Document dYD0Qgz13L9qjmKRyByD3RGL5

,4 > If) >I Mortality Experience of Workers in a Vinyl Chloride Monomer Production. Plant Patricia A. Buffler, Ph.D.; Susan Wood, MA.; Clayton Eifler, Ph.D.; Lucina Suarez, M.S.; ard Duane j. Kilian, MD. The evidence associating exposure to vinyl chloride with the risk of tumors of various sites, including lung cancer, is inconsistent In 1976 a mortality follow-up study 01464 white males employed in a vinyl chloride monomer (VCM) production plant since 1946 was conducted. Vital status was ascertained for 100% of the cohort Eight (28SM^of the 28 deaths observed were due to malignant neoj^Bis. No angiosarcomas or other liver tumors were observed. A statistically significant excess was noted for malignant neoplasms of the respiratory system (p=.03) The effect of smoking, duration of exposure to VCM, level of exposure, and the combined effect of duration and level 01 exposure were analyzed separately. A five-year latency requirement was maintained for all analyses except for the smoking analysis. Levels of exposure to VCM prior to 1971 were estimated from monitoring data available for the period 1971-75 by extrapolating the relative levels for job classifications backwards in time. Smoking histories were not available for 27.6% of the cohort. -When it was as sumed that all "unknowns"smoked, a significant excess of respiratory cancer was still observed (p = .05) When the minimum latency period of five years restricted analysis to the mortality experience aher five years from date of inrtial Exposure to VCM for the 314 employees satisfying this cri terion, the excess of respiratory cancer was moderate but not significant (p = .06). Both a longer duration and a higher level of exposure during the first five years following the-. tfate of initial exposure were associated with a statistically significant excess of respiratory cancer (p = .02 and p =.03, respectively). However, when duration and level of exposure were combined in an overall exposure index, the results were not significant (p=.07). The discrepancy in the results from the dos&response analyses may be due to the h ot 1ei$ School of Public Health. P.O Boi 20166. 77025 (Dr, Buffler and Ms, Wood and Suarez). Department of f*>entrve Medic me and Community Health. University of Texas Medical '*nch, Calveston. TX 77550 (Dr. Eifler), and Occupational Health and rtf'Cal Research. Dow Chemical Company, Freeport. TX 77541 (Dr Mian) Journal of Occupational Medicine/Vol. 21. No. 3/March 1979 potential error in the estimated levels and the small number of events observed, but the results do suggest that a relationship exists between exposure to VCM and respiratory cancer. The association of vinyl chloride and the develop ment of angiosarcoma of the liver' has been well documented. Vinyl chloride has also been described as a multi-system carcinogen causing tumors in the lung, cen tral nervous system, and hematopoietic systems.1'*. Most recently it has been suggested that vinyl chloride is a chemical mutagen and teratogen.* Early in 1974, three cases of angiosarcoma of the liver, an extremely rare tumor, were reported among workers in a vinyl chloride polymerization plant.4 All three workers had at one time been involved in the manual cleaning of polymerization reactors.7 In a recent report from Canada, this pattern was confirmed in 10 cases of angiosarcoma of the liver.* Seven of the workers cleaned reactors, and the rest were either operators or maintenance personnel. In itial exposures for all cases occurred before 1962. Following early reports, several epidemiologic studies of workers exposed to vinyl chloride were initiated. The first of these investigations was a proportionate mortality study by Monson and Peters.1. The study involved 161 deaths from two vinyl chloride plants: a plant producing vinyl chloride monomer (VCM), and the polymerization plant where the initial cases of angiosarcoma were iden tified. The deaths, which occurred between 1947 and 1974 among active and retired employees, were analyzed for excesses in cancer mortality. Monson and Peters reported a 50% excess in the proportion of deaths due to all malig nant neoplasms. Specific excesses were found for liver and biliary tract, lung and brain. These results suggested that vinyl chloride may be a multi-system carcinogen, although the excesses reported were based on propor- 195 rc/l! 2 ! (3); U '5, /-7 7?, f rv. t i fi donate mortality ratios, which are not specifically a polymer plant workers, and monomer plant workers. TheTM measure of risk. authors observed no relationship between length of In 1974, Tabershaw and Caffey5 published results of posure and risk of cancer. This particular analysis "^- an industry-wide historical prospective study of 8,384 however, has been challenged as methodologically inap---~ workers involved in the manufacture of vinyl chloride and propriate due to the misallocation of person-years.15 rl chloride polymers in 33 plants. The study was In 1975, the National Institute for Occupational Safety ricted to individuals who had at least one year of ex and Health reported results from a follow-up study of "v posure to vinyl chloride before [December 31, 1972. Ex1,294 workers at four polymerization plants, two of which cesses for- the specific causes of death were not statisti also produced vinyl chloride monomer.4 All workers in cally significant, except for cancer of the digestive cluded in the study had been exposed for five or more system, consisting primarily of previously identified years with at least ten years' latency period (ten years angiosarcomas of the liver. The excesses for cancers of since initial exposure). Significant risks for cancer of the the respiratory system and lymphomas, while not liver, lung, brain and central nervous system were found .. statistically significant were suggestive of possible risks for workers with 15 years of latency. An increased risk ; which appeared to be related to the dose of vinyl was also observed for lymphatic and hematopoietic chloride. Dose was measured by an average monthly ex cancers. , posure score determined for each worker and based on a By contrast, the