Incidence of lymphohematopoietic malignancies among styrene-exposed workers of the reinforced plastics industry.

Incidence of malignancies among styrene-exposed workers of the reinforced 1994;20:272-8. OBJECTIVES - The goal of this study was to determine the risk of lymphohematopoietic malignan cies for workers exposed to styrene. METHODS - This was a historical cohort study. The observed numbers of newly diagnosed cases of lymphohematopoietic malignancies in the study population were compared with expected numbers based on the national rates. The study took place in the Danish reinforced plastics industry, in which high exposure levels of styrene occur frequently in an environment free of most other suspected car cinogens. Altogether 36 525 male employees of 386 companies producing reinforced plastics and 14254 nonstyrene exposed employees of similar industries were studied. RESULTS - An insignificantly increased standardized incidence ratio (SIR) of 1.22 [95% confidence interval (95% CI) 0.88-1.65] was found for leukemia among all employees of the reinforced plas tics industry. Workers employed in the 1960s (the period with the highest recorded styrene levels) or in companies with the highest exposure probability showed increased SIR values of 1.54 (95% CI 1.04-2.19) and 1.38 (95% CI 0.75-2.32), respectively. Both estimates increased when a l O-year latency period was considered. CONCLUSIONS - An increased risk of leukemia was shown for workers in the early phase of the rein forced plastics industry in Denmark. If this association is not due to chance or confounding, the most likely cause is styrene exposure.

Styrene is a major component of plastics and synthetic rubbers . Small amounts of styrene are found in foods kept in plastic conta iners and also in the urban atmo sphere (1), while high expo sure levels occur in the plastics industry (2). Increased mortality from non-Hodgkin's lymphom a, Hodgkin's disease, and leukemia has been indicated in some studies (3-7) but has not been found in all (8)(9)(10)(11)(12). The objective of this study was to evaluate the incidence of Iymphohematopoietic malignan cies among employees exposed to styrene in the reinforced plastics industry, in which particularly high styrene levels are recorded in an environment free of most other suspected carcinogens.

Companies
All companies producing reinfor ced plastics in Denmark at any point between the early 1960s and 1988 were traced with assistance from the Work Inspection Service, the National Institute of Occupational Health, the Danish Plastic Federation, the Unskilled and Semiskilled Worker's Union, local authorities for protecti on of the external environment, and the Central Bureau of Statistics and from an examination of teleph one book s (13 ). Altogether 552 companies were identified.
On the basis of independ ent reviews of the production by two dealers of plastic raw materials, the companies were clas sified as ever or never producing reinforced plastics . The companies unknown to the dealer s were classified as having unknown production. The two dealers agreed independentl y on the production in all but 4 of 328 compani es known by both of them (kappa 0.94).
The companies were classified in a similar manner according to information on production given by 368 employers responding to a postal questionnaire, to which 184 employers did not reply. The employers and dealer s agreed on the product ion for 281 out of the 309 companies known by the dealers and mentioned in the questionnaires (kappa 0.72) (table I).
Most of the companies producing reinforced plastics were boat yards or manufa cturers of containers produced by hand lamination. According to the dealers the companies not involved in the production of reinforced plastics were produ cing wooden boats (N =17) or thermoplastics (N = 19) or were within the metal industry (N = 17) or were dealers of reinforced plastics goods (N = 31).
The companies were all rather small. Fifty percent had an average work force of fewer than five employees. Fifty percent of the companies were closed by the end of the stud y ( 1989).

Subjects
All employees of the 552 companie s were identified by means of unique personal ident ifica tion numbers in the Supplementary Pension Fund , which has administered a pension system for all Danish wage earners since 1964. A total of 53 720 male workers empl oyed at any point between 1964 and 1988 and resident in Denmark after I Janu ary 1970 was included in the study. (Six employees with invalid personal identification numbers were not included.) According to the dealers this figure repre sented over 90% of all reinforced plastics workers in Denmark between 1964 and 1988. Female workers (N = 10 799) were excluded from the study since the majority were not involved in the production of reinforced plastics.
