Diesel exhaust exposure and lung cancer risk

Exp. Patho!. 1989; 37: 32-38 VEB Gustav Fischer Verlag lena

Paper

Diesel exhaust exposure and lung cancer risk By P. BOFFETTA, R. E. HARRIS and E. L. WYNDER

Address for correspondence: Dr. E. L. WYNDER, American Health Foundation, 320 East 43rd Street, New York, NY 10017, USA Key words: diesel exhaust exposure, carcinogenicity; lung cancer risk, occupations; occupational risk, lung cancer; carcinogenicity, diesel exhaust

Summary The association between lung cancer and occupations with probable exposure to diesel exhaust has been examined in several studies in the past with inconclusive results. We report a case-control study among 2,584 cases and 5,099 hospital controls. The crude odds ratio for probable exposure was 1.31 (95 % CI 1.09, 1.57), but adjustment for smoking and other confounders reduced the estimate to 0.95 (95% CI 0.78,1.16). Similar results were observed for truck drivers, the only occupational category large enough for separate analysis. Data on self-reported exposure for a subset of 477 cases and 946 controls revealed a crude odds ratio of 1.45 (95 % CI 0.93, 2.27) which was reduced to 1.21 (95 % CI 0.78, 2.02) after controlling for smoking and other confounders. A duration-response relationship was suggested (P < 0.12) only for self-reported exposure. Cigarette smoking was the predominant condfounder in the analysis. Problems arising from biases and confounding in the present and in other previously published reports on the association between diesel exhaust exposure and lung cancer risk are discussed, and a review on the evidence of human carcinogenicity of diesel exhaust exposure is presented.

Introduction Epidemiologic studies of lung cancer risk in occupational groups likely to be exposed to diesel exhaust (DE) are conflicting and inconclusive. Four critical reviews of the subject have been published (1-4). Results presented here are based on data from an ongoing hospital based casecontrol study. Since a previous report (5), a more detailed assessment of DE exposure has been included in the study design, and a larger number of cases and controls have been collected for analysis.

Materials and Methods The data used in this analysis were derived from a large ongoing case-control study of tobaccorelated disases which has been in progress since 1969 as previously described (6). In the present study, a total of 2,548 male cases with newly diagnosed primary lung cancer and 5,099 suitably matched controls were examined to assess the association between DE exposure and lung cancer risk. Control subjects were matched to the cases based upon male sex, age to within 2 years, hospital, and year of interview. Two controls were ascertained per case. The control diagnoses included malignant neoplasms of the stomach (3.8 %), colon and rectum (15.3 %), prostate (10.2 %), skin (7.6%), hematopoietic system and soft tissues (13.3%), and other sites (7.3%); plus infectious 32

Exp. Patho!. 37 (1989) 1-4

diseases (2.6%), traumatic injuries (8.8%), benign prostatic hypertrophy (4.0%), benign skin lesions (6.8 %), and other non-malignant diseases unrelated to tobacco use (20.3 %). Cases and controls were interviewed by trained medical specialists using a structured questionnaire designed to elicit information on demographic factors, smoking history, medical history, consumption of alcoholic beverages, occupational history, etc. In 1985, new sections were added to assess dietary habits and vitamin consumption, and acquire more detailed information on tumor histology. The occupational section of the questionnaire was also expanded at that time to include record taking on the duration of exposure to 45 groups of chemicals including diesel exhaust during work or in hobby activities. For the purpose of analysis, occupations were classified a priori into categories: those with probable DE exposure (including truck and bus drivers, auto and heavy equipment mechanics, heavy equipment operators, railroad and subway workers, and garage and warehouse workers), those with possible DE exposure (including sanitation workers, firemen, sailors, machinists, taxicab drivers, and gas station attendants) and other occupations with minimal or no DE exposure. Risk associations were quantified by odds ratios and test-based 95 % confidence intervals (7). Dose-response effects were tested for linear trends according to the method by MANTEL (8). Multiple logistic regression models were used to simultaneously adjust effects of diesel exhaust for important other risk factors of lung cancer, viz. cigarette smoking and asbestos exposure (9).

