Lung cancer and diesel exhaust: an updated critical review of the occupational epidemiology literature

Figures & data

Figure 1.  Exposure-response of lung cancer and cumulative DME exposure among all cases and controls, women, non-smokers, those without working in jobs with known lung cancer risk; ORs adjusted for age, sex, study, ever employment in list “A” jobs, pack-years, time since quitting smoking (Olsson et al., 2011).

Table

N	Cases	Controls
	13,304	16,283
Average participations rates	82% (68–98)	67% (41–100)
% Non-smokers	6%	29%
% Former smokers	29%	39%
% Current smokers	65%	32%
% Exposed to diesel exhaust	42%	37%
% Employed in other high risk jobs	12%	8%

Figure 2.  Lung cancer risk by years worked among workers only exposed to low levels and high levels of diesel motor exhaust exposure (Olsson et al., 2011).

Table 1.  Proportion of workers exposed to diesel emissions in selected occupations and usual exposure coding, Montreal, Canada, 1979–1985 (Parent et al., 2007).

	% diesel-exposed	Diesel emissions
		Usual exposure coding
		Confidence*	Concentration*	Frequency^†
Motor transport workers	37	3	1	3
Bus drivers	91	3	2	1
Taxi drivers & chauffeurs	0		1 (1)	3 (3)
Truck drivers (heavy trucks)	39 (54)
Mechanics	28	3	2	2
Motor vehicle & repairers	29	3	2 or 3	3
Industrial trucks	0
Salesmen	1	2	2	1
Commercial travelers	0
Route drivers	0
Railway Transport Workers	72	3	1	3
Locomotive operators	25	3	1	3
Conductors & brake workers	82	3	1	3
Excavators and pavers	56	2 or 3	2	2
Excavating, grading & related occupations	95	2	2	2
Miners and quarrymen	79	2	2	3
Firefighters	95	3	2	2
Dockworkers	27	3	2	2

*Concentration and confidence levels: 1 = low, 2 = medium, 3 = high.

^†Frequency levels (% of normal workweek):1 = <5%, 2 = 5–30%, 3 = >30.

Table 2.  Comparison of ORs based on different DME exposure assessment and different relative scores for low and high exposure jobs (0 to 4) in the Peter et al. (2011) study of INCO countries.

Country	Expert assessment	Population-specific JEM	DOM-JEM-INCO	DOM-JEM pooled highest quartile
Intensity indices
Non-exposed	0	0 (<33% exposed)	0	0
Low	1	1	1	1
Medium	2	2	2	4
High	3	3
ORs: country (n)
Czech Repl (285)	2.12 (1.41–3.19)	1.85 (1.24–2.78)	1.89 (1.31–2.73)	1.16(0.64–2.13)
Hungary (361)	0.81 (0.56–1.18)	1.07 (0.74–1.56)	0.88 (0.61–1.27)	1.27 (0.76–2.11)
Poland (539)	1.18 (0.56–1.37)	1.20 (0.89–1.64)	1.15 (0.87–1.53)	1.77 (1.08–2.90)
Romania (181)	0.62 (0.35–1.09)	0.68 (0.40–1.16)	1.08 (0.64–1.82)	0.99 (0.40–2.48)
Russia (506)	0.89 (0.66–1.22)	0.87 (0.65–1.16)	0.74 (0.56–0.98)	1.17 (0.74–1.86)
Slovakia (346)	1.27 (0.84–1.93)	1.07 (0.72–1.59)	1.43 (0.97–2.12)	1.60 (0.80–3.18)
UK (192)	1.00 (0.61–1.63)	0.71 (0.43–1.18)	0.65 (0.92–1.08)	0.93 (0.59–1.46)

Figure 3.  Adjusted odds ratios of lung cancer in relation to occupational exposure to diesel engine emissions with 5-year latency and men ≥40 years; partial adjustment for confounders is age and province; full adjustments for confounders is age, province, pack-years, second-hand smoke, silica and asbestos (Villeneuve et al., 2011).

Table 3.  Percentage of job periods exposed and risk estimates for lung cancer between three methods of exposure assessment in the INCO studies in Czech Republic, Hungary, Poland, Romania, Russia, Slovakia, UK from Peters et al. (2011).