According to the dealers ' classification of the companies each worker was categorized into one of the following exposure categories (in order of priority): (i) exposed: ever employed in a comp any producing reinforced plastics during a period of production, (ii) unexposed : emplo yed in a company never producing reinforced plastics or in a company producing reinforced plastics but only outside years of production, (iii) exposure unknown: employed in a company with unknown production only. According to this Scand J Work Environ Health 1994, vol 20, no 4 classification 36 525 male workers were classified as exposed (table I). Similar categories were constructed from the employers' classifications. For the exposed employees the first and last year of exposed employment was recorded, and for all the other workers the first and last year of any employment was recorded. Duration of employment was calculated from the amount paid each year by the emplo yers to each emplo yee 's account in the Supplementary Pension Fund. For the exposed workers only payments recorded during exposed emplo yment were included . For 60% of the workers the payments corresponded to less than 12 months of emplo yment.
For each company the dealer s estimated the proportion of all employees workin g in the production of reinforced plastics. For 99 companies the estimate was between I and 49% of a work force of 23 688 emplo yees, and for 287 companies it was between 50 and 100% of a work force of 12 837 employees. If a mean proportion of 25% is applied for the companies empl oying less than 50% reinforced plastics workers and a mean proportion of 75% is applied for the companie s employing 50% or more, then an estimated total of 15550 [(0.25 ·23688) + (0.75· 12837)] reinforced plastics workers is obtained, corresponding to 43% of all emplo yees of companies producing reinforced plastics.
A total of 292 emplo yers report ed the prop ortion of all workhours spent producing reinfo rced plastics in their company. This measure was assumed also to indicate the employees ' probability of working with reinforced plastics. The dealer s' estimates correlated well with the emplo yers' report s (figure I).
The 12 837 male workers of companies with an estimat ed proportion of reinforced plastic s workers above 50% were included in an intern ational cohort study of cancer mortality amon g workers of the reinforced plastics industry (14).
In the Central Population Register, vital status for all worker s was reported by 31 December 1989 (table 2). In the Danish Cancer Register 1915 newly diagno sed cancer cases, including 161 cases of lyrn- a Compan ies were c lassifi ed according to ever or never produ cing reinforced plasti cs ; employees were cla ssif ied accord ing to ever or never employment during the perio d of productio n of reinforced pla st ics . Denmark between 1964 and 1988 as part of their surveillance routines (2). During 1964During -1970During , 1971During -1975During , and 1976During -1988 the mean styrene levels were 180, 88, and 43 ppm, respectively. The level thus decreased significantly over time. Of the measurements, 1814 were sampled from 128 companies included in the study, but the identity of the sampled workers was not available. The mean annual level of styrene was therefore calculated for each of the 128 companie s. Altogether 9335 workers were employed in these companie s during the year of measurement. These workers were selected for a separate analysis.  Years at observation were counted from 1 January 1970 or 1 January the year following the first year of employment if 1970 or later to ensure that only cancer cases occurring after the start of exposure were included, as no information on day or month of employment was available . Follow-up continued until the date of death, emigration, disappearance, or the end of the study on 31 December 1989, whichever came first. A mean follow-up of 10.9 years with a range of 0 to 20 years and a total of 584556 person-years were observed.

Stati stics
Standardized incidence ratios (SIR) were calculated from a comparison of the number of malignancies observed in the study population with the number of expected cases according to the national incidence rates, standardized for gender, age, and year of diagnosis . Ninety-five percent confidence intervals (95% CI) were calculated according to the Poisson distribution (15). Internal comparisons were made for the incidence rates with Poisson regression models (16). In tests for linear trend, the exposure variables were treated as continuous variables .
