Document 7ed1b4OZNrXV5R0wBN2n7G9o

American Journal of Epidemiology Copyright 2005 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 161, No. 3 Printed in U.S.A. DOI: 10.1093/aje/kwi034 Predictors of Lung Cancer among Asbestos-exposed Men in the p-Carotene and Retinol Efficacy Trial Mark R. Cullen1, Matt J. Barnett2, John R. Balmes3, Brenda Cartmel1, Carrie A. Redlich1, Carl A. Brodkin4, Scott Barnhart4, Linda Rosenstock5, Gary E. Goodman2,6, Sam P. Hammar7, Mark D. Thornquist2, and Gilbert S. Omenn8,9 1 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 Department of Medicine, University of California School of Medicine, San Francisco, CA. 4 Harborview Medical Center, Seattle, WA. 5 School of Public Health, University of California, Los Angeles, Los Angeles, CA. 6 Swedish Medical Center Cancer Institute, Seattle, WA. 7 Pathology Associates of Kitsap County/Diagnostic Specialties Laboratory, Bremerton, WA. 8 Departments of Internal Medicine and Human Genetics, University of Michigan Medical School, Ann Arbor, MI. 9 Department of Public Health, University of Michigan School of Public Health, Ann Arbor, MI. Received for publication February 17, 2004; accepted for publication August 20, 2004. Despite numerous published studies, debate continues regarding the risk of developing lung cancer among men exposed occupationally to asbestos, particularly those without radiographic or functional evidence of asbestosis. The p-Carotene and Retinol Efficacy Trial (CARET), a study of vitamin supplementation for chemoprevention of lung cancer, has followed 4,060 heavily exposed US men for 9-17 years. Lung cancer incidence for 1989-2002 was analyzed using a stratified proportional hazards model. The study confirmed excessive rates of lung cancer among men with radiographic asbestosis. Comparison of study arms revealed a strong, unanticipated synergy between radiographic profusion category and the active intervention. In the large subgroup of men with normal lung parenchyma on chest radiograph at baseline, there was evidence of exposurerelated lung cancer risk: Men with more than 40 years' exposure in high-risk trades had a risk approximately fivefold higher than men with 5-10 years, after adjustment for covariates. The effect in these men was independent of study intervention arm, but pleural plaques on the baseline radiograph and abnormal baseline flow rate were strong independent predictors of subsequent lung cancer. Residual confounding by subclinical asbestosis, exposure to unmeasured lung carcinogens, or differences in smoking are unlikely to explain these observations better than a carcinogenic effect of asbestos per se. asbestos; asbestosis; beta carotene; clinical trial [publication type]; lung neoplasms; occupational exposure; vitamin A Abbreviations: CARET, p-Carotene and Retinol Efficacy Trial; FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity; ILO, International Labour Organization. Numerous lung cancer cohort studies of men exposed occupationally to asbestos in various industries and trades have been published since Doll's groundbreaking observa tion in 1955 (1-12). Collectively, these studies have estab lished that asbestos causes lung cancer in humans, although the dose-response relation varies substantially from cohort to cohort. Toxicologists, environmental scientists, and epide- miologists have proffered competing explanations for observed differences, including differences in fiber type, fiber dimensions, and the presence of environmental cofac tors (5, 13-16). Substantial synergy with tobacco smoke has been supported in most studies (17-21). Although several case series on clinically characterized subjects have supple mented this epidemiologic literature (22-24), the utility of Correspondence to Dr. Mark Cullen, Department of Internal Medicine, Yale University School of Medicine, 135 College Street, Room 366, New Haven, CT 06510 (e-mail: mark.cullen@yale.edu). 