Diesel Engine Exhaust Exposure, Smoking, and Lung Cancer Subtype Risks. A Pooled Exposure-Response Analysis of 14 Case-Control Studies

ABSTRACT

Rationale: Although the carcinogenicity of diesel engine exhaust has been demonstrated in multiple studies, little is known regarding exposure-response relationships associated with different exposure subgroups and different lung cancer subtypes.

Objectives: We expanded on a previous pooled case-control analysis on diesel engine exhaust and lung cancer by including three additional studies and quantitative exposure assessment to evaluate lung cancer and subtype risks associated with occupational exposure to diesel exhaust characterized by elemental carbon (EC) concentrations.

Methods: We used a quantitative EC job-exposure matrix for exposure assessment. Unconditional logistic regression models were used to calculate lung cancer odds ratios and 95% confidence intervals (CIs) associated with various metrics of EC exposure. Lung cancer excess lifetime risks (ELR) were calculated using life tables accounting for all-cause mortality. Additional stratified analyses by smoking history and lung cancer subtypes were performed in men.

Measurements and Main Results: Our study included 16,901 lung cancer cases and 20,965 control subjects. In men, exposure response between EC and lung cancer was observed: odds ratios ranged from 1.09 (95% CI, 1.00-1.18) to 1.41 (95% CI, 1.30-1.52) for the lowest and highest cumulative exposure groups, respectively. EC-exposed men had elevated risks in all lung cancer subtypes investigated; associations were strongest for squamous and small cell carcinomas and weaker for adenocarcinoma. EC lung cancer exposure response was observed in men regardless of smoking history, including in never-smokers. ELR associated with 45 years of EC exposure at 50, 20, and 1 μg/m3 were 3.0%, 0.99%, and 0.04%, respectively, for both sexes combined.

Conclusions: We observed a consistent exposure-response relationship between EC exposure and lung cancer in men. Reduction of workplace EC levels to background environmental levels will further reduce lung cancer ELR in exposed workers. 

METHODS

Study Population

 Subjects from 14 hospital- and population-based lung cancer case–control studies in 13 European countries and Canada were pooled. A detailed description of the original study population is available elsewhere (Olsson et al., 2011). The current study updated the population with 3,663 cases and 4,805 controls from the TORONTO, CAPUA (Cáncer de Pulmón en Asturias), and ICARE (Investigation of Occupational and Environmental Causes of Respiratory Cancers) studies in Canada, Spain, and France, respectively (Supplementary Table 1). The project received ethical approvals from all participating countries and from the IARC institutional review board. More information about the SYNERGY project is available online at http://synergy.iarc.fr.


Job-Exposure Matrix and Exposure Assessment

A quantitative diesel engine exhaust JEM (DEE-JEM) was developed by C.G. and R.V. The DEE-JEM consists of EC exposure (in μg/m3) assigned to all 1,506 ﬁve-digit International Standard Classiﬁcation of Occupations (ISCO) (version 1968 or ISCO-68) (ILO, 2010) and was constructed based on 4,417 occupational EC measurements (data sources available in Supplementary Methods and Table 6). For occupations represented in the EC exposure measurements, the mean exposure concentrations were directly assigned. For occupations without measurement data, exposure concentrations from similar occupations with measurement data were assigned using expert decisions. An exposure probability factor was also assigned by expert decision to each exposed job (details on probability factors available in Supplementary Methods). The DEE-JEM was linked to study participant job histories by ISCO-68 occupations. Probability-weighted cumulative EC exposure (hereafter cumulative EC, expressed in μg/m3-years) was calculated as the sum of the product of exposure levels, probabilities, and duration (in years) across all reported job periods for each subject. The DEE-JEM is available upon request from the corresponding author.

Main Statistical Analysis

Separately for men and women, unconditional logistic regression models were used to calculate the odds ratios (ORs) and 95% confidence intervals (CIs) of lung cancer associated with various categorical EC exposure metrics, including ever/never exposure, duration of exposure (<10, 10–19, 20–29, and >29 yr), and cumulative exposure (quartiles of exposure distribution among controls: >0–22, 23–70, 71–178, >178 μg/m3-years). Trends were assessed using P values from the respective indices of EC exposure as continuous variables for all subjects and for exposed subjects only. Adjustments for the main analyses were determined a priori within the SYNERGY consortium and were identical with our previous occupational exposure publications (Olsson et al., 2011, 2017); these adjustments included study, age group (<45, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, and >74 yr), smoking [log(cigarette pack-years + 1)], smoking cessation prior to interview/diagnosis (current smokers; >0–7, 8–15, 16–25, and >25 yr; and never-smokers), and having been ever employed in occupations with known lung cancer risks (list A jobs ever/never; full list in Supplementary Table 7). First published in 1982, list A jobs include occupations with deﬁnite lung cancer risks according to the IARC Monographs; the list was updated in 1995 and 2000 to cover all IARC-reviewed agents up to volume 75 of the Monographs (Ahrens & Merletti, 1998; Mirabelli et al., 2001). Smokers were deﬁned as smoking more than one cigarette per day for more than 1 year. Smoking pack-years were calculated by summing the products of average daily smoking amount in 20-cigarette packs and smoking duration in years. Association between lung cancer and cumulative EC exposure as a continuous metric was assessed with a logistic linear regression model for men, women, and all subjects with identical adjustments as the categorical models. 

Models with various cumulative EC exposure lag times (i.e., omitting exposure in the last 5, 10, 15, or 20 years, or no omission at all) were constructed. According to minimized Akaike information criterion value, model ﬁt was the best when lag time was 10 years—therefore, only results from models with a 10-year lag are presented.
      
