Mortality and incidence of cancer among Swedish rubber workers, 1952-1981

HOLMBERG B. Mortality and incidence of cancer among Swedish rubber workers, 1952-1981. Scand J Work Environ Health 12(1986) 538- 544. The mortality and inci dence of cancer was studied among 8 734 workers from two Swedish rubber manufacturing companies. Mortality wasinvestigated from 1952to 1981and cancer incidence from 1959to 1980. The expected num bers of deaths were calculated from national statistics. No significant risk excesses were detected when the cohort was analyzed without consideration of employment time or latency period. However, the mor tality from coronary heart disease and the incidence of lung cancer were increased when the study period was limited to 2: 40 years since first employment. The standardized mortality ratio for coronary heart diseasecorrelated positivelywith employment duration . The mortality from asthma, bronchitis, and em physema was nonsignificantly increased. The incidence of bladder cancer was increased among indivi duals with heavy and long-term exposure in the weighing and mixing departments. Twenty-five percent of the individuals in the cohort were not Swedish citizens at the time of employment, and an analysis of the mortality and cancer incidence in this group showed a markedly increased lung cancer incidence for certain immigrant groups, probably mainly due to ethnic factors. The results indicate that ethnic fac tors must be considered in the analysis of occupational groups when a high proportion of the workers are immigrants.

Reprint requests to: Dr P Gustavsson, Department of Occupational Medicine, Karolinska Hospital, S-I04 01 Stockholm, Sweden.
Some rubber additives used in the manufa cturing process are mutagenic in bacterial test systems (18), and vulcanization ga ses also have mutagenic properties (10 , 17). Biological monitoring o f workers in the weighing a nd mixing departments sho wed an increased frequency o f chro moso me aberrations in pe ripheral lymphocytes (23), and an increased mutagenic activity in the urine (31). These observations indicate that rubber workers are exposed to mutagenic substances in the work en vironment , and the y are co nsistent with the epidemiologic findings o f increased cancer ris ks . The cancer hazard in the rubber industry was evaluated by the International Agency for Research on Cancer in 1982 (21) .
A retrospective cohort study from two Swed ish rubber manufacturing indust ries was initia ted in the middle of the 1970s, and the results of a follo w-up during the period 1961 to 1977 ha ve been reported earl ier (20). Th e cohort co m prised 13 114 individuals. Of the se 2 492 were un expo sed office workers, included as an internal referen ce group. The preliminary results showed an excess of malignancies in the lungs , gastrointestinal organs, urinary bladder, and skin among the exposed workers, as well as an increased mortality from coronary heart disease. However, several risk excesses were also detected among the unexposed office workers, a finding that made the interpretat ion o f the results so mewha t co m plica ted.
The follow-up of this cohort of rubber wor ker s ha s now been exte nded prospectively and retrospectively to cover the time period 1952 to 198I. Furthermore, the results have been anal yzed with regard to emplo yment time, latency period, calendar time , and the nationality of the cohor t members.

Subjects
Company records were used as a source from which to gather information about the workers' personal ident ification number , name, expo sure , employment period, and nationality at the time of first employment. All manufacturing workers, ie, workers in the weighing and mixing departments, calender workers, vulcanizers, pressing workers, tire builders, inspection and maintenance workers, cleaners, and storage handling workers, were included in the exposed group. All white-collar workers, ie, office workers and department head s, were included in the unexpo sed group . Persons chan ging their emplo yment stat us from manufacturing worker to office worker were included in the exposed group, the chang e being accounted for in the calculat ion of exposure time.
The original cohort consisted of 14611 persons, also including those who were lost to follow-up. The following restrictions or correction s were appli ed to the cohort: Indi vidual s with an emplo yment time of less than one year were excluded. Individuals leaving/work before 1952 were excluded since follo w-up was not possible via the national computerized registers befo re th is year. Howe ver, many individuals started work several decades earlier. Perso nal identification number s were checked, and duplicates in the registrations were corrected . After these procedures the cohort consisted of 12 212 persons (7 932 men and 4 280 women). The total numb er of person-years at risk in the exposed group was 179 894. At the start of employment 9 105 ind ividuals were Swedish citizens, and the rest were immigrant s, mainl y from the Nordic coun tries, Yugoslavia, and Greece.