	Expert assessment Peters et al. (2011)	Population-specific GEM Peters et al. (2011)	DOM-JEM (Peters et al. 2011) (used in pooled analysis)	DOM-JEM (Olsson et al. 2011)
Indices for rating intensity	0 = none; 1 = low; 2 = medium; 3 = high	0 = <33% exposed; exposed = >33% exposed to intensity ≥1	0 = non-exposed; 1 = low exposure; 2 = high exposure	0 = non-exposed, 1 = low; 4 = high
Asbestos
% exposed (range)	7% (4–21%)	8% (3–20%)	26% (16–35%)
OR (95% CI) (range)	1.12 (0.93–1.37) (0.77–2.23)	1.04 (0.86–1.26) (0.45–1.49)	1.19 (1.05–1.36) (0.96–1.5)
Silica
% exposed (range)	26% (10–54%)	31% (8–69%)	11% (4–26%)
OR (95% CI) (range)	1.26 (1.10–1.44) (1.00–1.76)	1.24 (1.08–1.43) (0.89–2.11)	1.26 (1.08–1.47) (1.10–1.83)
DME
% exposed (range)	16% (9–27%)	19% (12–28%)	22% (12–33%)
OR (95% CI) (range)	1.08 (0.94–1.25) (0.62–2.12)	1.05 (0.91–1.21) (0.71–1.85)	1.05 (0.92–1.20) (0.65–1.89)

Table

	Exposure Quartiles
	Unexposed	1	2	3	4
OR (95% CI)	1.0	0.74(0.52–1.05)	1.22 (0.90–1.65)	0.85 (0.57–1.26)	1.26 (0.90–1.79)

Table 4.  Associations of educational attainment with exposure to DME (Mohner et al., 1998).

Educational attainment	% exposure to DME	% of controls	Expected % controls (based distribution in general population)
No formal vocational training	24.7	10.0	16.9
Finished vocational training	13.0	68.6	71.5
University degree	3.7	21.4	11.6

Figure 4.  Adjusted odds ratios of lung cancer in relation to occupational exposures to diesel emissions, men 40+ years Canadian hospital-based case-control study (OR adjusted for age, province, pack-years, second-hand smoke, silica (Y/N), asbestos(Y/N)(p values refer to cumulative exposure) (Villeneuve et al., 2011).

Figure 5.  Adjusted odds ratios of lung cancer in relation to occupational exposures to gasoline emissions, men 40+ from Canadian hospital-based case-control study; ORs adjusted for age, province, pack-years, second-hand smoke, silica (Y/N) and asbestos (Y/N) (p values refer to cumulative exposure) (Villeneuve et al., 2011).

Figure 6.  Odds ratios for the association of lung cancer and exposure to diesel emissions in population-based case-control studies in Europe and Canada (Olsson et al., 2011) including subsets from Italy (Richiardi et al., 2006), Stockholm (Gustavvson et al., 2000) and Germany (Bruske-Hohlfeld et al., 2000), Canada (Villeneuve et al., 2011); Sweden (Boffetta et al., 2001); Finland (Guo et al., 2004), and Montreal, Canada (Parent et al., 2007).

Figure 7.  Odds ratios for the association of qualitative estimates of diesel exhaust exposure with all lung cancers and by histologic types from population-based case-control studies in Montreal, Canada (Parent et al., 2007), Sweden (Boffetta et al., 2001), Finland (Guo et al., 2004) and Canada (Villeneuve et al., 2011).

Table 5.  Summary of exposure-response slopes and HRs for surface and underground (UG) workers using untransformed log-linear and log transformed regression models over full exposure range and restricted exposure range (<1280 µg/m3 years) for UG cohort for workers with 15-year lags and excluding workers with <5-year tenure (Attfield et al., 2012).

	HR slope (95% CI) per µg/m^3 years, HR at highest µg/m^3 years in exposure range
	Surface	UG	UG
Exposure range (µg/m^3 years)	0–160	0–2000	<1280
Untransformed log-linear model	1.02 (1.0–1.03), 23.8	1.0001 (0.99–1.0003), 1.14	1.0014 (1.0007–1.002), 6.01
Log transformed	1.03 (0.75–1.4), 1.07	1.19 (1.04–1.37), 1.78	–

Primary results for authors’ conclusions.

Figure 8.  Proportional Hazards ratios on lung cancer mortality for 15-year lagged REC cumulative exposure of surface worker, UG workers and the complete cohort adjusted and unadjusted for worker location in DEMS cohort study (Tables 4, 5, 6 in Attfield et al., 2012).

Figure 9.  Proportional hazard ratios (HR) on lung cancer mortality for 15-year lagged REC cumulative exposure of surface workers, UG workers and complete cohort expanded categories excluding workers with ≤5-years tenure, Tables 4, 5, 6 in Attfield et al. (2012).

