Document Lpwa7vqbzwXBJ5qbm2LExv3Db

American Journal of Epidemiology Copyright 2005 by the Johns Hopkins Bloomberg School of Public Health All rights reserved; printed in U.S.A. Vol. 162, No. 9 DOI: 10.1093/aje/kwi285 Advance Access publication September 21,2005 Original Contribution Evidence for Excess Colorectal Cancer Incidence among Asbestos-exposed Men in the Beta-Carotene and Retinol Efficacy Trial Downloaded from aje.oxfordjournals.org by guest on May 10, 2011 Oluremi A. Aliyu1, Mark R. Cullen1, Matt J. Barnett2'3, John R. Balmes4,5, Brenda Cartmel1, Carrie A. Redlich1, Carl A. Brodkin3, Scott Barnhart3,6, Linda Rosenstock7, Leslie Israel8, Gary E. Goodman2,3,9, Mark D. Thornquist2,3, and Gilbert S. Omenn10,11 1 Yale Occupational and Environmental Medicine Program and the Cancer Center, Yale University School of Medicine, New Haven, CT. 2 Fred Hutchinson Cancer Research Center, Seattle, WA. 3 Occupational and Environmental Medicine Program, University of Washington, Seattle, WA. 4 Department of Medicine, School of Medicine, University of California, San Francisco, San Francisco, CA. 5 Center for Occupational and Environmental Health, University of California, Berkeley, Berkeley, CA. 6 Harborview Medical Center, Seattle, WA. 7 School of Public Health, University of California, Los Angeles, Los Angeles, CA. 8 Center for Occupational and Environmental Health, University of California, Irvine, Irvine, CA. 9 Swedish Medical Center Cancer Institute, Seattle, WA. 10 Departments of Internal Medicine and Human Genetics, School of Medicine, University of Michigan, Ann Arbor, MI. 11 Department of Public Health, School of Public Health, University of Michigan, Ann Arbor, MI. Received for publication September 29, 2004; accepted for publication June 1, 2005. The relation between asbestos exposure and colorectal cancer remains controversial. The authors of this 1984 2004 US study examined the association among 3,897 occupationally exposed participants in the Beta-Carotene and Retinol Efficacy Trial (CARET) for chemoprevention of lung cancer, followed prospectively for 10-18 years. When a Cox stratified proportional hazards model was used, risks of colorectal cancer were elevated among male heavy smokers exposed to asbestos. Their relative risk was 1.36 (95% confidence interval: 0.96, 1.93) when compared with that for CARET heavy smokers not exposed to asbestos, after adjusting for age, smoking history, and intervention arm. The presence of asbestos-induced pleural plaques at baseline was associated with a relative risk of 1.54 (95% confidence interval: 0.99, 2.40); colorectal cancer risk also increased with worsening pulmonary asbestosis (p = 0.03 for trend). A dose-response trend based on years of asbestos exposure was less evident. Nonetheless, these data suggest that colorectal cancer risk is elevated among men occupationally exposed to asbestos, especially those with evidence of nonmalignant asbestos-associated radiographic changes. asbestos; asbestosis; colorectal neoplasms; environmental exposure; prospective studies; randomized controlled trials; smoking Abbreviations: CARET, Beta-Carotene and Retinol Efficacy Trial; CI, confidence interval. In contrast to the findings for lung cancer, the relation between asbestos exposure and the risk of colorectal cancer is not universally accepted (1). Selikoff et al. (2) reported in 1964 a threefold excess mortality from colorectal cancer among asbestos insulators. Excess deaths due to colorectal cancer were also found in their larger study of insulation workers in 1979 (3). Subsequent epidemiologic studies have yielded conflicting results. Early cohort studies from Italy Correspondence to Dr. Mark R. Cullen, 135 College Street, New Haven, CT 06510 (e-mail: mark.cullen@yale.edu). 