study reported by Fox and Collier11 of - subjective estimate of relative levels of exposure for jobs 7,000 workers exposed to vinyl chloride in the British m- S~ within each plant. The authors quite correctly point out dustry did not indicate that cancers other than those of that exposure levels between plants may not have been the liver were associated with exposure to vinyl chloride 'V. comparable. In addition, the inclusion of many workers monomer. In this study initial exposures for a fairly high . with inadequate latency periods, some as short as one proportion of the workers occurred after I960, and year, may have obscured the exposure effects. This study follow-up by three of the eight factories was poor. In add*--^7 had been extended to more adequately evaluate latency tion, the determination and analysis of exposure were 2- and to include more complete information regarding the questionable in that a measure of level of exposure ascer* ." study cohort.' The final report is therefore based upon a tained retrospectively and subjectively by the par- total group of 10,173 employees from 37 plants, of whom ticipating industries was used rather than duration of ex- 9,677 (95.1 %) were successfully traced. While there were posure, some combination of level and duration, or.vsl' no statistically significant excesses of site specific specific job classifications. malignancies except for liver (angiosarcoma) and brain, Some background information on vinylidene chloride the risk of malignant neoplasms of the respiratory system (VDC). a structurally similar chlorinated hydrocarbon, is does appear to be associated with intensity of exposure-- also important in this review since exposures to VDC at ^^nndardized mortality ratio of 92 for the low intensity lower concentrations often occur simultaneously in the ^Regory compared to that of 141 for the high intensity VCM production area studied. VDC is a known hepato- ~ category. toxin14 and has been recently described as a genetically ' In a study presented at the New York Academy of active compound in the bacterial test systems used.11 Science conference on vinyl chloride, Nicholson et al1 The literature contains only one report on the health and described an excess in total mortality and in overall mortality experience of workers exposed to VDC without cancer mortality among 257 men with five or more years simultaneous exposure to vinyl chloride." In this study of of exposure occurring before 1963 in a polymerization 138 workers. Ott et al" reported no excess mortality due plant. In addition, they reported three deaths due to an to malignant neoplasms or adverse health effects at giosarcoma of the liver, one due to cancer of the brain, tributable to exposure to vinylidene chloride. two due to lymphoma and none due to lung cancer. Based upon this review of the literature, the evidence Two additional mortality studies' " resulted in essen associating vinyl chloride with the risk of tumors at tially negative results. One of these, a mortality follow-up various sites is strongly suggestive. However, some study of 594 employees at the Dow Chemical Company results, particularly those reported for lung cancer, are in polymer plant in Midland, Michigan, by Ott et al,' re consistent, and several questions remain regarding the vealed excesses in overall cancer only in the high ex strength of the association and the effects of varying posure category (> 220 ppm). In this study, evaluation of levels of exposure. The present study was undertaken to the effects of vinyl chloride was confounded by the fact further delineate the carcinogenic risk associated with ex that 72 of the 594 employees had also been exposed to posure to vinyl chloride by utilizing more definitive infor arsenic, a known carcinogen affecting the respiratory mation regarding duration and level of exposure, to system. Analysis of the mortality experience among the evaluate the effects at lower doses, and to address some cohort without the aresnic exposed workers indicated a of the methodological problems found in other statistically significant increase in mortality due to malig studies.'* " 10 Although it would be desirable to evaluate nant neoplasms, particularly lung cancer, among workers the specific effects of vinylidene chloride in the popula exposed to levels of vinyl chloride above 220 ppm. tion studied, unfortunately, it is not possible to separate Duck et al" found no significant excesses in overall or its effects from those of vinyl chloride. t specific mortality in a British Petroleum chemical *try population of 2,120 male workers exposed to Materials and Methods chloride. The analysis included a separate evalua Subjects of this investigation were persons employed in tion of mortality for autoclave workers (reactor cleaners)a, vinyl chloride monomer production plant of the Do*v Chen when this c vinyli study secut tweer Co the c since form, for < empl the ( techr, De emph 8th R provi' the N matic hospi death reviev Pe: infon and dassi chlor poter An in had i and jr of pt perio' randc ranke from altho over ing d. lot poter loade contr durin Much it war but b the high ; chlon pipeli opera proce Order Prepa work* avera often vinyl Empl, 196 R&S 100035 Mortality Experience in a VCM Production Plant/Bufller et al Journ; R&S 100036 Chemical Company. The facility began operations in 1948 when a small area of the chemical plant was dedicated to thi^aperation and to the simultaneous production of vil^Bene chloride. The criterion for inclusion in the study was that an employee had worked at least two con secutive months in the vinyl chloride department be tween August 1, 1948, and September 25,1975. Company personnel rosters were used to enumerate the cohort of persons who had worked in the department since 1948. Company records were utilized to compile in formation on date of birth, race, sex, and inclusive dates for each job and departmental assignment during employment. The vital status of employees who had left the company was determined by standard follow-up techniques. Death certificates were requested for all deceased employees. The certificates were coded according to