Letter-to-the-Editor RE: 'Diesel exhaust in miners study: how to understand the findings?' by Peter Morfeld
Debra Silverman, National Cancer Institute, NIH, DHHS
5 March 2013
We thank Dr. Morfeld for providing us the opportunity to clarify several points regarding the analyses of data from both the cohort (1) and case-control (2) components of the Diesel Exhaust in Miners Study (DEMS).
First, we agree with Dr. Morfeld's statement "The distinction between surface and underground work is obviously of major importance for an understanding of this study." In fact, to ignore location and simply estimate risk by exposure and adjusting for smoking would have led to erroneous results. However, we do not agree with Dr. Morfeld's statement "this factor `surface/underground work' remains unexplained." The smoking effect among surface-only workers shown in Table 2 (2) is similar to that observed in previous cohort studies of smoking and lung cancer (3), whereas the smoking effect among underground workers who smoke at least 1 pack per day is attenuated. As shown in Table 2 (2), after adjustment for diesel exposure, the risk among underground nonsmoking workers is virtually identical to that among surface-only nonsmoking workers (OR for 15-year lagged cumulative REC = 0.90), providing evidence that the observed difference in risk between surface only and underground workers is explained by smoking and diesel exposure. When smoking and location are included as separate variables in the model, estimates of risk are quite similar to those in Table 3 (2) (e.g., ORs for quartiles of 15-year lagged cumulative REC: 1.0 (referent), 0.87 vs. 0.74, 1.60 vs. 1.54, 2.97 vs. 2.83, respectively) with the P for trend = 0.001 in both models. Because the interaction between smoking and location was borderline significant, we included a cross-product term in the models to fully capture this interaction. The inclusion of the cross-product term had a negligible impact on the estimates of risk, however. Dr. Morfeld also misquotes us as hypothesizing that "high REC exposures are protective against lung cancer excess risks due to smoking." We wrote that the effect of smoking is "attenuated" in the presence of high levels of diesel exposure, which is not the same as "protection" since smoking does indeed cause lung cancer among underground workers who were heavily exposed to diesel exhaust.
Second, Dr. Morfeld questions whether we have a systematic underestimation of lung cancer deaths, especially at lower diesel exhaust exposure levels. The mortality follow-up was undertaken by NIOSH using typical and standard methodology and data from the National Death Index, Social Security Administration, and Internal Revenue Service. Other sources, such as postmasters, tracing agencies, and company records were also employed. Overall, the information from the various sources was internally very consistent. Third, Dr. Morfeld goes on to focus on the external vs. internal analysis results, finding it `perplexing' that they should differ. In response, we note that it is well established that external analysis is subject to factors including the healthy worker effect that are better addressed through internal analysis. In addition, as noted in the cohort paper (1), different patterns of lung cancer mortality by location, which were shown by the case-control analysis (2) to be caused by smoking, obscured the exposure-response relationship evident in the complete cohort. Clear exposure-response patterns were identified after stratification by location (Tables 4 and 5 of the cohort paper (1)), and were also observed after adjustment for location (Table 6 of the cohort paper (1)). Fourth, we were unable to comprehend what Dr. Morfeld is recommending with respect to analysis by worker location. We employed a time dependent stratification approach (e.g., a worker was surface-only until the first time they went underground, at which time they became an ever-underground worker). We cannot imagine that Dr. Morfeld was advocating using underground tenure as a primary exposure variable rather than using the available quantitative exposure data.
Fifth, it is generally accepted that retrospective exposure assessments in occupational studies such as DEMS have some imprecision (4). If, as is likely, the error is nondifferential, the exposure-response slopes may be attenuated. Considerable effort went into evaluating the exposure assessment in DEMS, including a separate paper on this topic (5). We also developed a number of alternative exposure metrics. These are discussed in the cohort paper (1) and were found not to impact the findings in any significant way.
Dr. Morfeld asks several detailed questions concerning our methods. In response: 1) We did not censor or otherwise treat deaths over age 85 differently from younger deaths (but only 3 of 200 lung cancer deaths in the cohort were >85 years and all available pathology reports for the lung cancer cases were evaluated by an expert pathologist to confirm lung cancer as the cause of death). 2) In the cohort analysis, all confounder variables were entered into the model together and each examined for their apparent influence on lung cancer mortality. In the case-control study, each potential confounder was added one-at-time to the conditional model containing an exposure metric, the cross product of smoking and location, employment in a high-risk occupation for at least 10 years, and a history of nonmalignant respiratory disease for at least 5 years to create a base model. We also ran full models that simultaneously included the base model plus the other potential confounders. 3) In the cohort analysis, observations with missing data for a particular analysis were dropped; in the case-control analysis, missing confounder data were included as a separate level. When we evaluated this approach by excluding missing confounder data from key analyses, observed patterns of risk were similar to those when missing data were included as a separate level. 4) Although workers were not permitted to smoke underground in the trona mines, patterns in exposure-response were similar by mine type, suggesting that differential smoking patterns across facilities did not impact the findings. 5) In the cohort analysis, there was no overall effect of dropping those whose age at first exposure was greater than 40 (the cumulative REC exposure hazard ratio dropped slightly while that for exposure intensity increased).
Despite the complexities of the cohort and case-control analyses, there is no doubt that excess risks of lung cancer were detected, and that these elevations were associated with increasing levels of diesel exhaust. We do not accept that our methods suffer from errors so egregious that the findings could be spurious. In fact, the scientific community seems to have had no difficulty in accepting our results, as reflected by the recent IARC review of the carcinogenicity of diesel exhaust, in which our findings played an important role in the evaluation process (6).
Debra T. Silverman
Michael D. Attfield
References
1. Attfield MD, Schleiff PL, Lubin JH, Blair A, Stewart PA, Vermeulen R, Coble JB, Silverman DT. The Diesel Exhaust in Miners Study: A cohort mortality study with emphasis on lung cancer. J Natl Cancer Inst 2012;104(11):869-883.
2. Silverman DT, Samanic CM, Lubin JH, Blair AE, Stewart PA, Vermeulen R, Coble JB, Rothman N, Schleiff PL, Travis WD, Ziegler RG, Wacholder S, Attfield MD. The Diesel Exhaust in Miners Study: A nested case-control study of lung cancer and diesel exhaust. J Natl Cancer Inst 2012;104(11):855-868.