868 Am J Epidemiol 2005;162:868-878 Colon Cancer among Asbestos-exposed Men 869 Downloaded from aje.oxfordjournals.org by guest on May 10, 2011 (4), Norway (5), and the United States (6, 7) were positive. More recently, Albin et al. (8) reported a strong relation between colorectal cancer risk and cumulative asbestos dose but no overall excess of colorectal cancer. Other positive cohort studies include those by Jakobsson et al. (9), Raffin et al. (10), and Szeszenia-Dabrowska et al. (11). Scandina vian and US case-control studies have also observed signif icant increases in asbestos-associated odds ratios (12-16). In a meta-analysis by Homa et al. (17) using published re ports of 20 amphibole asbestos-exposed cohorts, there was an elevated summary standardized mortality ratio (1.47, 95 percent confidence interval (CI): 1.09, 2.00). Other investigators have not found an association. In a study of British asbestos workers, Hodgson et al. (18) found a significant deficit of colon cancer mortality (stan dardized mortality ratio = 54); Gardner et al. (19) found the expected rate. Multiple cohort studies found no association (20-27). In a meta-analysis of 69 occupational cohorts, Goodman et al. (28) concluded that data for gastrointestinal cancers showed no evidence of a significant association with asbestos exposure and no dose-response effect. Proponents of an association have suggested that in creased risk occurs as a local response to inhaled asbestos fibers cleared from the lung and swallowed, eventually pen etrating the gastrointestinal mucosa and initiating tumor formation (29). Ehrlich et al. (30) reported the presence of asbestos bodies in the colon of an insulation worker with asbestosis and adenocarcinoma. Goldsmith (31), mean while, suggested that asbestos might act as a systemic car cinogen, noting that excess cancer at gastrointestinal sites parallels excess risk at other extrapulmonary sites. In any event, it is unclear whether synergism occurs with tobacco smoke, as for lung cancer, or putative dietary or other risk factors for colorectal cancer. We previously analyzed colorectal cancer incidence among the asbestos-exposed males followed prospectively as part of the Beta-Carotene and Retinol Efficacy Trial (CARET), a multicenter, randomized, double-blinded, placebo-controlled chemoprevention trial designed to as sess the effect of daily pharmacologic doses of vitamin A and beta-carotene on lung cancer incidence and mortality (32). Although the intervention was discontinued in 1996, 21 months ahead of schedule, when we recognized that the vitamins were associated with increased risk of lung can cer and increased total mortality (33, 34), CARET partic ipants continue to be followed. Notably, the intervention had no measurable effect on colorectal cancer incidence or mortality (33). MATERIALS AND METHODS Study participants For 10-18 years, the CARET trial for chemoprevention of lung cancer has followed 4,060 men occupationally exposed to asbestos as well as 14,254 heavy smokers (7,965 men and 6,289 women). The recruitment, enrollment, randomization, follow-up, and initial evaluations of the participants have been described in detail previously (32-37). Am J Epidemiol 2005;162:868-878 Two major cohorts were recruited at six centers in the United States: an asbestos-exposed cohort and a heavysmoker cohort (figure 1). Participants for the asbestosexposed cohort were recruited at five centers, four of which had large occupational health clinics. Subjects were referred by occupational and pulmonary physicians as well as by employers, unions, or lawyers or in response to public ad vertisements. Men were eligible for enrollment into the asbestos-exposed cohort during 1989-1993 if they were be tween 45 and 69 years of age, currently smoked or had quit smoking within the previous 15 years, and had been ex posed to asbestos as documented by the following criteria: 1) they had worked in one of eight CARET-specified high-risk trades with established, regular asbestos exposure (insulation, sheet metal, plumbing, plasterboard, ship fitting, ship electrical work, boiler making, or ship scaling) for at least 5 years, starting at least 15 years previously; or 2) they had a history of occupational asbestos exposure in any job or occupation and had evidence of chest radiograph changes-- pleural abnormalities or pulmonary fibrosis--consistent with a diagnosis of nonmalignant asbestos-related disease. In all, 3,244 men were enrolled as a result of these recruit ment efforts. We added 816 men who had been enrolled previously in the pilot phase only in Seattle, Washington, using similar criteria except for a wider age span of 45-74 years and no smoking requirement (32), for a total of 4,060 men. Thirty-four percent of this asbestos-exposed cohort qualified by virtue of work history alone, 21 percent quali fied by radiographic criteria alone, and 44 percent met both criteria. Excluded subsequently were 20 participants later found to be ineligible, eight whose smoking status