Mortality cancer morbidity among workers in a chemical factory.

S. Mor tality and cancer morbidity among workers in a chemical factory. Scand J Work Environ Health 12(1986) 545-551. A retrospective cohort study was performed on a group of 664 male workers employed for at least one month during the period 1942-1979 in a chemical factory. Both established and suspected carcinogens had been handled in the plant, primarilypiperazine, but also urethane, ethylene oxide, formal dehyde, and organic solvents. A significantly increased mortality, compared with the regional death rate, was observed in the cohort. The increase was mainly due to violent deaths and cardiovascular diseases. No rise in death rates was observed for asthma, bronchitis or emphysema, in spite of other evidence of a high risk of occupational asthma, due to exposure to piperazine. A statistically significant increase in cancer morbidity was observed for malignant lymphoma/myelomatosis when an induction latency time of at least 10 years was used. Furthermore, an increase in bronchial cancer was noted , but it was statisti cally significant only when an induction-latency time of at least 15 years was used. A case-referent study within the cohort did not reveal any significant association between any specific chemical exposure and cancer morbidity.

Increased cancer mortalit y has previously been shown among workers from the chemical indu stry (5,25). In a Swedish chemical plant several di fferent carci nogens ha ve been handl ed. Thu s, it was considered essent ial to study the mortalit y and cancer morbidity of a cohort of workers fro m th is factory.

Subjects and methods
The plant Several hund red chemicals, some of which are established or suspected carcinoge ns, had been handl ed in the plant since it was founded in th e I940s.
Pipera zine, a seco ndary heterocyclic a mine, which is an established inducer o f occupat ional asthm a (9,10,21,27), had been one of the ma in products in th e plan t since th e earl y 1950s. As many as a third of the workers heavily exposed to this compound may develop asthma (10). Furthermore piperazi ne nitrosates easily in human s into carcinogenous nitro sopiperazines. This act ion has been observed both afte r ingestion of a normal anthelmintic dose and afte r occupational exposure to piperazine (3,4). Due to techn ical improvements at the plant , the exposure levels decreased considerably during the late 1970s. Breathing zone sampling was per- formed in 1980. The time-weighted average (TW A) estimated for one wor k pro cess on four different occasio ns, a total of 38 h of sampling time , was 1.2 mg/m' (9). Urethane and formaldeh yde, both of which are carcinogenic in rodent s (15,17), were handled in the plant , the form er fro m 1958 to 1976 and the latter fro m the lat e 1950s on.
Benzyl chlo ride , which is mut agenic a nd possibly carcinogenic in rodents (16), was handled from 1970 to 1976. In the synthesis of piperonyl butoxide (1969)(1970)(1971)(1972)(1973)(1974), there may have been for mat io n of bis(chlo rometh yl)ether (BeME), which causes cancer in humans (15). Ho wever, no adequate air analysis of any of these substa nces was ever performed in the plan t.
Ethylene oxide, even in low concentrations, is genoto xic in humans (14), and in previous epid emiologic studies it has been associated with an increased incidence of cancer, especially leukemia (12,13). Thi s chemical had been hand led at th e plant since 1958. Breathing-zone samples (dur ing a tot al time of 32 h) performed since 1976 had a TWA of 1.5 ppm.
Epichlorohydrin , an alk ylating agent, is mutagenic in di fferent test systems and ca rcinogenic in rat s (17). It is also possibly a lung carcinogen in man (7,31). Epichlorohydrin had been used in the plant since 1974. Breath ing-zone sampling performed since 1979 (during a total tim e of 6 h a nd on different occa sions) had a TWA of 0.15 ppm. There is ho wever reason to believe that the exposure levels from 1974 to 1979 were substantially higher.
Different organic solvents, especially toluene, had been used all over the plant in numerous work opera-  Matched case-referent study within the cohort For all 29 subjects in the exposed cohort who obtained a cancer diagnosis between 1958 and 1983. and for cording to the lCD, seventh revision) diagnosed from 1958 to 1982 was obtained from the national Swedish and the southern Swedish regional tumor registers, and for tumors diagnosed in 1983 from the Southern Swedish Regional Tumour Register only. No tumor was identified solely in the national register. Each individual could have more than one tumor registered. Table 1 shows the distribution of person-years by age group and calendar year.
Expected mortality for the period 1961-1984 was calculated with the use of mortality rates (specific for calendar year, cause and five-yearage groups) for males in the county (Malmohus, excluding Malmo town; about 250000 male inhabitants). These rates were calculated from death and population counts obtained from the National Central Bureau of Statistics. Rates for 1983 and 1984 were not available, and so those for 1982 were repeated for these years. Similarily, yearly morbidity rates for cancer in the period 1958 to 1983 for the county were obtained from the Southern Swedish Regional Tumour Register. Date of death or emigration was used as the individual end point. No subjects had emigrated or died before the observation period began .
Cause-specific standardized mortality (or morbidity) ratio (SMR) values were calculated.
Dose-response relationships were analyzed through the linear modeling of trends in SMR with duration of employment (8). The correlation between age and length of employment affects the comparability of the SMR values from each employment stratum and may result in a negative bias for the SMR values, especially for workers employed for a long time. Thus the SMR for each employment stratum was itself standardized, and hence referred to as the SSMR, with respect to a single employment stratum by a system of weighting the age-specific SMR values according to the expected morbidity in the same age group of the standard stratum (28).
The term "significant" refers to a probability of -s 0.05. All tests were two-tailed, except in the analyses of dose-response relationship. Several hypotheses were addressed in this study. Increased mortality from nonmalignant respiratory diseases could be suspected, due to the asthma-inducing properties of piperazine. Furthermore, exposure to piperazine, as well as to urethane, formaldehyde, organic solvents, and ethylene oxide, may have increased the cancer morbidity. The two latter exposures were regarded as possible inducers of hematologic malignancies. With a reasonable induction-latency time for cancer taken into consideration, the exposures to epichlorohydrin, benzyl chloride, and BCME were too recent to be of any concern.

