Epidemiologic study of vinyl chloride workers: mortality through December 31, 1972.

A population of 10,173 men, employed in 37 plants, was identified as having worked for at least one year in jobs involving probable exposure to vinyl chloride monomer (VCM) prior to January 1, 1973. Of the 9677 men whose vital status was determined, 707 were known to have died. For 699, death certificates were obtained. The standardized mortality ratio (SMR) for all causes was 89, that for all malignancies was 104. The only type of malignancy found in significant excess was in the category of malignant neoplasms of the brain and other parts of the nervous system; 12 deaths occurred where 5.9 were expected, for an adjusted SMR of 203. There were slight but inconclusive upward trends in all malignancies, and for malignancies of the respiratory tract, digestive tract, and central nervous system associated with reported levels of maximum exposure to VCM. When groups in whom less than 20 years had elapsed from the first exposure were compared with those with 20 or more elapsed years, and 25 or more elapsed years, no significantly different SMR's were detected for major primary sites of malignancy. Plans for an updated study of mortality, to include deaths in the period 1973-1979 are briefly discussed.

The epidemiologic studies of vinyl chloride workers summarized in this report were carried out during the period June 15, 1973 through December, 1976 by Tabershaw-Cooper Associates, Inc., and Equitable Environmental Health, Inc., for the Manufacturing Chemists Association (MCA) (now the Chemical Manufacturers Association).
An initial report, dealing with 8,384 workers from 34 plants, was prepared May 3, 1974 (1). A summarized version (2) was published in 1974. The study population was subsequently increased and follow-up was improved. After an interim report in 1976 (3), a final report based on 10,173 workers was prepared in January, 1978 (4). In all of these studies the observation period ended December 31, 1972.

Participating Plants
In mid-1973, the MCA identified 43 plants in the United States, belonging to 19 companies, which either produced vinyl chloride monomer (VCM) or used it in the production of poly(vinyl chloride) *2150 Shattuck Ave., Suite 401, Berkeley, California 94704.
October 1981 (PVC). Of these, 34 were included in the initial study; four were excluded because they had been in operation less than 5 years, one had stopped production in 1966, and in others information on job histories or exposures was deficient. Three plants were subsequently added to the original 34, so the 1978 report included 37 plants. Of these, 11 produced only VCM, 18 produced only PVC, three produced both, and five plants produced homopolymers and copolymers, with or without VCM and PVC.
The geographical distribution of those in the study, as shown in Table 1, indicates a disproportionate number of workers from the South, particu-

The Study Population
The objective was to include all male employees whose work had involved exposure to vinyl chloride for at least one year prior to December 31, 1972. The designation of jobs which involved exposure to VCM was made by staff members at individual plants or by a corporate industrial hygienist. In approximately two thirds of the study population, TCA staff copied personnel records on individuals who were identified as having been in exposed jobs. In the remaining portion of the population, detailed information on exposed individuals was provided by plant personnel. The methods used in doing this are described in the complete reports cited earlier.

Study Period
The period of time in which the work-force of a plant was included in the study depended upon the date it began making or using VCM and also upon the earliest date when personnel records were complete for all employees, if that was later than the foregoing. This was done to eliminate periods when there was differential record retention of workers terminated, deceased, or retired. The end of the study period was December 31, 1972.

Estimates of Exposure
In each plant, every job and location with VCM exposure was graded in terms of probable exposure. Originally, a job history form was designed in the expectation that the exposures could be quantified in parts per million. This proved impossible in practice. However, for each plant, jobs and locations involving the highest exposures could be classified as "high", and other jobs classified as "medium" or "low" relative to the "high". It is recognized that this subjective classification is of questionable validity in categorizing the past and present exposure of a given worker. From the number of months spent in jobs with classifications of 3 (high), 2 (medium) and 1 (low), a number of exposure categories were developed for use in later 102 Criterion Unit Duration of exposed employment Months Interval from beginning of exposure to end of Months observation Estimated maximum level to which an individual High was exposed for at least 12 months, Medium classified as high, medium, or low. Low Integrated or cumulative exposure, crediting 1 1 for each month at low, 2 for each 2 month at medium and 3 for each month 3 at high exposure to VCM Exposure index (EI) = Cumulative score/average number of months analyses. As shown in Table 2, individual exposures in various papers were classified in a number of ways, including the maximum level at which an individual had been exposed for at least 12 months, an integrated or cumulative exposure, and an exposure index based on the cumulative score divided by the number of months.

