A nested case–control study of lung cancer and DE in a cohort of 12 315 workers in eight non-metal mining facilities included 198 lung cancer deaths and 562 control subjects. The case–control study evaluated the exposure–response relationship between DE and lung cancer mortality after adjustment for cigarette smoking and other potential confounding factors that were unavailable in the cohort study.

The results showed a strong and consistent relationship between quantitative exposure to DE and increased risk of dying from lung cancer. Among heavily exposed workers, the risk of dying from lung cancer was approximately three times greater than that among workers in the lowest quartile of exposure.

Exposure to DE may cause lung cancer in mine workers.

Data on smoking and other potential confounders were derived mainly from next-of-kin interviews. Retrospective assessment of DE exposure may result in some misclassification, leading to imprecision in exposure estimates.

From the Editors

Odds ratios for potential confounders (except cigarette smoking) and lung cancer risk are shown in Table 1. A statistically significant increased risk of lung cancer was observed for workers employed at least 10 years in occupations at high-risk for lung cancer (OR = 1.75, 95% CI = 1.06 to 2.91) (Table 1) and those with a history of nonmalignant respiratory disease for at least 5 years before death/reference date (OR = 2.15, 95% CI = 1.21 to 3.82) (Table 1). The elevated risk among those with nonmalignant respiratory disease less than 5 years before death may have been reflective of the early stages of lung cancer. Statistically nonsignificant increased risks were observed for workers who had a family history of lung cancer, smoked cigars for 10 or more years, lived with two or more smokers, exercised less than once per day, and had a vocational school education. Statistically nonsignificant decreased risks were observed among workers who were overweight or obese and who smoked at least 10 pipefuls of tobacco per week (Table 1).

Several non-diesel exposures present at very low levels (ie, levels not typically associated with risk in epidemiological studies) at the study mining facilities were not statistically significantly related to lung cancer risk in our study (Table 1). Levels of radon underground at the study mines were low (ie, arithmetic mean ≤0.02 Working Levels). The odds ratio for workers in the top quartile of cumulative radon exposure was 1.32 (95% CI = 0.76 to 2.29), and workers in quartiles 2 or 3 had little or no increased risk (Table 1). No consistent trend in risk with increasing cumulative radon exposure was apparent (P _trend = .220). Little or no increased risk was observed for possible exposure to asbestos, silica, and PAHs from non-diesel sources, which was consistent with the low measured mean air levels of these potential confounding variables (Table 1) (14). Workers in the top quartile of cumulative respirable dust exposure had an elevated risk (OR = 1.31, 95% CI = 0.70 to 2.46), but workers in quartiles 2 or 3 had no increased risk (Table 1). Factors with statistically nonsignificant increased or decreased risks had little or no confounding effect on estimates of risk from diesel exposure (ie, changed point estimates by ≤10%) and were not included in the final models.

Table 2 shows the effect of cigarette smoking overall and cross-classified by location of employment (ie, surface only and ever underground). Overall, for both surface-only and ever underground workers combined, the risk of lung cancer was statistically significantly associated with smoking status (never, former, current smoker) and smoking intensity (former smoker of ≥2 packs per day vs never smoker: OR = 5.40, 95% CI = 2.23 to 13.06; current smoker of ≥2 packs per day vs never smoker: OR = 12.41, 95% CI = 5.57 to 27.66) (Table 2). We also observed an interaction between cigarette smoking and location of employment, after adjustment for cumulative REC, lagged 15 years (P _interaction = .082). The lung cancer risks associated with moderate (1 to <2 packs per day) and heavy smoking (≥2 packs per day) were higher among workers who only worked at the surface than among those who ever worked underground for both current and former smokers. For example, the odds ratio for current smokers of one to less than two packs per day who worked only at the surface was 13.34 (95% CI = 4.50 to 39.53) compared with an OR of 4.51 (95% CI = 1.50 to 13.58) for those who ever worked underground (Table 2). Because the effect of smoking appeared to be diminished among underground workers compared with that among surface workers, we included the cross classification of location of employment, smoking status, and smoking intensity in all models used to estimate lung cancer risk by diesel exposure (Tables 1, 3, and 7; Figure 1), unless noted otherwise. It is also noteworthy that among never smokers, underground and surface-only workers had similar risks after adjustment for 15-year lagged cumulative REC (OR = 0.90; 95% CI = 0.26 to 3.09) (Table 2), suggesting that the risk experienced by surface-only workers was mainly due to smoking.