Using the lung cancer risk from our linear continuous exposure model with all subjects, we calculated lung cancer excess lifetime risks (ELRs) at age 80 associated with 45 years of occupational EC exposure at 50, 20, and 1 μg/m3 using life-table methods accounting for all-cause mortality outlined by Vermeulen and colleagues (Vermeulen et al., 2014). The selected exposure levels at 50, 20, and 1 μg/m3 represented recommended limit values from the following: 1) the German Committee for Hazardous Substances in 2017 based on a study on lung irritation after controlled human exposure (AGS, 2017); 2) the U.S. National Institute of Occupational Safety and Health in 2003 that was later withdrawn (NIOSH, 2003); and 3) the Health Council of the Netherlands in 2019 based on exposure–response estimates from Vermeulen and colleagues (Health Council of the Netherlands, 2019; Vermeulen et al., 2014), respectively. 2008 European data on mortality from all causes and lung cancer were used in our calculations (European Commission, 2008).

Extended Analysis for Male Subjects 

To further investigate the exposure–response relationship between EC exposure and lung cancer in men, stratiﬁed analyses were performed to calculate lung cancer ORs associated with cumulative EC exposure categories with different major lung cancer subtypes and smoking histories. In addition, nonparametric thin-plate regression splines were created, as implemented in the R package mgcv, to visualize the shape of the exposure–response relationships between EC exposure and lung cancer subtypes in men. The number of basis functions was limited to three (k = 3), and the smoothing parameter was estimated using the relative maximum likelihood method. Spline model results were truncated at the 99th percentile of EC exposure to emphasize results with greater data support.
      
Additive interactions of cigarette smoking and EC exposure on lung cancer and subtype risks in men were assessed by calculating the excess risks due to interaction (RERI) using ORs from our logistic models as deﬁned by Rothman and Greenland (Rothman & Greenland, 1998) and as implemented in the epi.interaction package in R. RERI values measure departure from additivity, with 0 representing no interaction on the additive scale (Knol et al., 2011). Interactions in men on the multiplicative scale were assessed using P values obtained from the cross products of smoking and EC exposure in the adjusted logistic models.
 
Statistical analyses were conducted using SAS (version 9.3; SAS Institute) and R (version 3.6).





RESULTS

A total of 37,866 subjects (16,901 cases; 20,965 controls) were included in our ﬁnal analyses (Table 1). Among the lung cancer cases there were 4,752 adenocarcinomas, 810 large cell carcinomas, 2,730 small cell carcinomas, 6,503 squamous cell carcinomas, 2,012 other lung cancers, and 94 cases without subtype information. 
      
      
In men, we observed elevated ORs for subjects with ever occupational exposure to EC (OR, 1.22; 95% CI, 1.15–1.29; Table 2). Increasing trends in lung cancer risks in men were associated with increases in both exposure duration and cumulative exposure (P trends < 0.01). Elevated male lung cancer ORs were also observed in the lowest categories of exposure duration (1–9 yr; OR, 1.07; 95% CI, 1.00–1.16) and cumulative exposure (>0–22 μg/m3-years; OR, 1.09; 95% CI, 1.00–1.19). In our female population, we observed no associations between lung cancer and different EC exposure metrics.
      

Our continuous EC exposure models show that a 1 μg/m3-year increase in cumulative exposure was associated with an increase in lung cancer OR by a factor of 1.00001 (95% CI, 0.9987–1.00131) for women. The corresponding results for men and for all subjects were identical: lung cancer OR increased by a factor of 1.00034 (95% CI, 1.00021–1.00048) per μg/m3-years increase in cumulative EC exposure. Lung cancer ELRs associated with lifetime occupational EC exposure at 50, 20, and 1 μg/m3 were 3.0%, 0.99%, and 0.04%, respectively, for both sexes combined.
      
By lung cancer subtype, increasing cumulative EC exposure was associated with increasing ORs of squamous cell (P trend < 0.01) and small cell carcinomas (P trend = 0.02) in men (Table 3). For squamous cell carcinoma, all categories of cumulative EC exposure were associated with elevated ORs in men, including the lowest exposure (OR, 1.13; 95% CI, 1.01–1.26). The highest risks for both adenocarcinoma (OR, 1.23; 95% CI, 1.09–1.39) and large cell carcinoma (OR, 1.31; 95% CI, 1.02–1.67) were also observed in men in the group with the highest exposure.

Results from the nonparametric spline analyses for male subjects show monotonic increases in cancer risks for overall lung cancer and all four of the included subtypes (Figure 1). Among the lung cancer subtypes, squamous cell and small cell carcinomas show the strongest association with cumulative EC exposure, followed by large cell carcinoma and adenocarcinoma.
      
      
      
In our analyses stratiﬁed by smoking status, exposure–response associations between cumulative EC exposure and lung cancer were observed in men regardless of smoking history (Table 4). Lung cancer risks were similar for men in the highest EC exposure group who were never-smokers (OR, 1.41; 95% CI, 1.04–1.88), former smokers (OR, 1.47; 95% CI, 1.31–1.65), and current smokers (OR, 1.40; 95% CI, 1.24–1.57). Superadditive joint effects of smoking and EC exposure were observed in men for overall lung cancer and for all four cancer subtypes (Table 5). Suggestive super-multiplicative joint effects of smoking and EC exposure were observed for large cell carcinoma in men (P = 0.05).
      
