Mortality of workers exposed to styrene in the manufacture of glass-reinforced plastics.

Epidemiologic studies have suggested an increased risk of leukemia and lymphoma among workers exposed to styrene. In a further exploration of this possible hazard, an analysis was conducted of the mortality among 7,949 men and women employed during 1947-1984 in eight British companies manufacturing glass-reinforced plastics. The subjects were identified from company files and traced to the end of 1984 through National Health Service and National Insurance records. The overall mortality in the cohort was less than in the national population (693 deaths observed, 830.1 expected) as was mortality from cancer (181 deaths observed, 223.7 expected). In particular, there was a deficit of deaths from lymphoid and hemopoietic cancer (6 observed, 14.9 expected). The small excess of lung cancer (89 deaths observed, 80.1 expected) was not statistically significant and can probably be attributed to chance. Among 3,494 hand laminators (the job with the highest exposure to styrene) there was one death from lymphoma and none from leukemia. The findings do not exclude the possibility that styrene is a human carcinogen, but give no support to the hypothesis that it causes leukemia and lymphoma.

Styrene ha s been produced commercially for mo re than 50 years and is used mainly in the pr oduction o f pla stics, resins , and synthetic rubbers . In 1977 wo rld wide production was estimated at 7 million tons (4). Th e highest exposures to the compound usually occur in industry, but contact with styrene is not confined to the workplace . Styrene-based resins are sold for domestic use as woodfillers, a nd low levels of styrene have been recorded in foods packaged in pol ystyrene containers.
The discovery that vinyl chlo ride cau ses an giosarcoma of the liver rai sed concern that sty rene, with its som ewhat similar chem ical struc ture, might also be carcinogeni c. Suspicions ha ve been increased by th e demonstration that st yrene is mutagenic (5) , po ssibl y through metabolic ac tivation to styrene oxide (9). Moreover, increased rates of chromosom e aberrat ion s have been found in wo rkers exposed to styrene (3,8). On the other hand, animal tests of carcinogenicity have been inconclusive (5).
Several epidemiologic inve stigations have suggested high rates of leukemia or lymphoma among people working with styrene (2, 6, II) . However, the total number of reported cases is small, and a confounding effect of other chemicals such as benzene cannot always be excluded . There is thus a need for additional epidemiologic data on this possible ha zard.
The best opportunity to study the chronic health effects of styrene is in th e manufacture of glassrein forced pla stic s, since exposures are higher than in other industries and the range of potential confounders is smaller. We have examined the mortality and cancer incidence o f employees at eight Brit ish companies ma king a va riety of products from glassrein for ced pla stic s.

Subjects and methods
Det ails of the com pan ies surv eyed and th e cri te ria by which the st udy cohort was defined are given in table I . We identi fied subjects from per sonnel and wage records and abstracted each person ' s name, address, sex , date of birth, and National Insurance number, as fa r as the information was a va ilable. We also recorded a history of the j obs which ea ch subj ect had held du rin g his o r her employment , incl uding the dates of sta rting and finishing wo rk.
With help from the management a nd staff of the facto ries, and prior to our analysis of mortality and cancer incide nce , we cla ssified jobs according to th eir potential for exposure to styrene. Four grades of expo sure were dist ingu ished -high (hand lam inators), moderate (subjects who regularly ent ered areas o f glas s-reinforced plastics production or worked close to lam inating operations although not activel y engaged in lam inating) , low (subjects who occasionally entered areas of gla ss-reinforced plasti cs production or exper ienced a co nsta nt but low-le vel expo sure at a situation remote from the laminating operation), and background (no exposure other than the background level in th e factory as a whole) . No hygiene da ta were ava ilable for the early years of th e study period , but since 1975 styr ene concentrations had been mea sur ed at factories A , B, C , E, and F. On the basis of th ese mea surements we estima te that our high exposure category corresponds to an 8-h time-weight ed average exposure of 40-100 ppm. Wher e personal respiratory protection was provided, it took the form of particulate filters and would not have affected exposure to styrene. Of the other po ssible hazard s to which the study population was exposed , the most imp or tant were glass fiber, acetone, methyl ethyl ketone, organ ic pero xides, and (at factories C and F onl y) asbesto s.
Memb ers of the cohort were traced through the National Health Service (NHS) Central Register and Nation al Insu ran ce Ind ex, and their vital status on 31 December 1984 was det ermined . The Office of Population Cen suses and Sur veys pro vided us with copies of the death certificates of subj ects who had died , with the underlying cause of death coded to the nin th revision of th e Int ern ational Cla ssification of Diseases. In addition it notified us of any cancers that ha d been registe red in the study population during the period of follo w-up .
The mortalit y of the cohort was compared with that of the national population of En gland and Wales by the person -years method and was examined in relation to length and gra de of pot ential exposure to styrene.
Confidence intervals and statistical significance were calculated by th e usual technique s based on the Po isson distribution.