Trends in risk with increasing levels of diesel exposure are either statistically significant or of borderline significance (P _trend ≤ .08) for all metrics (both unlagged and lagged) (Table 3). The strongest gradient in risk was seen for 15-year lagged cumulative REC (P _trend = .001). The odds ratio for workers in the top quartile of 15-year lagged cumulative REC exposure (ie, ≥536 μg/m³-y) was 2.83 (95% CI = 1.28 to 6.26) compared with workers in the lowest quartile. When the top exposure quartile was split at the median (ie, 1005 μg/m³-y), the risk continued to rise (P _trend over all five exposure levels = .002); odds ratios were 2.53 (95% CI = 1.06 to 6.04) and 3.20 (95% CI = 1.33 to 7.69) for workers in the top quartile with cumulative REC exposures below and above the median of the quartile, respectively.

We observed a statistically significant gradient in risk with increasing number of years exposed to diesel exhaust among all workers (P _trend = .043), although an elevated odds ratio occurred only in the highest duration category. The odds ratio for workers exposed to diesel exhaust for 15 or more years was 2.09 (95% CI = 0.89 to 4.90) compared with surface workers with negligible or bystander exposure (Table 3).

We also examined risk among all subjects who ever worked underground (Table 4) and among those who worked only at the surface (Table 5). Among underground workers, we observed statistically significant trends in risk with increasing average REC intensity, unlagged (P _trend = .01) and lagged 15 years (P _trend = .001), and with increasing cumulative REC, lagged 15 years (P _trend = .004) (Table 4). Among surface workers, in contrast, no consistent positive gradient in risk with increasing diesel exposure was apparent (Table 5), probably due to the small number of subjects (53 case subjects and 100 control subjects) and the low levels of diesel exposure experienced by surface workers. Because of the increased precision gained by estimating odds ratios based on all subjects, our primary estimates of risk are based on surface and underground workers combined (Table 3).

We stratified the combined results (Table 3) on whether the subject had self-reported diesel exhaust exposure from a job outside the study mining facility (eg, ever employed as a long-haul truck driver) (data not shown). No systematic differences in risk were apparent among subjects with or without occupational diesel exposure outside the study facility (P _interaction between cumulative REC, lagged 15 years, and outside occupational diesel exhaust exposure = .222).

Use of protective equipment did not appear to modify the observed associations between diesel exhaust exposure and lung cancer. However, most information on protective equipment use was obtained from next-of-kin interviews, resulting in a large number of workers with unknown data (59 case subjects and 129 control subjects). Subjects who reported having used protective equipment appeared to experience risks similar to the estimates for all workers combined (Table 3). For example, among workers who used protective equipment, odds ratios for 15-year lagged cumulative REC exposures of less than 3 μg/m³-y, 3 to less than 72 μg/m³-y, 72 to less than 536 μg/m³-y, and 536 μg/m³-y or more were 1.0 (referent), 0.31 (95% CI = 0.04 to 2.23; 16 case subjects and 42 control subjects), 1.76 (95% CI = 0.11 to 27.91; 10 case subjects and 23 control subjects), and 3.66 (95% CI = 0.26 to 52.09; 20 case subjects and 31 control subjects), respectively.