Figure 10.  Proportional hazard ratios for lung cancer and cumulative REC among surface workers with 15-year lags and exclude workers with <5-year tenure in expanded categories, untransformed and log transformed regression models, Tables 5 and S8 from Attfield et al. (2012).

Figure 11.  Proportional hazard ratios (HR) on lung cancer mortality for 15-year lagged REC cumulative exposure in UG workers excluding workers <5-years tenure; Expanded categorical model plus log-linear regressions for full and restricted exposure range <1280 μg/m³-years; log transformed model with full exposure range (Table 4) (Attfield et al., 2012).

Table

	15-year lags		Unlagged
HR at penultimate Exposure category	Exclude <5-year tenure	No exclusions	Exclude <5-year tenure	No exclusions
	5.01	2.42	4.09	2.62

Figure 12.  Proportional hazards ratios (HR) for lung cancer by 15-year lagged and unlagged REC cumulative exposure (µg/m³-years) for the complete cohort adjusted for worker location using expanded categorical cutpoints (Tables S5 and S6) (Attfield et al., 2012).

Figure 13.  Proportional hazards ratios for lung cancer mortality by unlagged REC cumulative exposure (µg/m³-years) among Underground workers with expanded categories, untransformed (HR = 1.01) and log transformed regression (HR = 1.15) models for UG workers from Tables S6 and S8, Attfield et al. (2012).

Figure 14.  Proportional hazards ratios (HR) for lung cancer among underground workers by 15-year lags REC cumulative exposure (µg/m³-years) in expanded category untransformed (HR = 1.03/1000 μg/m³-year) and log transformed (HR = 1.07) regression models from their Tables S5 and S7 in Attfield et al. (2012).

Table

A posteriori	Association?	A priori	Association?
15-year lags, Exclude <5 year tenure, restrict REC ≤1280 µg/m^3-years, untransformed HR = 4.06(2.11–7.8)/10^3 REC	Yes, p <0.001	15-year lags, No tenure exclusion, No exposure restriction, untransformed HR = 1.03 (0.83–4.3)/10^3 REC, log transform HR = 1.07 (0.97–1.19)	No, p = 0.82, p = 0.17

Table 6.  Summary of exposure-response slopes and HRs for underground (UG) workers using untransformed and log transformed regression models over full cumulative exposure range for unlagged and 15-year lags (Table S7) (Attfield et al., 2012).

	Untransformed HR (95% CI), p, HR at 1000 → 4000 REC	Log transformed HR (95% CI), p, HR at 1000 → 4000 REC	Untransformed <1280 REC HR (95% CI), p, HR at 1000 →1280 REC, Exclude <5-years tenure
Un lagged	1.01 (0.89–1.14), (p = 0.89), 2.7 → 57 (35–96)	1.15 (1.0–1.31), (p = 0.046), 9.2 → 11.4 (8.3–16)
15-year lag	1.03 (0.83–1.28), (p = 0.12), 2.8 → 62 (28–165)	1.07 (0.97–1.19), (p = 0.08), 7.9 → 9.6 (7.8–12.4)	4.06 (2.1–7.8), (p =<0.001), 58 → 181 (15–21676)

These are primary results for authors’ conclusions following the procedures outlined in the protocols (NCI/NIOSH 1996; NCI/NIOSH 1997). UG workers with restricted exposure range (<1280 µg/m³-years) and excluding workers <5-years tenure (Figure 11 or 15) are included for comparison to demonstrate biased results produced by this a posteriori analysis.

Figure 15.  HR for lung cancer and 15-year lagged cumulative REC for UG workers in expanded categories as referent: Attfield primary results = untransformed HR=4.06/1000 REC (exclude <5-year tenure, restrict <1280 REC) versus a priori results of untransformed HR=1.03/1000 REC and log transformed HR = 1.07 (no exposure restriction or tenure exclusions) Tables 4 and S7 in Attfield et al. (2012).

Table 7.  Summary of increases in E-R slopes that always occur when deleting the high exposure category (>1280 µg/m³ years) (From Table S7 in Attfield et al., 2012).

