METHODS

Subjects

The cohort, the procedures for data collection and the use of isocyanates in the UK flexible polyurethane foam industry have been described previously¹; the cohort comprises 8288 male and female production employees from 11 participating factories in England and Wales. The definition of the cohort is complex although all study subjects had some employment in the period 1958–79 and were employed for a minimum period of six months. The cohort is a mixture of entry cohorts (n = 5321) and census cohorts (survivor populations) (n = 2967); the latter comprising those workers in employment on the earliest dates for which complete personnel records were still available. There were different dates for different factories; table 1 summarises the factory subcohorts.

The Office for National Statistics (ONS) has continued to supply the principal investigator with copies of death certificates (underlying cause coded to the 9th revision of the International Classification of Diseases (ICD-9)), cancer registration particulars (for the more limited period 1971–94), and emigration details relevant to study subjects. On the closing date of the survey (31 December 1998), there were 1704 (20.6%) decedents, 163 (2.0%) workers had emigrated, and 146 (1.8%) workers were untraced. The mean period of follow up (date of hire to the end of the study period) was 26.3 years.

A total of 1436 study subjects were still employed at the end of 1986, of which 241 employees had been employed at two factories now closed. Work histories were left unchanged for these 241 employees. Company personnel departments provided updated work histories for the period 1987–97 for the remaining 1195 employees. The additional work histories were all coded in terms of the job dictionary containing 4612 unique factory/department/job entries used in the earlier analysis.¹ Guided by the occupational hygiene data (1978–86) collected for the Medical Research Council isocyanate study⁶ each of these entries had been classified under one of four headings. A total of 189 entries had been classified by an occupational hygienist as “higher” exposure jobs (eight hour time weighted average exposure (TWA) for TDI was greater than 4 ppb or excursions above 10 ppb occurred on most days), 701 entries had been classified as “lower” exposure jobs (eight hour TWA for TDI was 1.5–4 ppb or occasional excursions above 5 ppb occurred), 94 entries were unclassifiable, and the remaining 3628 entries were classified as jobs involving minimal/zero exposure. “Higher” exposure jobs included block plant operative, polyester moulding process worker, and curing chargehand; “lower” exposure jobs included block stores controller and maintenance fitter; minimal/zero exposure jobs included sewing machinists, pillow fillers, stock clerks, and latex workers. The current ACGIH eight hour TLV for TDI exposure is 5 ppb; the short term exposure level (STEL) is 20 ppb.

External standard (SMR)

Expected numbers of deaths were calculated by applying sex, age, and period specific mortality rates for England and Wales to corresponding person-years-at-risk (pyr). Each study subject contributed pyr from the end of the six months minimum period of employment or the factory census date (see table 1), whichever was later, to the closing date of the study, death, emigration, or date last known alive, whichever was the earliest. SMRs were calculated as the ratio of observed deaths to expected deaths, expressed as a percentage. These procedures were carried out using the PERSONYEARS programme.⁷ Data were censored at age 85 years for three reasons. Firstly, published mortality rates are only available for the “open ended” age group ≥85 years and the distribution of the cohort pyr by single years of age might be very different from that of the general population; secondly, the reliability of cause of death particulars is probably poorer at later ages; and thirdly, any study subjects incorrectly classified as traced alive at the end of the study would have a disproportionate effect on the expected numbers for the open ended age group. There were a total of 200 262 pyr available for the updated analysis (1951–88: males 93 983 pyr, females 41 679 pyr; 1989–98: males 44 578 pyr, females 20 022 pyr). All tests of statistical significance were two tailed.

Internal standard

A number of variables were considered to have the potential for influencing mortality within the cohort: sex, attained age, calendar period, type of cohort (entry cohort/census cohort), period from commencing employment, cumulative duration of “higher” exposure to diisocyanates and cumulative duration of “lower” exposure to diisocyanates. Each variable was categorised into a number of levels (see table 2). The final models were constructed such that there was at least one death observed at each level of each variable. Any adjustments were made before statistical modelling was carried out.