Figure 1 shows category-specific odd ratios (square symbol), with confidence intervals omitted for clarity, and fitted odds ratios for 15-year lagged average REC intensity and cumulative REC using various continuous models. To provide additional points for graphing the exposure–response curve based on categorical data (Figure 1), we expanded the number of cut points (cut points for average REC intensity, lagged 15 years: <2, 2 to <4, 4 to <8, 8 to <16, 16 to <32, 32 to <64, 64 to <128, 128 to <256, and ≥256 μg/m³; cut points for cumulative REC, lagged 15 years, were similarly defined but multiplied by a factor of 10 to account for duration of exposure: <20, 20 to <40, 40 to <80, 80 to <160, 160 to <320, 320 to <640, 640 to <1280, 1280 to <2560, and ≥2560 μg/m³-y; Supplementary Table 2, available online). Odds ratios increased with 15-year lagged average REC intensity and leveled off above 20–80 μg/m³ (Figure 1, A for the full range and Figure l, B for average REC intensity under 128 μg/m³). For the full range, the odds ratio pattern was best explained by a one-parameter power model (deviance = 5.3, P = .022), whereas for the restricted range, the power and linear models were comparable (deviance = 2.8, P = .092 and deviance = 3.2, P = .075, respectively). A similar increasing pattern of odds ratios was observed for cumulative REC exposure, lagged 15 years (Figure 1, C for the full range and Figure l, D for cumulative REC under 1280 μg/m³-y), with a leveling off of risk for exposures above 1,000 μg/m³-y and perhaps a decline in risk among the most heavily exposed workers. The two-parameter linear-exponential model (dotted line) was the best fitting model for the full range (relative to the null model, deviance = 12.2, P = .002) (Figure l, C); for the restricted range, the best models were the one-parameter linear model (dashed-dotted line) (deviance = 15.6, P < .001) and the two-parameter linear-exponential model (dotted line) (deviance = 16.0, P < .001) (Figure 1, D) (Supplementary Table 1, available online). We carried out similar model comparisons using the unlagged exposure metrics (Supplementary Table 3, available online). However, our evaluation of optimal lag intervals (Supplementary Figure 1, available online) suggested that the unlagged approach led to exposure misclassification because recent exposures may not have had sufficient time to contribute to lung cancer risk and thus resulted in generally poorer fit of the various models.

The combined effect of diesel exposure and intensity of cigarette smoking is shown in Table 6. Among the 14 case subjects and 178 control subjects who never smoked, odds ratios by tertile of cumulative REC, lagged 15 years, were: 1.0 (referent), OR = 1.47 (95% CI = 0.29 to 7.50), and OR = 7.30 (95% CI = 1.46 to 36.57). Risk also increased with increasing level of diesel exposure among smokers of less than one and one to less than two packs per day. In contrast, risk decreased with increasing levels of diesel exposure among smokers of at least two packs per day. Similarly, risk associated with smoking intensity was modified by diesel exposure. Among workers in the lowest tertile of cumulative REC, lagged 15 years, smokers of at least two packs per day had a risk 27 times that of nonsmokers, whereas among those in the highest tertile of cumulative REC, heavy smokers had about 2.5-fold the risk of nonsmokers. The P _interaction between level of diesel exposure and cigarette smoking was .086.

We evaluated lung cancer risk by quantitative level of diesel exposure for each type of mining facility (Table 7). Too few workers were employed in the one salt and the one limestone mining facility to estimate risk for these types separately. For workers in both potash and trona mining facilities, risk tended to increase with increasing levels of average REC intensity and cumulative REC exposure. Trends were more consistent among potash miners, perhaps reflecting more stability in odds ratios resulting from twice as many case subjects in the potash as in the trona facilities (Table 7).

This case–control study nested within a cohort of miners showed a strong and consistent relation between quantitative exposure to diesel exhaust and increased risk of dying of lung cancer. To our knowledge, this is the first report of a statistically significant exposure–response relationship for diesel exposure and lung cancer based on quantitative estimates of historical diesel exposure with adjustment for smoking and other potential confounders. We observed increasing trends in risk with increasing exposure to diesel exhaust for both average REC intensity and cumulative REC exposure, unlagged and lagged 15 years, with the strongest gradient in risk with cumulative REC, lagged 15 years. We further observed a gradient of increasing risk within the top quartile of 15-year lagged cumulative REC exposure for workers below and above the median of the quartile. The associations between diesel exposure and lung cancer were apparent for workers employed in either the potash or trona facilities (too few workers were employed in the one salt and one limestone mine to estimate risk separately). The consistency of findings for both potash and trona facilities is noteworthy because smoking was prohibited in the trona facilities but not in potash or the other facilities in the study. Reports by next of kin or study subjects of workers’ use of protective equipment within the study mining facilities and workers’ additional occupational exposure to diesel exhaust outside the study facilities had little or no impact on our findings.