Cohort study
The cohort included all 664 male workers born in 1895 or later and employed at least one month during the period from 1942 to 30 April 1979 (figure 1). As the company records, containing name, date of birth, native country, address, and start and end of employment, were incomplete for the period before 1965, information on 120of the subjects was obtained through interviews with a previous staff manager, and with foremen or workers employed for a long time. Verification of these data was obtained by direct contact with the 98 men who were still alive and with the nearest living relative of the 22 deceased subjects. There were only 60 immigrant workers, of which 32 were of Danish origin, in the cohort.
Vital status of the whole cohort was determined up to 31 December 1984.
All death certificates were obtained. They had been coded according to the International Classification of Diseases (ICD) by the National Swedish Central Bureau of Statistics, which is responsible for the coding of all Swedish death certificates. All codes were transformed to the eighth revision of the ICD. A total of 60 010 of the death certificates was based on forensic or clinical autopsy. Information on tumors (coded ac-three additional subjects with cancer diagnosed in 1984, two referents without cancer and matched for age (mean dif ference 4.2 years, referents being younger), first time of emplo yment (mean difference 2.2 years, referents being employed later), and length of employment (mean difference 1.8 years, re ferents being employed for a short er time) were selected from the cohort.
In a second analysis, for each case of malignant lymphoma /myelomatosis, bladder cancer, or bronchial cancer , two additional referents were selected according to the same matching criteria from the pool of referents primarily mat ched with the other cancer cases.
Information on lifetime smoking habits and previous employment was collected by interviews with the subjects, or, if they were deceased, with their nearest living relative. Their employment was assessed regarding possible chemical exposure. Furthermore previous employment was also checked throu gh employee records from a rubber factory, two paint-manufactur ing plant s, a roto gravure printing office , and another chemical factor y (producing inor ganic acids) in the same city.
For all but thr ee o f the cases and two of the referents, one of the foremen, employed since the 1950s at the chemical plant under study, performed a " blinded" estimation of worktime at various work processes involving possible exposure to the established or suspected carcinogens. As both organic solvents and formaldehyde had been used in numerous different work operations at the plant, it was not feasible to evaluat e each worker's individual exposure to these agents.
With the conditional likelihood procedure, the matched sets were anal yzed with a score test (6). All sta ted p-values are two-tailed.

Mortality
The overall mort ality ratio for the period 1961 to 1984 was statistically significantly increa sed (88 observed deaths versus 68.7 expected, SMR 128, P = 0.02) (table 2). As to cause-specific mortality, ther e was a significantly increased risk of malignant lymphoma and myelomatosis (SMR 327, p = 0.04). No statisticalIy significantly increased risk for death from asthma, bronchitis, or emph ysema was ob served (SMR 204, p = 0.18). On the other hand, there was an increased risk of violent death (SMR 190, p = 0.01).
When at least six months of employment and at least 10 years of induction-latency time was used, the overall mortality rat io was still slightly elevated (SMR 130), but , probabl y due to the diminished number of personyears under observation, it was not statistically significant. The risk for death due to malignant lymphoma and myelomatosi s did, however, increase (SMR 479, p = 0.03). Furthermore the risk for death fro m bronchial cancer was increased , but was still not statisti cally significant (SMR 228, P = 0.07).