Results Table I shows the distributions of cases and controls by age, race, and education. Since age was a matching variable, there are only minimal differences in the age distributions of cases and controls. The case group also contributed higher frequencies of blacks and subjects with lower educational levels than the control group (P < 0.05). Data on smoking in the cases and controls are given in table 2. As expected, there is a strong dose-response relationship between the intensity of smoking and lung cancer risk in both current smokers and exsmokers . In addition, there was a significant elevation in lung cancer risk for pipe and/or cigar smokers (OR = 1.9, P < 0.05). These results are therefore consistent with the doseresponse relationships between lung cancer risk and tobacco smoking observed in many other studies (6). Fig. I shows the odds ratios for occupations with probable DE exposure versus self-reported DE exposure with and without adjustment for cigarette smoking. Two points are evident. First, the odds ratio for self-reported exposure to DE is slightly higher than for occupational exposure. Most importantly, adjustment of both sets of odds ratios for cigarette smoking reduces the estimates to levels near unity. Odds ratios for occupations with possible DE exposure (not shown) were also approximately 1.0. These results imply that the observed elevations in unadjusted odds ratios for DE exposure could merely be the result of confounding with the dominant lung cancer risk factor, cigarette smoking. Fig. 2 illustrates the effect of duration of exposure for different categories of subjects. Separate results are given based upon self-reported exposure, occupations with probable exposure, and truck drivers. These results suggest an effect (P < 0.12) with long-term duration (> 30 years) of self-reported exposure. However, the sample sizes are extremely small in the subgroups reporting the longest durations of exposure. Furthermore, it should be emphasized that none of the odds ratios deviate significantly from 1.0. It is also noteworthy that no dose-response effect is suggested for duration of truck driving, the only occupation with a sample sufficient for separate analysis. Information on broad histologic type (Kreyberg types I and II) were available on all subjects. After adjustment for smoking and other confounders, the ORs for occupations with probable exposure were 0.89 (95 % Cl = 0.70, 1.13) for 1,468 cases of Kreyberg type I, and 0.83 (95% Cl = 0.61, 1.13) for 963 cases of Kreyberg type II (fig. 3). The corresponding ORs for self-reported DE exposure were also not significantly elevated in Kreyberg types I and II. 3

Exp. Pathol. 37 (1989) [-4

33

Table 1. Distribution of cases and controls by age, race, and education. Cases

Controls

n

%

n

%

Total

2,584

100.0

5,099

100.0

Age < 44 45-54 55-64 65-74 75+

199 543 1034 712 96

7.7 21.0 40.0 27.6 3.7

396 1068 2067 1395 173

7.8 20.9 40.5 27.4 3.4

Race White Other

2363 221

91.4 8.6

4793 305

94.0 6.0

495 424 710 664 288

19.2 16.4 27.5 25.7 11.2

626 727 1304 1560 879

12.3 14.3 25.6 30.6 17.2

Years of education 0-8 9-11 12 13-16 17+

Table 2. Tobacco-smoking and lung cancer risk. Smoking habit

Cigarettes per day

Cases n

Never smoker

Controls %

Odds ratio a

%

n

84

3.8

1199 23.7

1.0 - -

Current smoker Current smoker Current smoker

1-20/day 21-40/day 41 +/day

430 799 270

16.8 31.2 10.5

703 13.9 600 11.9 135 2.7

8.7 (7.0, 10.9) 19.0 (15.4, 23.4) 28.5 (22.2, 36.8)

Exsmoker Exsmoker Exsmoker

1-20/day 21-40/day 41 +/day

324 431 180

12.3 16.8 7.0

1078 21.3 704 13.9 246 4.9

4.2 (3.4, 5.4) 8.7 (7.0, 11.0) 10.4 (8.1, 13.6)

53

2.1

Pipe/cigar smoker a

392

7.8

1.9 (1.3, 3.2)

95 % CI in parentheses.

Discussion This investigation revealed no overall increase in lung cancer risk for subjects who were employed in occupations likely to have DE exposure relative to subjects never employed in such occupations. Furthermore, the only specific occupation with enough cases and controls to allow a

separate analysis, truck driving, was not associated with any increase in risk. Risk analysis by duration of exposure also failed to yield statistically significant results, although the trend test for self-reported duration of DE exposure approached significance (P < 0.12). Interestingly, subjects who self-reported exposure to DE had consistently higher point estimates of risk than those in DErelated occupations (overall adjusted ORs: 1.21 vs 0.95), which suggests the possibility of recall bias. 34

Exp. Pathol. 37 (1989) 1-4

2.4

Crude Adjusted For Smoking

2.2 20 1.8 Crude

I

1.6 :£

rJ)

1.4

(J)

1.2

i:i:

>

~ a;

1.0

0:

Adjusted For Smoking

T

I

08 0.6 0.4 0.2 00

Self-Reported Exposure 477 Cases 946 Controls

Occupations With Probable Exposure 2.584 Cases 5,299 Controls

AHF, 1989

Fig. 1. Diesel exhaust exposure and lung cancer. 40

Occupational Exp.

40

Seff-Reported Exp.