Follow-Up
As is customary in historical prospective studies, all who had left employment were traced when possible. Methods included form letters sent by mail and use of retail credit follow-up. For the first report, there was insufficient time to utilize Social Security Administration records, but for subsequent reports such follow-up was used.
In the first, or 1974 report, 85% of the study population was located; for the final or 1978 report, the percentage had been increased to 95%.
The mortality calculations were based only on those who were successfully traced, which is equivalent to assuming that mortality among those not found was the same as among those who were found. This usually, but not always, results in some overestimation of mortality.

Calculation of Standardized Mortality Ratios
Each worker in the study, i.e., everyone whose vital status was ultimately known, was considered to have been under observation from the date on which he attained a year of exposed employment or from the date when his plant's records were complete, whichever came later. Observation periods ended December 31, 1972, or on the date of death, whichever occurred first.
Observed deaths were classified by cause according to the 7th (1955) revision of the International Environmental Health Perspectives Classification of Diseases. The expected number of deaths by cause were calculated by using age and cause-specific mortality rates for United States males with the same birth years and age distribution for the years 1950, 1955, 1959, 1965, 1967 and 1970. Deaths for which death certificates could not be obtained were assumed to have the same cause distribution as the death certificates that were obtained. Appropriate increases were made in calculated SMR's.
The statistical significance of the deviation of each SMR in the study population from an expected value of 100 was tested by a method derived from Chin Leong Chiang (5). The formula for determining the standard error of the SMR was 100 x SMR S T No. expected deaths If an observed SMR differed from 100 by more than 1.96 standard errors, it was regarded as significant at the 5% level; if it differed by more than 2.57 standard errors, it was regarded as significant at the 1% level. SMR's based on fewer than five observed deaths were usually not tested for significance. Table 3 summarizes the numbers of individuals, success of follow-up, person-years and deaths in successive phases of the study. The number of deaths per 1000 man-years of observation, which in general reflects the age distribution of the work force, suggests that the proportion of older individuals increased in the study population as it was expanded and follow-up improved. Even so, the 5.88 deaths per 1000 man-years indicates that a relatively young population was being observed; the U.S. male population 20 and above has about 11 Table 3. Numbers of individuals, success of follow-up, personyears, and deaths analyzed in successive phases of the study. Report

Duration of Exposure to VCM
In the first report, only 15.2% of those studied had begun exposure prior to 1950 (22 years before end of the observation period). In the third report, 33.4% had had 20 or more years for observation since exposure began. This resulted from finding more early employees by improved follow-up, and the inclusion of an older group from one plant in the augmented population.