3. Blot WJ, Fraumeni JF, Jr. Lung and pleura. In: Schottenfeld D, Fraumeni JF Jr. eds. Cancer Epidemiology and Prevention. 2nd Ed. New York: Oxford University Press, 1996:637-665.
4. Stewart PA, Coble JB, Vermeulen R, Blair A, Lubin J, Attfield M, Silverman DT. Response to Borak et al., 2010 on the Diesel Exhaust in Miners Study. Ann Occup Hyg 2011;55(3): 343-346.
5. Stewart PA, Vermeulen R, Coble JB, Blair A, Schleiff PL, Lubin JH, Attfield M, Silverman DT. The Diesel Exhaust in Miners Study: V. Evaluation of the exposure assessment methods. Ann Occup Hyg 2012 Mar 1 Epub.
6. Lamia Benbrahim-Tallaa, Robert A Baan, Yann Grosse, Beatrice Lauby-Secretan, Fatiha El Ghissassi, Veronique Bouvard, Neela Guha, Dana Loomis, Kurt Straif. Carcinogenicity of diesel-engine and gasoline-engine exhausts and some nitroarenes. Lancet Oncol 2012;13(7):663-664.
Affiliations of authors: Occupational and Environmental Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD (DTS); Surveillance Branch, Division of Respiratory Disease Studies, National Institute for Occupational Safety and Health, Morgantown, WV (MDA).
Competing interests
None
Author reply
Peter Morfeld, Institut for Occupational Epidemiology and Risk Assessment of Evonik Industries
23 April 2013
I'd like to thank Drs Silverman and Attfield for their profound reply [1] to my commentary [2] and their in-depth explanations of specific aspects of the Diesel Exhaust in Miners Study [DEMS, 3, 4]. It was my motivation to instigate discussions about this important and impressive epidemiological project. Thus, I highly appreciate that the leading DEMS authors responded to my commentary in such a great detail. The Editors of JOMT invited me to follow-up on this discussion and I will respond to the points as they were numerated in Silverman and Attfield [1].
1) The DEMS findings about the effect of long-term respirable elemental carbon (REC) exposure estimates on lung cancer mortality depended critically on the adjustment by location (surface vs. underground work). Cox models that followed the analysis plan did not show any convincing association between REC exposure estimates and lung cancer mortality. Only after subdividing the cohort into 'surface only workers' vs. 'ever underground workers' or after adjusting for a 'surface only/ever underground work' indicator the Cox models returned pronounced dose/response relationships with REC exposure estimates. Silverman and Attfield [1] agreed. They confirmed that the distinction between surface and underground work is of major importance for an understanding of DEMS. Of note, they did not agree with my statement that "this factor `surface/underground work' remains unexplained." They argued that "the observed difference in risk between surface only and underground workers is explained by smoking and diesel exposure". Unfortunately, this statement appears to be wrong. If their statement were true we expected to see that the location variable could be substituted by a combination of the smoking and exposure variables without affecting the REC exposure estimates. Thus, models that follow the original study protocol and included no location factor but smoking information should return large REC exposure effect estimates. In addition, the location variable should show a clearly reduced and non-significant effect in models after adjusting for smoking and exposure. Based on the DEMS papers, my understanding is that this is not true. If Drs. Silverman and Attfield [1] still disagree I'd like to ask them to reveal a) the REC exposure coefficients (with CIs and p-values) returned by conditional logistic regression with and without adjustment for location, while always controlling for smoking and b) to show the location coefficient (with CI and p-value) after adjustment for smoking and REC exposure (smoking and location variables should be entered into the models as separate variables so that we can distinguish their effects on lung cancer mortality).
Silverman and Attfield [1] made an observation that they interpreted as evidence for identical risks among surface and underground workers after appropriate adjustment:: "As shown in Table 2 [2], after adjustment for diesel exposure, the risk among underground nonsmoking workers is virtually identical to that among surface-only nonsmoking workers (OR for 15-year lagged cumulative REC = 0.90), providing evidence that the observed difference in risk between surface only and underground workers is explained by smoking and diesel exposure." However, the 95%-confidence interval of the cited odds ratio (OR) spanned from 0.26 to 3.09. This can hardly be interpreted as evidence in favor of the null hypothesis.
There are other interpretations of the DEMS authors about smoking that are difficult to follow. Möhner [5] noted in his Letter to the Editor: "According to the authors, there was no evidence of an increased smoking prevalence in those employees working underground. Yet a comparison of the smoking data from controls by location of employment (Table 2 [2]) yielded a significantly higher percentage of never smokers and significantly lower percentage of heavy smokers (>2 packs per day) among surface-only workers than among ever-underground workers (34.2% vs. 22.1% and 6.3% vs. 13.6%, respectively). Although underground workers smoked significantly more, were more burdened with exposure to other occupational lung carcinogens (see for example standardized mortality ratios (SMR) for pneumoconiosis, Supplementary Table 2 [1]) and were much more exposed to DME than surface workers, their lung cancer mortality is lower. We would be grateful to the authors if they could provide readers with additional data so that the studies' conclusions can be comprehended. In particular, we would like to know, how the SMR reflect differences by exposure parameters." The authors did not reply to this. Silverman et al. [4] wrote: "a comparison of confounder data derived directly from living and from next of kin for deceased control subjects revealed comparability of responses". This statement appears to be odd when applied to smoking. Pallapies et al. [6] noted: "Silverman et al. [4] obtained smoking data in cases from next of kin only. However, they obtained such data from living control subjects as well as from next of kin of other control subjects. This may have led to considerable bias, as different percentages of 'current smokers' from direct versus next of kin interviews (11 vs. 23 %) demonstrate".
In this respect I like to repeat point 4 in Morfeld [2] that remained unanswered: "The REC exposure risk estimates differed with location [1]: the authors reported a twenty times higher excess risk per g/m3-y on 'surface only' in comparison to 'ever underground'. After taking logs of exposure the 'surface only' REC coefficient was about twice the one calculated for workers 'ever underground'. The researchers [1] tested the differences of the effect estimates (significant on the log scale, not significant on the linear scale) but neither reported how they performed the tests nor did they show any details of the results. The usual way to perform such a test is to add an interaction term 'REC exposure x location' to the models. It is surprising that the authors [1,2] did not present such interaction models.