the fith Revision of the ICDA and reviewed by a nosologist provided by the Environmental Epidemiology Branch of the National Cancer Institute. Pathologic or clinical infor mation was requested from the attending physician or hospital named on the death certificate for all cancer deaths. Clinical and pathology reports received were reviewed by the UTMB Pathology Department Personnel monitoring data at the plant since 1971 and information from the Department of Industrial Hygiene and plant supervisors allowed the grouping of all classifications with respect to potential exposure to vinyl chloride. Seven job classification groups with similar for exposure within each group were identified. An^Hhstrial hygienist and a panel of five persons who had Tong-term experience with the production process and job assignments ranked the seven categories in terms of potential for exposure to vinyl chloride during the periods 1948-70 and 1971-75. The categories were listed randomly and each member of the panel independently ranked the seven groups. The rankings were consistent from one time period to the other, suggesting that although actual levels of exposure may have changed over time, relative levels had not and the recent monitor ing data could be extrapolated backwards in time. lob classifications considered to have a relatively high potential for exposure included control lab personnel, loaders, and production personnel (control operators). A control lab worker sampled the product at several stages during the production process and analyzed it for purity. Much of this sampling is now automated. In earlier years it was common for the control lab worker to deliberately, but briefly, expose himself to vapors by methods such as the "sniff" test for sample purity, thereby incurring very, high short-term exposures. Loaders are exposed to vinyl chloride in the process of connecting or disconnecting Pipelines to tank cars and tank trucks. The control operators are responsible for monitoring the production process, performing minor maintenance procedures in order to insure proper functioning of equipment, and prefMj^p equipment for major repairs. Maintenance ^or^^Pgenerally have lower eight-hour time weighted average exposures than do the control operators, but they often experience relatively high short-term exposures to rinyl chloride while repairing worn equipment. Employees with the lowest exposures are those in super Table 1. -- Follow-up Status of 464 Workers in a VCM Production Plant, August 1, 1948 * September 25. 1975. Still employed at company No longer employed, alive Retired Released Deceased Died while employed Died after leaving company Died while retired Unknown status Total 291 145 27 118 28 17 11 0 0 464 visory positions followed by persons in the development lab. The assignments of supervisory personnel are such that they are not required to spend extended periods of time in the production area or to be physically close to the source of vapor emissions. Development lab person nel usually work with small quantities of vinyl chloride in evaluating the product or production process. The time weighted average exposures to vinyl chloride by job classification, based upon available monitoring data, were averaged for the period 1971-75 and are summa rized in Appendix A. Standardized mortality ratios (SMR's) were computed in the analysis of the mortality experience of the cohort. Expected numbers of deaths for the study population were calculated by applying 1950-59 and 1960-69 agecause specific death rates for white males in Texas to the observed distribution of person-years of observation, categorized into five-year age groups. Significance testing is based on the assumption of a Poisson distribution for the observed number of deaths, utilizing a one-sided test of significance. In the analysis of the data, the effect of smoking, duration of exposure, level of exposure and the combined effect of duration and level of exposure were each considered separately, although a five-year latency requirement was maintained for all analyses except that of the effect of smoking. Results Four hundred eighty-one males were identified for in clusion in the study population. Evaluation of mortality risks was restricted to white males due to the small number of nonwhite males (17) in the cohort. The as certainment of vital status for the 464 white males as of the cutoff date was 100% complete, and is described in Table 1. Follow-up investigation was required for 129 in dividuals who were no longer employed. Eleven of the total 28 deaths occurred among this group of 129 workers. Table 2 shows the distribution of the 28 deaths by underlying cause. Eight deaths (28.5%)were due to malig- Table 2. -- Distribution of 28 Deaths Observed Among Workers in a VCM Production Plant by Cause of Death. Causes ol Death (ICDA 8th Revision) Cancer (140-209) Hean (390-458) Accident (E800-E999) Other (038.9, 330.4 and 513) All causes No. of Death 8 10 7 3 28 Journal of Occupational Medicine/Vol. 21. No. 3/March 1979 197 Table 3. -- Case Summaries ot Eight Cancer Deaths Observed Among Workers in a VCM Production Plant. Underlying Cause ^^.of Death Revision ICOA) Date of Death Age at Deaths (Years) Date of Initial Exposure to VCM Age at Initial Exposure (Years) Years ot VCM Exposure Interval from Date of Initial Exposure to Date of Death (Years) Smoking Pathologic History Confirmation I '1 4. Cancer of lung 4-21-74 60 11-15-48 34 (162.1) 2. Primary carcinoma lunos (162.1) 12-01-58 53 10-18-48 43 3. Carcinoma ot lung (162.1) 5-21-73 58 8-31-51 36 4. Alveolar cell - carcinoma (162.1) 11-14-71 52 2-11-57 38 5. Malignant mediastinal tumor unclassified with generalized metastasis (163.1) 1-28-63 21 1-15-62 20 6. Carcinoma of colon (153.8) 4-12-71 48 11-17-52 29 7. Carcinoma ot lip metas. to lung & neck (140.9) 9-03-66 31 8-01-62 27 8. Metastatic squamous cell cancer. palate (145.1) 9-28-71 57 4-18-49 35 7.3 25.4 10.1 10.2 21.7 21.8 10.9 14.8 1.0 1.0 .6 18.4 4.1 4.1 6.0 22.4 Yes Yes -- X-ray report presumptive Yes No reports available Yes No reports available Yes Yes -- Biopsy Unk Yes -- Autopsy Yes No -- Biopsy report not received