NIOSHTIC-2 Publications Search

Re: The Diesel Exhaust in Miners Study: a nested case-control study of lung cancer and diesel exhaust and a cohort mortality study with emphasis on lung cancer response.

Silverman DT; Attfield MD
J Natl Cancer Inst 2012 Dec; 104(23):1848-1849
We would like to thank the Journal for the opportunity to clarify several points regarding the findings from the Diesel Exhaust in Miners Study (DEMS) that have been raised by several industry consultants and others. First, the observed higher lung cancer risks among surface-only workers compared with those among ever-underground workers in the cohort analysis have resulted in some misinterpretation of our findings by Boffetta, Hesterberg et al., and McClellan and confusion regarding the role of smoking as an effect modifier and confounder of the diesel exhaust (DE)/lung cancer association in this study. As Table 2 of the case-control article shows, the smoking effect among surface-only workers is similar to that observed in previous cohort studies of smoking and lung cancer {providing evidence of the high quality of the smoking data), whereas the smoking effect among moderate-to-heavy smokers is attenuated among ever-underground workers. In addition, Table 2 reveals that, after adjustment for differences in level of diesel exposure, the risk among underground nonsmoking workers is virtually identical to that among surface-only nonsmoking workers (odds ratios = 0.90), suggesting that the risk experienced by surface-only workers (who have much lower levels of DE exposure than underground workers) was mainly due to smoking. The confounding effect of smoking in this study cannot be understood by simply comparing smoking prevalences among surface vs underground workers, as suggested by Boffetta and Moehner et al. Rather, it is the relationship between smoking status/intensity and quantitative levels of DE that is important. Although surface- only workers have a lower prevalence of smoking than underground workers, smoking is inversely related to DE among underground workers, resulting in negative confounding by smoking. Boffetta labels the attenuation of the smoking risk in the presence of heavy DE as "hardly credible." As we noted in the case-control article, this observation will require confirmation because little is known about the interaction between cigarette smoking and diesel exhaust exposure on lung cancer risk. It is interesting, however, that this finding is similar to a recent observation in Xuanwei, China, where the effect of tobacco on lung cancer risk was weak in the presence of heavy indoor exposure to smoky coal and became stronger with diminished smoky coal concentrations after the installation of venting. It is possible that carcinogens in DE, smoky coal, and cigarette smoke operate, in part, in the same metabolic pathway in the body. If so, they may compete with each other, resulting in a saturation of the pathway, thus diminishing the effects of each component. Tse and Yu suggest that the attenuated smoking effect in the presence of heavy DE may have been the result of residual negative confounding by cigarette smoking. As we stated in the case-control article, we cannot rule out the presence of residual confounding by smoking. It is unlikely, however, that the attenuated smoking effect among underground workers is explained by residual confounding. Of the 10 case subjects and 28 control subjects who smoked two or more packs per day and were in the highest DE tertile in Table 6, nine case subjects and 25 control subjects smoked two to less than three packs per day, suggesting good control for cigarette smoking. Finally, the 50% increased risk estimated for individuals with respirable elemental carbon (REC) exposures of 2-6 ug/m3 over a lifetime was based on all workers in the case-control study, not just underground workers, as asserted by Boffetta. Boffetta alleges that the results reported in the articles do not follow the analytical plan described in the DEMS protocol, particularly with regard to the use of lagging in the exposure-response analyses. This is inaccurate. As McClellan acknowledges, we did include "initial (ie, a priori defined)" analyses. These are the categorical analysis findings, which were based on all of the data, without restriction by tenure. Because it is well known that short-tenured workers are often transient workers, smoke more, have greater health problems, and are more likely to have taken jobs with high exposures to known hazards, we believed that it was appropriate to focus on the tenure-restricted analysis in the cohort article. With respect to lagging, the protocol specifies that "lagged estimates of exposure will be used to eliminate the effect of later exposures in examining the exposure-response relationship." Boffetta does not mention that the results for unlagged exposure were made available in the Supplementary