Characteristics of UG cohort	Change in slope (HR per 1000 µg/m^3 years) from entire exposure range to exposures <1280 µg/m^3 years
No lag, 0 tenure exclusions	HR = 1.01 (p = 0.89) to 2.37 (p < 0.004)
No lag, <2 year tenure excluded	HR = 1.02 (p = 0.71) to 2.98 (p < 0.001)
No lag, <5 year tenure excluded	HR = 1.05 (p = 0.47) to 4.07 (p < 0.001)
No lag, <10 year tenure excluded	HR = 1.07 (p = 0.39) to 4.90 (p < 0.001)
15 year lag, 0 tenure exclusions	HR = 1.03 (p = 0.82) to 2.79 (p < 0.001)
15 year lag, <2 year tenure excluded	HR = 1.04 (p = 0.72) to 3.19 (p < 0.001)
15 year lag, <5 year tenure excluded	HR = 1.07 (p = 0.59) to 4.06 (p < 0.001)
15 year lag, <10 year tenure excluded	HR = 1.10 (p = 0.49) to 5.19 (p < 0.001)

Figure 16.  Lung cancer odds ratios for smoking status/smoking intensity for surface workers (REC = 0-8 µg/m³) versus UG workers (REC = 1–423 µg/m³) in case-control study (Silverman et al., 2012, Table 2).

Figure 17.  HRs and ORs of lung cancer and 15-year lagged cumulative REC among UG workers from cohort (Table 4 in Attfield et al., 2012), adjusted ORs with smoking removed from model (ORs at cohort exposure cutpoints) (page 9 from Silverman et al., 2012).

Figure 18.  Relative risks of lung cancer with cumulative REC exposure among surface workers with and without 15-year lags in cohort study (15-year lags, Table 5 in Attfield et al., 2012) and case-control study (unlagged and 15-year lags) Table 5 in Silverman et al. (2012).

Figure 19.  Lung cancer odds ratios for 15-year lagged cumulative REC for quartile, expanded categorical with split highest exposure category and continuous linear-exponential regression models for entire study population in case-control study (Tables 3, S1 in Silverman et al., 2012).

Figure 20.  Odds ratios (ORs) and 95% confidence intervals for unlagged cumulative REC exposure and lung cancer in quartile and linear-exponential (β = 0.0016, λ = −0.00042) regression models in DEMS nested case-control Tables 3 and S3 in Silverman et al. (2012).

Figure 21.  Odds ratios for lung cancer and cumulative REC in expanded category and linear-exponential (β = 0.0043, λ= −0.00056) models lagged 15-years and unlagged linear-exponential (β = 0.0016, λ = −0.00042) regression models from Tables S1, S2 and S3 in Silverman et al. (2012)

Figure 22.  Lung cancer and cumulative REC with linear-exponential models unlagged with (a) full REC (β = 0.0016, λ = −0.00042) and (b) REC <1280 μg/m³-year (β = 0.0025, λ = −0.00011) versus 15-year lagged models with (c) full REC (β = 0.0043, λ = −0.00056) and (d) REC <1280 μg/m³-year (β = 0.0025, λ = +0.00053) from Tables S1 and S3 in Silverman et al. (2012).

Figure 23.  Distribution of 200 lung cancer cases in complete cohort by 15-year lagged and unlagged average REC intensity (µg/m³) with no tenure restrictions (Tables S5 and S6 from Attfield et al., 2012).

Table 8.  Association of smoking prevalence by cumulative REC exposure among controls. Similar results obtained for cases and total cases + controls (calculated from Table 6 (Silverman et al., 2012).

	0–8 µg/m^3 years	8–304 µg/m^3 years	≥304 µg/m^3 years
Non-smokers	10.5%	13.2%	8.0%
<1 pack/day	7.3	8.7	6.9
1–2 packs/day	13.9	15.3	11.2
≥2 packs/day	3.9	3.9	5.0
Unknown	4.5	4.1	2.1

Figure 24.  Adjusted and crude lung cancer ORs stratified by confounders from employment in other high risk occupations and history of respiratory disease (Table 1 in Silverman et al., 2012).

Figure 25.  Crude and adjusted ORs for lung cancer and unlagged cumulative REC; ORs adjusted for 24 smoking × mine location, history of respiratory disease >5-years and history of high risk job for lung cancer >10-years; Crude ORs calculated from Table 3 in Silverman et al., 2012.

Figure 26.  Crude and adjusted ORs for lung cancer and 15-year lagged cumulative REC; ORs adjusted for smoking × mine location interaction; >5-year history of respiratory disease and >10-year history of high risk jobs for lung cancer; crude ORs calculated from Table 3 in Silverman et al. (2012).

Figure 27.  Crude and adjusted ORs for lung cancer and 15-years lagged cumulative REC for expanded exposure categories; ORs adjusted for smoking × mine work location, >5-year history of respiratory disease, and >10-year history of high risk jobs for lung cancer; crude ORs calculated from Table S2 in Silverman et al. (2012).