Unk No reports available -V Unk Yes -- Autopsy pe" resp Wit. sign all p app resp Effe t: may post peril elap niqu dent or c year.1 the r empi verse nant neoplasms, four of which were confirmed upon review of autopsy, biopsy or x-ray reports by the UTMB Pathology Department Case summaries for the eight cancer deaths are shown in Table 3. No angiosarcomas or other liver tumors were observed. The eight persons who died of cancer were initially exposed to vinyl chloride 0 1 963, and the four with lung cancer, prior to 1958. *i of exposure to vinyl chloride ranged from seven years to 22 years in the cases of lung cancer, and the inter val from date of initial exposure to date of death ranged from 10 years to 25 year*. Six of the 28 deaths reported here, including two of the eight deaths due to malignant neoplasms (malignant teratoma and alveolar cell car cinoma), occurred among a special subgroup of 165 workers exposed to 1,4-dioxane. These data were reported in an earlier mortality study of workers exposed to 1,4-dioxane.J1 Table 4a shows the observed and expected numbers of deaths by cause. For overall mortality, the standardized mortality ratio was 11% lower than expected. The total number of observed deaths due to malignant neoplasms was not significantly different from the expected (8 ob served vs. 5.19 expected). There is, however, a statistically significant difference between observed and expected for malignant neoplasms of the respiratory system (5 vs. 1.73, P = .03). Table 4a. -- Observed and Expected Numbers of Deaths Among 464 White Males in a VCM Production Plant, August 1, 1948 to September 25, 1975. Cause of Death All causes ^tealignant neoplasms MBnam neoplasms of the respiratory system *p= .032. one-tailed test Observed 28 8 5 Expected 31.63 5.19 1.73 SMR 89 154 289* 198 R&S 100037 Effect of Smoking The excess in mortality due to respiratory cancer >- necessitates a consideration of the effect of smoking as._^ an explanatory variable. Differential patterns of smoking among the vinyl chloride workers as compared to the T' Texas white male reference population might account for *T this excess. The case summaries indicate that four of the five workers who died of respiratory cancer had a history of smoking; the smoking status of the fifth worker is unknown (Table 3). In addition, smoking histories are not available for a large proportion (27.6%) of the 464 white males in the total cohort. Because of the missing data on smoking status, it is difficult to identify the effect that smoking patterns may have on the results obtained. The potential effect of smoking on the expected mor tality was examined, however, by noting the consequence of an assumed pattern of smoking for the "smoking unknown" category. Assuming the availability of standard age specific rates for the two smoking categories, the ex pected mortality, under the conservative assumption that all those in the "smoking unknown" category were actual ly smokers, was computed. Unfortunately, an appropriate standard set of age specific rates according to smoking status was not readily available, but a reasonable set of rates was constructed from available information. The construction of these standard rates is described in Ap- Table 4b. -- Observed and Expected Numbers of Deaths Occurring Five or More Years Past Initial Exposure Among 314 White Male* in a VCM Production Plant Prior to September 25, 1970. Cause ot Death Observed Expected SMR All causes All malignant neoplasms Malignant neoplasms of the respiratory system 22 25.18 B7 6 4.34 138 4 1.49 268^ *p = .06, one-tailed test Mortality Experience in a VCM Production Plant/Buflleretal Fifof ex: spent years to a r It i tion o lookir posur minat minirr in the of wf posun long I, other These nique, period Dur ing da exposr posure the fir; then c exposl the mr Person deaths date o. groups deaths the mir determ results statistic Journal pendix B. Under the extreme assumption that ail unknowns" actually smoked, the expected number of respiratory cancer deaths for the 464 white males is 1.98. With five deaths observed, this excess is of borderline significance (p = .05). Since it is extremely unlikely that all ^K>ns in the "unknown" category smoked, smoking to be an unlikely explanation for the excess respiratory cancer mortality. Effect of Five-Year Latency The latency period for occupationally induced cancers may range from five to 20 years from the date of first ex posure. In the above mortality comparisons, the ex periences of all workers, regardless of the length of time elapsed since initial exposure, are included. This tech nique may mask the effects of exposure in that a suffi cient latency period may not have accrued prior to death or observation. When the SMR's in Table 4a were recalculated utilizing a minimum latency period of five years from the date of initial exposure to vinyl chloride, the resulting SMR for malignant neoplasms for the 314 employees satisfying.this criterion was slightly lower, 268 versus 289 (Table 4b). Effect of Duration of Exposure Fifty-four percent of the cohort had less than two years of exposure to vinyl chloride. The average length of time spent in a vinyl chloride area for all 464 workers was 4.6 years. The values ranged from a minimum of two months to a maximum of 26.9 years. It is important to determine whether increased duratioi^xposure is associated with higher mortality. When Iooct^for such a relationship, bias may occur if the ex posure and observation periods overlap.20 Death may ter minate exposure prior to the satisfaction of some minimum exposure requirement, so that some deaths fall in the category of short duration of exposure, regardless of whether the death was causally related to the ex posure. In addition, a long duration of exposure implies a long latency period in which malignancies possibly due to other chemicals in the environment may be observed. These problems can be avoided by the following tech nique, which separates the exposure and observation periods. Duration of exposure during the first five years followmg date of initial exposure was noted for each individual exposed prior to September 25, 