260 Am J Epidemiol 2005;161:260-270 Predictors of Lung Cancer in Asbestos-exposed Men 261 available data for clinical management of the millions of formerly exposed survivors, and the even larger numbers with environmental exposure (25), remains limited. Using various strategies for retrospective exposure quantification, these studies have confirmed high risk in the most heavily exposed persons, especially those with radiographically apparent asbestosis, but findings differ for those with lesser exposure. Studies of the predictive value of pleural plaques have yielded contradictory results (26-28). The predictive role of lung function in asbestos-exposed subjects remains unexplored. Particularly controversial is the risk of lung cancer among persons who do not have clinical evidence of asbestosis. As Cullen (13) has reviewed in detail, some investigators have suggested that such subjects remain at excess risk for lung cancer proportional to their exposure (4, 26, 29-32). Others have argued that lung cancer in this setting is a complication of asbestosis, not a primary consequence of exposure in the absence of fibrosis (33-35). If confirmed, such a theory would downgrade estimates of lung cancer risk among persons exposed to asbestos at doses too low to cause fibrosis, and would provide the basis for reassurance to more heavily exposed persons whose radiographs remained normal or demonstrated only nonmalignant pleural changes. To address some of these questions, we analyzed outcomes among 4,060 asbestos-exposed men followed prospectively for 9-17 years. These men were recruited from diverse jobs and occupations involving asbestos exposure as participants in the P-Carotene and Retinol Efficacy Trial (CARET). CARET is a multicenter randomized, double- blinded, placebo-controlled trial of the efficacy of daily pharmacologic doses of vitamin A and P-carotene in preventing lung cancer among heavy smokers and asbestos workers that was begun in the mid-1980s (36). The interven tion was discontinued 21 months ahead of schedule in 1996 when accrued evidence proved that the vitamins did not prevent lung cancer and strongly suggested increased risk of lung cancer, an effect most pronounced among the asbestos workers (37, 38). Follow-up of participants has continued to the present. In this analysis, we addressed the following questions: 1. Among men with asbestosis based on standardized (Inter national Labour Organization (ILO)) reading of a plain chest radiograph, what is the relative risk of lung cancer, adjusted for both intensity and timing of smoking? 2. Among men without radiographic evidence of fibrosis of the lung parenchyma, is there evidence of increased lung cancer risk? If so, how does this risk vary by occupational history, the presence or absence of pleural plaques, and baseline lung function? 3. Do any of these characteristics interact with the interven tion vitamins? Because CARET recruitment was designed to select partici pants at the highest possible risk of lung cancer, men with less than 5 years in a heavily exposed occupation and those who had never smoked or had quit smoking more than 15 years before study entry were excluded (except in a small pilot phase); women were also excluded, since too few were deemed potentially eligible. For this reason, we are unable to Am J Epidemiol 2005;161:260-270 address the issue of risk among women, nonsmokers, or men with brief or low-level asbestos exposures. MATERIALS AND METHODS Recruitment and enrollment The strategies used for recruitment, enrollment, random ization, and follow-up of the 7,965 male heavy smokers and 4,060 asbestos-exposed men in the CARET study have been previously described (39, 40), as have the baseline character istics of the asbestos-exposed subcohort (41). Asbestosexposed men were eligible for enrollment during the years 1989-1993 if they were between 45 and 69 years of age, currently smoked or had quit smoking within the previous 15 years, and had documentation of exposure. Exposure was documented by 1) employment in a trade with established regular asbestos exposure and published risk of asbestosrelated diseases--insulation, sheet metal work, plumbing, plasterboard application, shipfitting, ship electrical work, boilermaking, or ship scaling--for at least 5 years, starting at least 15 years previously, or 2) occupational asbestos expo sure in any job or occupation and evidence of radiographic changes consistent with a diagnosis of nonmalignant asbestos-related disease. This included 1) benign pleural disease, defined as thickening or fibrotic plaques on pleural surfaces of the lung bilaterally, and/or 2) asbestosis, defined as diffuse lung scarring based on increased profusion of small irregular shadows bilaterally. Asbestos-exposed men were recruited at five US study centers: Seattle, Washington; Baltimore, Maryland; Portland, Oregon; New Haven, Connecticut; and San Francisco, California. A total of 3,244 men were enrolled; most were referred by occupational health clinics, employers, unions, or compensation lawyers or responded to advertisements. Previously, between 1985 and 1988, 816 men had been enrolled in a pilot phase of the study in Seattle using similar criteria but a wider age span (45-74 years) and no smoking requirement. All participants signed consent forms that had been reviewed by a human subjects protection committee before baseline evaluation. Baseline evaluation Prior to randomization to daily receipt of P-carotene (30 mg) plus retinyl palmitate (25,000 IU) or placebo, partici pants in the asbestos-exposed cohort had a plain chest radio graph and standardized spirometric measurements taken using methods published previously (41). Briefly, radio graphs were interpreted by a single trained "B-reader" at each study center, using ILO standard films to quantify the profusion of small irregular shadows in the lung paren chyma, using a 12-point scale from 0/- to 3/+. Radiographs rated 0/-, 0/0, and 0/1 are considered to provide evidence that lung fibrosis is scant or absent (i.e., normal), whereas those rated 1/0 or higher are considered to provide evidence for asbestosis in exposed men. To evaluate consistency in the readings from center to center, the Seattle center panel reread a sample of 48 films from the individual sites; exact agree ment was obtained for 44-50 percent of the films, and agree ment within one minor category was obtained for 79-91 262 Cullen et al. percent. There was 88-93 percent agreement on the presence or absence of pleural change (41). After complete enrollment and randomization, the asbestos-exposed cohort included 34 percent men who qual ified by virtue of their work history alone (5 or more years in one of the designated trades), 21 percent who qualified by radiographic criteria alone, and 44 percent who met both types of entry criteria. Approximately 0.5 percent of those randomized were found upon review to be ineligible and were subsequently excluded from the analysis, along with 133 participants from the pilot study who were lifelong nonsmokers and 10 with inadequate baseline information. This left 3,897 participants for these analyses. Of these subjects, 79 percent had incurred their primary asbestos exposure in one of the designated trades; 21 percent had worked at one or more of 420 different asbestos-exposed jobs in over 260 industries. Because of the dearth of quanti tative exposure data from most workplaces, no effort was undertaken to further classify participants by exposure dose, fiber type, or distribution of fiber sizes. Duration of exposure was used instead as a surrogate for dose where appropriate. Radiographic parenchymal changes indicating asbestosis (ILO grade >1/0) were found in 39 percent of the partici pants; 47 percent had pleural abnormalities diagnosed at baseline. To facilitate internal comparisons within the CARET asbestos cohort and external comparison with the only appropriate reference group--the CARET smoker cohort-- we further subdivided the asbestos-exposed cohort by smoking status, eligibility criteria, and baseline radiographic findings (figure 1). Of the asbestos-exposed men, 1,839 also met the more rigorous criteria for eligibility in the heavysmoker arm of CARET: quitting no less recently than 6 years, having at least 20 pack-years of accumulated tobacco smoking, and being at least 50 years of age at enrollment. This subgroup is identified as the "smoker-eligible partici pants" to distinguish them from the larger cohort of all asbestos-exposed participants, of whom 2,195 had less than 20 pack-years of smoking, had quit more than 6 years before, or were younger than age 50 at randomization. A second subcohort has been designated "work-history-eligible." These 3,067 men met the formal occupational exposure criterion for entry, that is, more than 5 years in one of the designated high-risk trades beginning more than 15 years prior to randomization. Excluded from this subcohort were the 830 men exposed in other occupations, who were enrolled on the basis of asbestos-related radiographic changes. Within the work-eligible subcohort, a subset of participants had normal lung parenchyma at baseline (ILO grade <1/0). These 2,089 