These positive findings are consistent with those of the cohort analysis of underground workers in the same study population (12). However, estimates of risk for underground workers in the case–control analysis were somewhat higher than those based on the cohort analysis. For example, the odds ratios by quartile of the 15-year lagged cumulative REC exposure in the case–control analysis were 1.0, 2.11, 3.48, and 5.90 (for cohort cut points, <108, 108 to <445, 445 to <946, and ≥946 μg/m³-y, respectively), compared with hazard ratios of 1.0, 1.50, 2.17, and 2.21 from the cohort analysis (12). The lower point estimates from the cohort analysis may be partly due to negative confounding from cigarette smoking because current smoking was inversely related to diesel exposure in underground workers (36% and 21% current smokers in lowest vs highest cumulative REC tertile, respectively). Odds ratios for underground workers in the case–control analysis using the same cohort cut points dropped to 1.0, 1.94, 2.42, and 3.75, respectively, when smoking was removed from the model.

The continuous models suggest a steep slope at the low end of the exposure–response curve followed by a leveling, or perhaps even a decline, in risk among the most heavily exposed workers. A plateauing of exposure–response curves has been reported in studies of other occupational exposures and cancer risk (19). Possible biological explanations for a plateauing effect include saturation of metabolic activation and enhanced detoxification or greater DNA repair efficiency at higher exposure levels. Alternatively, nondifferential misclassification of diesel exposure may be greater at higher exposures, obscuring further increases in risk.

We observed an increased lung cancer risk associated with diesel exposure as was seen among German potash miners (11), as well as among other diesel-exposed occupational groups including truck drivers (6,7), railroad workers (8,20), dockworkers (9), and bus garage workers (10). The German potash miners study (11) found elevated risk with increasing estimated cumulative total carbon exposure (another surrogate for diesel exposure), although the trend was not statistically significant. Relative risks were 1.0, 1.13, 2.47, 1.50, and 2.28 for exposure quintiles (ie, <1.29, 2.04, 2.73, 3.90, >3.90 mg/m³-y, respectively) (11). Some differences between the German study (11) and this study are that this study is considerably larger (US miners: 198 lung cancer deaths out of a total of 278 041 person-years; German miners: 61 lung cancer deaths out of 152 557 person-years), and the US miners had a longer latent period for the development of lung cancer than the German miners because diesel technology was introduced earlier in the US study mines (1947–1967) than in the German mines (1969). Finally, in this study, an intensive effort was undertaken to characterize diesel exposure levels over time by incorporating changes in size of the diesel equipment, numbers of equipment, and air flow rates exhausted from the mines based on information collected from the facilities. Our information indicated that these factors varied considerably over time (14). In the German study, the investigators relied on reports from local engineers and industrial hygienists that working conditions were constant over past years. However, in contrast to this study, no past industrial hygiene measurements were available to confirm this assumption.

We observed an attenuation of the effect of cigarette smoking among study subjects who were exposed to high levels of diesel exhaust as estimated by REC (Table 6). This finding mirrors a recent observation from a study in Xuanwei, China (21), where lung cancer rates are high because of unvented indoor burning of coal for heating and cooking in homes (22). The effect of tobacco on lung cancer risk in that study was weak in the presence of heavy indoor exposures from smoky coal but became stronger with installation of venting, which greatly diminished smoky coal air concentrations (21,22). Little is known about the effect of the interaction between cigarette smoking and diesel exhaust exposure on lung cancer risk. If our observation of attenuation of the smoking effect in the presence of high levels of diesel exhaust is confirmed, several possible mechanistic explanations are apparent. First, at high levels of diesel exhaust exposure, PAHs, nitro-PAHs, and related compounds could compete with the metabolic activation of PAHs in tobacco smoke, leading to enzyme saturation. For example, PAHs in complex mixtures have been shown to have less than additive genotoxic effects at higher exposure levels (23). Second, constituents of diesel exhaust may suppress enzymes that activate or induce enzymes that detoxify carcinogens in tobacco smoke. For example, diesel exhaust particles have been shown to reduce activity of CYP2B1, which plays a role in the activation of certain tobacco-specific nitrosamines (24). Also, diesel particulate matter has been shown to reduce the initiation of skin tumors in Sencar mice treated with the potent PAH dibenzo[a,l]pyrene, possibly through inhibition of enzymes that carry out its metabolic activation (25).