Figure 28.  Crude and adjusted ORs for lung cancer and unlagged cumulative REC among underground (UG) workers; ORS adjusted for smoking × mine location, >5-years respiratory disease and >10 years history high risk job for lung cancer; crude ORs calculated from Table 4 in Silverman et al. (2012).

Figure 29.  Crude and adjusted ORs for lung cancer and 15-year lagged cumulative REC among underground (UG) workers; ORs adjusted for smoking × mine work location, >5-year respiratory disease and >10-year history high risk job for lung cancer; crude ORs from Table 4 in Silverman et al. (2012).

Figure 30.  HR versus crude OR for Lung cancer and 15-year lagged REC cumulative exposure in expanded categories for complete cohort: HRs from Table S5 in Attfield et al. (2012) and crude ORs calculated from Table S2 in Silverman et al. (2012).

Table 9.  Summary of diesel-exposed workers in occupational cohorts with quantitative or semi-quantitative estimates of cumulative exposure to diesel exhaust and analysis of exposure-response trends.

	Coal miners (Johnston et al., 1997)	Truckers (Steenland et al., 1998)	Railroad engineers/conductors (Laden et al., 2006)	Potash miners (Neumeyer-Gromen et al., 2009)	Non-metal miners (Silverman et al., 2012)
Exposure units	Respirable g/m^3 h from historical PM & NO2 samples	µg/m^3 years EC based on emissions, 1990 samples = transitional DE; actual exposure = TDE	Intensity-years (emission × HP × fuel consumption) = g/mile emissions; ecological	mg/m^3 years total carbon, 1992 samples	µg/m^3 years REC
Latency	Potential 35 years. follow-up; but ~50% latency short	Essentially all <20-years; inaccurate estimates of “dieselization”	Hired 1945–1949, both steam and diesel; ~70% diesel only with adequate latency	Latency >20-years. 80–90%;	Presumably adequate
Lags/lung cancer deaths	Zero and 15-year lags/; both non-significant after adj pit/632 cases	Zero years; 994 cases 1982–1983, 1085 controls	5-years; 2396 cases among engineers/conductors, 918 unexposed (clerks, signal maintainers), 880 shopworkers	Zero years 61 cases	Zero years 198 cases, 562 controls
Exposure-response	Cox: unlagged HR = −0.0296/0.57 mg/m^3 year diesel PM; 15-year lag HR = 0.1451	Logistic: log = 0.1797(0.0696) (p = 0.01); linear = 0.000352 (0.000155) (p = 0.02)	Quintile, Cox proportional hazards model on workers hired 1945–1949	Cox regression, RR = 1.07(0.87–1.31)	Linear-exponential change deviance = 4.7, β = 0.0016, λ = −0.00041, p = 0.09
Limitations	About 60% may have inadequate latency	Exposure bases on air pollution; extrapolate from 1990 transitional DE; low lung cancer risk in referents; low exposure (~5 µg/m^3 EC); µg/m^3 years extrapolated from vehicle miles for non-drivers;	No increased risk among higher exposed shopworkers (E-R not analyzed); Lack of E-R trends; exposure misclassification (no measurements)	E-R slope reduced to 1.02 when adjusted for uranium confounding. Significant lung cancer deficit in referent group;	Incorrect adjustments for smokingand unreliable REC make E-R unreliable
Authors’ conclusion	“Limited evidence” entirely dependent on high exposures at Colliery Q	“Positive and significant increase …risk with increasing estimated cumulative exposure” among truck drivers; regard results with “caution” and considered “exploratory”	Evidence supports carcinogenicity of DE; “no evidence of increasing risk” by intensity-years.;	Non-significant E-R support diesel hypothesis	Supports DE hypothesis with “steep increase in risk” at low-moderate levels followed by plateau or decline in risk at high exposures
Conclusion	Does not support diesel hypothesis	Indeterminate	Indeterminate based on lack of E-R trends and lack of alternative hypotheses	Does not support diesel hypothesis; no E-R trend	Indeterminate based on unreliable exposure and biased E-R trends

Figure 31.  Exposure-response trends of Coal Miners (Johnston et al., 1997), Truckers (Steenland et al., 1998), Railroad engineers/conductors (Laden et al., 2006); Potash miners (unadjusted for uranium confounding) (Neumeyer-Gromen et al., 2009); US non-metal miners (crude and adjusted ORs) (Silverman et al., 2012).

