Objective To examine the risk of lung cancer among men associated with exposure to diesel engine emissions incurred in a wide range of occupations and industries.
Methodology 2 population-based lung cancer case–control studies were conducted in Montreal. Study I (1979–1986) comprised 857 cases and 533 population controls; study II (1996–2001) comprised 736 cases and 894 population controls. A detailed job history was obtained, from which we inferred lifetime occupational exposure to 294 agents, including diesel engine emissions. ORs were estimated for each study and in the pooled data set, adjusting for socio-demographic factors, smoking history and selected occupational carcinogens. While it proved impossible to retrospectively estimate absolute exposure concentrations, there were estimates and analyses by relative measures of cumulative exposure.
Results Increased risks of lung cancer were found in both studies. The pooled analysis showed an OR of lung cancer associated with substantial exposure to diesel exhaust of 1.80 (95% CI 1.3 to 2.6). The risk associated with substantial exposure was higher for squamous cell carcinomas (OR 2.09; 95% CI 1.3 to 3.2) than other histological types. Joint effects between diesel exhaust exposure and tobacco smoking are compatible with a multiplicative synergistic effect.
Discussion Our findings provide further evidence supporting a causal link between diesel engine emissions and risk of lung cancer. The risk is stronger for the development of squamous cell carcinomas than for small cell tumours or adenocarcinomas.
- Lung cancer
- occupational exposure
- diesel exhaust
- general expertise
- organ system
- disease type
- public health
- risk assessment
Statistics from Altmetric.com
- Lung cancer
- occupational exposure
- diesel exhaust
- general expertise
- organ system
- disease type
- public health
- risk assessment
What this paper adds
The International Agency for Research on Cancer classified diesel engine emissions as probably carcinogenic to humans (group 2A) in 1989 based on limited evidence in humans and sufficient evidence in experimental animals.
Our findings provide further evidence supporting a causal link between diesel engine emissions and risk of lung cancer.
The risk is stronger for the development of squamous cell carcinomas than for small cell tumours or adenocarcinomas.
Lung cancer is the most common malignancy worldwide after skin cancer.1 Tobacco smoking is by far the main determinant of lung cancer, accounting for approximately 90% of the cases among men. Estimates of the fraction of lung cancers attributable to occupational exposures have varied widely, from 5% to 15%.2 ,3
Diesel engine emissions (also referred to as diesel exhaust) are highly complex mixtures that vary widely depending on engine type, fuel type and operating conditions. The components of exhaust most often quantified in occupational setting are particles, carbon monoxide and nitrogen oxides, but polycyclic aromatic compounds and aldehydes have also been measured in work environments.4 The International Agency for Research on Cancer classified diesel engine emissions as probably carcinogenic to humans (group 2A) in 1989, based on limited evidence in humans and sufficient evidence in experimental animals.4
Since then, several epidemiological studies have found an increased risk of lung cancer among exposed workers, either in specific occupations and industries, such as truck and bus drivers, railroad workers, maintenance workers,5 ,6 and miners7 or in a wide range of occupations.8–10 However, some authors have hypothesised that these positive associations could reflect biases, including inadequate control of confounding, and that a causal link between diesel engine emissions and lung cancer cannot be confirmed.11–14
In the early 1980s, we carried out a population-based case–control study in Montreal, Canada, to explore the possible associations between hundreds of occupational substances and multiple cancer sites, including lung cancer (labelled here as study I). In the late 1990s, we carried out a similar study in the same region, this time focusing on respiratory cancers (labelled here as study II). These two investigations offered the possibility of examining the effect of occupational exposures at different levels and in a wide range of occupations. The purpose of the present study was to examine the risk of developing lung cancer associated with occupational exposure to diesel engine emissions under conditions of exposure experienced in diverse occupational settings, while properly controlling for major confounders, such as tobacco smoking and other occupational exposures. As secondary aims, we examined effect–measure modification by smoking and whether the risk differs by major histological type. Results concerning the association between diesel exhaust and lung cancer from study I have been previously reported8; the present analysis includes both studies and thereby approximately doubles the number of study subjects, and it entails some modifications in the analytic strategy that will be mentioned below.