Silverman and Attfield [1] wrote that I misquoted them as hypothesizing that "high REC exposures are protective against lung cancer excess risks due to smoking." I am not sure whether this comment is on substance or on semantics. I just wanted to make clear that Silverman et al. [4] listed mechanisms that they believe could be true and may reduce lung cancer risk from smoking because of elevated diesel motor emission exposures, e.g., "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]pyrene, possibly through inhibition of enzymes that carry out its metabolic activation [25]."
2) I'd like to thank Drs Silverman and Attfield [4] that they presented the ORs when not adjusting for an interaction of location and smoking. This analysis demonstrated that an inclusion of the cross-product term was not necessary to estimate the effect of REC exposure. Unfortunately, they did not show other results I asked for: ORs for REC exposure after controlling for only those variables used in SMR calculations, and ORs for smoking in underground and surface workers without adjustment for REC exposure [2] Furthermore, like Möhner [5] I showed interest in lung cancer SMRs across REC exposure categories to judge whether the study has pronounced deficits of lung cancer mortality in low categories of exposure and whether there are any significant SMR excesses in the highest exposure categories. Thus, the authors did not present data to rule out my hypothesis that the lung cancer mortality structure is strange and similar to the leukemia pattern observed in NCI's formaldehyde cohort [7, 8]. Silverman and Attfield [1] confirmed that their mortality follow-up procedure was the same or very similar to the method applied in the NCI formaldehyde cohort. Of note, NCI formaldehyde researchers had missed about 1000 deaths out of 9500 which led to distorted exposure-response findings [9]. Silverman and Attfield [1] did present no arguments that DEMS may not suffer from the same problems.
3) Silverman and Attfield [1] wrote: "Dr. Morfeld goes on to focus on the external vs. internal analysis results, finding it `perplexing' that they should differ. In response, we note that it is well established that external analysis is subject to factors including the healthy worker effect that are better addressed through internal analysis." I'd like to emphasize that I was not talking about exposure. I found it perplexing (and still find it perplexing) that such a large location effect popped up after adjusting for REC exposure and went unnoticed otherwise. The idea of Silverman and Attfield [1] that this observation may be associated to the healthy worker effect appears to be off the track.
4) Silverman and Attfield [1] "were unable to comprehend what Dr. Morfeld is recommending with respect to analysis by worker location. We employed a time dependent stratification approach (e.g., a worker was surface-only until the first time they went underground, at which time they became an ever-underground worker)." I will try to explain it differently. The binary variable 'location', as defined and used in the DEMS analysis, does not tell us for each person-year whether the miner was underground or on surface. Here is why: Assume that a miner worked underground up to year x but then changed to a surface job and stayed on surface until year x+s. Then the variable 'location', as defined and used in the DEMS analysis, is set to 'underground' during the years x to x+s although the miner was on surface. In contrast to what had been done in DEMS I recommended and I still recommend that the variable 'location' should correctly reflect for each person-year whether the miner was underground or on surface. According to the description of the exposure assessment process a REC estimate was allocated to every person-year. And because REC exposures differed between surface and underground work a reliable exposure assessment should take account of workers' location in every person- year. Thus, the variable 'location' can be defined in the suggested way (if not, the DEMS exposure estimates are obviously inaccurate). Note that it is improbable that in this large cohort miners only changed from surface to underground work but never vice versa. Because the variable 'location' is so important in DEMS this variable should be defined in a better way than Silverman and Attfield [1] did.
Silverman and Attfield [1] wrote "We cannot imagine that Dr. Morfeld was advocating using underground tenure as a primary exposure variable rather than using the available quantitative exposure data." I never recommended using underground tenure as a primary exposure variable. This appears to be a misunderstanding.
5) Silverman and Attfield [1] wrote: "it is generally accepted that retrospective exposure assessments in occupational studies such as DEMS have some imprecision [4]. If, as is likely, the error is non-differential, the exposure-response slopes may be attenuated. Considerable effort went into evaluating the exposure assessment in DEMS, including a separate paper on this topic [5]. We also developed a number of alternative exposure metrics. These are discussed in the cohort paper [1] and were found not to impact the findings in any significant way." This does not justify taking the exposure data as fixed and assuming 'no exposure errors' when modeling risks. Monte Carlo sensitivity and Bayesian bias analyses are well developed and these procedures are able to cover the involved uncertainty. I will provide two examples that may help to understand what kind of potential problems exist in DEMS. The REC exposure assessment was mainly based on a correlation between CO and REC measurements [10]. The authors admittted "that the observed coefficient derived from this cross-sectional study might not apply longitudinally to past conditions." This induces problems because the DEMS analysis relies on life-long cumulative exposures. The relevance of such uncertainties can be measured by appropriate bias analyses. Moreover, the coefficient chosen by DEMS researchers to transfer CO values into REC values was not supported by the data [11]. Thus, this may have distorted the risk estimates. Again, this motivates to perform bias analyses.
6) I'd like to thank Drs. Silverman and Attfield [1] that they were open to clarify even some of the minor issues that I raised (sub-issues 1 to 5 in Silverman and Attfield [1]). Sub-issue 3: I am not clear whether the authors wanted to say that the case-control study was analyzed twice (two different ways to deal with missing data)? Sub-issue 4: I had interest in the smoking patterns. Can the patterns be presented? One issue remained without response: the suggestion of a curvilinear modeling.
I'd like to add that the paper describing the re-analysis of the German potash miner cohort was accepted for publication [12]. Silverman et al. [4] stated:"We observed an increased lung cancer risk associated with diesel exposure as was seen among German potash miners [11]." This statement is no longer evident because of the null results reported in Möhner et al. [12] after taking account of prior exposures in Uranium mining.