1970, counting only ex posures incurred prior to that date. Individuals surviving the first five years after the date of initial exposure were then classified into two groups according to duration of exposure in the first five years. The groups were divided at the median value for duration of exposure, 2.29 years: Person-vears of observation and the expected numbers of oeaths for the period following the first five years since date of initial exposure were then calculated for the two groups This procedure removes observed and expected deaths occurring among workers before completion of n^imum fiveyear latency period. Fig 1 illustrates the tiet^iation of these intervals for three individuals. The tesuT^^f t|ijs analysis are shown in Table 5. There is a Jlatistically significant excess of deaths due to respiratory Journal of Occupational Medicine/Vol. 21. No. 3/March 1979 cancer in the longer exposure group (4 observed vs. 1.05 expected, p = .023). Effect of Level of Exposure In order to further explore the relationship between ex posure to vinyl chloride and cancer mortality, one can consider a second dimension of exposure: the estimated level of exposure, or concentration. As previously men tioned, levels of exposure to vinyl chloride in the popula tion studied have decreased substantially in recent years, but based upon subjective evidence, the relative poten tial for exposure has not changed extensively. Therefore, the time weighted averages of exposures to vinyl chloride for the period 1971-75 were extrapolated backward in time to obtain an estimate of minimum exposure levels for the various job categories. Average exposure indices were obtained for each individual for the five-year period following initial exposure by multiplying the extrapolated exposure level for each job classification by the time spent in that job during the initial five-year period. These products were summed over all jobs in the initial period, then divided by the total time exposed to vinyl chloride in this initial five-year period. Only exposures before September 25,1970 were counted, thereby allowing for a five-year latency period as defined above. Categories of high and low level exposure were defined by the median extrapolated level of exposure for the 314 individuals alive and under observation at the end of the initial fiveyear period. The mortality experience observed subse quent to the initial five years for the categories of high and low average levels of exposure is illustrated in Table iExposure Storting Date [ August 1,1948 i Individual 6 Jonuory 1964 Observatiion Cut off Dale Sept.25,5ept.2S ` 1970 1975 t ' Individuoi! A t-- ^L- -- --"X Initial by t Ooservotion period , period Indio! 5 yr. Observation period period mdividuoi C Initial 5yr. penod exposure no exposure Fig 1. -- Example of determination ol duration ol exposure during five years following date of initial exposure to vinyl chloride for three employees exposed prior to date allowing lor 5-year latency (September 25, 1970), The initial date of exposure for Individual A was January 1, 1964, prior to September 25, 1970 (the cutoff date minus 5 years). He died on January 1,1975. Since Individual A was continuously exposed for the entire five-year interval from the date of his initial exposure, the person-years of observation after January 1, 1969, and his death would be assigned to the long exposure category, >2.29 years. By contrast, Individual B was exposed for only two years during the five-year interval trom the date ol his initial exposure, contributing approximately eight person-years to the short exposure category. Individual C represents an additional category of observations: those individuals tor whom an initial five-year interval was not completed prior to September 25, 1970 (the study cutotl date minus 5 years) and were therefore excluded from these analyses. R&S 100038 199 Table 5. -- Observed and Expected Deaths Among 314 White Miles in a VCM Production Plant by Initial Five-Year Duration o( Exposure.* Initial 5-Year Duration (Years) Short (< 2.29) Long ( >2.29) No. Of Persons 157 157 PorsonYears 1123 2227 Malignant Naetesmt All Matignant ot the AH Causes Neoplasms Respiratory System Obs. Exp. SMR Obs. Exp. SMR Obs. Exp. SMR 7 7.75 90 1 1.32 76 0 15 17.43 86 5 3.02 166 4 .45 0 1.05 381t * For 314 individuals exposed before September 25.1970 and exposures incurred before that date, counting deaths occurring five or more years after initial exposure t p = .023'. one tailed test 6. Again, there is a statistically significant excess of respiratory cancer in the group with a high average level of exposure (3 deaths observed vs. .68 expected, p = .032). Effect of Duration and Level of Exposure Finally, a dose-response analysis was carried out for a cumulative exposure index of information regarding both duration and level of exposure. This index, for a given in dividual and period of exposure, is the product of the duration of the period of exposure and the time weighted average level as previously defined. The mortality ex perience occurring five years after initial exposure for in dex groups of high and low level exposures, as defined on the basis of the first five years from initial exposure, is given in Table 7. The excess in the high level exposure group is not statistically significant (p = .07). died before termination, they would have been included in the study. Removing from the analysis those persons who terminated employment creates a tendency to overestimate the SMR, as noted in Table 8. A second potential source of bias in the analysis with limited follow-up stems from the possibility that persons ter minating employment before retirement may differ in cer tain demographic or environmental characteristics related to mortality. Lastly, the elimination of a signifi cant number of individuals from the study reduces the precision of the estimate of the SMR. This is particularly undesirable when small numbers of deaths are involved. Despite these problems, it is not uncommon to find this type of limited follow-up analysis in the literature. To avoid the various sources of error in this type of analysis, it is preferable to strive for a complete cohort as was done in this study.