participants constituted the "workeligible ILO <1/0" subcohort. Follow-up Prior to discontinuation of the active intervention in January 1996, participants were contacted by their local study center three times per year and evaluated in person at least once. During the active phase of the trial, lung function tests were repeated annually but not chest radiographs. After active intervention was stopped, contact remained local but FIGURE 1. Subcohorts delineated within the asbestos exposure cohort of the p-Carotene and Retinol Efficacy Trial (CARET). "Smoker-eligible" participants met separate age and smoking criteria for inclusion in the smoking arm of the CARET study. "Work-historyeligible" participants met entrance criteria for having worked at least 5 years in a designated trade involving high asbestos exposure. Among these persons, a subset of approximately two thirds had no evidence of lung parenchymal asbestosis (International Labour Organization (ILO) grade 0) on the chest radiograph at the time of randomization in CARET. was reduced to an annual phone call until April 2000. Since then, all follow-up has been performed annually by mail and phone by the staff of the CARET Coordinating Center in Seattle. All reports of incident cancer or cancer mortality are confirmed by the CARET Endpoints Committee through review of clinical records and pathology reports. The current analysis included follow-up information that was complete through October 31, 2002. Statistical analysis Stratified Cox proportional hazards models were used to obtain lung cancer relative risk estimates and 95 percent confidence intervals. The time axis in the models was the time to diagnosis of lung cancer or the date on which the participant was last known to be alive. Mean length of follow-up was 10 years among asbestos-exposed partici pants and 9 years in the heavy-smoker cohort. The assump tion of proportional hazards was assessed with timedependent covariates (the product of log-transformed time and the factor of interest) and examination of log cumulative hazard plots. In all models, data were stratified on enroll ment period (pilot or efficacy phase) to account for differ ences in baseline hazards due to changes in eligibility criteria. Models included adjustment for intervention arm assignment (vitamin A + p-carotene vs. placebo), baseline smoking status (current smoker vs. former smoker), packyears of smoking (<40, 40-60, or >60), and age (<55, 55-64, or >65 years). Age and pack-years were modeled as grouped linear variables, constructed by assigning ordinal scores to the categories, and fitted as continuous variables. Analyses restricted to the asbestos-exposed workers were further stratified on enrollment center (except in smokereligible comparisons, because only two of the six CARET study centers recruited participants for both the asbestos and Am J Epidemiol 2005;161:260-270 Predictors of Lung Cancer in Asbestos-exposed Men 263 TABLE 1. Demographic characteristics of participants by exposure population and lung cancer status, p-Carotene and Retinol Efficacy Trial, 1989-2002 Characteristic All asbestos-exposed subjects** * Asbestos-exposed cohort Work-historyeligiblet Work-history-eligible with ILOf grade <1 Smoker-eligible^ Smoker cohort# Total LCf cases No. of participants 3,897 241 Mean age (years) (mean (SDf)) Age group (years) (no. (%)) 57 (7) 61 (6) <55 1,585 (41) 46 (19) 55-64 >65 Smoking status at enrollment (no. (%)) Current smoker 1,616 (41) 114 (47) 696 (18) 81 (34) 1,548 (40) 127 (53) Former smoker 2,349 (60) Years since quitting smoking (former smokers) (no. (%)) 114 (47) 0^ 636 (27) 38 (33) 5-6 365 (16) 18 (16) 7-9 351 (15) 16 (14) 10-14 595 (25) 31 (27) >14 Mean pack-years of smoking (mean (SD)) Pack-years of smoking (no. (%)) <40 41-60 402 (17) 11 (10) 43 (24) 55 (26) 2,051 (53) 79 (33) 1,127 (29) 82 (34) >60 719 (18) 80 (33) Mean years of asbestos exposure (mean (SD)) Mean years in a high-risk trade (mean (SD)) 27 (10) 19 (13) 29 (11) 22 (14) Total 3,067 56 (7) LC cases 197 60 (7) 1,360 (44) 1,205 (39) 502 (16) 42 (21) 92 (47) 63 (32) 1,235 (40) 103 (52) 1,832 (60) 94 (48) 480 (26) 274 (15) 281 (15) 468 (26) 329 (18) 31 (33) 13 (14) 13 (14) 28 (30) 9 (10) 42 (24) 54 (26) 1,652 (54) 874 (28) 541 (18) 65 (33) 69 (35) 63 (32) 27 (10) 29 (10) 24 (10) 27 (11) Total 2,089 55 (7) LC cases 97 60 (7) 1,036 (50) 764 (37) 289 (14) 25 (26) 45 (46) 27 (28) 814 (39) 48 (49) 1,275 (61) 49 (51) 331 (26) 187 (15) 196 (15) 324 (25) 237 (19) 15 (31) 6 (12) 9 (18) 15 (31) 4 (8) 40 (24) 56 (29) 1,186 (57) 556 (27) 347 (17) 32 (33) 30 (31) 35 (36) 26 (10) 29 (10) 23 (10) 27 (10) Total 1,839 59 (6) LC cases 165 61 (5) 531 (29) 980 (53) 328 (18) 26 (10) 92 (56) 47 (28) 1,105 (60) 113 (68) 734 (40) 52 (32) 476 (65) 258 (35) 35 (67) 17 (33) 53 (23) 61 (27) 618 (34) 678 (37) 543 (30) 38 (23) 58 (35) 69 (42) 28 (10) 30 (10) 19 (14) 22 (15) Total 7,924 58 (5) LC cases 542 61 (5) 2,557 (32) 4,150 (52) 1,217 (15) 81 (15) 322 (59) 139 (26) 5,134 (65) 404 (75) 2,790 (35) 138 (25) 2,020 (72) 108 (78) 770 (28) 30 (22) 53 (22) 61 (22) 2,522 (32) 3,016 (38) 2,386 (30) 97 (18) 222 (41) 223 (41) 3 (9) 4 (10) <1 (<1) <1 (<1) * Excludes 163 participants: never smokers (n = 133), participants who did not meet the asbestos eligibility criteria (n = 24), participants with an unknown number of pack-years of smoking (n = 8), and/or participants who were missing a baseline radiograph (n = 2). t Excludes 993 participants: 830 who were eligible on the basis of radiographic findings only, 133 never smokers, 24 who did not meet the asbestos eligibility criteria, eight with missing data on pack-years of smoking, and two who did not have a baseline radiograph. f ILO, International Labour Organization; LC, lung cancer; SD, standard deviation. ILO classification of radiographic category 0. Excludes 1,971 participants: 978 with a profusion rating greater than 0/1,830 who were eligible on the basis of radiographic findings only, 133 never smokers, 24 who did not meet the asbestos eligibility criteria, eight with missing data on pack-years of smoking, and two who did not have a baseline radiograph. H Excludes 2,195 participants who did not meet the heavy-smoker eligibility criteria (<6 years since quitting smoking, >20 pack-years of smoking, and age 50-69 years). Also excludes 24 participants who did not meet the asbestos eligibility criteria and two participants who were missing a baseline radiograph. # Excludes 12 participants who did not meet the heavy-smoker eligibility criteria (see above footnote) and 29 participants who met the asbestos eligibility criteria. heavy-smoker cohorts) and included additional adjustment for years since quitting smoking (as a categorical variable: current smoking or 0-4, 5-9, 10-14, or >14 years since quit ting). Time since last smoking was excluded from the final smoker-eligible model, because almost 80 percent of the smoker-eligible participants had smoked within 2 years of enrollment. Asbestos-related predictors (years in a high-risk trade, years since last asbestos exposure, primary trade, presence of pleural plaques, and ILO grade) were fitted simultaneously in analyses including the asbestos cohort only. For the smoker-eligible analyses, the asbestos-related variables were assessed in separate Cox models, with CARET's male Am J Epidemiol 2005;161:260-270 heavy-smoker population serving as the referent in each. The association of lung function with lung cancer incidence was examined in the work-eligible ILO <1/0 subcohort by modeling forced expiratory volume in 1 second (FEV1)/ forced vital capacity (FVC) and percent predicted FEV1 and FVC as categorical variables (>80 percent, 70-79 percent, 60-69 percent, or <60 percent), adjusted for work history and presence of pleural abnormalities. Because lung function may be in the causal pathway between asbestos exposure and lung cancer incidence (42), we also examined estimates unadjusted for work history (data not shown); no appreciable differences were observed. We performed likelihood ratio tests to test for linear trend across categories where appro- 264 Cullen et al. TABLE 2. Radiographic predictors of lung cancer incidence among asbestos-exposed participants in the smoker-eligible subcohort,* ^-Carotene and Retinol Efficacy Trial, 1989-2002 Risk population Heavy smoker Asbestos-exposed Pleural abnormality^ Heavy smoker Negative Positive Profusion rating Major categories Heavy smoker 0/- to 0/1 1/0 to 1/2 2/1 to 2/3 3/2 to 3/+ Minor categories Heavy smoker 0/0/0 0/1 1/0 1/1 1/2 2/1 2/2 2/3 3/2 3/3 3/+ Total no. No. of lung cancer cases Relative riskt 95% confidence interval 7,924 1,839 542 165 1.00 1.24 1.04, 1.47 7,924 953 886 542 71 94 1.00 1.05 0.82, 1.34 1.44 1.15, 1.79 7,924 1,007 769 47 16 542 69 83 