Final remark
Silverman and Attfield [1] wrote as a final comment: "Despite the complexities of the cohort and case-control analyses, there is no doubt that excess risks of lung cancer were detected, and that these elevations were associated with increasing levels of diesel exhaust. We do not accept that our methods suffer from errors so egregious that the findings could be spurious. In fact, the scientific community seems to have had no difficulty in accepting our results, as reflected by the recent IARC review of the carcinogenicity of diesel exhaust, in which our findings played an important role in the evaluation process". I do have to insist that pronounced and clear-cut lung cancer excess risks were only observed after adjustment for location. Anyhow, Silverman and Attfield [1] were right when they reported on the evaluation of a working group who met at IARC, Lyon in June 2012 [13]. I attended that meeting as an observer. Indeed, the group of invited experts took the exceptionally large REC exposure estimates of the DEMS for granted without mentioning the potential problems involved. I have to admit that the frustrating commentary published by another observer is appropriately describing the downsides of the evaluation process at this IARC monograph meeting [14]. John Gamble reminded the reader that the Preamble of the IARC monographs [e.g., 15] defines clearly what the working group should do: "When an important aspect of a study that directly impinges on its interpretation should be brought to the attention of the reader, a Working Group comment is given in square brackets". Regarding DEMS the working group had no square brackets added to mention the discussed weaknesses of the study. They did so although the weaknesses were brought to their attention, even in published form [2]. Another striking example is linked to the European-Canadian pooled case-control study [16]. In the abstract the authors described a positive association between diesel exhaust exposure and lung cancer risk in a smoking-adjusted analysis. The working group repeated these statements of the authors and noted in a square bracket that this finding is unlikely to ["be explained by bias or confounding"]. Tim Lash, another observer and well-known for his high-level work on bias adjustment in epidemiological studies [e.g., 17], commented on this square bracket during the epidemiological subgroup meetings and also in the plenary meeting of the full working group. Lash noted that this pooled case-control study was very large and one of the rare investigations that studied lung cancer risks among never smokers. He made clear that a lack of association was found among never smokers, possibly the group that could provide the least confounded risk estimates after Diesel exhaust exposure [Table 3, 16]. Contrary to the statement of the working group Lash explained that the positive result after smoking adjustment could be due to residual confounding, in particular in the light of the null finding within never smokers. Again, this comment went unnoticed by the working group. The square bracket statement about Olsson et al. [16] was not changed and this important information about never smokers was not brought to the attention of the monograph readers. Thus, the fact that the working group at the IARC meeting did not critically assess weaknesses of the DEMS should not be given much weight. Of note, a well designed re-analyses of the DEMS cohort and case-control studies may return results different from those reported by the IARC working group and in the original papers [3, 4]. Such complementary research is in the planning stage and it may help to understand the important DEMS project in more detail.
References
1. Silverman DT, Attfield MD. RE: "Diesel exhaust in miners study: how to understand the findings?" by Peter Morfeld. J Occup Med Toxicol 2013:http://www.occup-med.com/content/7/1/10/comments (13th March 2013, date last accessed).
2. Morfeld P. Diesel exhaust in miners study: how to understand the findings? J Occup Med Toxicol 2012;7(1):http://www.occup-med.com/content/7/1/10 (13th March 2013, date last accessed).
3. Attfield M, D., Schleiff PL, Lubin JH, Blair A, Stewart PA, Vermeulen R, Coble JB, Silverman DT. The diesel exhaust in miners study: a cohort mortality study with emphasis on lung cancer. J Natl Cancer Inst 2012;104(11):869-83.
4. Silverman DT, Samanic CM, Lubin J, H., Blair AE, Stewart PA, Vermeulen R, Coble JB, Rothman N, Schleiff PL, Travis WD, et al. The diesel exhaust in miners study: A nested case-control study of lung cancer and diesel exhaust. J Natl Cancer Inst 2012;104(11):855-68.
5. Möhner M. To the editor: The impact of selection bias due to increasing response rates among population controls in occupational case-control studies. Am J Respir Crit Care Med 2012;185(1):104-6.
6. Pallapies D, Taeger D, Bochmann F, Morfeld P. Comment: Carcinogenicity of diesel-engine exhaust (DE). Arch Toxicol 2013;87(3):547-9.
7. Hauptmann M, Lubin JH, Stewart PA, Hayes RB, Blair A. Mortality from lymphohematopoietic malignancies among workers in formaldehyde industries. J Natl Cancer Inst 2003;95(21):1615-23.
8. Beane Freeman LE, Blair A, Lubin JH, Stewart PA, Hayes RB, Hoover RN, Hauptmann M. Mortality from lymphohematopoietic malignancies among workers in formaldehyde industries: the national cancer institute cohort. J Natl Cancer Inst 2009;101(10):751-61.
9. Marsh GM, Youk AO, Morfeld P, Collins JJ, Symons JM. Incomplete follow-up in the National Cancer Institute's formaldehyde worker study and the impact on subsequent reanalyses and causal evaluations. Regul Toxicol Pharmacol 2010;58:233-6.
10. Vermeulen R, Coble JB, Yereb D, Lubin JH, Blair A, Portengen L, Stewart PA, Attfield M, Silverman DT. The diesel exhaust in miners study: III. Interrelations between respirable elemental carbon and gaseous and particulate components of diesel exhaust derived from area sampling in underground non-metal mining facilities. Ann Occup Hyg 2010;54(7):762-73.
11. Crump K, Van Landingham C. Evaluation of an exposure assessment used in epidemiological studies of diesel exhaust and lung cancer in underground mines. Crit Rev Toxicol 2012;42(7):599-612.
12. Möhner M, Kersten N, Gellissen J. Diesel motor exhaust and lung cancer mortality: reanalysis of a cohort study in potash miners. Eur J Epidemiol 2013:http://rd.springer.com/article/10.1007/s10654-013-9784-0 (13th March 2013, date last accessed).
13. Benbrahim-Tallaa L, Baan RA, Grosse Y, Lauby-Secretan B, El Ghissassi F, Bouvard V, Guha N, Loomis D, Straif K, International Agency for Research on Cancer Monograph Working G. Carcinogenicity of diesel-engine and gasoline-engine exhausts and some nitroarenes. Lancet Oncol 2012;13(7):663-4.
14. Gamble JF. IARC evaluations of cancer hazards: Comment on the process with specific examples from volume 105 on diesel engine exhaust. J Clinic Toxicol 2012;2:http://www.omicsonline.org/2161-0495/2161-0495-2-e106.digital/2161-0495-2-e106.html (13th March 2013, date last accessed).
15. IARC. Carbon black, titanium dioxide, and talc. Lyon: International Agency for Research on Cancer; 2010.
16. Olsson AC, Gustavsson P, Kromhout H, Peters S, Vermeulen R, Bruske I, Pesch B, Siemiatycki J, Pintos J, Bruning T, et al. Exposure to diesel motor exhaust and lung cancer risk in a pooled analysis from case-control studies in Europe and Canada. Am J Respir Crit Care Med 2011;183(7):941-8.
17. Lash TL, Fox MP, Fink AK. Applying quantitative bias analysis to epidemiologic data. Dordrecht: Springer Science+Business Media, LLC; 2009.
Competing interests
The author is member of the research committee and scientific advisory group of EUGT (http://www.eugt.org/) and received research grants from EUGT for a project on the effectivity of low-emission zones.