was un known, two whose radiographs were missing, and 133 from the pilot study who were lifelong nonsmokers, leaving data on 3,897 participants available for these analyses. Because of the diversity of the asbestos exposure settings among the men in the asbestos-exposed cohort in construc tion, shipbuilding, and manufacturing and the fact that most exposure occurred long before study entry, no formal effort was undertaken to further classify participants by exposure dose, fiber type, or distribution of fiber sizes. Duration of exposure and severity of radiographic changes were used as crude surrogates of exposure dose instead, since all had postero-anterior and lateral chest radiographs before ran dom assignment (37). Radiographic changes were indepen dently assessed at each center by a B-reader, a radiologist, or a chest physician trained and certified in using the system of the National Institute for Occupational Safety and Health. International Labour Organization standard (1980) films were used to assess for each of two independent patterns of radiographic change typical of asbestos. Pleural reaction--bilateral thickening or plaque, with or without calcification--was rated as present or absent. Profusion throughout the lung fields of small irregular shadows was separately rated on a progressive 12-point scale: 0/--, 0/0, 0/1, 1/0, 1/1, 1/2, 2/1, 2/2, 2/3, 3/2, 3/3, 3/+, where the first number represents the major category and the second a modifier akin to plus or minus for letter grades. Category 1/0 changes and higher, typical for pulmonary asbestosis, were found in 39 percent of the participants at baseline; 47 percent had asbestos-associated pleural abnormalities. Colon Cancer among Asbestos-exposed Men 871 Downloaded from aje.oxfordjournals.org by guest on May 10, 2011 This paper includes complete follow-up information through December 2003. Statistical analysis To examine the relation between asbestos exposure and colorectal cancer incidence, the following analyses were performed: 1) comparison of the colorectal cancer risk for the smoker-eligible subcohort of the asbestos-exposed cohort with that for the heavy-smoker cohort, using radio graphic changes as crude surrogates for asbestos exposure dose; 2) comparison of risk within the asbestos-exposed cohort, using radiographic changes as surrogates for expo sure dose; 3) comparison of risk within the work-historyeligible subcohort, using years working in a high-risk trade, years since first exposure to asbestos, and specific trade as surrogates for exposure dose; and 4) survival analysis post diagnosis comparing colorectal cancer cases in the smokereligible subcohort with cases in the heavy-smoker cohort. Stratified Cox proportional hazards models were used to obtain colorectal cancer relative risk estimates and 95 per cent confidence intervals. All models were stratified on en rollment period (pilot phase vs. full study). Models included adjustment for age (as a linear variable), baseline smoking status (current, former), pack-years of smoking (<40, 41-60, >60), and intervention assignment (vitamin A + beta-carotene, placebo). Comparisons restricted to the asbestos-exposed workers were further stratified on enroll ment center and included additional adjustment for years since quitting smoking. Only two of the six CARET study centers recruited participants for both the asbestos and heavy-smoker cohorts. Thus, in the comparison between the asbestos-exposed smoker-eligible subcohort and the non-asbestos-exposed heavy-smoker cohort, study center could not be evaluated as a potential confounder. Adjustment for body mass index and dietary intake of calcium, fiber, fat, and percentage of energy from fat as potential confounders had little effect, so they were excluded from the final models. Radiographic findings (presence of pleural reaction, Inter national Labour Organization profusion score) and work history (years working in a high-risk trade, years since first exposure, specific trade) were examined independently as surrogates of asbestos exposure. Radiographic variables were fit simultaneously in models including the asbestosexposed cohort and its subcohorts only; to avoid overfitting bias, no adjustment for years working in a high-risk trade was made in these analyses. A similar approach was used to examine the association between asbestos-related work his tory measures and