Truck Drivers

4.0

26 Cases 53 Controls

35 Cases 49 Controls

23 Cases 27 Controls

c; c:

:;,; ~30

30

30

20

2.0

III oll III Cl


~

"0 20

~ ~ ~

'0

~

:3III

~

-1

10

"0 "0

o

00

o

I

I 1-15

I 16-30

,30

I

00

I

0

I 1-15

I 16-30

Duration of Exposure

I

00

·30

I o

I 1-15

I 16-30

I

.30

(Years)

Fig.2. Lung cancer risk and duration of diesel exhaust exposure. Confounding is a major methodological problem in the study of weak associations (10, 11). It is therefore not surprising that the crude association we observed between DE exposure and lung cancer was strongly confounded with the dominant risk factor of the disease, cigarette smoking. The fact that a smaller dilution effect was shown for self-reported exposure to DE raises concern about the validity of this variable. The association between DE and lung cancer exposure might also be confounded by other life style 3*

Exp. Pathol. 37 (1989) 1-4

35

1.4

Oi c: :;:

1.3

E

1.2

0

(f) o(l Q)

Ol

«

1.1

0

1.0

"U;

09

lL Q)

::J

'0 ~

0.8

0

~

II:

0.7

if)

""

0

0.6 0.5 '-.

-".--

---..

--..

Kreyberg Kreyberg Type I Type II n = 1.488 n = 969 Lung Cancer

Fig.3. Lung cancer risk and diesel exhaust exposure by histologic type. 2.0

o

1.6

1.2

0-

:;:; Cll

cc

Ul

o ""

ell

o

oJ

~

0.8

I',

1',0

0.4

oI -0.4

II

£(e'l

00 0 0

c;

I',

o

,0, 1',. @O 0

0

"

0.5 0 01',

I',

. ? c; o

o

o

o Unadjusted

for Smoking

°

I',

I',

0l',£;

0

-0.8

-1.2

o

I',

,

1".,0

Standard Error

I',

o 6 Adjusted for Smoking

-1.6

-2.0

Fig. 4. Studies of diesel exhaust exposure and lung cancer risk: funnel plot. factors, such as diet. There is some evidence from epidemiological studies that certain dietary

factors influence lung cancer risk. Specifically, high consumption of fats and low consumption of vitamin A have been reported to be associated with increased rates of lung cancer (12, 13). Truck drivers have been found to have a different diet from subjects in other occupations (14), and these results are confirmed in the present study (table 3). Truck drivers, subjects employed in other occupations with probable DE exposure, and subjects who self-reported DE exposure drank more alcohol and coffee, ,and consumed beef, butter, and fried foods more frequently than other 36

Exp. Pathol. 37 (1989) 1-4

Table 3. Average mean consumption of alcohol, coffee, and selected food items by exposure to diesel exhaust (DE) (a). Self-reported exposure to DE

Usual occupation Food item

Alcohol (b) Coffee (c) Beef (d) Eggs (d) Fried foods (d) Cheese (d) Butter (d) Salads (d) Vegetables (d) Fruits (d) Meal in fast food restaurant (d)

No exposure to DE

Probable exposure to DE

Truck driving

No

Yes

261 3.0 23.2 12.3 6.5 11.2 13.8 17.9 27.0 25.3 12.6

282 3.6* 27.7* 13.6 7.7 13.2 16.2 13.55** 24.6* 24.0 13.1

198 3.7 25.8 13.6 8.4 11.0 15.1 11.5** 22.3* 19.4* 14.5

255 3.0 23.2 12.3 6.4 11.5 13.9 17.6 27.0 25.4 12.4

378* 3.4 26.0 12.7 8.9* 10.5 14.9 15.6 26.0 23.3 15.0

(a) *, ** Denote significant differences from the nonexposed referent group at the 0.05 and 0.01 levels of probability, respectively; (b) Grams/week; (c) Cups/day; (d) Servings/month.