Journal of Occupational Medicine and Toxicology
Letter-to-the-Editor RE: 'Diesel exhaust in miners study: how to understand the findings?' by Peter Morfeld
5 March 2013
We thank Dr. Morfeld for providing us the opportunity to clarify several points regarding the analyses of data from both the cohort (1) and case-control (2) components of the Diesel Exhaust in Miners Study (DEMS).
First, we agree with Dr. Morfeld's statement "The distinction between surface and underground work is obviously of major importance for an understanding of this study." In fact, to ignore location and simply estimate risk by exposure and adjusting for smoking would have led to erroneous results. However, we do not agree with Dr. Morfeld's statement "this factor `surface/underground work' remains unexplained." The smoking effect among surface-only workers shown in Table 2 (2) is similar to that observed in previous cohort studies of smoking and lung cancer (3), whereas the smoking effect among underground workers who smoke at least 1 pack per day is attenuated. As shown in Table 2 (2), after adjustment for diesel exposure, the risk among underground nonsmoking workers is virtually identical to that among surface-only nonsmoking workers (OR for 15-year lagged cumulative REC = 0.90), providing evidence that the observed difference in risk between surface only and underground workers is explained by smoking and diesel exposure. When smoking and location are included as separate variables in the model, estimates of risk are quite similar to those in Table 3 (2) (e.g., ORs for quartiles of 15-year lagged cumulative REC: 1.0 (referent), 0.87 vs. 0.74, 1.60 vs. 1.54, 2.97 vs. 2.83, respectively) with the P for trend = 0.001 in both models. Because the interaction between smoking and location was borderline significant, we included a cross-product term in the models to fully capture this interaction. The inclusion of the cross-product term had a negligible impact on the estimates of risk, however. Dr. Morfeld also misquotes us as hypothesizing that "high REC exposures are protective against lung cancer excess risks due to smoking." We wrote that the effect of smoking is "attenuated" in the presence of high levels of diesel exposure, which is not the same as "protection" since smoking does indeed cause lung cancer among underground workers who were heavily exposed to diesel exhaust.
Second, Dr. Morfeld questions whether we have a systematic underestimation of lung cancer deaths, especially at lower diesel exhaust exposure levels. The mortality follow-up was undertaken by NIOSH using typical and standard methodology and data from the National Death Index, Social Security Administration, and Internal Revenue Service. Other sources, such as postmasters, tracing agencies, and company records were also employed. Overall, the information from the various sources was internally very consistent. Third, Dr. Morfeld goes on to focus on the external vs. internal analysis results, finding it `perplexing' that they should differ. In response, we note that it is well established that external analysis is subject to factors including the healthy worker effect that are better addressed through internal analysis. In addition, as noted in the cohort paper (1), different patterns of lung cancer mortality by location, which were shown by the case-control analysis (2) to be caused by smoking, obscured the exposure-response relationship evident in the complete cohort. Clear exposure-response patterns were identified after stratification by location (Tables 4 and 5 of the cohort paper (1)), and were also observed after adjustment for location (Table 6 of the cohort paper (1)). Fourth, we were unable to comprehend what Dr. Morfeld is recommending with respect to analysis by worker location. We employed a time dependent stratification approach (e.g., a worker was surface-only until the first time they went underground, at which time they became an ever-underground worker). We cannot imagine that Dr. Morfeld was advocating using underground tenure as a primary exposure variable rather than using the available quantitative exposure data.
Fifth, it is generally accepted that retrospective exposure assessments in occupational studies such as DEMS have some imprecision (4). If, as is likely, the error is nondifferential, the exposure-response slopes may be attenuated. Considerable effort went into evaluating the exposure assessment in DEMS, including a separate paper on this topic (5). We also developed a number of alternative exposure metrics. These are discussed in the cohort paper (1) and were found not to impact the findings in any significant way.
Dr. Morfeld asks several detailed questions concerning our methods. In response: 1) We did not censor or otherwise treat deaths over age 85 differently from younger deaths (but only 3 of 200 lung cancer deaths in the cohort were >85 years and all available pathology reports for the lung cancer cases were evaluated by an expert pathologist to confirm lung cancer as the cause of death). 2) In the cohort analysis, all confounder variables were entered into the model together and each examined for their apparent influence on lung cancer mortality. In the case-control study, each potential confounder was added one-at-time to the conditional model containing an exposure metric, the cross product of smoking and location, employment in a high-risk occupation for at least 10 years, and a history of nonmalignant respiratory disease for at least 5 years to create a base model. We also ran full models that simultaneously included the base model plus the other potential confounders. 3) In the cohort analysis, observations with missing data for a particular analysis were dropped; in the case-control analysis, missing confounder data were included as a separate level. When we evaluated this approach by excluding missing confounder data from key analyses, observed patterns of risk were similar to those when missing data were included as a separate level. 4) Although workers were not permitted to smoke underground in the trona mines, patterns in exposure-response were similar by mine type, suggesting that differential smoking patterns across facilities did not impact the findings. 5) In the cohort analysis, there was no overall effect of dropping those whose age at first exposure was greater than 40 (the cumulative REC exposure hazard ratio dropped slightly while that for exposure intensity increased).
Despite the complexities of the cohort and case-control analyses, there is no doubt that excess risks of lung cancer were detected, and that these elevations were associated with increasing levels of diesel exhaust. We do not accept that our methods suffer from errors so egregious that the findings could be spurious. In fact, the scientific community seems to have had no difficulty in accepting our results, as reflected by the recent IARC review of the carcinogenicity of diesel exhaust, in which our findings played an important role in the evaluation process (6).
Debra T. Silverman
Michael D. Attfield
References
1. Attfield MD, Schleiff PL, Lubin JH, Blair A, Stewart PA, Vermeulen R, Coble JB, Silverman DT. The Diesel Exhaust in Miners Study: A cohort mortality study with emphasis on lung cancer. J Natl Cancer Inst 2012;104(11):869-883.
2. Silverman DT, Samanic CM, Lubin JH, Blair AE, Stewart PA, Vermeulen R, Coble JB, Rothman N, Schleiff PL, Travis WD, Ziegler RG, Wacholder S, Attfield MD. The Diesel Exhaust in Miners Study: A nested case-control study of lung cancer and diesel exhaust. J Natl Cancer Inst 2012;104(11):855-868.
