A, Dill P, Sceenland K, Suruda A. Cancer mortality in worker exposed to 23*7,S-trtrachlorodibeniopdioxin. N Engl J Med 1991;324:212-216, 3. Internationa) Agency for Research on Cancer. 1ARC Mono graph* on the Evaluation of the Carcinogenic Risk of Chemical* to Man, Vol 15. Some Fumigants, the Herbicide 2,4-D and 2,43-T, Chlorinated Dibenzodioxins and Miscellaneous Indus trial Chemicals. Lyon: lARC, 1977. 4- Zack J, Suskind R. The mortality experience of workers exposed to tetrachlorodibenzodioxin in a trichlorophenol process acci dent. ) Occup Med 1980:22:11-14. 5. Zack JA, Gaffey WR. A mortality study of workers employed at the Monsanto Company plant in Nitro, West Virginia. Environ Sd Rea 1983;26:575-591. 6. Cook RR, Townsend MS, O tt MG, SUverstein LG. Mortality of employees exposed to 2,3,7,8-tetrachlorodibenzo-p-dtoxin (TCDD). J Occup Med 1980;22:530-532. 7. O tt MG, Holder BB, Olson BS. A mortality analysis of employ ees engaged In the manufacture of 2,4,5-trlchlorophenoxy acetic acid j Occup Med l980;22;47-50. 8. Cook RR, Bond GG, Olson RA, Ott MG, Gondek MR. Evaluation ,of the mortality experience of worker* exposed to die chlorinated dioxin*. Chemo&pherc 1986;15:1769-1776. 9. O u MG, Olson RA, Cook RR, Bond GG. Cohort mortality study ofchemical worker* with potential exposure to the higher chlorinated dioxins. J Occup Med 1987;29:422-429. 10. Bond GG, Molaren EA, Lipps TE, Cook RR. Update of mor tality among chemical workers w ith potential exposure to the higher chlorinated dioxins. J Occup Med I9S9;31: 121-123. 11. Zober A, Messerer P, Huber P. Thlrty-feur-year mortality fol low-up of BASF employees exposed to 2,3,7,8-TCDD after the 1953 accideL k it Arch Occup Environ Health 1990*62:139- 157. 12. Kodba R, Keyes D, Beyer J, Carreon R, Wade C, Dittcnbcr D, Kalnlns R, Franson L, Park C, Bernard S, Hummel R, Hum&on C. Results o f a two-year chronic toxicity and oncogenicity study of2,3,7,8-TCDD in rats. Toxicol Appl Pharmacol 1978^46:279- 303. 13. National Toxicology Program. Carcinogenesis btaa&aay of 2,3,7-tetrachlorod&cnro-p-dioxin (CAS No. 1746'01-6) In Os- bomc-Mendel rats and B6C3Fi mice (gavage study). DHHS Pub. No. (NIH) 82-1765. Washington DC: Nadonal Institute* of Health. 1982. 14. Hardcll L. Soft tissue sarcomas and exposure to phenaxyaectic adds and cancer; clinical observation (in Swedish). Lakartidnlngen 1977;74:2753-2754. 15. Fingerhut MA, Halperin WE, Honchar PA, Smith AB, G roth DH, Russell WO. An evaluation of rpons of diemn exposure and soft tissue sarcoma pathology in U.5. chemical worker*. Soand J Work Environ Health 1984;10:299-303. 16. Fingerhut M, Halperin W, Marlow D, Piacitelli L, Honchar P, Sweeney M, Greife A, Dill P, Steenland K, Suruda A Mortality among U.S. workers employed in the production of chemicals contaminated with 2,3,7,8'trtrachlorodibenzo-p-dloxin (TCDD). NIOSH Final Report PB91125971. Springfield, VA National Technical Information Service, 1990. 37. Percy C, Stande E, Glckler L. Accuracy of death certificates and ils effect on cancer mortality statistica. Am J Public Health 1981;71:242-250. 72 . Epidemiology January 1992, Volume 3 Number 1