In table 3, the SIR values are presented for lymphohem atopoietic malignancies by the production and proportion of reinforced plastics workers of each company according to the dealers' reports. Among the exposed workers of companies producing reinforced plastic s the SIR values for non-Hodgkin's lymphoma and leukemia were slightly higher than those obtained for the total study population (observed 42, SIR 1.33, 95% CI 0.96-1.80 and observed 42, SIR 1.22,95% CI 0.88-1.65, respectively). Non-Hodgkin's lymphoma showed a significantly increased risk among employees of companies with 1-49% reinforced plastic s worker s. The increase was seen in particular during the first 10 years after the start of employment. For both Hodgkin's disease and leukemia the SIR values were higher for employe es of companies with 50-100% reinforced plastics workers, but none of these risk estimates significantlyexceeded 1. For Hodgkin's disease the in- crease was confined to the first 10 years after start of employment, while for leukemia the increase was confined to the period of 10 years or more after the start of employment (observed 10, SIR 1.92,95% CI 0.92-3.52). Alllymphohematopoietic malignancies, except multiple myeloma, showed increased risks for workers of companies assumed to produce reinforced plastics but the production was unknown to the dealers. The estimates were based on small numbers and only non-Hodgkin ' s lymphoma showed a significantly increased SIR value, however, during the first 10 years of follow-up. The risk of leukemia was significantly increased among the exposed workers starting employment in compan ies producing reinforced plastics during the 1960s. This was in particul ar the case among employee s followed 10 years or more after the start of employment (observed 25, SIR 1.69,95% CI 1.09-2.49) (table 4). Among the workers starting employment in the 1960s in a company with 50% or more reinforced plastics workers, an SIR of 2.02 (observed 6,95% CI 0.74---4.40) was found (data not presented). The risk pattern for Hodgkin' s disease by year of first employment was similar to that of leukemia, but not significant. The occurrence of non-Hodgkin's lymphoma showed little difference across the three time periods. In the Poisson model the incidence ratio for leukemia showed a significantly declining trend during subsequent periods 1964-1970, 1971-1975, and 1976-1 988 (P = 0.02). When time since first employment was included in the model, the slope was reduced to an insignificant value .
In table 5 the SIR values for Iymphohematopoietic malignancies in exposed worker s is presented according to the time since first employment and the duration of employment, as calculated from the contributions to the Supplementary Pension Fund. Non-Hodgkin' s lymphoma showed increased SIR values within the first 10 years after the start of employment, while no increase in risk was seen when 10 years of latency was taken into account. A statistically signific ant increased risk of leukemi a was found after 10 years of latency (observed 32, SIR 1.57, 95% CI 1.07-2.22). The increase was confined to workers employed for less than one year.
Direct comparisons with the unexposed population by Poisson regre ssion modeling showed rate ratios close to the results present ed.    In the subpopulation of employee s of companies for which styre ne measurements were available, Hodgkin ' s disease showed an SIR of 2.48 (observed 5,95 % CI 0.80-5.78) for employee s of companies with a recor ded mean styrene level above 50 ppm and an SIR of 1.37 (observed 2, 95% CI 0.17-4.95) for emp loyees of compan ies with a mean level below 50 ppm. The SIR values of the other Iymphohematopoietic malignancies were not increase d in this population and did not correlate with increasing styrene level (data not shown) .

Discussion
A slightly increased incidence of leukemia was found among male employees in the reinforced plastics industry . The incidence was significantly increased among employees in the early years of production.
When account was taken for an appropriate latency period , a significa ntly increased risk was also observed. The estimates were slightly higher for workers empl oyed in comp anies where the majority worked with reinforced plastics than in companies where less than 50 % were employed in this production. No increase was, on the other hand, seen in the subpopulation of workers of companies with available measurements of styrene. The risk pattern s of non-H odgkin ' s lymphoma and Hodgkin' s disease were less consistent. No indication s of an increa sed risk of multiple myeloma were found.
Among the workers of the 82 companies unknown to the dealers, increased risks for leukem ia and lymphomas were obser ved. We traced the owners of 59 of these companies, and 30 verifie d a production of reinforced plastics, while 29 denied ever having produced this material. The risk of lymphomas and leukemi a was increased for the exposed workers in the 30 companies (observed 4, SIR 3.33, 95% CI 0.91-8.46), which may have been due to higher exposure levels of styrene in less known companies. On the other hand, an increa sed risk for leukemia (observed 5, SIR 3.63, 95% CI 1. 18-8.46), but no excess for lymphomas, was found for the presumably unexposed workers of the 29 companies. Th is finding could indicate an aggregation of nonoccupational risk factors for leukemia among workers in less known companies.
Only a few women took part in the production of reinforced plastics according to both the employers and the dealers of plastic raw material s, and female employees were not included. We have earlier presented a significantly reduced risk of non-Hodgkin ' s lymphoma (observed 1, SIR 0.16, 95% CI 0.00-0.88) for the women of this popul ation, while the observed number of cases of Hodgkin' s disease, multiple myeloma, and leukemia among the women were close to the expected values based on the national rates (13).
Almost all of the employees of the Danish reinforced plastics industry between 1964 and 1988 were included in this study, and it can be assumed that virtually all diagnosed cancers were involved. Selection bias was unlikely since the loss to follow-up was negligible (less than 2%) and mainly due to emigration.