Both studies followed a case–control design. In brief, study I was conducted from 1979 to 1986 and included men aged 35–70 years diagnosed with cancer at any of the 19 sites.15 ,16 Study II was conducted from 1996 to 2001 and included men and women aged 35 –75 years diagnosed with a lung malignancy. Both studies included patients with incident histologically confirmed cancers identified across all major Montreal area hospitals, living in the Montreal area and restricted to Canadian citizens. Both studies also included a series of population controls randomly selected from electoral lists. In both studies, controls were frequency matched by age, sex (only applicable for study II) and area of residence (electoral district of about 40 000 individuals) to all cancer cases for study I and to lung cancer cases for study II. Results are presented here for men only since the prevalence of occupational exposure to diesel engine emissions was very low among women in our study population.
In study I, 1082 lung cancer cases and 740 eligible population controls were identified and attempts were made to interview them. Of these, 857 (79%) cases and 533 (72%) population controls completed the interview. In study II, 858 eligible male cases and 1024 eligible male controls were identified, and 86% and 70% of these, respectively, agreed to participate and completed the interview. Seven subjects were subsequently excluded due to incomplete information on smoking history, leaving a total of 739 cases and 894 controls for study II. Ethical approval was obtained for both studies from each participating hospital and university. All participating subjects provided informed consent.
In study I and study II, over 82% and 76% of individuals, respectively, responded for themselves, whereas surrogate respondents (proxies) provided information for the other participants. Interviews included a structured section that requested information on socio-demographic and lifestyle characteristics, including ethnicity, family income and smoking history, and a semistructured section that elicited a detailed description of each job held by the subject in his working lifetime. For each job held, a trained interviewer asked the subject about the company, its products, the nature of the work site, the subject's main and subsidiary tasks, and any additional information (eg, equipment maintenance, use of protective equipment, activities of co-workers) that could provide clues about work exposures and their intensity. Supplementary questionnaires were used to assist interviewers with detailed technical probing for some occupations, including among others: bus, truck and car drivers; industrial and auto mechanics; machinists; miners; firefighters and concrete-construction workers.17 A team of chemists and industrial hygienists examined each completed questionnaire and translated each job into a list of potential exposures using a checklist of 294 agents, including diesel engine emissions.
Combining the two studies, more than 28 000 jobs were evaluated. The final exposure codes attributed to a participant were based on consensus among the coders. Chemical coders were blind with regard to the subject's disease status. For each substance considered present in each job, the coders noted three dimensions of information, each on a three-point scale: their degree of confidence that the exposure had actually occurred (possible, probable, definite), the frequency of exposure in a normal work week (<5%, 5%–30%, >30% of the time) and the relative level of concentration of the agent (low, medium, high). Unfortunately, it proved impossible to reliably estimate absolute concentration values corresponding to the relative levels coded.
As diesel engine exhausts are ubiquitous in the urban environment, non-exposure to diesel exhaust was interpreted as exposure up to the level that can be found in the general environment, including people living in urban areas and including people working outdoors but not in close proximity to diesel engines. Among those considered exposed above background level, benchmark occupational circumstances were established to correspond to low, medium and high concentrations, and each job was coded with respect to these benchmarks. The ‘low’ concentration benchmark comprised truck, taxi and bus drivers in urban areas; ‘medium’ concentration benchmark comprised train crews and roundhouse workers in railway transport, workers in open-door garages with diesel engine maintenance, underground parking garage attendant and toll-booth workers; ‘high’ concentration benchmark comprised diesel engines maintenance mechanics working with doors closed and to underground mine workers who worked near diesel engines. These benchmarks were indicative and the experts were free to score a given job in a way that took account of the specifics of the job, the industry, the era and the particular characteristics of the workplace and work habits. Although a subject's job title was certainly a factor in attributing exposure, the details of the subject's activities were taken into account in assessing the exposure. This has been elaborated in Parent et al.8
Unconditional logistic regression18 was used to estimate ORs of disease and 95% CIs for each occupational factor, while adjusting for the following potential confounders: age, socioeconomic status measured by family income, schooling level, ethnic–cultural background (French, Anglo, other), respondent status (self, proxy), ever occupational exposure to asbestos, crystalline silica, cadmium, chromium VI compounds or nickel compounds and tobacco smoking. After comparison of several parameterisations of the smoking variables in our data sets, we selected the comprehensive smoking index,19 which integrates duration, intensity and time since quitting smoking, and best captured the confounding nature of smoking history.