We also observed a weakening of the diesel exhaust effect among heavy smokers (ie, smokers of at least two packs per day), which is necessarily implied by the observation of a weakening of the effect of smoking at least two packs per day among workers heavily exposed to diesel exhaust. It has previously been reported that coal dust burden in the lungs of coal miners is reduced among smokers, which may be attributable to increased coal dust clearance (26), and it is possible that diesel exhaust particulate deposition may be reduced in the lungs of smokers by a similar process. Although little experimental evidence is available to date to support and explain effect modification of diesel exposure by smoking, it is theoretically possible by one or more of the mechanisms described above.

If the observed interaction between smoking and diesel exhaust represents a real effect, then the generalizability of our estimates of risk for diesel exposure to other populations depends not only on the level of exposure to diesel exhaust but also on the distribution of smoking status and intensity in the population. For example, estimates of lung cancer risk in a population of never smokers with diesel exposures similar to those of the miners in this study would be 1.0, 1.47, and 7.30 for individuals with cumulative REC, lagged 15 years, of less than 8 μg/m³-y, 8 to less than 304 μg/m³-y, and 304 μg/m³-y or more, respectively. In contrast, the overall study population, which included 29% never smokers, had lower odds ratios of 1.0, 1.12, and 2.40 for the same tertiles of cumulative REC exposure, lagged 15 years, respectively (data not shown). In fact, the proportion of never smokers in this study population is substantially lower than the 51% reported for the US population of men aged 18 years or older (27), suggesting that diesel-related estimates of lung cancer risk in the US population may be higher than the overall risk estimates reported here because the proportion of never smokers in the US population is higher than in this study cohort.

Our study has several major strengths including its relatively large size, which provided adequate statistical power to detect a statistically significant exposure–response relationship, adequate latent period for the development of lung cancer, detailed exposure assessment that enabled us to evaluate risk based on quantitative historical exposure to REC, subjects with a wide range of diesel exposure and with underground workers experiencing exposure levels considerably higher than that of other occupationally exposed groups in previous studies, a high interview participation rate for both case subjects and control subjects, and the ability to control for confounding from smoking and other lung cancer risk factors. Two main limitations are also apparent. First, the data on smoking and other potential confounders were derived mainly from next-of-kin interviews. Although a comparison of confounder data derived directly from living and from next of kin for deceased control subjects revealed comparability of responses, we cannot completely rule out the possibility of residual confounding. Second, as in most epidemiological studies of cancer that rely on retrospective exposure assessment, estimates of diesel exposure in this study undoubtedly had some imprecision despite considerable effort to minimize misclassification. This imprecision is likely to result in nondifferential misclassification of exposure, which would tend to bias the estimates of risk toward the null (28). Thus, the true estimates of lung cancer risk associated with diesel exhaust may, in fact, be higher than those reported here.

In sum, our results provide further evidence supporting a causal effect of diesel exhaust exposure on lung cancer mortality in humans. We observed a statistically significant exposure–response relationship after we adjusted for possible confounding from smoking and other established and hypothesized lung cancer risk factors. The exposure–response curve showed a steep increase in risk with increasing exposure at low-to-moderate levels followed by a plateauing or perhaps a decline in risk among heavily exposed subjects.

Our findings are important not only for miners but also for the 1.4 million American workers and the 3 million European workers exposed to diesel exhaust (29), and for urban populations worldwide. Some of the higher average elemental carbon levels reported in cities include Los Angeles (4.0 μg/m³) (30), the Bronx (a borough in New York City) (6.6 μg/m³) (31), nine urban sites in China (8.3 μg/m³) (32), Mexico City (5.8 μg/m³) (33), and Estarreja, Portugal (11.8 μg/m³) (34). Environmental exposure to average elemental carbon levels in the 2-6 μg/m³ range over a lifetime as would be experienced in highly polluted cities approximates cumulative exposures experienced by underground miners with low exposures in our study. Because such workers had at least a 50% increased lung cancer risk, our results suggest that the high air concentrations of elemental carbon reported in some urban areas may confer increased risk of lung cancer. Thus, if the diesel exhaust/lung cancer relation is causal, the public health burden of the carcinogenicity of inhaled diesel exhaust in workers and in populations of urban areas with high levels of diesel exposure may be substantial.

The research was funded by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, Division of Cancer Epidemiology and Genetics and the National Institute for Occupational Safety and Health, Division of Respiratory Disease Studies.