The EPICURE computer program was used to provide both pyr and numbers of deaths for all combinations of all levels of the selected variables,⁸ and to carry out statistical modelling by means of Poisson regression,⁹ calculating point estimates of rate ratios (relative risks) for categories of work history variables with and without adjustment for other variables. More importantly, the statistical significance of any tendency for risks to increase or decrease progressively across levels of cumulative duration of exposed employment was assessed by means of linearly weighted p values for trend; the three levels for each of the two work history variables (see table 2) were scored 1–3.

RESULTS

External standard

Table 3 shows observed and expected numbers of deaths both for individual sites of cancer and for broad categories of non-cancer deaths for male and female employees. In males, there were significantly increased SMRs for all causes (Obs 1298, SMR 107, p < 0.05), other and unspecified leukaemias (Obs 4, SMR 507, p < 0.05), and non-malignant diseases of the respiratory system (Obs 140, SMR 120, p < 0.05). Findings for all leukaemias were unexceptional (Obs 8, SMR 87) as were those for lung cancer (Obs 134, SMR 107). In females, a significantly increased SMR was shown only for lung cancer (Obs 35, SMR 181, p < 0.01). Findings for female cancer of the pancreas in the new period of follow up (1989–98) were unexceptional (Obs 3, SMR 125) whereas a further excess was shown for female lung cancer (Obs 19, SMR 184, p < 0.05).

Table 4 shows, for male and female employees, observed and expected numbers of deaths for all causes, lung cancer, and non-malignant diseases of the respiratory system by successive 10 year periods from commencing employment. There was no indication that SMRs for lung cancer were increased more in later periods of follow up compared with earlier periods, in either group of employees (p value for trend: men, p = 0.60; women, p = 0.76). In females, however, there was a significant trend of SMRs for non-malignant diseases of the respiratory system increasing with period from hire (p = 0.02). There was no corresponding trend for males (p = 0.79).

Table 5 shows observed and expected numbers of deaths both for individual sites of cancer and for broad categories of non-cancer deaths for those 1782 male employees with some period of “higher” or “lower” diisocyanate exposed employment. There were no significantly increased SMRs for any cause of death. An SMR of 117 was found both for lung cancer (Obs 36) and for all neoplasms (Obs 107). A similar analysis was carried out for the small exposed group (n = 90) of female workers (not shown in table). There were no deaths from cancers of the pancreas (Exp 0.11), lung (Exp 0.49), or kidney (Exp 0.04), or from non-malignant diseases of the respiratory system (Exp 0.60).

Table 6 shows observed and expected numbers of cancer registrations (1971–94) for male and female employees. Study subjects dying before 1971 or untraced at the ONS do not feature in these analyses. In males, there was no site of cancer for which registrations were significantly increased. In females, there were significantly increased standardised registration ratios (SRRs) for cancer of the larynx (Obs 3, SRR 500, p < 0.05) and for lung cancer (Obs 31, SRR 199, p < 0.001). In the new period of follow up (1987–94) there were no further cases in females of cancer of the larynx (Exp 0.30) or cancer of the kidney (Exp 1.14). An excess SRR persisted for female lung cancer (Obs 14, SRR 170).

Internal standard

Table 7 shows risks of mortality for lung cancer, non-malignant diseases of the respiratory system, and all causes for two durations of “higher” exposure employment relative to risks in the never highly exposed category and corresponding analyses for duration of “lower” exposure employment. Durations of lower and higher exposure are separate time dependent variables and have been analysed simultaneously. The first column of relative risks were adjusted for sex, age, and calendar period. The second column of relative risks includes additional simultaneous adjustment for type of cohort (entry cohort/survivor population) and period from hire. None of the relative risks was significantly different from unity, and there were no significantly positive or negative trends of risks either with duration of lower exposure employment (all causes, p > 0.50; lung cancer, p = 0.47; non-malignant diseases of the respiratory system, p = 0.32) or with duration of higher exposure employment (all causes, p = 0.20; lung cancer, p > 0.50; non-malignant diseases of the respiratory system, p > 0.50). None of the lung cancer cases had more than five years of higher exposure employment. The type of cohort had almost no influence on overall mortality; the rate ratio for all causes mortality in the census cohorts compared with the entry cohorts was 1.04 (95% CI 0.93 to 1.17).