Results
Th e study population comprised 8 354 subjects, bu t 405 (4.8 070) could not be included in the analysis becau se their sex, date of birth, or date of first employmen t was not avail able from the compan y records. A further 210 subjects could not be traced in the files of the NHS Central Register or Na tio nal Insurance Index and could onl y be follo wed to the date on which they were last emplo yed by one of the compa nies under investigation . One hundred and seventy-one members of the cohort emigrated during the study period and were followed to the time of emigration . Of the 7 949 subjects whose records were suitable for anal ysis, 6 638 were men and I 311 were women.
Details of the follow-up achieved at individual companies are given in table 1. The personnel records held by company D often lacked important items of information , and as a co nsequence the proportion of its employees excluded fro m the analysis or untraced was mu ch higher th an elsewhere. In view of the pos sible biases resulting from thi s loss, th e findings for compan y D are presented separat ely from tho se of the other compa nies. • Six subjects each work ed at two co mpanies -fou r at factories C and E, one at fact ories C and H, and one at factories B and D.
Of the subjects included in the analysis, 5 434 worked in jobs entailing exposure to styrene at levels above background. Table 2 shows the distribution of these men and women according to the length of their exposure to styrene , and also according to the most highly exposed job in which each worked. A total of 2 458 subjects worked in exposed occupations for at least 12 months, and 3 494 were at some time employed in jobs with high exposure.
At the companies where the follow-up was reasonably complete (A-C, E-H), the overall mortality was less than that of the national population (637 deaths observed , 768.8 expected) as was mortality from cancer (167 deaths observed, 208.1 expected) and from the other major disease categories (table 3). The findings for company D were similar, apart from a small and statistically nonsignificant excess of death s due to digestive disease (5 deaths observed, 1.6 expected). Table 4 shows the mor tality of the cohort from specific cancers, and tabl es 5, 6, 7, and 8 give a breakdown of the cancer deaths according to grade and duration of exposure to styrene, time since first ex-posure , and calendar period of first exposure. Several tumors occurred more frequently than would have been expected from the nat ional rate s, but in no instance was the risk statistically significant at a 5 <J lo level.
The most prominent excess was for lung cancer (89 deaths ob served in the entir e cohort , 80. I expected). The death rates from bro nchial carcinoma were the highest for subjects with mode rate and high exposur e to styrene, although the dose-response relationship was not entirely consistent (table 5). The raised rate was concentrated particularly among workers with one to nine years of employment in an exposed jo b (table 6). Among subjects exposed for ten or more years, there was no excess. Nor did the rates of lung cancer increase with time since first exposure (table 7). The highest standardized mortality ratio was found for worker s first exposed after 1969 (table 8). When the expected mortality rates were corrected by the 1968-1978 standardized mortality ratios for lung cancer in the local authority areas in which each factory was situated , the total numb er of expected lung cancer deaths was virtually unchan ged at 80.0. a Two thousand five hundred and fi fte en subj ects wi th only background exposure have been excluded. Five subjec ts wo rked in exposed jobs at two compan ies.  In addition one subject died from cancer of the mouth, one from unspecified digestive cancer, one from testicular cancer, and ten from carcinomatosis with unknown primary. C In addition one subject died from cancer of the small intestine.
Deaths from lymphoma, myeloma, and leukemia all occurred less frequently than expected (table 4). Altogether six subjects died from lymphoid and hemopoietic malignancies. In addition, a further eight such cancers were registered among members of the cohort who were still alive at the end of the follow-up period or who had died from other causes. If these cases are added to the six deaths, the total is still less than the 14.9 deaths expected. Only one of the six observed deaths occurred in a subject with high exposure to styrene -a hand laminator from Company D who died of non-Hodgkin's lymphoma.