8 5 1.00 0.98 1.42 2.35 4.42 0.76, 1.26 1.12, 1.79 1.17, 4.72 1.83, 10.69 7,924 542 1.00 2 0 0.00 446 26 0.89 0.60, 1.32 559 43 1.05 0.77, 1.44 510 52 1.34 1.01, 1.79 207 24 1.52 1.01, 2.29 52 7 1.72 0.82, 3.63 25 4 2.17 0.81, 5.81 15 4 3.45 1.29, 9.24 7 0 0.00 7 1 1.82 0.26, 13.0 7 3 5.78 1.85, 18.05 2 1 15.62 2.17, 112.3 p value 0.02 0.01 0.71 0.001 <0.0001H 0.89 0.003 0.02 0.001 <0.0001H 0.57 0.75 0.04 0.04 0.15 0.12 0.01 0.55 0.002 0.006 * 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. Also excludes 41 heavy smokers: 12 who did not meet the heavy-smoker eligibility criteria and 29 who met the asbestos eligibility criteria. t Risk estimate from a Cox regression model with stratification by cohort and adjustment for age (<55, 55-64, or >65 years), smoking status at baseline (current smoker or former smoker), pack-years of smoking (<40, 40-60, or >60), and intervention arm. Presence of bilateral pleural thickening or plaques on radiograph at baseline, with or without calcification. Density of small irregular shadows in the lung fields using the International Labour Organization's 12-point rating scale. H Test for trend using a grouped linear variable. priate and to test whether the associations between asbestosrelated measures and lung cancer risk were modified by age, smoking status, pack-years, and intervention assignment. Intervention assignment-specific relative risk estimates were obtained from models that included cross-product interac tion terms between intervention arm and each predictor of interest. The association between pack-years and years of working in a high-risk trade was assessed with Pearson and Spearman correlations; only the Pearson estimates are presented, since differences between the two were negli gible. All significance tests were two-sided. Analyses were performed using SAS, version 8.2 (SAS Institute, Inc., Cary, North Carolina). RESULTS A total of 241 cases of lung cancer were observed among the 3,897 men in the asbestos-exposed cohort, giving an Am J Epidemiol 2005;161:260-270 Predictors of Lung Cancer in Asbestos-exposed Men 265 TABLE 3. Radiographic predictors of lung cancer incidence among all asbestos-exposed participants,* ^-Carotene and Retinol Efficacy Trial, 1989-2002 Profusion ratingt 0/- or 0/0 0/1 1/0 to 1/2 2/1 to 2/3 3/2 to 3/+ Total no. 1,181 1,184 1,424 84 24 Overall No. of LCf cases RRf, 40 1.00 69 1.48 113 1.93 12 2.90 7 6.76 95% CIf 0.99, 2.22 1.32, 2.84 1.47, 5.73 2.80, 16.3 Active Total no. No. of LC cases RR 609 17 1.00 574 27 1.40 728 66 2.72 36 7 5.25 15 6 12.45 95% CI 0.76, 2.60 1.57, 4.73 2.12, 13.0 4.47, 34.6 Placebo Total no. No. of LC cases RR 572 23 1.00 610 42 1.55 696 47 1.37 48 5 1.65 9 1 1.92 95% CI 0.92, 2.61 0.82, 2.31 0.61,4.48 0.25, 14.9 p for trendy p for interaction# <0.0001 <0.0001 0.001* * * 0.25 * Excludes 163 participants: 133 never smokers, 20 who did not meet the asbestos eligibility criteria, eight with missing data on pack-years of smoking, and two who did not have a baseline radiograph. t Density of small irregular shadows in the lung fields using the International Labour Organization's 12-point rating scale. f LC, lung cancer; RR, relative risk; CI, confidence interval. Risk estimate from a Cox regression model with stratification by cohort and study center and inclusion of the following covariates: age (<55, 55-64, or >65 years), years since quitting smoking (current smoking or 0--4, 5-9, 10-14, or >14 years), pack-years of smoking (<40, 40-60, or >60), trade (eight specified high-risk trades and an "other" category), intervention arm (active vitamins or placebo), years of working in a high-risk trade (<10, 11-20, 21-30, 32-40, or >40), presence of pleural abnormality (yes/no), and years since last asbestos exposure (0, 1-5, 6-10, 11-15, or >15). ! Test for trend using a grouped linear variable. # Test of interaction for intervention arm assignment x profusion rating coded as a grouped linear variable. overall incidence rate of 5.9 per 1,000 person-years (95 percent confidence interval: 5.2, 6.7). The demographic, smoking, and occupational histories of men in the cohort and the lung cancer cases, as well as the three subcohorts (figure 1), are shown in table 1. Also shown for comparison are the same parameters for the 7,924 male heavy smokers; 29 were found to have asbestos exposure sufficient to be eligible for the asbestos cohort