- lts With Limited Follow-up is of methodologic importance to the evaluation of ts from studies with incomplete follow-up, as well as to the conduct of future studies, to compare the results obtained with complete follow-up to those obtained with limited follow-up, that is, by "standard" techniques utiliz ing individuals who are easy to locate (current employees and retirees, and deaths occurring in these two groups on ly). Table 8 shows that the results with limited and with complete follow-up are similar when comparing overall mortality between the two groups, but that results differ somewhat when comparing mortality due to malignant neoplasms, specifically malignant neoplasms of the respiratory system. There are several sources of error in herent in the SMR's reported with limited follow-up. First of all, individuals removed from the analysis were certain to have survived up to their termination date. If they had Discussion One of the most challenging problems in a cohort study of the type presented here is the delineation of ex posure. Although no historical documentation of ex posure to vinyl chloride exists prior to 1970, it was reported that during the early period of production (1948-1960) exposures in the range of several hundred ppm (200-500 ppm) were not uncommon. During the 1950's and early 1960's the standard for exposure to vinyl chloride (threshold limit value. TLV) was 500 ppm." In 1961, based on chronic toxicity testing, the Dow Chemical Company voluntarily reduced their exposure standard to a TWA of 50 ppm (100 ppm ceiling)11 In 1974 the perma nent OSHA standard for exposure to vinyl chloride was reduced from 50 ppm to 1 ppm for an eight-hour period.'* As previously noted, workers in the cohort under study were simultaneously exposed to VCM vapors and varying Table 6. -- Observed and Expected Deaths Among 314 White Males in a VCM Production Plant by Estimated Relative Level of Exposure for Initial Five-Year Exposure Interval, * Avg. Laval ol Exposure During Initial 5 Years') Low High No. of Persons 160 154 PersonYears 1374 1977 Malignant Neoplasms All Malignant of the All Causes Neoptasms Respiratory System Obs. Exp. SMR Obs. Exp. SMR Obs. Exp, SMR 10 13.56 74 2 2.38 84 1 .82 122 12 11.62 103 4 1.95 205 3 . 68 441f For 314 individuals exposed belore September 25. 1970. exposures incurred during the live-year interval trom date ot initial exposure, and counting deaths occurring tive or more years after initial exposure t Based on 1971-75 monitoring data + p = .032, one-tailed test 200 Mortality Experience in a VCM Production Plant/Buffler el a* R&S 100039 Table 7, -- Observed and Expected Deaths Among 314 White Males in a VCM Production Plant by Exposure Index for initial Five-Year Exposure Interval.* Inltnl 5-Year Exposure Indexf No. of Persons PersonYears Malignant Neoplasms All Malignant of the All Causes Neoplasms Respiratory System Obs. Exp. SMR Obs. Exp. SMR Obs. Exp. SMR Low 157 1143 9 9.40 95 3 1.62 185 1 .56 180 High 157 2207 13 15.79 82 3 2.72 110 3 .94 319J For 314 individuals exposed betore September 25. 1970, exposures incurred during the live-year interval Irom date of initial exposure, and counting deaths occurring five or more years after initial exposure t Based on 1971-75 monitoring data t p= .07, one-tailed test concentrations of vapors from the production of VDC. ethylene dichloride, methyl chloroform and ethyl chloride. A weakness of most mortality studies of chemical indistry employees is that workers may have been exposed to many other chemicals while working in other areas of the production facility or in other chemical companies before or after the period of specific observa tion.We were able to determine that no members of the cohort had been exposed to arsenic or asbestos while they were employed at the Dow Chemical Company. In view of the lack of data regarding levels of exposure prior to 1971, an assumption was made that while levels of exposure were higher before 1971, the ratios of levels for any two job classifications remained approximately constant. It is believed that reasonable estimates of the relative levels of exposure prior to 1971 were made by backwards extrapolation of post-1971 exposure data. It is also important to note that initial exposure to vinyl chloride occurred in mid-career (3-4-43 years of age) for all four persons who died of lung cancer (Table 3\ In this regard, the lack of information on previous employment or occupational exposures for these individuals adds to the difficulty of interpreting the significance of these statistical findings. To pursue the question of whether this excess may be causally related to VC exposure, a dose-response relation ship was examined. Dose was considered to have two dimensions, duration and level of exposure. In testing for a dose-response relationship, a comparison of mortality may be biased if the periods of exposure and observation overlap. We avoided this source of bias by looking at an initial five-year exposure period and a subsequent obser vation period, thereby also allowing for a five-year laten cy period past initial exposure. With regard to duration of exposure in the first five years following initial exposure, there was a statistically significant excess of respiratory cancer occurring after the initial five-year period (4 versus 1.05 expected. p = .02) among workers in the longer ex posure group. No such excess was observed for workers in the shorter exposure group. In the three analyses of dose-response performed in this study, the most accurate and objective measure of dose available is duration of exposure, which ignores the concentration component of dose. On the other hand, the cumulative index takes into account both duration and level, is based on data extrapolated into the past, and assumes that the relative levels of exposure did not change over time. If this assumption were false, any ex isting dose-response relationship would be obscured, had the true cumulative indices been known. The use of the time weighted average levels as a measure of dose is not so heavily dependent on this assumption. Possible