colorectal cancer incidence among the work-history-eligible subcohort. Participants eligible on the basis of radiographic findings only were excluded from this analysis to avoid bias due to a potential underreporting of years of exposure in this subgroup. In this paper, results of this analysis are presented both unadjusted for specific trade--to assess possible differences by trade in the inten sity of asbestos exposure per year--and adjusted, assum ing that variation may reflect nonasbestos, trade-specific exposures. For the comparison to the heavy-smoker cohort, the asbestos-related variables were assessed in separate models, Am J Epidemiol 2005;162:868-878 with the heavy-smoker cohort serving as the referent in each model. Likelihood ratio tests were performed to test for linear trend across categories and to test whether associa tions between asbestos-related measures and colorectal can cer risk were modified by age, smoking status, pack-years of smoking, and intervention assignment. No test of interaction was statistically significant. Kaplan-Meier estimates were calculated to examine postdiagnosis survival by risk popu lation; log-rank tests were performed on differences in sur vival curves. All significance tests were two sided. For our analyses, we used SAS software, version 8.2 (SAS Institute, Inc., Cary, North Carolina). RESULTS There were 85 incident cases of colorectal cancer ob served among the 3,897 participants in the asbestos-exposed cohort. Among the 7,924 men in the heavy-smoker cohort, there were 123. Crude incidence rates and 95 percent confi dence intervals for colorectal cancer in the asbestos-exposed and heavy-smoker cohorts were, respectively, 2.0 (95 per cent CI: 1.6, 2.5) and 1.6 (95 percent CI: 1.3, 1.9) per 1,000 person-years. Incidence rates did not differ by intervention arm overall or within each cohort. The crude incidence rates for the smoker-eligible and work-history-eligible subcohorts of the asbestos-exposed cohort were 2.2 (95 percent CI: 1.6, 3.0) and 2.1 (95 percent CI: 1.6, 2.6), respectively. Demo graphic, smoking, occupational, and dietary histories for the asbestos-exposed and heavy-smoker cohorts are presented in table 1. All parameters except asbestos exposure and smok ing were similar. The same parameters for the smoker-eligible and work-history-eligible subcohorts are shown in table 2. Work-history-eligible participants spent a higher average number of years working in high-risk trades but had similar durations of exposure to asbestos overall. Since no appropriate external comparison group for these volunteer study participants was available, we used the CARET heavy-smoker cohort as a non-asbestos-exposed comparison group for the asbestos-exposed cohort and its smoker-eligible subcohort. Tests for homogeneity revealed that the heavy-smoker cohort and smoker-eligible subcohort were indistinguishable regarding all measurable factors ex cept asbestos; although some heavy smokers had shortduration asbestos exposures (table 1), almost none was working in high-risk trades after age 29 years, who would have qualified as asbestos exposed, so these heavy smokers were excluded (refer to the Materials and Methods section). Table 3 shows the crude incidence rates and results of the adjusted analyses for the asbestos-exposed smoker-eligible subcohort versus the unexposed heavy-smoker cohort. When we adjusted for smoking history, age, and interven tion arm, the asbestos-exposed smoker-eligible subcohort had a 36 percent higher rate of colorectal cancer compared with the heavy-smoker cohort, although it was not statisti cally significant (95 percent CI: 0.96, 1.93). Asbestosexposed participants in the smoker-eligible subcohort who had pleural abnormalities had a 54 percent increased risk of colorectal cancer compared with participants in the heavysmoker cohort (p = 0.05). There was also a significant trend Colon Cancer among Asbestos-exposed Men 873 TABLE 2. Demographics of the asbestos-exposed smoker-eligible subcohort and the work-history-eligible subcohort, Beta-Carotene and Retinol Efficacy Trial: Seattle, Washington; Irvine, California; New Haven, Connecticut; San Francisco, California; Baltimore, Maryland; and Portland, Oregon, 1985-2004* Variable No. of participants Age (years) <55 55-64 >65 Smoking status at enrollment Current Former Years since quitting smoking (no.) (former smokers) 0-4 5-6 7-9 10-14 >14 Pack-years of smoking (no.) <40 41-60 >60 Years of asbestos exposure (no.) Years working in a high-risk trade (no.) Body mass index (kg/m2) Calcium intake (mg/day) Fat intake (g/day) Energy from fat (%) Fiber intake (g/day) Smoker-eligible subcohorty (asbestos exposed) Total Colorectal cancer cases Mean (SD) No. % Mean (SD) No. % 1,839 42 59 (6) 62 (4) 531 29 4 10 980 53 23 55 328 18 15 36 1,105 60 734 40 24 57 18 43 476 65 258 35 13 72 5 28 53 (23) 28 (10) 19 (14) 28.3 (4.8) 790 (478) 85 (38) 38 (8) 16(7) 618 34 678 37 543 30 55 (29) 30 (9) 15 36 12 29 15 36 23 (14) 29.9 (3.5) 784 (534) 82 (38) 39 (8) 16(2) Work-history-eligible subcohortZ (asbestos exposed) Total Colorectal cancer cases Mean (SD) No. % Mean (SD) No. % 3,067 71 56 (7) 61 (7) 1,360 44 13 18 1,205 39 30 42 502 16 28 39 1,235 40 1,832 60 24 34 47 66 42 (24) 27 (10) 480 26 274 15 281 15 468 26 329 18 1,652 874 541 54 28 18 46 (24) 28 (10) 10 21 49 10 21 16 34 7 15 35 49 20 28 16 23 24 (10) 28.7 (4.6) 810 (489) 84 (38) 38 (8) 16(7) 25 (11) 29.6 (4.1) 797 (502) 81 (41) 38 (8) 15(7) * Some percentages do not total 100 because of rounding. y Excludes 2,221 asbestos-exposed participants: 2,199 who did not meet the heavy-smoker eligibility criteria, 20 who did not meet the asbestos eligibility criteria, and two who did not have a baseline radiograph. Z Excludes 993 participants: 830 who were eligible on the basis of asbestos-related radiographic changes and did not meet the occupational exposure criterion, in addition to the 163 participants dropped previously (refer to table 1). SD, standard deviation. Downloaded from aje.oxfordjournals.org by guest on May 10, 2011 was similar to that shown in table 3 but did not achieve statistical significance. As an alternative to radiographic change as a measure of dose, we looked at the effect of years of exposure in a highrisk trade, years since first exposure to asbestos, and specific trade on the risk of colorectal cancer among the workhistory-eligible subcohort (eligible for the study based on a work history of 5 or more years in one of the jobs defined as high-risk trades). Table 5 shows the relative risks for colorectal cancer in this subcohort, adjusting for age, years since quitting smoking, pack-years of smoking, intervention arm, years working in a high-risk trade, and time since first asbestos exposure; models with and without adjustment Am J Epidemiol 2005;162:868-878 for trade are presented to explore the possibility of tradespecific heterogeneity. Although the trend for years in a high-risk trade was not statistically significant, more than 10 years in a high-risk trade carried a progressively increas ing risk of colorectal cancer with increasing number of years of exposure up until 30 years; those participants with 21-30 years of exposure had a 74 percent increased risk compared with those with less than 10 years of exposure. After 30 years in a high-risk trade, the risk progressively decreased, those with more than 40 years of exposure having a risk lower than that for those with less than 10 years of exposure. Time since first asbestos exposure had no predictive effect on the risk of colorectal cancer. Although the numbers were 874 Aliyu et al. TABLE 3. Colorectal cancer incidence among asbestos-exposed participants in the smoker-eligible subcohort and the non-asbestos-exposed heavy-smoker cohort, Beta-Carotene and Retinol Efficacy Trial: Seattle, Washington; Irvine, California; New Haven, Connecticut; San Francisco, California; Baltimore, Maryland; and Portland, Oregon, 1985-2004 Total no.* At-risk population Heavy-smoker cohort (non-asbestos-exposed) Smoker-eligible subcohort (asbestos-exposed) Pleural abnormality^ Heavy-smoker cohort (non-asbestos-exposed) Smoker-eligible pleura negative Smoker-eligible pleura positive Radiographic profusion rating category (major categories)# Heavy smoker 0/- to 0/1 1/0 to 1/2 2/1 to 2/3 3/2 to 3/+ 7,924 1,839 7,924 953 886 7,924 1,007 769 47 16 Colorectal cancer cases (no.) RRt 95% CIt,Z p value 123 1.00 0.09 42 1.36 0.96, 1.93 123 1.00 0.17 18 1.17 0.71, 1.92 24 1.54 0.99, 2.40 123 1.00 0.03** 20 1.20 0.75, 1.93 19 1.44 0.89, 2.34 2 2.47 0.61, 10.0 1 3.92 0.54, 28.2 * Excludes 2,221 asbestos participants: 2,199 who did not meet the heavy-smoker eligibility criteria, 20 who did not meet the asbestos eligibility criteria, and two who did not have a baseline radiograph. Also excludes 