Table 6 also shows the same analyses with employment histories “lagged” by 10 years—that is, periods of lower or higher exposure employment are not considered until a further 10 years have elapsed. None of the relative risks was significantly different from unity, and there were no significantly positive or negative trends of risks either with duration of lower exposure employment (all causes, p > 0.50; lung cancer, p > 0.50; non-malignant diseases of the respiratory system, p = 0.37) or with duration of higher exposure employment (all causes, p > 0.50; lung cancer, p = 0.46; non-malignant diseases of the respiratory system, p > 0.50).

DISCUSSION

This study has found overall increased rates of lung cancer (for both mortality and morbidity) in female workers from the flexible polyurethane foam industry but unexceptional rates of lung cancer in male workers. Similar findings were shown in the first report,¹ and it follows that the increased lung cancer rates in females are unlikely to be chance findings. There was no indication, however, that these increased rates were caused by exposure to isocyanates. All female lung cancer cases occurred in women without any period of isocyanate exposed employment, and there was no indication in the total study population (males and females) that lung cancer risks were associated with duration of isocyanate exposed employment. The possible role of nine other occupational exposures (amine catalysts, non-flammable solvents, flammable solvents, polyurethane dust, latex, rubber, curled hair/coir fibre, feathers, foam handling) in the pattern of lung cancer risks experienced by females in this cohort was investigated in the first report¹; none of these exposures was implicated. Such other occupational exposures in the industry under study offer an unlikely explanation for the increased rates, given that female lung cancer SMRs were marginally higher in the first two 10 year periods following hire than they were in later periods of follow up. It might be expected that the opposite pattern would be shown if an occupational lung cancer hazard had been present. Unfortunately, it has not been possible to evaluate the role of smoking habits or dietary factors¹⁰ in these findings.

Some attention was given in the original report¹ to findings in female workers for cancers of the pancreas, larynx, and kidney. There have been no further cases of cancers of the larynx or kidney in the new period of follow up for cancer incidence data (1989–94), and the overall SRRs for both cancer sites in female workers are no longer significantly increased. In a similar vein, the original increased rates for cancer of the pancreas (mortality and incidence) have not been maintained in the new period of follow up and the overall SMR and SRR for this site of cancer in female workers are no longer significantly increased.

In female workers there was a significant trend of SMRs for non-malignant diseases of the respiratory system increasing with decade from hire. Similar findings were not found for male workers and in the more sophisticated analyses, risks for this disease grouping were not found to be associated with duration of isocyanate exposed employment. Given the large number of statistical comparisons that have been carried out, some chance findings are to be expected, and the positive trend for female workers may well be such a finding. Only three deaths were recorded as being caused by asthma and none of these individuals had had any period of higher exposure employment.

The cohort definition is complex which is unattractive. Nevertheless, there was no evidence of the presence of census cohorts introducing major selection effects into the analysis. There are limitations to be attached to the use of the exposure surrogates used in these analyses. These surrogates take no account of any changes in intensity of job specific exposures over time. While exposure conditions in the 1980s and 1990s were similar, it may be the case that concentrations in earlier periods were higher. In addition, it needs to be recognised that this cohort study, along with the Swedish and US studies, relate to an industry that operates with well controlled exposure conditions, and that all the available human studies are “low” exposure studies. Nevertheless this study has been unable to link cancer risks or risks from non-malignant diseases of the respiratory system to occupational isocyanate exposure.