3. Blot WJ, Fraumeni JF, Jr. Lung and pleura. In: Schottenfeld D, Fraumeni JF Jr. eds. Cancer Epidemiology and Prevention. 2nd Ed. New York: Oxford University Press, 1996:637-665.
4. Stewart PA, Coble JB, Vermeulen R, Blair A, Lubin J, Attfield M, Silverman DT. Response to Borak et al., 2010 on the Diesel Exhaust in Miners Study. Ann Occup Hyg 2011;55(3): 343-346.
5. Stewart PA, Vermeulen R, Coble JB, Blair A, Schleiff PL, Lubin JH, Attfield M, Silverman DT. The Diesel Exhaust in Miners Study: V. Evaluation of the exposure assessment methods. Ann Occup Hyg 2012 Mar 1 Epub.
6. Lamia Benbrahim-Tallaa, Robert A Baan, Yann Grosse, Beatrice Lauby-Secretan, Fatiha El Ghissassi, Veronique Bouvard, Neela Guha, Dana Loomis, Kurt Straif. Carcinogenicity of diesel-engine and gasoline-engine exhausts and some nitroarenes. Lancet Oncol 2012;13(7):663-664.
Affiliations of authors: Occupational and Environmental Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD (DTS); Surveillance Branch, Division of Respiratory Disease Studies, National Institute for Occupational Safety and Health, Morgantown, WV (MDA).
Competing interests
None
Author reply
23 April 2013
I'd like to thank Drs Silverman and Attfield for their profound reply [1] to my commentary [2] and their in-depth explanations of specific aspects of the Diesel Exhaust in Miners Study [DEMS, 3, 4]. It was my motivation to instigate discussions about this important and impressive epidemiological project. Thus, I highly appreciate that the leading DEMS authors responded to my commentary in such a great detail. The Editors of JOMT invited me to follow-up on this discussion and I will respond to the points as they were numerated in Silverman and Attfield [1].
1) The DEMS findings about the effect of long-term respirable elemental carbon (REC) exposure estimates on lung cancer mortality depended critically on the adjustment by location (surface vs. underground work). Cox models that followed the analysis plan did not show any convincing association between REC exposure estimates and lung cancer mortality. Only after subdividing the cohort into 'surface only workers' vs. 'ever underground workers' or after adjusting for a 'surface only/ever underground work' indicator the Cox models returned pronounced dose/response relationships with REC exposure estimates. Silverman and Attfield [1] agreed. They confirmed that the distinction between surface and underground work is of major importance for an understanding of DEMS. Of note, they did not agree with my statement that "this factor `surface/underground work' remains unexplained." They argued that "the observed difference in risk between surface only and underground workers is explained by smoking and diesel exposure". Unfortunately, this statement appears to be wrong. If their statement were true we expected to see that the location variable could be substituted by a combination of the smoking and exposure variables without affecting the REC exposure estimates. Thus, models that follow the original study protocol and included no location factor but smoking information should return large REC exposure effect estimates. In addition, the location variable should show a clearly reduced and non-significant effect in models after adjusting for smoking and exposure. Based on the DEMS papers, my understanding is that this is not true. If Drs. Silverman and Attfield [1] still disagree I'd like to ask them to reveal a) the REC exposure coefficients (with CIs and p-values) returned by conditional logistic regression with and without adjustment for location, while always controlling for smoking and b) to show the location coefficient (with CI and p-value) after adjustment for smoking and REC exposure (smoking and location variables should be entered into the models as separate variables so that we can distinguish their effects on lung cancer mortality).
Silverman and Attfield [1] made an observation that they interpreted as evidence for identical risks among surface and underground workers after appropriate adjustment:: "As shown in Table 2 [2], after adjustment for diesel exposure, the risk among underground nonsmoking workers is virtually identical to that among surface-only nonsmoking workers (OR for 15-year lagged cumulative REC = 0.90), providing evidence that the observed difference in risk between surface only and underground workers is explained by smoking and diesel exposure." However, the 95%-confidence interval of the cited odds ratio (OR) spanned from 0.26 to 3.09. This can hardly be interpreted as evidence in favor of the null hypothesis.
There are other interpretations of the DEMS authors about smoking that are difficult to follow. Möhner [5] noted in his Letter to the Editor: "According to the authors, there was no evidence of an increased smoking prevalence in those employees working underground. Yet a comparison of the smoking data from controls by location of employment (Table 2 [2]) yielded a significantly higher percentage of never smokers and significantly lower percentage of heavy smokers (>2 packs per day) among surface-only workers than among ever-underground workers (34.2% vs. 22.1% and 6.3% vs. 13.6%, respectively). Although underground workers smoked significantly more, were more burdened with exposure to other occupational lung carcinogens (see for example standardized mortality ratios (SMR) for pneumoconiosis, Supplementary Table 2 [1]) and were much more exposed to DME than surface workers, their lung cancer mortality is lower. We would be grateful to the authors if they could provide readers with additional data so that the studies' conclusions can be comprehended. In particular, we would like to know, how the SMR reflect differences by exposure parameters." The authors did not reply to this. Silverman et al. [4] wrote: "a comparison of confounder data derived directly from living and from next of kin for deceased control subjects revealed comparability of responses". This statement appears to be odd when applied to smoking. Pallapies et al. [6] noted: "Silverman et al. [4] obtained smoking data in cases from next of kin only. However, they obtained such data from living control subjects as well as from next of kin of other control subjects. This may have led to considerable bias, as different percentages of 'current smokers' from direct versus next of kin interviews (11 vs. 23 %) demonstrate".
In this respect I like to repeat point 4 in Morfeld [2] that remained unanswered: "The REC exposure risk estimates differed with location [1]: the authors reported a twenty times higher excess risk per g/m3-y on 'surface only' in comparison to 'ever underground'. After taking logs of exposure the 'surface only' REC coefficient was about twice the one calculated for workers 'ever underground'. The researchers [1] tested the differences of the effect estimates (significant on the log scale, not significant on the linear scale) but neither reported how they performed the tests nor did they show any details of the results. The usual way to perform such a test is to add an interaction term 'REC exposure x location' to the models. It is surprising that the authors [1,2] did not present such interaction models.
Silverman and Attfield [1] wrote that I misquoted them as hypothesizing that "high REC exposures are protective against lung cancer excess risks due to smoking." I am not sure whether this comment is on substance or on semantics. I just wanted to make clear that Silverman et al. [4] listed mechanisms that they believe could be true and may reduce lung cancer risk from smoking because of elevated diesel motor emission exposures, e.g., "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]pyrene, possibly through inhibition of enzymes that carry out its metabolic activation [25]."