41 heavy smokers: 12 who did not meet the heavy-smoker eligibility criteria and 29 who met the asbestos eligibility criteria. t RR, relative risk; CI, confidence interval. Z Estimates from Cox proportional hazards model stratified on enrollment period (pilot or efficacy phase) and adjusted for age, smoking status at baseline (current, former), pack-years of smoking (<40, 41-60, >60), and intervention arm. Test for heterogeneity. { Presence of bilateral pleural thickening or plaques on radiography, with or without calcification. # Density of small irregular shadows in the lung fields using the International Labour Organization 12-point rating scale. ** Test for trend using group linear covariates. Downloaded from aje.oxfordjournals.org by guest on May 10, 2011 FIGURE 2. Postdiagnosis colorectal cancer survival in two sub populations of the Beta-Carotene and Retinol Efficacy Trial (Seattle, Washington; Irvine, California; New Haven, Connecticut; San Francisco, California; Baltimore, Maryland; and Portland, Oregon, 1985-2004). Cases in the asbestos-exposed smoker-eligible sub cohort (n = 42, solid line) were compared with cases in the non asbestos-exposed heavy-smoker cohort (n = 123, broken line). small and confidence intervals were wide, there was a sug gestion of possible trade-associated differences (table 5). DISCUSSION The results of this large, longitudinal cohort study suggest an increased risk of colorectal cancer among men with radio graphic evidence of nonmalignant asbestos-related disease. This risk was most clearly observed when the (asbestosexposed) smoker-eligible subcohort was compared with the (non-asbestos-exposed) heavy-smoker cohort (table 3). In this comparison, there appeared to be a dose-response relation when profusion score on chest radiograph was used as a surrogate for dose. We believe that this comparison is most appropriate because it controls for smoking and those unmeasured behaviors likely associated with heavy smok ing. The dose trend, although similar, was not significant for the internal analysis of the whole asbestos-exposed cohort (table 4) when the same surrogates for exposure were used--radiographic changes--suggesting that the trend seen in table 3 was anchored by the large nonexposed comparison group. In each comparison, however, both the Am J Epidemiol 2005;162:868-878 Colon Cancer among Asbestos-exposed Men 875 TABLE 4. Radiographic predictors of colorectal cancer incidence among the asbestos-exposed cohort, Beta-Carotene and Retinol Efficacy Trial: Seattle, Washington; Irvine, California; New Haven, Connecticut; San Francisco, California; Baltimore, Maryland; and Portland, Oregon, 1985-2004 Pleural abnormality Negative Positive Radiographic profusion rating category# 0/- to 0/1 1/0 to 1/2 2/1 to 2/3 3/2 to 3/+ Radiographic abnormality Negative Parenchymal changes Pleural abnormality Parenchymal + pleural Total no.* 2,050 1,847 2,365 1,424 84 24 1,325 725 1,040 807 Colorectal cancer cases (no.) 34 51 48 33 3 1 21 13 27 24 RRt,Z 95% CIt p value 1.00 1.40 0.88, 2.23 0.151 1.00 1.12 1.41 1.38 0.70, 1.80 0.42, 4.75 0.18, 10.6 0.49** 1.00 1.21 1.47 1.62 0.59, 2.48 0.81, 2.66 0.85, 3.09 0.451 * Excludes 163 participants: 133 never smokers, 20 who did not meet the asbestos eligibility criteria, eight for whom information on pack-years was missing, and two who did not have a baseline radiograph. t RR, relative risk; CI, confidence interval. Z Estimates were derived from a multivariate Cox proportional hazards model stratified on enrollment period (pilot or efficacy phase) and study center and included the following covariates: age, years since quitting smoking (current smokers, 0-4, 5-9, 10-14, >14), pack-years of smoking (<40, 41-60, >60), intervention arm (active vitamins, placebo), occupational trade (eight study-specific high-risk trades and an "other'' category), presence of pleural abnormality, and profusion rating (<1/0, 1/0-1/2, 2/1-2/3, >2/3). Presence of bilateral pleural thickening or plaques on radiography, with or without calcification. 