Cancer morbidity
During the period 1958 to 1983,32 cancers, in 29 subjects, were registered in the cohort compar ed with 30.8 expected (SMR 104, p > 0.5) (table 3). The morbid ity rat io for malignant lymphoma and myelomato sis was significantly increased (SMR 270, p = 0.04). An increase in bladder cancer was also observed (SMR 200), but it did not reach statis tical significance.
When at least six months of employment and at least 10 years of induction-latency time was used, the overall Table 2. Obs erved (0 ) and ex pected (E) mortality and causes of death in the cohort in 1961-1984. (Rest ricted cohort = t ime of em ploymen t of '" 6 mon ths and an ind uc tio n-late ncy time of '" 10 years, SMR = stand ard ized mortal ity ratio)   There was a statistically significant linear doseresponse relationship between duration of employment and morbidity for all cancer (p < 0.0001) (table 4). Table 5 summarizes the characteristics of the cases and referents. There was no overrepresentation of smokers among the cases or of previous possible chemical exposure. Thirty-eight percent of the cases had been exposed to piperazine on more than singular occasions, compared with 31 070 of the referents (p = 0.38). Thirty-one percent of the cases had been exposed to urethane in the same manner, compared with 16 % of the referents (p = 0.14) . Only 3 070 of the cases had been exposed to ethylene oxide and none to benzyl chloride, BCME, or epichlorohydrin.

Matched case-referent study in the cohort
The cases of malignant lymphoma/myelomatosis, bronchial cancer, and bladder cancer did not differ from their referents with respect to smoking habits and previous possible chemical exposure (table 5). It is, however, noteworthy that one of the subjects with bladder cancer had, during the period 1943 to 1965, been mixing rubber in a rubber factory , and another 548 had worked for five months in 1957 as a storeman in the same plant.

Validity
No preemployment health examination was performed before 1972, when the company founded its medical health service . In addition the occurrence of occupational asthma at the plant had not been given any particular attention by the medi cal health service until 1979. It is therefore unlikely that an y bias in the selection of employees into the cohort affected mortalit y from nonmalignant respiratory disea ses.
A more vita l problem is that information on 18 % of the subjects in the cohort was not obtained from the company record, but through other means. Ten of these 120 subjects later developed cancer. However, none of these cases was diagnosed during employment. Thus there is no reason to believe that subjects with cancer were more likely to be recalled by the persons interviewed.
An underascertainment of short-term employees may hav e affected the mortality patterns, as such a group might have a higher than average mortality. However, Table 5. Exposure of the 32 subjects with cancer ("cases ") and their 64 referents to possible risk factors for cancer. (N number of subjects for which informat ion was available)

All cancers Lymph omas ' Bronchial cancers
Bladder cancers 32  77  62  88  5  80  20  90  5  100  20  90  5  100  20  80   Previous chemical  exposure   History  32  41  64  38  5  40  20  45  5  40  20   30   5  20  20  35  Records  32  34  64  45  5  60  20  75  5  20  20  35  5  40  20  35  Exposure in  pres ent plant  Piperaz ine  29  38  62  31  4  50  19  26  5  40  20  30  4  25  19  37  Urethan e  29  31  62  16  4  50  20  20  5  40  20  10  4  25  19 16 Benzyl chl oride cros s-sectional estimates of the number of employees in the plant in different year s, as derived from the cohort (figure 1), fits very well with independent data from the company. Thus such an underascertainment, which would bias our results in a " conservative" way, mu st be very limited. Diagno stic accuracy for tumors, as well as for causes of death, is important for the validity of the study. Information on cau se of death was obtained for all the deceased wo rkers in the cohort, and 60 % of the death certificates were ba sed on clinical or forensic autopsy, which is in accordance with the frequency of autopsies of deceased males in the county during the per iod 1969 to 1978 (30).