Exposure data was available only at a company level, and eve n though most of the companies were small, not all emplo yees were exposed. The dealers estimated that an average of about 40% of the workers classified as expo sed actually had worked producing reinforced plastics, and from this estimate we can expect to detect only about 40% of the "true" excess rate ratios ( 17). Exposure data were collected from several independent sources, and agreement was good. However, a threefold excess risk for non-Hodgkin ' s lymph oma and leuk emia was see n for 1774 employees of companies producing reinforced plastics according to the dealers but not acco rding Scan d J Work Env iron Health 1994, vol 20, no 4 to the employers. This finding may indicate differential reporting of exposure by a few employers. For this reason the analysis was restricted to the exposure data obtained from the dealers.
Few causes of Iymphohematopoietic malignancies are known. Benzene and radiation cause leukemia but are not used in the reinforced plastics industry. Confoundin g by factors related to occupat ional career and the life-styles of the emplo yees is a possibility (18), but perhaps unlikel y, since internal comparisons among worker s within the enrolled companies did not change the results very much.
The finding of an increa sed risk of leukemi a among short-term employees « I year) but not among long-te rm employees (~I year) does not appear to corroborate a causal associ ation , given that duration of employment calculated from the payment records in the Supplementary Pension Fund indicates dose of exposure. The phenomenon of high risk for short-term workers is well known from other occupational cancer studies and is explained by a higher mobility among short-term workers. Thi s higher mobility increases their probability for attracting carcinogenic exposures in other industries or for having more unfavorable life-styles (19). Although this explanation is a possibility, it is perhaps less likely since a compariso n between the exposed and unexposed short-term employees did not yield lower rate ratios for leukemia (observed 27, rate ratio 1.89,95% CIO .78-4.59).
The contributions to the Supplementary Pension Fund were assumed to give the most precise estimates of durati on of emplo yment, as no exact employment dates were known. When we instead estimated duration of emplo yment by the extent of employment (the differen ce between last and first year of employment), the risk difference between the short-term and long-term employees diminished, but the same pattern was found. Howe ver, the est imates of duration of employment from the Supplementary Pension Fund did not correlate well with information on duration of employment obtain ed in a questionnaire from a sample of 671 employees of eight companies. According to the data, as much as 40% of the workers classified as short-term worker s by the Supplementary Pension Fund were classified as long-term workers in the questionnaire, while 13% of the long-te rm workers were classified as shortterm workers.
The risk estimates observed for employees of companies with available measurements of styrene showed inconsiste ncy with the results obtained for the total exposed population, except for Hodgkin ' s disease. The findings could be due to random variation, as the wide confidence inter val indicates, or they could be a result of a high proportion of longterm workers (according to the Supplement ary Pension Fund) in the cross-sections of workers employed durin g years of recorded measurements (lengthbiased sampling). The short-term workers showed an SIR value for leukemia of 2.6 3 (obs e rv ed 3 , 95 % CI 0.54-7.69) when a IO-year latency period was consi de re d .
The mean follow-up of 10.9 years limited the ability to e luc idate fully the ri sks of lymphohematopoietic malignancies and m ay in part ex p lain why increas ed rat e rati os were found m ainly among the expo sed in th e early y e ars. The reinforced pl ast ic s industry is free of most sus pected carcinogens other th an sty re ne , wh ic h was th e re ason fo r selecting th is indus try fo r the study . Durin g th e 1960s the ex p os ure le vels to styre ne were h igh, and the year of firs t exp osure is probably the be st in dicator of e xposu re level to styre ne. A sm all but sig nifi c antly increased ri sk of leukemia was fo und for workers of the 1960s, and th e risk estimate in creased when an appropri at e lat ency period was implemented. On the other h and, th e data did not fit the e xp ected dose-response pattern based on duration of em p loy m ent, perhaps becau se o f an inappropr iat e pro xy measure o f "dose ," In co ncl us io n, the stu dy sho ws an in creased ri sk of leukemia for workers in th e earl y ph a se of th e rein fo rced p last ic s industry in Denmark. If this association is not due to c ha nce or confoundi ng , th e mo st lik el y cause is sty re ne ex posure. The lack o f ind ivi dua l w ork hi stori es weakens, how ever, th e stre ng th o f th e concl us ions to be draw n. E fforts sho uld be mad e to co llect individual ex pos ure histori es in largescale st udies to clari fy th e pre se nt e d fi ndi ngs .