An exposure lag-period of 5 years was used to construct exposure indices. As an a priori approach, subjects were classified into one of three possible exposure categories based on the probability, timing, duration and degree of exposure. The first category consisted of ‘never exposed’ subjects. The second consisted of subjects for whom the degree of confidence that the exposure actually occurred was coded as ‘possible’ by the hygienists; these subjects were classified into the ‘uncertain exposure’ category. The remaining subjects, whose exposure to diesel exhaust was rated as probable or definite, were considered as ‘exposed’ for these analyses. We further subdivided those ‘exposed’ into two exposure groups: ‘substantial exposure’ was assigned to subjects who had been exposed to medium or high concentrations, during more than 5% of their work week, and for 5 years or more, whereas ‘non-substantial exposure’ was assigned to the remaining exposed subjects.
Other cumulative exposure indices were calculated using different combinations of weights to the exposure dimensions frequency, concentration, duration and latency. None of these indices showed better goodness-of-fit (data not shown). In addition to analysing study I and study II separately, we carried out analyses on the pooled data sets; for this, we added to the model a covariate for study (I or II), and we tested for effect modification of the diesel effect by study.
Table 1 shows the distribution of subjects according to selected socio-demographic characteristics. In both studies, cases were more likely than controls to have French ancestry, fewer years of education, lower mean family income and were more likely to have had a proxy responding for them. As expected, the proportion of current smokers and the intensity of smoking were higher among cases than among controls. It is noteworthy that the prevalence of smoking decreased from study I to II, especially among controls, and the proportion of former smokers increased, reflecting the smoking habits trend of the last decades in North America.
Exposure to diesel exhaust occurs in many different jobs and occupations. In both studies combined, out of a total of 18 304 jobs classified according to the Canadian Classification and Dictionary of Occupations20 and evaluated for possible exposure to diesel exhaust, 1988 (10.8%) were considered as exposed jobs. In our study population, the main occupations that accounted for diesel-exposed workers were: truck drivers (15.0% of all subjects exposed to diesel exhaust in both studies combined), mechanics (5.5%), bus drivers (4.9%), excavators (4.7%), firefighters (2.7%), industrial and construction machinery mechanics (2.5%) and rail transport mechanics (2.0%). This ranking of prevalence of exposure does not necessarily reflect the ranking in terms of concentration of exposure. Indeed, of the occupations listed above, the last four tended to have higher concentrations attributed than the preceding ones.
Table 2 shows the lifetime prevalence of occupational exposure to diesel engine emissions. Lifetime prevalence of exposure was higher in study II than in study I. For all dimensions, exposure was higher among cases than controls.
Table 3 presents the adjusted ORs for the relation between lung cancer and occupational exposure to diesel exhaust in each study. Point estimates were very similar between studies, and there was no interaction in a model in which we added a term for interaction between study and diesel exhaust (p=0.413). Thus, we also present the pooled results. An increased risk for lung cancer was found among those ever exposed to diesel exhaust (for pooled estimates, OR 1.34, 95% CI 1.1 to 1.7). We did not find a clear trend for duration or average time per week of exposure, but there was increasing OR with increasing concentration level. For those subjects exposed at the substantial level, the OR was 1.80 (95% CI 1.3 to 2.6).
