Methods

The UK Office of National Statistics supplied data on deaths by underlying cause among men aged 16–74 years in England and Wales for 1991–2000, including details of age, last occupation held and social class. Occupations were coded according to the Office of Population Censuses and Surveys (OPCS) 1990 Classification of Occupation7 and causes of death according to the International Classification for Diseases, 9th Revision.8

From the supplied data we estimated expected numbers of deaths by applying age-specific proportions of deaths by cause in the general population to the total number of deaths at each age in each occupational group. Expected numbers were adjusted for social class.

We abstracted data on pneumonia (ICD-9 480–483, 485, 486) for occupations with exposure to metal fume defined a priori4 6 (welders, moulders and coremakers, furnace operatives, sheet metal workers and other metal manufacturers) and on asthma (ICD-9 125) for a selection of occupations commonly reported to European surveillance schemes as at special risk of sensitiser-induced occupational asthma (eg, workers in textiles, plastics, health care, agriculture, animal handling, baking, spray painting, woodworking and electrical assembly)9 10 11 12 and/or in which deaths from occupational asthma have been documented (table 1).13 14 15 16 17

In compiling the final choice of occupational codes for asthma, we focused on occupations where we judged exposure to sensitisers to be fairly common and excesses of mortality to be plausibly related to occupation rather than to chance. Two authors (KTP and PC) chose the codes independently, with any differences of opinion being resolved by consensus. Choices were made blind to data on mortality by occupation.

Observed and expected numbers were compared for each cause of death and proportional mortality ratios (PMRs) calculated with 95% confidence intervals (CI) based on the Poisson distribution. Deaths from pneumonia in occupations involving metal fume exposure were analysed separately for men aged 16–64 years and 65–74 years, as well as overall. To estimate attributable deaths from pneumonia in workers exposed to metal fumes, we calculated the difference between observed and expected numbers of deaths at age 16–64 years, since the increase in risk is known to be short-lived and unlikely to extend beyond retirement.4 5 However, given the known persistence of airways irritability in cases of occupational asthma and the potential for attributable deaths to occur after retirement, we counted excess deaths from asthma at ages 16–74 years.

Results

Among men of working age (16–64 years) who had worked in occupations with exposure to metal fumes, there were excesses of mortality from pneumococcal and lobar pneumonia (ICD-9 481: 54 deaths vs 27.3 expected) and from other pneumonias (ICD-9 480, 482–3, 486: 71 deaths vs 52.4 expected), especially viral pneumonia (PMR 299) and other bacterial pneumonia (PMR 237) (table 2). However, no excess was found from these causes at older ages or from bronchopneumonia (ICD-9 486) at any age.

Among men of working age, excesses of pneumococcal and lobar pneumonia were noteworthy in welders (PMR 242), in moulders and coremakers (PMR 300) and in sheet metal workers (PMR 268) (table 3).

Assuming the excess of attributable deaths was confined to men of working age dying of pneumococcal and lobar pneumonia and pneumonias other than bronchopneumonia, table 2 suggests an attributable mortality from occupational metal fume (excess of observed vs expected) of 45.3 deaths. This compared with an estimated 62.6 excess of observed vs expected deaths from asthma over the same time interval among men aged 16–74 years from the occupations listed in table 4 in the online supplement (overall PMR 110, 95% CI 102 to 118), among whom synthetic fibre makers (PMR 740), other textile processing operatives (PMR 254), dental technicians (PMR 230) and other health professionals (PMR 223) showed the most noteworthy relative increases in mortality.

Discussion

In keeping with national analyses of occupational mortality covering earlier periods, we found an excess of deaths from pneumonia among male welders during 1991–2000. Following the model of one previous analysis,4 we also confirmed that the excess risk extended to other workers with occupational exposure to metal fumes, was principally confined to deaths from lobar pneumonia and was limited to men below the normal retirement age of 65 years. As in an earlier report, the absence of an association with bronchopneumonia provided evidence against confounding by smoking. Assuming a short-term reversible effect of work, the estimated excess of deaths among 16–64-year-olds was about 45 over the decade in question, and was material in comparison with (although lower than) the 63 deaths we ascribed to occupational asthma.

This analysis and interpretation has several limitations. Death certification over the study period is likely to have been almost complete, but with the potential for misclassification of cause of death. For example, some deaths from occupational asthma in bakers could have been ascribed to chronic obstructive pulmonary disease, a recognised problem in recording asthma deaths in general.18 This would lead to an underestimate of mortality from asthma in the occupations selected for study, but in both the observed and expected counts, so would not necessarily bias an estimate of risk difference. On the other hand, part of the excess asthma mortality in the occupations in table 4 may have arisen by chance rather than as a consequence of occupational sensitisation (the risk ranking of occupations by mortality did not correspond exactly to what might be expected from statistics on morbidity11); we did not have access to detailed clinical data or post-mortem findings on individuals. Deaths from chemical pneumonitis might have been ascribed to unspecified pneumonia, but large-scale misdiagnosis as pneumococcal or lobar pneumonia (the main outcome) seems unlikely, and there is empirical evidence against it from earlier investigations.4

Misclassification of exposure is another possibility. Death certificates in the UK record the last held occupation which, for some subjects, may have preceded death by several years. Cases of occupational asthma may have arisen during earlier employments that were changed because of chronic symptoms or health advice (unhealthy worker selection bias). Under-attribution of deaths from occupational asthma could thus arise in the occupations chosen for analysis. This is less likely in relation to deaths from pneumonia as fatal cases of the illness will typically have had a rapid onset, with less opportunity to occasion a change of employment.

Other ways in which exposure misclassification can arise include heterogeneity of exposure experience in workers who share a common job title and the selection of occupations for inclusion in our analysis. All welders are likely to have been exposed to metal fumes within their job and all bakers to flour dust, but only a proportion of textile process workers will have been exposed to reactive dyes and, similarly, only a minority of coach painters and vehicle body repairers to isocyanates.

Other factors being equal, the inclusion of an occupation in which a minority of workers have relevant exposure will not bias estimates of risk difference (observed minus expected numbers), but the larger the pool of occupations selected, the greater the potential for plays of chance or unrecognised confounding to conceal genuine differences measured on an absolute scale. The risk of including too many occupations in an analysis, however, has to be weighed against the potential to include too few and thereby omit occupations in which cases may arise. Table 4, although based on an appraisal of current awareness, may not have included every occupation in which an attributable case of fatal asthma arose. In fact, as Leigh et al point out, “no data are available that unambiguously count the number of deaths nationally from occupational asthma”,19 and the methodology for doing so is not yet well established. Thus, for example, a Swedish registry survey which highlighted differences in asthma mortality by occupation could not estimate the occupation attributable mortality in the absence of a rule for counting attributable deaths.20

The estimates for asthma in this paper are therefore tentative and do not include the additional toll arising from work-aggravated asthma. However, they serve to illustrate that the mortality risk of occupational exposure to metal fumes is not inconsiderable relative to another well-recognised respiratory hazard of public health significance and concern.

Occupational and work-aggravated asthma have other well-advertised and important consequences in terms of chronic ill health and hospitalisation episodes. To a degree, infectious pneumonia induced by exposure to metal fumes is also a source of morbidity. Only a minority of cases are fatal, but the increased risk extends to non-fatal cases admitted to hospital with serious illnesses.5

We conclude that the hazard of occupational exposure to metal fumes is an ongoing problem, still detectable after more than five decades, and the hazard deserves far more priority than it presently receives.