1 Test for heterogeneity. # Test for trend using group linear covariates. ** Density of small irregular shadows in the lung fields using the International Labour Organization 12-point rating scale. Downloaded from aje.oxfordjournals.org by guest on May 10, 2011 presence of pleural plaques and the International Labour Organization profusion score appear to be predictive of co lorectal cancer risk. Smoking is unlikely to confound this association because it had no independent effect on colorec tal cancer risk. Our results were less impressive when we used years work ing in a high-risk trade as surrogates of exposure (table 5). There are several possible reasons. For one, all participants in the work-history-eligible subcohort had significant expo sure to asbestos because they worked in a high-risk trade for at least 5 years; even those we classified as least exposed by using the surrogate measures of exposure dose were heavily exposed compared with men in the heavy-smoker cohort, so the range of exposures is limited. A second pos sibility for the weaker association is that the use of years in a high-risk trade leads to substantially greater misclassification than using radiographic change as the exposure marker. Al ternatively, the effect may be limited to those men exposed heavily enough to have radiographic changes or a differential susceptibility to asbestos effects manifested by the abnormal radiographs. Another possibility for the observed results is that selec tive pressures have operated in the population. The analysis within the work-history-eligible subcohort revealed that Am J Epidemiol 2005;162:868-878 years in a high-risk trade predicted colorectal cancer risk up to 30, beyond which the colorectal cancer rate started to drop. This finding may be due to a survival effect, in which those most heavily exposed died preferentially of lung can cer, mesothelioma, or other diseases strongly associated with asbestos exposure. It may also indicate a healthyworker effect, whereby those more physically active or whose body mass is lower--protective factors for colorectal cancer--remained in the trades for longer periods. That adjustment for body mass index did not alter the results weighs against such an interpretation. Alternatively, it is possible that results of the comparison between the (asbestos-exposed) men in the smoker-eligible subcohort and those in the (non-asbestos-exposed) heavysmoker cohort are spurious. Since smoking, unlike asbestos, is not a strongly suspected risk factor for colorectal cancer, there may be a greater tendency for health-care providers to more aggressively screen for colorectal cancer among asbestos-exposed workers, resulting in detection bias. In fact, many asbestos-exposed cohort participants had been advised to receive screening for colorectal cancer at occu pational health clinics and through targeted educational programs. If there were differential detection of colorectal cancer between these two groups, however, we would expect Colon Cancer among Asbestos-exposed Men 877 Downloaded from aje.oxfordjournals.org by guest on May 10, 2011 "healthier'' than comparable subjects in the general popu lation or, alternatively, more worried about their health. For this reason, all comparisons we made--even the ``external'' comparison with the heavy-smoker cohort--were ``inter nal'' to the CARET population. As such, our results should not be extrapolated uncritically to women, men with lower levels of exposure to asbestos, or nonsmokers. Despite these limitations, this study has many important strengths. We were able to identify an appropriate compar ison group with nominal exposure to asbestos. Participants were identified and enrolled in the study before they devel oped colorectal cancer. Characterization of participants at baseline was exhaustive and standardized, including consis tent interpretations of the occupational and smoking histo ries and chest radiographs. The asbestos-exposed cohort of CARET is very diverse and likely representative of the many occupationally exposed men for whom the risk of colon cancer is a clinically relevant issue. In conclusion, we have provided new evidence consistent with the hypothesis that asbestos exposure leads to in creased risk of colorectal cancer. An apparent dose-response relation was observed for those men with radiographic changes, similar to that seen in this CARET population for lung cancer (39), although colorectal cancer occurs only about half as often. Unlike our lung cancer results, no clear effect was evident on colon cancer risk for those without pleural changes or asbestosis on radiograph, raising the question of whether chest radiographic findings reflect dose or host susceptibility to asbestos in both the lungs and the gastrointestinal tract. Neither selection bias nor residual confounding by diet or other risk factors appears to be a better explanation for our observations than a causal link, especially for those with asbestos-associated changes on radiograph who have also been heavy smokers. We con clude that, for such men, previous recommendations for colorectal cancer screening appear well founded. ACKNOWLEDGMENTS This study was supported by National Cancer Institute grant U01 CA63673. Conflict of interest: none declared. REFERENCES 1. Gamble JF. Asbestos and colon cancer: a weight of evidence review. Environ Health Perspect 1994;102:1038-50. 