Sensitivity analyses were conducted separately among self-respondents and proxies. The analysis among self-respondents showed even higher ORs than those among all subjects. Indeed, among self-respondents, the OR for ever exposed was 1.52 (95% CI 1.2 to 1.9) and for substantial exposure, it was 2.08 (95% CI 1.4 to 3.0). Among proxy respondents, the OR for ever exposed was 0.79 (95% CI 0.5 to 1.3) and for substantial exposure, it was 0.78 (95% CI 0.3 to 1.9). The interaction term between self/proxy respondent and diesel exhaust was statistically significant (p=0.024). These findings are compatible with the notion that there was a great deal of measurement error in the assessment of exposure among proxy respondents, presumably related to the lower validity of the reported job history descriptions. Because subjects of French ancestry comprised about two-thirds of the study base and they have distinctive genetic and lifestyle/cultural characteristics, we also conducted a sensitivity analysis restricted to this ethnic group. The results did not differ materially from those shown in table 3 (data not shown).
Table 4 presents the results for the major histological types, both studies combined. For adenocarcinoma, we did not find an increased risk, not even among subjects exposed at the substantial level (OR 1.17, 95% CI 0.7 to 1.9). The highest risk among histological types was found for squamous cell carcinoma: OR 1.67 (95% CI 1.3 to 2.2) for ever exposure and OR 2.09 (95% CI 1.3 to 3.2) for those exposed at the substantial level.
In table 5 we present OR estimates for the joint effects of diesel emissions and tobacco smoking. In the top part of the table, we present results for never- and ever smokers. Since the number of cases who never smoked is very low and the risk estimates imprecise, we also conducted analyses combining non-smokers with low smokers (up to 15 pack-years of lifetime cumulative exposure). The bottom part of table 5 shows the diesel-related ORs within these new smoking strata. Within each smoking stratum portrayed in this table, there is a tendency for the OR to increase with increasing exposure to diesel exhaust. These findings are more compatible with a multiplicative rather than an additive joint effect of the two exposures (OR for the interaction term =1.15, 95% CI 0.5 to 2.7), but the estimates are not precise enough to rule out an additive effect.
Although results from study I were previously published,8 we include them in this paper to improve statistical power. However, for the present paper, we implemented some changes in analytic strategy. First, in study I, we had access to both population controls and cancer controls, and in the previous paper, we reported results with each set of controls. We found an excess risk of lung cancer when using population controls, but no excess when using cancer controls. In the present paper, we have reported only the results using population controls because they are conceptually identical with the controls used in study II and because the results using the population controls were homogeneous between the two studies. These similarities justify the pooling of the two sets of data. Still we do not wish to imply that the null results from study I with cancer controls are uninformative. In that paper, we have presented the pros and cons of the alternative control groups.8 We admit that the null results from study I when using cancer controls fail to support the results when using population controls. A second change had to do with gasoline engine exhaust. In the previous paper, we evaluated both diesel and gasoline engine emissions and used as the unexposed reference group, those workers unexposed to both diesel and gasoline exhaust. In that paper, we found no association between gasoline engine emissions and lung cancer. For the present paper, we focused exclusively on diesel engine emissions and used as the unexposed reference group only those workers unexposed to diesel exhaust.
We noted that the lifetime prevalence of exposure to diesel exhaust increased considerably from study I to study II. The reason for this increase is not perfectly clear, but there are probably three elements: (1) there was probably a real increase in the prevalence of diesel-powered engines, particularly for motor vehicles, over the era from 1950s to 1980s,21 ,22 (2) the ability of our experts to discern the possibility of diesel exhaust exposure grew as they learned more about the local area industries between the two studies and (3) there was some change between studies in the meaning of the benchmarks, in that for a given level (low, medium, high) the absolute exposure levels were probably lower in study II than in study I. One implication of these elements is that it is likely that average exposure levels among workers coded as exposed to diesel exhaust were probably somewhat lower in study II than in study I, but we cannot speculate by how much. Indeed, the OR estimates (table 3) were somewhat higher in study I than in study II.
The findings from the two studies when using population controls support the hypothesis that occupational exposure to diesel engine emissions is causally linked to lung cancer, and across a wide variety of occupations that present levels of exposure probably lower than those of the classic cohort studies in which excess lung cancer risks were previously found.