Methods
The follow-up procedure included the mat ching of the personal identification number s with comput erized registers at the Swedish National Central Bureau of Statistics. Register s of the total living population , emigrations and immigrations since 1968, and death s since 1952 were used. Mortality was investigated for the period 1952 to 1981. Cancer cases in the cohort were identified for the period 1959-1980 by matching with the National Cancer Register . Indiv iduals not traced in these registers were considered lost to follow-up and were excluded. Indi vidual s who had emigrated were also excluded fro m the statistical analy sis.
The number of death s, cancer cases, emigrations , and indi vidual s lost to follow-up are presented in ta- Th e computation of expected values was based on national stat istics for mortality and cancer incidence. The computations were performed accord ing to the person-year method, counting man -years at risk specific for gender, calend ar year, and five-year age class. The number of person-years in each stratum was multiplied with the disease rates in the corresponding stratum of the national statistics, and the number of expected cases was added over strata. The summary risk measure obtained was the standardized mortality (or morbidity) rat io (SMR) computed as (observed/expected ) x 100. Confidence limits for the SMR were computed with an exact method (28). Significance testing was ba sed on the addition of probabilities derived from the Poisson distribution , and two-sided tests were used . Laten cy time was calculated as the time period from first employment unt il the start of risk calculation . The computations were performed with a computer prog ram developed at the University of Linkeping (1) for epidemiologic calculations.
Non-Swedish citizens at the time of first employment comprised 25 % of the total ; this level is higher than the nation al average. Since cause-specific mortality and incidence of cancer vary between ethn ic groups, the majority of the statistical analyses were performed on the Swedish citizens in the cohort (tables 2-10). Mortalit y and incidence of cancer among the immigrants was analyzed separately (table 11).

Results
The observed total mortality was lower than the expected (table 2) due to the so-called "healthy worker effect." Cause-specific mortality was calculated for the exposed men (table 3). There was a slightly increased mortality from asthma, bronchitis, emphysema, and pancreatic cancer. None of these risk elevations was statistically significant. The total cancer incidence corresponded well with the expected (table 4). Site-specific cancer incidence was calculated for the exposed men (table 5). No significant risk excesses were observed.
In the analysis of the mortality and incidence of cancer during different time periods after first employment (table 6), the SMR values for total numbers of deaths, coronary heart disease, and total cancer incidence increased with increasing follow-up time. There was a significantly increased incidence of lung cancer after 40 years or more since first employment. The SMR for asthma, bronchitis, and emphysema fluctuated, the risk during follow-up between 20 and 40 years after first emplo yment being nearly statistically significant [95 0J0 confidence inter val (CI) for the SMR Table 6. Mort al ity and cance r inc idence by tim e since first employment. Ris k calculati on up to and inc luding 79 years of age.
(SMR = standardi zed mortal ity ratio) , P < 0.05. Other causes of death being 98-388]. The number of cases of pancreatic cancer and bladder cancer was smail, the SMR fluctuated, and there was no time trend. For a further evaluation of the observed risk excesses the relation between exposure time and SMR was investigated, with a latency requirement of 40 years Table 9. Incidence of cancer among workers in the weighing and mixing departments, 1959-1980, as divided into two subgroups. Risk calculation up to and including 79 years of age.
(SMR = standardized mortality ratio)  (table 7). The mortality from diseases of the circulatory system was positively correlated to duration of exposure. An increased incidence of lung cancer was observed in all three exposure-time categories, but the SMR was significantly raised only among those exposed for at least 30 years. The time trend could not be evaluated because of the small number of observed cases.
A group of exposed individuals had worked in the weighing and mixing departments, "mixers." The SMR was computed for mixers with a latency requirement of 10 years (subgroup I) and for a subgroup first employed before 1951, exposed more than five years, and with a latency requirement of 20 years (subgroup 2) (table 8). The latter subgroup was more heavily exposed since the exposure time was longer and the exposure occurred in a time period when dust control was less developed. The SMR was higher among the more heavily exposed for a majority of the causes of death, but it was not significantly raised for any of the studied diagnoses. The cancer incidence for the mixers is presented in table 9. Again the SMR was higher in the more highly exposed group when compared to that of all mixers. Four cases of bladder cancer were observed versus 1.08 expected, corresponding to an SMR of 368 (p < 0.05).
The mortality and cancer incidence by time of first employment was analyzed (table 10). There was an excess of lung cancer only among those employed before 1951. However, it should be pointed out that this is another reflection of the findings in the latency time analysis (table 6), since only those first employed before 1941 had a latency of 40 years within the study period, and therefore they were all included among those first employed before 1951.
Among the office workers, an increased incidence of tumors of the lung and larynx was found (II observed versus 4.46 expected lung tumors and 3 observed versus 0.59 expected cases of laryngeal cancer). However, no relation to latency from first employment was found in this group, in contrast to the exposed group.
The mortality and incidence of cancer among the exposed male immigrants was analyzed (table II). The Swedish national statistics were used as a reference, and the results must be interpreted with caution since ethnic factors, as well as occupational exposures, may have influenced the results. Immigrants from Finland had an increased total mortality, an increased risk for violent death, and an increased lung cancer incidence. Immigrants from Denmark had a raised cancer incidence and lung cancer incidence. Immigrants from Yugoslavia and other immigrant groups did not exhibit significant excess risks.