2. Selikoff IJ, Churg J, Hammond EC. Asbestos exposure and neoplasia. JAMA 1964;188:22-6. 3. Selikoff IJ, Hammond EC, Seidman H. Mortality experience of insulation workers in the United States and Canada, 1943 1976. Ann N Y Acad Sci 1979;330:91-116. 4. Puntoni R, Vercelli M, Merlo F, et al. Mortality among ship yard workers in Genoa, Italy. Ann N Y Acad Sci 1979;330: 353-77. 5. Hilt B, Langard S, Andersen A, et al. Asbestos exposure, smoking habits, and cancer incidence among production and Am J Epidemiol 2005;162:868-878 maintenance workers in an electrochemical plant. Am J Ind Med 1985;8:565-77. 6. Zoloth S, Michaels D. Asbestos disease in sheet metal work ers: the result of a proportional mortality analysis. Am J Ind Med 1985;7:315-21. 7. Seidman H, Selikoff IJ, Gelb SK. Mortality experience of amosite asbestos factory workers: dose-response relationships 5 to 40 years after onset of short-term work exposure. Am J Ind Med 1986;10:479-514. 8. Albin M, Jacobsson K, Attewell R, et al. Mortality and cancer morbidity of asbestos cement workers and referents. Br J Ind Med 1990;47:602-10. 9. Jakobsson K, Albin M, Hagmar L. Asbestos, cement, and cancer in the right part of the colon. Occup Environ Med 1994;51:95-101. 10. Raffin E, Villadsen E, Lynge E. Colorectal cancer in asbestos cement workers in Denmark. Am J Ind Med 1996;30:267-72. 11. Szeszenia-Dabrowska N, Wilczynska U, Szymczak W, et al. Environmental exposure to asbestos from indiscriminate use of industrial wastes. Int J Occup Med Environ Health 1998;11: 171-7. 12. Fredricksson M, Bengston NO, Hardell L, et al. Colon cancer, physical activities and occupational exposures. A case-control study. Cancer 1989;63:1838-42. 13. Gerhardsson de Verdier M, Plato N, Steineck G, et al. Occu pational exposures and cancer of the colon and rectum. Am J Ind Med 1992;22:291-303. 14. Vineis P, Ciccone G, Magnimo A. Asbestos exposure, physical activity and colon cancer: a case control study. Tumori 1993; 79:301-3. 15. Kang SK, Burnett CA, Freund E, et al. Gastrointestinal mor tality of workers in occupations with high asbestos exposures. Am J Ind Med 1997;31:713-18. 16. Goldberg MS, Parent ME, Siemiatycki J, et al. A case-control study of the relationship between the risk of colon cancer in men and exposures to occupational agents. Am J Ind Med 2001;39:531-46. 17. Homa DM, Garabrant DH, Gillespie BW. A meta-analysis of colorectal cancer and asbestos exposure. Am J Epidemiol 1994;139:1210-22. 18. Hodgson JT, Jones RD. Mortality of asbestos workers in England and Wales 1971-81. Br J Ind Med 1986;43:158-64. 19. Gardner MJ, Winter PD, Pannett B, et al. Follow-up study of workers manufacturing asbestos cement products. Br J Ind Med 1986;43:726-32. 20. Garabrant DH, Peters RK, Homa DM. Asbestos and colon cancer: lack of association in a large case-control study. Am J Epidemiol 1992;135:843-53. 21. Tulchinsky TH, Ginsberg GM, Iscovich J, et al. Cancer in ex asbestos cement workers in Israel, 1953-1992. Am J Ind Med 1999;35:1-8. 22. Dement JM, Okun AH. Follow-up study of chrysotile asbestos textile workers: cohort mortality and case-control analyses. Am J Ind Med 1994;26:431-47. 23. Anderssen A. Mortality and cancer morbidity in workers from an aluminum smelter with prebaked carbon anodes--part II: cancer morbidity. Occup Environ Med 1994;52:250-4. 24. Giaroli C, Belli S, Bruno C, et al. Mortality study of asbestos cement workers. Int Arch Occup Environ Health 1994;66:7-11. 25. Levin JL, McLarty JW, Hurst GA, et al. Tyler asbestos work ers: mortality experience in a cohort exposed to amosite. Occup Environ Med 1998;55:155-60. 26. Demers RY, Burns PB, Swanson GM. Construction occupa tions, asbestos exposure, and cancer of the colon and rectum. J Occup Med 1994;36:1027-31.