Strengths of this work include the large number of lung cancer cases, the collection of detailed lifetime job histories, the labour-intensive expert assessment of exposure and the collection of data on smoking and other covariates, including other occupational carcinogens. In addition, we were able to examine effects of occupational exposures over two periods, and the risks were comparable, and we were able to examine risks specific to different histological types. Risk estimates presented in this manuscript refer to exposures incurred in a broad spectrum of occupations and entailed a wider range of exposure levels and circumstances than those typically found in industrial cohort studies. Having collected detailed lifetime smoking histories and using the best possible parameterisation, we were able to carefully control for the possible confounding effect of the major determinant of lung cancer. In addition, we were also able to control for occupational carcinogenic co-exposures, such as asbestos, crystalline silica, and chromium VI, cadmium, and nickel compounds.
Limitations of our study include exposure misclassification that inevitably occurred in attempting to retrospectively estimate whether or not each subject was exposed, lack of quantitative data on exposure levels and limited statistical power to estimate risk among never-smokers.
Previous studies conducted among workers with high levels of exposure to diesel engine emissions found a consistent increased risk of lung cancer linked to the exposure. These studies were conducted among non-metal miners,7 ,23 potash miners,24 as well as other heavily exposed groups, such as truck drivers,25 ,26 railroad workers,27 ,28 dockworkers29 and bus garage workers.30 We believe that the mean concentrations experienced by our study subjects, even those we labelled as ‘substantial’ exposure, were lower than those of some previously studied cohorts.
Our findings are consistent with those of case–control investigations that used expert-based assessments of occupational exposures.8 ,9 ,26 ,28 ,31 All these studies found an increased risk of lung cancer associated with exposure to diesel engine emissions. Our results are also in line with the positive associations reported in studies that assessed diesel exhaust exposure using either self-reports32 or job exposure matrices.10 ,33
We found a stronger association with squamous cell carcinomas than with small cell carcinomas, and we failed to find any link between diesel exhaust and adenocarcinomas. This specificity of associations supports the view that our positive results are unlikely to be explained by information, selection or confounding bias. If any of these biases played a major role, it should influence in a similar way the risk estimation for all histological types. Stronger associations between diesel exhaust and squamous cell carcinomas were also found by Villeneuve et al 9 and by Boffetta et al.31
Few studies have assessed the joint effects of diesel exposure and smoking, and the results are not consistent. Some authors found a stronger effect of diesel exhaust among smokers compared with non-smokers10; others found that the effect of each of these exposures was attenuated in the presence of high levels of the other7 and some others found an additive effect.34 We clearly found an increased risk due to diesel exhaust in each stratum of smoking. As for the nature of the interaction between smoking and diesel, the pattern of results seems to support more a multiplicative than an additive model, but wide CIs preclude any strong inferences in this regard.
In summary, our findings provide further evidence supporting a causal link between diesel engine emissions and risk of lung cancer. The risk is stronger for the development of squamous cell carcinomas than for small cell tumours or adenocarcinomas.
Exposure assessment methods were expertly developed and implemented by Michel Gérin, Louise Nadon, Ramzan Lakhani, Denis Bégin and Benoit Latreille. The study would not have been possible without the able participation of a large number of research assistants and interviewers, including Marie-Claire Goulet, Jerome Asselin and Sally Campbell.
Funding The study was funded by a number of agencies, including the Health Canada, the National Cancer Institute of Canada, the Medical Research Council of Canada and the Canadian Institutes for Health Research. M-EP is the recipient of a Fonds de la recherche en santé du Québec (FRSQ) salary award. JS was the recipient of a Canada Research Chair and holds the Guzzo-SRC Research Chair in Environment and Cancer.
Competing interests None.
Patient consent Signed by all participants. For deceased subjects, it was signed by next of kin.
Ethics approval Ethical approval was obtained for both studies from each participating hospital and university.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement We have no data sharing policy at the moment.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.