Smok ing
Errors in both the Register of Causes of Death and the tumor registers have been detected through quality studies (19,30). A few such errors were present in thi s cohort. One case of colon cancer, according to the death certificate, was missing in the tumor registers, and one case of stomach cancer was falsely registered as bronchial cancer in the Register of Cau ses of Death.
Only three of the 29 subjects with can cer diagnosed during the period 1958 to 1983 were submitted to the hosp ital by the company's medical health service, and none of the cases were detected through an y health screening pro gram . Thus there is no ba sis for suspecting that the increa sed incidences of tumors were due to a mo re int ense med ical surveillance of the emplo yees in the cohort than of the regional population.
Th e cohort ha s been geographically stable, as until Jul y 1986 only 86 subjects had moved from Malm ohus County to other parts of Sweden. Thus, as there are substa nt ia l regional difference s in incidence rate s for causes of death (30), as well as for cancer morbidit y (23), it was rea sonable to employ county rates instead of national ones.
Smoking is a po ssible confounding factor for bronchial cancer. Smoking hab its were, through a questionnaire on occupational asthma (10), known for 507 of the 644 subjects in the cohort. Fifteen percent were lifelon g nonsmokers, 55 % had a life consumption of 1-20 pack-years (pack -year = cigarettes per day multiplied by years of smoking, divided by 20), and the remaining 30 % had a life consumption of more than 20 pack-years. This proportion of smokers is simila r to that of Swedish ind ust rial wor kers in general (24), but higher than among the adult population in Malrnohus Co unty (29). This finding may explain so me of the increased morbidity from bronchial cancer.
On the assumption of a mean smoking period of 30 years, the risk of bronchial cancer, compared with that of non smokers, is about five times higher for "moderate" smokers and about 15 times higher for "heavy" smokers (20). On th e basis of these figures, the abundant smoking in the cohort, compared with that of the general population , could explain an SMR of about 185 (2).

Mortality
An interesting observation was that , in spite of the increased overall mortality ratio of the cohort, no rise in deaths from asthma , bronchitis, or emph ysema was fou nd . Thus , even if as many as 30 % of the workers most hea vily exposed to piperazine in the plant experien ced bronchial asthma (10), it did not affect the mortality ratio . However, it is of course po ssible that the occurrence of occupational asthma in some subjects may ha ve contributed to death from cardiovascular diseases. The follow-up period is also still sho rt.
In most cohort studies of indu str ial workers no increased overall mortality ratio is found , and this state has usually been ascribed to the "healthy worker effect" (22). Thus the increased overall mortality in this cohort is somewhat remarkable. The increase was not mainly due to cancer, but rather to violent deaths (no occupational accident s) and cardiovascular diseases. Thus the observed overris k was probably due to the polluted and worn -out plant, which may have cau sed a socioeconomic selection impact on the workforce .

Cancer morbidity
The most essential result of this study was the significant increase of malignant lymphoma/myelomatosis in the cohort. It is important that all three cases of malignant lymphoma were of non-Hodgkin's type, ie, B-Iymphocytetumors, and thus related to the two cases of myelomatosis. It is thus justified to treat them as one entity. The finding is strengthened by the fact that chemists (18) and other subjects exposed to chemicals, including organic solvents (11), have earlier been claimed to have an overrisk from this kind of tumor. Moreover, during the preparation of this manuscript, a fourth case of non-Hodgkin's lymphoma was diagnosed in the cohort. This subject had been employed in the present plant during the preceding 15 years, and had been substantially exposed to piperazine but not to urethane.
The present case-referent study did not give any evidence as to etiologic agents of the malignant lymphoma and myelomatosis. However, partly due to the few cases of cancer and partly due to the complexity of the exposure, this method is insensitive. Some of the workers had solely been exposed to one chemical during their employment. However, others had been exposed to several, eg, 29 0J0 of tho se exposed to piperazine had also been exposed to urethane. Furthermore the influence of the exposure to organic solvents, previously associated with lymphoma (II), could not be evaluated.
Part of the increased risk for bronchial cancer may be explained by the abundant smoking habits in the cohort. On the other hand, the positive relationship between length of induction-latency time and morbidity ratio may favor an association with the work environment. In any case, an increased incidence of bronchial cancer was not due to BCME or epichlorohydrin, as none of the cases had ever been exposed. There was a prevalent exposure to formaldehyde in the plant. A possible risk of bronchial cancer has earlier been noted for workers exposed to that agent (1). The same is true as regards the present study.
The plant involves a complicated pattern of varying exposure to many chemicals over time. It is thus not possible to know whether the observed tumor clusters are solely due to historical exposure or whether one or several carcinogens are still operating. In order to determine if carcinogenic exposure is still present, a study on cytogenetic effects and the excretion of mutagens in the urine has been initiated in the plant. The results of that study may enlighten us with regard to this question.