2) I'd like to thank Drs Silverman and Attfield [4] that they presented the ORs when not adjusting for an interaction of location and smoking. This analysis demonstrated that an inclusion of the cross-product term was not necessary to estimate the effect of REC exposure. Unfortunately, they did not show other results I asked for: ORs for REC exposure after controlling for only those variables used in SMR calculations, and ORs for smoking in underground and surface workers without adjustment for REC exposure [2] Furthermore, like Möhner [5] I showed interest in lung cancer SMRs across REC exposure categories to judge whether the study has pronounced deficits of lung cancer mortality in low categories of exposure and whether there are any significant SMR excesses in the highest exposure categories. Thus, the authors did not present data to rule out my hypothesis that the lung cancer mortality structure is strange and similar to the leukemia pattern observed in NCI's formaldehyde cohort [7, 8]. Silverman and Attfield [1] confirmed that their mortality follow-up procedure was the same or very similar to the method applied in the NCI formaldehyde cohort. Of note, NCI formaldehyde researchers had missed about 1000 deaths out of 9500 which led to distorted exposure-response findings [9]. Silverman and Attfield [1] did present no arguments that DEMS may not suffer from the same problems.
3) Silverman and Attfield [1] wrote: "Dr. Morfeld goes on to focus on the external vs. internal analysis results, finding it `perplexing' that they should differ. In response, we note that it is well established that external analysis is subject to factors including the healthy worker effect that are better addressed through internal analysis." I'd like to emphasize that I was not talking about exposure. I found it perplexing (and still find it perplexing) that such a large location effect popped up after adjusting for REC exposure and went unnoticed otherwise. The idea of Silverman and Attfield [1] that this observation may be associated to the healthy worker effect appears to be off the track.
4) Silverman and Attfield [1] "were unable to comprehend what Dr. Morfeld is recommending with respect to analysis by worker location. We employed a time dependent stratification approach (e.g., a worker was surface-only until the first time they went underground, at which time they became an ever-underground worker)." I will try to explain it differently. The binary variable 'location', as defined and used in the DEMS analysis, does not tell us for each person-year whether the miner was underground or on surface. Here is why: Assume that a miner worked underground up to year x but then changed to a surface job and stayed on surface until year x+s. Then the variable 'location', as defined and used in the DEMS analysis, is set to 'underground' during the years x to x+s although the miner was on surface. In contrast to what had been done in DEMS I recommended and I still recommend that the variable 'location' should correctly reflect for each person-year whether the miner was underground or on surface. According to the description of the exposure assessment process a REC estimate was allocated to every person-year. And because REC exposures differed between surface and underground work a reliable exposure assessment should take account of workers' location in every person- year. Thus, the variable 'location' can be defined in the suggested way (if not, the DEMS exposure estimates are obviously inaccurate). Note that it is improbable that in this large cohort miners only changed from surface to underground work but never vice versa. Because the variable 'location' is so important in DEMS this variable should be defined in a better way than Silverman and Attfield [1] did.
Silverman and Attfield [1] wrote "We cannot imagine that Dr. Morfeld was advocating using underground tenure as a primary exposure variable rather than using the available quantitative exposure data." I never recommended using underground tenure as a primary exposure variable. This appears to be a misunderstanding.
5) Silverman and Attfield [1] wrote: "it is generally accepted that retrospective exposure assessments in occupational studies such as DEMS have some imprecision [4]. If, as is likely, the error is non-differential, the exposure-response slopes may be attenuated. Considerable effort went into evaluating the exposure assessment in DEMS, including a separate paper on this topic [5]. We also developed a number of alternative exposure metrics. These are discussed in the cohort paper [1] and were found not to impact the findings in any significant way." This does not justify taking the exposure data as fixed and assuming 'no exposure errors' when modeling risks. Monte Carlo sensitivity and Bayesian bias analyses are well developed and these procedures are able to cover the involved uncertainty. I will provide two examples that may help to understand what kind of potential problems exist in DEMS. The REC exposure assessment was mainly based on a correlation between CO and REC measurements [10]. The authors admittted "that the observed coefficient derived from this cross-sectional study might not apply longitudinally to past conditions." This induces problems because the DEMS analysis relies on life-long cumulative exposures. The relevance of such uncertainties can be measured by appropriate bias analyses. Moreover, the coefficient chosen by DEMS researchers to transfer CO values into REC values was not supported by the data [11]. Thus, this may have distorted the risk estimates. Again, this motivates to perform bias analyses.
6) I'd like to thank Drs. Silverman and Attfield [1] that they were open to clarify even some of the minor issues that I raised (sub-issues 1 to 5 in Silverman and Attfield [1]). Sub-issue 3: I am not clear whether the authors wanted to say that the case-control study was analyzed twice (two different ways to deal with missing data)? Sub-issue 4: I had interest in the smoking patterns. Can the patterns be presented? One issue remained without response: the suggestion of a curvilinear modeling.
I'd like to add that the paper describing the re-analysis of the German potash miner cohort was accepted for publication [12]. Silverman et al. [4] stated:"We observed an increased lung cancer risk associated with diesel exposure as was seen among German potash miners [11]." This statement is no longer evident because of the null results reported in Möhner et al. [12] after taking account of prior exposures in Uranium mining.