Discussion
This mortality study is statistically more powerful than previously reported surveys of styrene workers, with more than 200 deaths expected among the highly exposed group of hand laminators. Our estimates of the exposure levels associated with this occupation were based upon measurements made since 1975 at five of the eight factories studied, and they are in broad agreement with those which have been found in similar 2 circumstances by other investigators (10,12,13,15). There is no reason to believe that work conditions were better in the earlier years of the study period or at the factories for which no hygiene data were available.
A possible weakness of the investigation is the incomplete follow-up achieved, particularly at factory D. This failure was due largely to deficiencies in records made at the time of employment -for example omission of full first names or date of birth -and as such is unlikely to have been biased in respect of subsequent mortality. Nevertheless, we felt it safer to analyze the results for factory D separately. Of the 7 376 subjects from the other seven companies, 96.7 % were both traced and suitable for analysis.
The death rates of the cohort were lower than those of the national population for all the major disease categories, including cancer. This deficit probably reflects a healthy worker effect, although there was little increase in the standardized mortality ratios when allowance was made for a latency of 10 or 20 years from first exposure to styrene.
Our a priori suspicions of carcinogenicity centered on the risk of lymphoma and leukemia suggested by Table 5. Mortality from selected cancers by grade of exposure to styrene -Companies A-C and E-H." (0 = observed number of deaths, SMR = standardized mortality ratio)  earlier studies, but in our cohort there were fewer deaths than expected from lymphoid and hemopoietic malignancy. Nor were these cancers concentrated among subjects with the longest and highest exposure to styrene. A precise analysis of cancer incidence (as opposed to mortality) was not possible because of the incompleteness of cancer registration in England and Wales, especially during the early years of the study period. However, if the known occurrences of lym-phoma and leukemia among living subjects and subjects who died from other diseases are added to the deaths caused by these tumors, the count of cases still falls short of the number of deaths expected in the cohort. The apparent discrepancy between this finding and the positive results of some previous studies may be a chance effect, but it could also be due to unrecognized confounding exposures in the chemical and plastics industries. In the only other reported survey of mortality among glass-reinforced plastics workers (10), no cases of lymphoma or leukemia were found.
As might be expected, several cancers showed standardized mortality ratios that were greater than 100, but in no instance was the excess mortality statistically significant at a 5 070 level. Only two of the observed associations were based on more than three deaths. Four subjects died from ovarian cancer when 2.7 deaths would have been expected. However, two of these women had only background exposure to styrene, and a third had been exposed for less than a year.
Of more interest is the excess mortality attributed to bronchial carcinoma. At companies where followup was acceptably complete, there were 83 deaths from lung tumors, the standardized mortality ratio being 112. The excess was apparent at five of the factories studied (A, B, C, E, and H) and could not be explained by concomitant exposure to asbestos, which was confined to factories C and F. Unfortunately, no data were available on the smoking habits of the study population, and interpretation must therefore be speculative. It is worth noting, however, that the mortality of the cohort from respiratory disease was well below that of the national population. This finding suggests that the smoking rates were not excessively high. Analysis by level of exposure to styrene gave some indication of a dose-response relationship, but against this evidence must be weighed the lack of an increase in mortality with time since first exposure and the absence of a risk of lung cancer in most previous studies of styrene workers (1, 2, 6, 7, 11). An increased rate of lung tumors has been reported in one other investigation -a study of boatbuilders using glass-reinforced plastics in the United States -but it was statistically nonsignificant and limited to subjects with minimal styrene exposure (10).
On balance it seems unlikely that styrene is a cause of lung cancer. If the excess mortality from the tumor did not occur by chance, it is probably due to a confounding exposure in the glass-reinforced plastics industry. From the limited range of agents encountered in the production of glass-reinforced plastics, perhaps the most likely candidate would be glass fiber dust. On the other hand, no clear excess of lung cancer has been demonstrated in studies of glass wool manufacturers (14). Further follow-up of the cohort may clarify the picture.
The findings of this survey are broadly reassuring. In a substantial group of workers, many with high exposure to styrene, we found no overall excess of mortality from cancer and, more specifically, a deficit of deaths due to lymphoma and leukemia. However, although larger than previous investigations, the study had only limited power to detect cancers with a long latency. After allowance for a latency period of 20 years from first exposure only five cancer deaths would have been expected among the hand laminators exposed to styrene for at least 12 months. The carcinogenicity of styrene is thus still in question.