Materials that were released with the cohort article (Supplementary Tables 4, 6, and 7, available online), as well as in the case-control article (Table 3). In contradiction to McClellan, significant exposure-response was also seen for unlagged cumulative REC and average REC intensity in the cohort study, as well as in the case-control study for unlagged average REC intensity . Moreover, using calendar time since entry to the study and adjusting for age in the proportional hazard analysis with no restriction on tenure led to virtually the same finding as those in the main cohort article based on age as the underlying time variable and restricting on tenure (Supplementary Table 17). In addition, birth year was included in all of the cohort models and was not added later, as stated by Moehner et al. Hesterberg et al. imply that we provided a biased review of lung cancer risk and DE exposure in our articles. We included both the 2010 review by Gamble in addition to the earlier review by Hesterberg et al.. However, we chose not to include results on the coal miner study cited by Hesterberg et al. because of the absence of exposure to DE in these miners. We dispute the suggestion that our exposure assessment is subject to "substantial" uncertainties. We agree that, as in all retrospective assessments, some uncertainty is to be expected. However, we believe that the criticisms made by McClellan and Hesterberg et al. are founded, at least in part, on a critical misunderstanding of our exposure assessment. These correspondents take issue with our use of carbon monoxide (CO) for back extrapolation of DE levels. In this connection, as we have previously explained, the correlation between the various emission components from individual engines over different cycles is not relevant. Rather, our focus was on the relationship between the contaminants as measured in different locations in the study facilities. In this, we would expect the level of both CO and REC, as measured over the work shift (as opposed to any instant in time), to depend on the extent of diesel engine usage in the different locations. This is a totally different concept and in no way reflects a flaw or weakness in our approach. In addition, our factor analysis findings showed that CO and other gaseous components loaded on a presumed "diesel exhaust" factor most strongly. This finding indicates that CO was a useful proxy for DE, and because CO was the component of DE most frequently measured historically in the study facilities (>6800 measurements), it was the best DE surrogate for the historical DE exposure extrapolation in this study. This large body of historical information was applied to over 1100 REC measurements collected for the epidemiologic investigation. The argument made by Boffetta that the historical values are underestimated is a speculation without analytical foundation. As we indicated in our discussion, it is possible that nondifferential misclassification of DE may have been greater at higher exposures (higher exposures were more likely to have occurred in the earlier years of dieselization). This may have led to a plateauing of risk at the high end of the exposure-response curve, not an overestimation of slope, as Boffetta suggests. We also disagree with Boffetta's comment that we presented little on quantitative bias analysis in our articles. Our analyses included several alternative exposure estimates to evaluate how robust our findings were to variations in assumptions and methods; we found that the results were little changed by choice of estimator. Finally, as McClellan and Spallek and Morfeld point out, the exposures in DEMS are largely based on emissions from older diesel engines, rather than those from engines designed with new technology. As long as exposure from older engines with less effective control technologies dominates human exposure to diesel exhaust worldwide, the estimates of diesel-induced lung cancer risk contained in our articles will continue to be relevant and will provide important data for risk assessment across the full range of exposures.
Diesel-exhausts; Diesel-emissions; Emission-sources; Mining-industry; Mining-equipment; Mine-workers; Lung-function; Lung-disease; Lung-cancer; Lung; Cancer; Case-studies; Respiration; Respiratory-system-disorders; Pulmonary-system; Pulmonary-system-disorders; Pulmonary-disorders; Pulmonary-cancer; Exposure-levels; Smoking; Risk-factors; Underground-mining; Surface-mining; Surveillance
Debra Silverman, NCI, Occupat & Environm Epidemiol Branch, Div Canc Epidemiol & Genet, Bethesda, MD 20892
Publication Date
Document Type
Email Address
Fiscal Year
Identifying No.
Issue of Publication
NIOSH Division
Priority Area
Source Name
Journal of the National Cancer Institute
Page last reviewed: March 25, 2022
Content source: National Institute for Occupational Safety and Health Education and Information Division