ex planations for the fact that statistical significance was observed in two of these analyses, but not the levelduration index, may be due to the potential error in the estimated levels, and in the small numbers of events observed. It should be mentioned that our inability to detect an increase in mortality in the low level exposure groups does not necessarily indicate that no increase ex ists, but may be due to the low power associated with small expected numbers, and a longer latency period at lower doses. A more appropriate statistical technique for these types of dose-response analyses might be a comparison of the increase in the high level (or long) exposure group to that in the low level (or short) exposure group. Statistical inference on the ratio of the true underlying SMR's for the two exposure groups can be carried out conditional on the number of deaths in the two groups combined.37 However, when the numbers of deaths are small, the power of such a comparison is very low, and may even be zero. Because of this inefficiency, it was deemed ap propriate not to report significance levels, but rather to note that although statistical significance was not ob served, the probability of observing such was very low. Throughout this study, as is common in occupational studies, the measure of mortality used is the $MR. There are, however, several problems with the use of SMR's that should be kept in mind when interpreting such results, especially when comparisons of SMR's are made. Limited 1oliow-up Complete follow-up Table 8. -- Observed and Expected Deaths Among White Males in a VCM Production Plant by Type ol Follow-Up. No. of Persons PersonYears Overall Malignant Malignant Neoplasms Mortality Neoplasms of the Respiratory System Dbs. Exp. SMR Obs. Exp. SMR Obs. Exp. SMR 335 3067 18 20.55 88 6 3.43 175 5 1.16 431* 464 5313 28 31.63 89 8 5.19 154 5 1.73 289t * p= .007. one-tailed test t p = .032 Journal ol Occupalional Medicine/Vol. 21, No. 3/March 1979 R&S 100040 201 ,, Although the SMR is an adequate measure of excess mor tality as compared to the mortality of a standard popula tion. the comparison of two SMR's depends not only on the differences in the age specific mortality rates of each 10. On MC, Langner RR, and Holder BB: Vinyl chloride exposure tn a controlled industrial environment. Arch Envron Health 30333-339,1975. 11. Duck BW, Carter JT, and Coobes E): Mortality studv of workers in a polyvinylchloride production plant Lancet 2.1197*1199.1975. 12. Wagoner IK and Infante PF: Vinyl chloride and mortality? Letters to wp from the standard, but also on the age specific the Editor. Lancet 2:194-195,1976. *Ijlation weightings of the two groups. Thus, it is entire- 13. Pox A| and Collier PF: Mortality experience of workers exposed to possible, if the age distributions are vastly different, vinyl chloride monomer in the manufacture of polyvinyl chlonde in Creat Britain. Br / Ind Med 34:1-10,1977. that the two study groups might have equal age specific 14. Jenkins LI. Trabulus Ml. and Murphy SD; Biochemical effects of mortality rates but somewhat different SMR's. Despite 1,1-dichloroethvlene in rats: Comparison with carbon tetrachloride and 1, the problems in the use of SMR's, no alternative was con sidered because of the small numbers involved. 2-dichloroethylene. Toxicol Appl Pharmacol 23301-510,1972. 15. McCann J. Choi E. Yamasaki E. and Ames BN: Deteaion of car cinogens as mutagens tn the Salmonella/microsome test Assay of 300 Conclusions chemicals. Proc Natl Acad Set 723135-5139,1975. 16. Baruch H, Malaveille C. Montesano R, and Tomato L Tissue In view of our inability to detect a significant doseresponse relationship, we cannot state that the observed excess in respiratory cancer deaths is due to exposure to vinyl chloride. However, the fact that excesses were seen in the group with the longer duration of exposure in the mediated mutagenicity of vinylidene chloride and 2-chiorobutadiene m S. Typhimurium. Nature 255641-653.1975. 17. Ott MC, Fishbeck Wa. Townsend MS, and Schneider ): A health study of employees exposed to vinylidene chloride. / Occup Med 18:735-738,1976. 18. Pasternak BS and Shore RE: Statistical methods of assesskig rk initial five years, and in the group with the higher average estimated exposure levels in the initial five years, suggests that a relationship may exist It is unfortunate that data were not available regarding following exposure to environmental carcinogens, in Environmental Health: Quantitative Methods, A. Whrttemore (Ed.) Philadelphia: Society for Industrial and Applied Mathematics, 1977. pp 49-71. 19. Liddell FDK. McDonald |C. and Thomas DC Methods of cohort analysis; Appraisal by application to asbestos mining. I Royal Sat Soc the levels of exposure experienced 20 to 30 years ago, 140-469491, 1977. when the four workers with respiratory cancer were first exposed to vinyl chloride. It is also unfortunate that data were not available regarding exposures to other chemicals during these earlier decades. 20. Enterline PE: Pitfalls in epidemiological research. I Occup Med 18.150-156.1976. 21. Buffler PA, Wood SM. Suarez L. and Kilian D|: Mortality folkwwup of workers exposed to 1,4-dioxane, I Occup Med 20255-259,1978. 22. Kahn HA: The Dom Study of smoking and mortality among U-S. The results of this study can only be considered in con junction with other studies, both past and future. In view of inconsistent reports to date, more studies of the specific dose relationships and confounding exposures veterans: Report on eight and One-half years of observation. .National Cancer Institute Monograph No. 19. 1-125, U3. Department of Health Education, and Welfare. January 1966. 23. U3. Department of HEW: Changes in cigarette smoking habits be tween 1955 and 1966. Public Health Service Publication No. 1000, Series needed. 10, No. 59. U.S. Government Printing Office. Washington, D.C. 24. ACCIH: Threshold limit values for 1960. AMA Arch Envirn Health Tntj study was funded bv a research award from Dow Chemical Co , U.S.A, The authors wish to acknowledge the support and assistance of the various Departments of Dow- Chemical Co., Texas Division and U.S.A.. without whom this studv could not have been conducted the Occupational Health and Medical Research Department, the Industrial Hygiene Department, the In dustrial Medicine Department and the Personnel Department The critical 1:62. 