Olsson AC, Gustavsson P, Kromhout H, Peters S, Vermeulen R, Brüske I, Pesch B, Siemiatycki J, Pintos J, Brüning T, Cassidy A, Wichmann HE, Consonni D, Landi MT, Caporaso N, Plato N, Merletti F, Mirabelli D, Richiardi L, Jöckel KH, Ahrens W, Pohlabeln H, Lissowska J, Szeszenia-Dabrowska N, Zaridze D, Stücker I, Benhamou S, Bencko V, Foretova L, Janout V, Rudnai P, Fabianova E, Dumitru RS, Gross IM, Kendzia B, Forastiere F, Bueno-de-Mesquita B, Brennan P, Boffetta P, Straif K. (2011). 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 183:941–948.

 PubMed Web of Science ®Google Scholar

Parent ME, Rousseau MC, Boffetta P, Cohen A, Siemiatycki J. (2007). Exposure to diesel and gasoline engine emissions and the risk of lung cancer. Am J Epidemiol 165:53–62.

 PubMed Web of Science ®Google Scholar

Villeneuve PJ, Parent MÉ, Sahni V, Johnson KC; Canadian Cancer Registries Epidemiology Research Group. (2011). Occupational exposure to diesel and gasoline emissions and lung cancer in Canadian men. Environ Res 111:727–735.

 PubMed Web of Science ®Google Scholar

Mohner M, Pohlabeln H, Jockel KH, Wichmann HE. (1998). Lungenkrebsrisko durch berufliche Exposition Weiterfuhrende Ansatze zur Modellierung und Adjustierung. Lungenkrebsrisiko durch berufliche Exposition. K. Jockel, I. Bruske-Hohlfeld and H.-E. Wichmann. Landsberg: Ecomed: 241–251.

 Google Scholar

Richiardi L, Mirabelli D, Calisti R, Ottino A, Ferrando A, Boffetta P, Merletti F. (2006). Occupational exposure to diesel exhausts and risk for lung cancer in a population-based case-control study in Italy. Ann Oncol 17:1842–1847.

 PubMed Web of Science ®Google Scholar

Gustavsson P, Jakobsson R, Nyberg F, Pershagen G, Järup L, Schéele P. (2000). Occupational exposure and lung cancer risk: a population-based case-referent study in Sweden. Am J Epidemiol 152:32–40.

 PubMed Web of Science ®Google Scholar

Boffetta P, Dosemeci M, Gridley G, Bath H, Moradi T, Silverman D. (2001). Occupational exposure to diesel engine emissions and risk of cancer in Swedish men and women. Cancer Causes Control 12:365–374.

 PubMed Web of Science ®Google Scholar

Guo J, Kauppinen T, Kyyrönen P, Lindbohm ML, Heikkilä P, Pukkala E. (2004). Occupational exposure to diesel and gasoline engine exhausts and risk of lung cancer among Finnish workers. Am J Ind Med 45:483–490.

 PubMed Web of Science ®Google Scholar

Attfield MD, Schleiff, PL, Lubin JH, Blair A, Stewart PA, Vermeulen R, Coble JB, Silverman DT. (2012). The diesel exhaust in miners study: a cohort mortality study with emphasis on lung cancer. J Natl Cancer Inst 104: 1–15.

 Google Scholar

Chatfield C. (1995). Model uncertainty, data mining and statistical inference. J Roy Stat Soc (Ser A) 158: 419–466.

 Web of Science ®Google Scholar

Johnston A, Buchanan D, Robertson A. (1997). Investigation of the possible association between exposure to diesel exhaust particulates in British coalmines and lung cancer. Technical Memorandum Series. Edinburgh, Scotland: Institute of Occupational Medicine, 1–105.

 Google Scholar

Steenland K, Deddens J, Stayner L. (1998). Diesel exhaust and lung cancer in the trucking industry: exposure-response analyses and risk assessment. Am J Ind Med 34:220–228.

 PubMed Web of Science ®Google Scholar

Garshick E, Laden F, Hart JE, Smith TJ, Rosner B. (2006). Smoking imputation and lung cancer in railroad workers exposed to diesel exhaust. Am J Ind Med 49:709–718.

 PubMed Web of Science ®Google Scholar

Neumeyer-Gromen A, Razum O, Kersten N, Seidler A, Zeeb H. (2009). Diesel motor emissions and lung cancer mortality–results of the second follow-up of a cohort study in potash miners. Int J Cancer 124:1900–1906.

 PubMed Web of Science ®Google Scholar