Final remark
Silverman and Attfield [1] wrote as a final comment: "Despite the complexities of the cohort and case-control analyses, there is no doubt that excess risks of lung cancer were detected, and that these elevations were associated with increasing levels of diesel exhaust. We do not accept that our methods suffer from errors so egregious that the findings could be spurious. In fact, the scientific community seems to have had no difficulty in accepting our results, as reflected by the recent IARC review of the carcinogenicity of diesel exhaust, in which our findings played an important role in the evaluation process". I do have to insist that pronounced and clear-cut lung cancer excess risks were only observed after adjustment for location. Anyhow, Silverman and Attfield [1] were right when they reported on the evaluation of a working group who met at IARC, Lyon in June 2012 [13]. I attended that meeting as an observer. Indeed, the group of invited experts took the exceptionally large REC exposure estimates of the DEMS for granted without mentioning the potential problems involved. I have to admit that the frustrating commentary published by another observer is appropriately describing the downsides of the evaluation process at this IARC monograph meeting [14]. John Gamble reminded the reader that the Preamble of the IARC monographs [e.g., 15] defines clearly what the working group should do: "When an important aspect of a study that directly impinges on its interpretation should be brought to the attention of the reader, a Working Group comment is given in square brackets". Regarding DEMS the working group had no square brackets added to mention the discussed weaknesses of the study. They did so although the weaknesses were brought to their attention, even in published form [2]. Another striking example is linked to the European-Canadian pooled case-control study [16]. In the abstract the authors described a positive association between diesel exhaust exposure and lung cancer risk in a smoking-adjusted analysis. The working group repeated these statements of the authors and noted in a square bracket that this finding is unlikely to ["be explained by bias or confounding"]. Tim Lash, another observer and well-known for his high-level work on bias adjustment in epidemiological studies [e.g., 17], commented on this square bracket during the epidemiological subgroup meetings and also in the plenary meeting of the full working group. Lash noted that this pooled case-control study was very large and one of the rare investigations that studied lung cancer risks among never smokers. He made clear that a lack of association was found among never smokers, possibly the group that could provide the least confounded risk estimates after Diesel exhaust exposure [Table 3, 16]. Contrary to the statement of the working group Lash explained that the positive result after smoking adjustment could be due to residual confounding, in particular in the light of the null finding within never smokers. Again, this comment went unnoticed by the working group. The square bracket statement about Olsson et al. [16] was not changed and this important information about never smokers was not brought to the attention of the monograph readers. Thus, the fact that the working group at the IARC meeting did not critically assess weaknesses of the DEMS should not be given much weight. Of note, a well designed re-analyses of the DEMS cohort and case-control studies may return results different from those reported by the IARC working group and in the original papers [3, 4]. Such complementary research is in the planning stage and it may help to understand the important DEMS project in more detail.
References
1. Silverman DT, Attfield MD. RE: "Diesel exhaust in miners study: how to understand the findings?" by Peter Morfeld. J Occup Med Toxicol 2013:http://www.occup-med.com/content/7/1/10/comments (13th March 2013, date last accessed).
2. Morfeld P. Diesel exhaust in miners study: how to understand the findings? J Occup Med Toxicol 2012;7(1):http://www.occup-med.com/content/7/1/10 (13th March 2013, date last accessed).
3. Attfield M, D., Schleiff PL, Lubin JH, Blair A, Stewart PA, Vermeulen R, Coble JB, Silverman DT. The diesel exhaust in miners study: a cohort mortality study with emphasis on lung cancer. J Natl Cancer Inst 2012;104(11):869-83.
4. Silverman DT, Samanic CM, Lubin J, H., Blair AE, Stewart PA, Vermeulen R, Coble JB, Rothman N, Schleiff PL, Travis WD, et al. The diesel exhaust in miners study: A nested case-control study of lung cancer and diesel exhaust. J Natl Cancer Inst 2012;104(11):855-68.
5. Möhner M. To the editor: The impact of selection bias due to increasing response rates among population controls in occupational case-control studies. Am J Respir Crit Care Med 2012;185(1):104-6.
6. Pallapies D, Taeger D, Bochmann F, Morfeld P. Comment: Carcinogenicity of diesel-engine exhaust (DE). Arch Toxicol 2013;87(3):547-9.
7. Hauptmann M, Lubin JH, Stewart PA, Hayes RB, Blair A. Mortality from lymphohematopoietic malignancies among workers in formaldehyde industries. J Natl Cancer Inst 2003;95(21):1615-23.
8. Beane Freeman LE, Blair A, Lubin JH, Stewart PA, Hayes RB, Hoover RN, Hauptmann M. Mortality from lymphohematopoietic malignancies among workers in formaldehyde industries: the national cancer institute cohort. J Natl Cancer Inst 2009;101(10):751-61.
9. Marsh GM, Youk AO, Morfeld P, Collins JJ, Symons JM. Incomplete follow-up in the National Cancer Institute's formaldehyde worker study and the impact on subsequent reanalyses and causal evaluations. Regul Toxicol Pharmacol 2010;58:233-6.
10. Vermeulen R, Coble JB, Yereb D, Lubin JH, Blair A, Portengen L, Stewart PA, Attfield M, Silverman DT. The diesel exhaust in miners study: III. Interrelations between respirable elemental carbon and gaseous and particulate components of diesel exhaust derived from area sampling in underground non-metal mining facilities. Ann Occup Hyg 2010;54(7):762-73.
11. Crump K, Van Landingham C. Evaluation of an exposure assessment used in epidemiological studies of diesel exhaust and lung cancer in underground mines. Crit Rev Toxicol 2012;42(7):599-612.
12. Möhner M, Kersten N, Gellissen J. Diesel motor exhaust and lung cancer mortality: reanalysis of a cohort study in potash miners. Eur J Epidemiol 2013:http://rd.springer.com/article/10.1007/s10654-013-9784-0 (13th March 2013, date last accessed).
13. Benbrahim-Tallaa L, Baan RA, Grosse Y, Lauby-Secretan B, El Ghissassi F, Bouvard V, Guha N, Loomis D, Straif K, International Agency for Research on Cancer Monograph Working G. Carcinogenicity of diesel-engine and gasoline-engine exhausts and some nitroarenes. Lancet Oncol 2012;13(7):663-4.
14. Gamble JF. IARC evaluations of cancer hazards: Comment on the process with specific examples from volume 105 on diesel engine exhaust. J Clinic Toxicol 2012;2:http://www.omicsonline.org/2161-0495/2161-0495-2-e106.digital/2161-0495-2-e106.html (13th March 2013, date last accessed).
15. IARC. Carbon black, titanium dioxide, and talc. Lyon: International Agency for Research on Cancer; 2010.
16. Olsson AC, Gustavsson P, Kromhout H, Peters S, Vermeulen R, Bruske I, Pesch B, Siemiatycki J, Pintos J, Bruning T, et al. Exposure to diesel motor exhaust and lung cancer risk in a pooled analysis from case-control studies in Europe and Canada. Am J Respir Crit Care Med 2011;183(7):941-8.
17. Lash TL, Fox MP, Fink AK. Applying quantitative bias analysis to epidemiologic data. Dordrecht: Springer Science+Business Media, LLC; 2009.
Competing interests
The author is member of the research committee and scientific advisory group of EUGT (http://www.eugt.org/) and received research grants from EUGT for a project on the effectivity of low-emission zones.