1960. 25. Totkelson TR, Oven F, and Rowe VK: The toxierty ot vinyl chlonde as determined by repeated exposure of laboratory animals. Am Ind Hyg Assoc I 22354-361. 1961. 26. Department of Labor, Occupational Safety and Health Administra tion: Exposure of vinyl chloride. Federal Register 3935890-35898. Part II. review oi this report bv Dr Ralph Cook. Dow Chemical Co. USA and the October 4,1974 assistance provided bv the U'TmB Pathoiogx Department, the Universih ot Texas at Houston Education and Research Computation Center, and the En vironmental EpiOemioiQgv Branch oi the National Cancer Center m reviewing 27. Ederer F and Mantal N- Confidence limits on the ratio of two Poisson variables. Am I Epidemiol 100,165-167,1974. and processing data acknowledged This paper was issued as the Dow Technical Release 6-600-487-78 Appendix A References 1 Monson RR and Peters |M. Proportional mortality among vinyl chloride worisers. Lancet 2:397-398. 197a 2. Tabershaw IR and Caffev WR: Mortality study of workers in the manufacture of vinyl chloride and its polymers. / Occup Med 16:509-518, 1974, 3 Nicholson WJ Hammon EC, Seidman H. and Selikoff IJ Mortality experience of a cohort of vinyl chloride-polyvinyl chloride workers Ann NY Acad Sci 246.225-230.1975 4 SVaxweiler RJ, Stringer W. Wagoner |K, and tones I, Neoplastic risk among workers exposed to vinyl chloride Ann NY Acad So 2714048. 1976 5 Infante P. Wagoner IK. McMichael A), et al Cenetic risks of vinyl chloride Lancet 1 734-735. 1976. 6 Creech JL and Johnson MN. Angiosarcoma of liver in the manufac ture of polyvinyl chloride. I Occup Med 16 1 50-151, 1974. 7. Block |: Angiosarcoma of the liver following vinyl chlonde exposure JAMA 229:53-54. 1974, ^Delorme F and Theriault C: Ten cases of angiosarcoma of the liver in flvngan. Quebec. I Occup Med 20 338-340. 1978, 9. Equitable Environmental Health, Inc.: Epidemiologic studv of vinyl chloride workers: Final report submined to Manufacturing Chemists Association, January, 1978 Estimated Time Weighted Average (8-Hour) Exposure to Vinyl Chloride by Job Classification, 1971-1975 Percent of Total Average Person-Years Job Classification 8+1 r. TWA Exposed 1. Control lab personnel 22.3 4.7 2. Development lab personnel 4.1 5.7 3. Production personnel Control A 7.7 16.6 Control B 2.8 7.5 Control C 4.4 8.9 Class 3 operators -- .8 Total 5.7 33.8 4. Loaders and plant men Class 1, 2 operators Head packaging operator 7.1 12.3 15.2 M Material handling operator .2 .1 202 Mortality Experience in 2 VCM Production Plant/Buftler et si R&S 100041 Packaging operator, service technician Total 5. Certain supervisory positions Production super intendent, assistant production superintendent, production engi neer, R&D engineer, engineer, safety engi neer, Sr, production engineer. Parts technician, Sr. Manager Assistant engineering technician General superintendent, section superinten dent, superintendent supervisor, assistant superintendent plan ning engineer, material control clerk, plant assistant head clerk, chief material handling technician, shipping coordinator, mainte nance coordinator. maintenance engineer Total 4aintenance personnel Boilermaker, apprentice Welder, apprentice Machinist, apprentice. helper, crew leader Pipefitter, apprentice. helper Utility man Instrument technician Production foreman, shift foreman Maintenance foreman Utility crew leader, rotating shift foreman, foreman Total -- 7.4 1.9 2 .3 .0 -- 1.8 1.4 2 1.1 1.7 2 3.2 4.3 .4 -- 2.5 .1 16.4 93 .1 2 .0 1.9 11.6 12 1.1 1.7 42 12 3.7 5.5 2 52 24.0 7, Other personnel Electrician, apprentice Loading supervisor, technical foreman, coverer, janitor, elec trical foreman Total All groups combined 33 ___ 3.8 5.4 2.0 13 33 100.0 For iomt job cUsjHiejuorw there were no monitoring diu tvailabie lor the mervil 1971-75 The value assigned was determined by avengesg at mentwm data available lor job classifications withm the group to which the specilic rob claudica tion was assigned The averages computed were weighted accenting to tout personyears in the cohort spent in each job claudication. Appendix B* Despite the lack of suitable standard age specific rates for smokers (present or past) and nonsmokers (never smoked), it is possible to construct a reasonable set of rates from available information. The following two simplifying assumptions are made: 1. Age specific relative risks for smokers compared to non smokers from the study by Dorn11 for U.S. white male veterans for the years 1954 to 1962, are applicable to the Texas white male populations of 1950-59 and of 196069. 2. The age specific percentages of smokers for U.S. males for the years 1955 and 19661J reflect the correspon ding percentages for the Texas white male population in the time periods 1950-59 and 1960-69, respectively. Using these two assumptions, a set of age specific rates for the Texas white male population according to smok ing status was constructed for use as a standard, in the following manner. Let Mi= cause specific mortality rate for age group i M*-- cause specific mortality rate for smokers in age group i Mi= cause specific mortality rate for nonsmokers in age group i s.= proportion of population in age group i who smoke r,= age specific relative risk for smokers com pared to nonsmokers Then M, = SiM* + (1-s.) M,, = M[Sir, + (1 - s.)] Mm = M./fsA + (1-s.)]; M -- nm*. Thus, the values for r. and s. are found for assumptions 1 and 2 above, and the desired mortality rates by smoking status are obtained. Truth Truth rests on several conditions. Among other things, truth rests on a regard for relevant facts, an intelligent assembling of them and on knowing how the facts matter. Truth also rests on knowing what is important and what is not, on judgmental capacity and on courage. -- from "When Values are Substituted for Truth" by | Bennett, m Thp Wj// Sire*f lOurnil. |ofy 25. 1978, Journal of Occupational Medicine/Vol. 21, No. 3/March 1979 r&S 100042 203