Discussion
There was no significantly increased mortality or cancer incidence among the rubber workers when the cohort was analyzed without regard to latency or exposure time. However, the mortality from coronary heart disease and the incidence of lung cancer was increased when 40 years had elapsed since first employment. The SMR for coronary heart disease correlated positively with employment time. The finding of a dose-response relationship indicates that the risk excess was due to occupational factors. An increased risk for coronary heart disease was reported in two earlier studies (2, 26), but has not been found in the majority of studies on rubber workers . The finding of a risk excess only after more than 40 years since first employment may be interpreted in the following two ways: first, the causative agent might not have been present after 1940, and, second, the toxic agent might take four decades or more to mani fest its effect as increased mortality. A median latency time of 38.5 years was described for lung cancer caused by exposure to asbestos (4), and a similar relationship between lung cancer and rubber work is possible. No definite conclusions can be dra wn on this point , and it remain s unknown if the causative factor has been present in more recent time periods or not.
The mortality from asthma, bronchitis, and emphysema (leo 490-493) was increased in the time period 20-40 years after first employment. The increase was of borderline statistical significance. An increased frequenc y of chronic bronchitis and reduced pulmonary function was earlier reported among rubber workers (11,12,13), but increased mortality from lung diseases has not been reported earlier.
Bladder cancer is well documented as an occupational cancer hazard among rubber workers . In this study, a risk excess was found only among individuals who had worked in the weighing and mixing departments before 1951. Similar results were obtained in a case-referent study from the rubber industry in the United States, where an excess risk of bladder cancer was noted especially among worker s in a mixing department and among calender workers (7). It seems probable that the special exposure conditions in these departments are related to the bladder cancer excess, especially since an increased mutagenic activity in urine has been found among workers in mixing departments (31). However, excess risks for bladder cancer may exist also in other departments, even if not detected in this study.
The present results, obtained after an extension of the follow-up period to 1952-1981, differ on some points from the earlier report covering the time period 1961-1977 (20). The proportion of individuals lost to follow-up has been reduced by the restrictions made with regard to nationality, employment time, and employment period. Increased frequencies of liver and pancreas tumors and of malignant melanomas were observed among the exposed workers in the earlier report. After the extension of the follow-up period no excess risks were noted for these tumor sites. The excess risk for respiratory malignancies observed among unexposed men was unaffected by the extension. The exposure classification was rechecked and found to be correct. The excess remain s unexplained, but could partially be due to smok ing habits.
The number of observed cases of leukemia corresponded well with the number expected; this finding contrasts with the results of several American studies of rubber workers. Benzene was earlier used as a solvent in the rubber ind ustry but was abandoned in the Swedish industry before 1950. It seems as if benzene has been used in more recent years in the American rubber industry (22), and its use could explain the epidemiologic finding .
The analysis of mortality and cancer incidence among the immigrants revealed some high excess risks. Immigrants from Finland had an increased total mortality and an increased incidence of lung cancer and violent deaths. It should be pointed out that the Swedish national stati stics were used as reference. Therefore, the results reflect a comb inat ion of occupational and ethnic factors. Regarding the first, it is possible that the immigrants got the "worst" jobs and therefore had a higher exposure. However , no corresponding risk excesses were observed for immigrants from non-Scandinavian countries. Regarding the latter, the Finnish population has a higher incidence of lung cancer (33) and coronary heart disease (35) than the Swedish population, and the increased risks can, to a large extent, be explained by ethnic factors. Since no individual exposure data are available, this question remains unanswered, but the results are important from a methodological point and show that the ethnic composition of a cohort must be accounted for in the analysis.
The cohort comprised individuals of different job categorie s and , with the exception of mixers, a jobspecific analysis was not possible . Addit ional excess risks, related to uncommon exposure factors, cannot be ruled out. The conclus ions that can be drawn concerning the risks associated with the present exposure situation in the rubber industry are limited . Exposure levels have been reduced in comparison to the situation in the 1930sand 1940s. The long latency time for the diseases under study makes it impossible to evaluate epidemiologically the effects of exposures during more recent decades. However, the results from gen-543 otoxicologic investigations made during recent years corroborate the epidemiologic findings and indicate that a further reduction of exposure is warranted.