subjects. In addition, truck drivers ate more meals at fastfood restaurants, and ate less salads, vegetables, and fruits than subjects in the unexposed categories. The influence of these dietary factors on the cancer risk associated with smoking, asbestos, and DE exposure is currently being examined. The evidence on human lung carcinogenicity due to DE exposure, as derived from the literature, is inconclusive (15). Unfortunately, many of the positive studies have not been controlled for important confounders, such as cigarette smoking and exposure to asbestos or other occupational carcinogens. On the other hand, negative studies often lack the statistical power to detect a weak association. Fig. 4 summarizes the results derived from published epidemiological studies (15). This diagram is a special scatter plot wherein the log of the estimated odds ratios for DE exposure and lung cancer risk are plotted against the standard errors. In theory, the pattern should depict a funnel pointing to the log of the true odds ratio on the vertical axis. In this case, the funnel appears to point to approximately zero, or perhaps only very slightly above zero, therefore providing little evidence of a positive association. The overall evidence from the literature is controversial regarding the role of DE exposure in human lung carcinogenesis. As is the case for many weak epidemiologic associations, the assessment of past exposures to DE based on questionnaire-derived data may not be adequate to accurately detect small increases in risk. Results from the present analysis confirm the prudence of a cautious approach to this problem. No clear increase in risk for truck drivers or other occupational groups was shown, while the modest excess in risk associated with self-reported exposure to DE could likely have resulted from reporting bias. In order to definitively characterize the relationship between diesel exhaust exposure and lung cancer risk, better measures of exposure will be required. One potentially useful measure would be to characterize the level of risk by intermediate molecular outcomes, such as the formation of 1-6dinitropyrene DNA adducts in lung tissue of exposed subjects. Tools of molecular biology may therefore enhance the epidemiologist's ability to accurately detect weak disease associations. Exp. Pathol. 37 (1989) [-4

37

References 1. HIGGINS, I. T. T.: Air pollution and lung cancer: diesel exhaust, coal combustion. Prevo Med. 1984; 13: 207-218. 2. STEENLAND, K.: Lung cancer and diesel exhaust: a review. Am. J. Ind. Med. 1986; 10: 177-189. 3. WVNDER, E. L., HIGGINS, I. T. T.: Exposure to diesel exhaust emissions and the risk oflung and bladder cancer. Dev. Toxico!. Environ. Sci. 1986; 13: 489-501. 4. FRASER, D.: Lung cancer risk and diesel exhaust exposure. Public Health Rev. 1986; 14: 139-171. 5. HALL, N. E. L., WVNDER, E. L.: Diesel exhaust exposure and lung cancer: a case-control study. Environ. Res. 1984; 34: 77-86. 6. WVNDER, E. L., STELLMAN, S. D.: Comparative epidemiology of tobacco-related cancers. Cancer Res. 1977; 37: 4608-4622. 7. MIETTINMEN, O. S.: Estimability and estimation in case-referent studies. Am. J. Epidemio!. 1976; 103: 226-235. 8. MANTEL, N., HAENSZEL, W.: Statistical aspects of the analysis of data from retrospective studies of disease. JNCI 1959; 22: 719-748. 9. HARRELL, F. E.: The LOGIST procedure. In: SUGI supplemental library user's guide, version 5 edition. SAS Institute, Cary, NC 1986; 269-293. 10. WVNDER, E. L.: Workshop on guidelines to the epidemiology of weak associations: introduction. Prevo Med. 1987; 16: 139-141. II. STELLMAN, S. D.: Confounding. Prevo Med. 1987; 16: 165-182. 12. BVERS, T., GRAHAM, S.: The epidemiology of diet and cancer. Adv. Cancer Res. 1984; 41: 1-69. 13. WVNDER, E. L., GOODMAN, M. T., HOFFMANN, D.: Lung cancer etiology: challenges of the future. ' Carcinogenesis 1985; 8: 39-62. 14. - MILLER, S.: Motor exhaust-related occupations and bladder cancer. Cancer Res. 1988; 48: 1989-1990. 15. BOFETTA, P., HARRIS, R. E., WVNDER, E. L.: A case-control study on occupational exposure to diesel exhaust and lung cancer risk. Under review, Cancer Research.

Exp. Patho!. 1989; 37: 38 VEB Gustav Fischer Verlag Jena

Book Review

Tumorlokalisationsschliissel International Classification of Diseases for Oncology (lCD..Q) Topographischer Tell edited by G. WAGNER (3rd, revised edition). 96 pages, with 44 figures IX, DM48,-, ISBN: 3540-19386-3. Springer-Verlag, Berlin-Heidelberg-New York-London-Paris-Tokyo-Hong Kong 1988. The book contains the key in German language for the understanding of the tumor localisation with the aim to standardise the documentation of malignant tumors. It is founded on the topographic code of the Manual of Tumor Nomenclature and Coding (MOTNAC) of the American Cancer Society basing upon the enlarged 8th revision of the International Classification of Diseases adapted for use in the United States (IDCA). The key for the tumor localisation has proved its value in the recent years and, therefore, the code numbers of this third edition are largely unchanged. - The topographic figures represented as schematic drawings are enlarged and, partly, precised. - Each code number is presented at the figures, so that an exact topographic documentation is guaranteed as a prerequisite for the homogeneous understanding of data. This is the reason why this third, revised edition belongs into F. BOLeK, lena the hand of each doctor faced with tumor patients. 38

Exp. Patho!. 37 (1989) 1-4