Standardized Mortality Ratios
Standardized mortality ratios (SMR's) for selected causes of death are shown in Table 4, based on the 352 deaths analyzed in Report 1 and the 707 deaths analyzed in Report 3. Results in the first report led Tabershaw and Gaffey (1) to conclude that vinyl chloride may be associated with cancer of a number of sites, notably digestive cancer, respiratory cancer, cancer of other and unspecified sites (primarily those of the central nervous system) and lymphomas. This was based not on statistically significant excesses in each category, but upon apparent trends when different levels and durations of exposure were compared.
With the enlarged study group, the SMR for malignancies in the entire population dropped slightly, as did the SMR's for malignancies of the buccal cavity and pharynx, digestive tract and respiratory tract. However, tumors of the brain and central nervous system, when examined separately, still appeared to be in excess.
A number of analyses were done in the third report in an attempt to sharpen the focus on work exposures.
There appeared to be a slight but definite trend in the SMR's for all malignancies, malignancies of the digestive tract, the respiratory tract, and for other and unspecified sites with increasing levels of estimated maximum exposure (Table 5). However, the numbers of expected deaths were relatively few in some categories and the groups differed widely in age distribution as manifested by deaths per 1000 person-years.
To reduce dilution of the study population by men whose exposures had begun only recently, a separate analysis was carried out on those whose exposures to VCM had begun 20 years or more prior to 1972, and on those whose exposures had begun 25 years or more prior to 1972 (Table 6).
Another analysis was made of a population of men who had worked in plants producing only PVC (where VCM exposures were presumably high), whose exposures had begun 20 years or more before the end of the study period and who had been reported as having medium or high VCM exposures for a year or more (Table 7). This group experi-enced 210 deaths where 249.7 had been expected. The pattern of mortality from malignancies was not appreciably different from that of the total study group.
In summary, increasing the study population and bSignificant at 5% level. improving follow-up did not strengthen the suggested associations between VCM exposure and malignancies other than those caused by hepatic angiosarcoma, as will be pointed out later, and a suggested association with tumors of the brain and central nervous system.

Angiosarcomas
Nine angiosarcomas are known to have occurred in the U.S. during the study period, i.e., prior to 12/31/72. As shown in Table 8, eight of these were found in the study, but only three were coded as angiosarcoma on the death certificate. However, four others were coded as tumors of the digestive tract. Unfortunately, two were coded 230x so as to fall out of the category for malignant tumors of the GI tract, and one was coded as cirrhosis of the liver.
As shown in Table 8, the angiosarcoma which was not found was in a man who had died in 1961. We have not determined how he failed to be in the study population. The years of exposure for the eight cases ranged from 4 to 23 years, while elapsed time from beginning of exposure to death ranged from 16 to 24 years.

Tumors of the Central Nervous System
The 12 tumors of the brain had been diagnosed on death certificates as follows: glioblastoma multiforme, 4 (1 confirmed by autopsy); astrocytoma, 2 (2 au-  topsies); ependymoma of the 4th ventricle (autopsy); "malignant brain tumor" or "carcinoma of the brain," 5 (with no autospies). This group is currently being made the subject of more rigorous review. The information obtainable from our records, summarized in Table 9, is insufficient to prove or disprove a cause-and-effect relationship between occupational exposure and these tumors.

Conclusions
A study of 707 deaths in a population of 9677 men who had worked for one year or more in jobs involving exposure to vinyl chloride and whose vital status had been determined as of December 31, 1972, did not show a significant excess of deaths due to malignancies. There did appear, however, to be a significant excess of tumors of the brain and central nervous system, based on 12 such deaths. There also continued to be slight but inconclusive trends toward higher SMR's for deaths from digestive tract and respiratory tract tumors associated with maximum levels ofpast exposure. No striking changes in malignancy patterns were apparent when analyses were directed toward individuals in whom 20 to 25 years had elapsed since first exposure. The results suggest that, except for a proven association with hepatic angiosarcoma and a strongly suggestive association with central nervous system tumors, vinyl chloride probably is not associated with significant excess cancers of other sites.
It should be emphasized that the epidemiologic study summarized in this report was planned, the populations defined, and analysis under way before cases of hepatic angiosarcoma had been diagnosed in workers exposed to vinyl chloride (6). An update is scheduled with inclusion of additional deaths in the cohort during the years 1973 through 1979. The study can be improved by a separate analysis of data from the plants which began operations before 1960, and by separating, insofar as possible, exposures to vinyl chloride monomer, polyvinyl chloride, and various copolymers. It is also hoped that criteria for defining exposure and for rating levels of exposure can be improved to permit better indices of integrated exposure.