and were excluded. As anticipated, men in the asbestos-exposed smokereligible subcohort had a significantly higher rate of lung cancer than their counterpart smokers (table 2). Baseline radiographic status was a strong predictor, as is shown in the same table. Participants with pleural plaques or other asbestos-related pleural changes had a 44 percent higher risk of lung cancer than the unexposed heavy smokers; the subgroup without plaques did not have significantly higher risk. Parenchymal changes on radiograph were associated with progressively increasing relative risk, evident at the lowest level of pulmonary fibrosis by radiograph: ILO profu sion ratings 1/0-1/2. The relative risk of men in this group as compared with the heavy smokers without asbestos exposure was 1.42. Men with advanced asbestosis (grade 3) had 4.4 times' greater risk; those with grade 2 changes upon radio graph had intermediate risk. To better estimate risk among men with normal lung parenchyma on the baseline radiograph, we performed an internal analysis of the entire asbestos-exposed cohort. Table 3 shows the relative risk of lung cancer in the entire cohort, by ILO category. After adjustment for age, pack-years of smoking, years since quitting smoking, presence of pleural abnormalities, and high-risk trade exposure and type, the risk of lung cancer in participants with radiographs graded 0/1 (borderline normal) was 48 percent higher than in those Am J Epidemiol 2005;161:260-270 with ILO grade 0/0--an excess of borderline statistical significance. While the numbers of participants in the higher categories dwindled, there was an increase in risk with rising category, as in the smoker-eligible subcohort. Table 3 demonstrates that when data were stratified across treatment arms, there was a strong and previously unexpected "synergy" between asbestosis grade and the active intervention arm (p < 0.0001). Among participants receiving placebo, only a small, nonsignificant risk gradient was seen with rising profusion score, whereas among those taking active vita mins, the gradient was steep and significant, rising to over 12-fold in those with advanced (ILO grade >3/2) radio graphic changes. To evaluate the predictive significance of exposure history per se independently of radiographic change, we assessed the subcohort of participants who were eligible for the study on the basis of having worked for 5 or more years in one of the preassigned "high-risk trades," excluding the 830 partic ipants who had entered the study on the basis of an asbestosrelated radiographic abnormality at baseline. Among these exposure-eligible participants, there was a significant predictive effect of years in the trade (table 4), with risk rising to 2.3 from the least exposed categories to the most exposed. Intervention arm did not modify this finding. The effect of radiograph in this subcohort mirrors that seen in the full cohort. Variation by year of last exposure shows signifi cantly increased risk among persons last exposed 5 or more years prior to randomization, which is consistent with likely heavier exposure per year in earlier years. Given the uncertainty about whether asbestos exposure, as opposed to asbestosis, presents a quantifiably increased risk of lung cancer, we examined the exposure surrogate-- years of exposure in a high-risk trade--in the subgroup of 266 Cullen et al. TABLE 4. Predictors of lung cancer incidence among work-history-eligible asbestos-exposed participants,* p-Carotene and Retinol Efficacy Trial, 1989-2002 Overall Active Placebo Years in a high-risk trade <10 11-20 21-30 31-40 >40 p for trendy p for interaction# Pleural abnormality*** * Negative Positive p for main effect! p for interaction# Profusion ratingtt 0/- to 0/1 1/0 to 1/2 2/1 to 2/3 3/2 to 3/+ p for trend! p for interaction# Years since last asbestos exposure 0 1-5 6-10 11-15 >15 p for trend! p for interaction# TMno. 1LC^t +cas;e^s RRt,t 95% CIt no. LC cases RR 95% CI no. LC cases RR 95% CI 410 776 1,003 722 156 19 1.00 219 11 1.00 191 8 1.00 39 1.26 0.72, 2.21 390 23 1.42 0.67, 3.01 386 16 1.10 0.46, 2.59 59 1.60 0.92, 2.77 491 25 1.26 0.60, 2.67 512 34 1.94 0.88, 4.30 59 1.68 0.92, 3.08 362 30 1.32 0.61,2.88 360 29 1.96 0.85, 4.54 21 2.31 1.12, 4.76 72 9 1.86 0.70, 4.93 84 12 3.05 1.15, 7.89 0.02 0.40 0.008 0.14 1,769 86 1.00 878 40 1.00 891 46 1.00 1,298 111 1.43 1.06, 1.94 656 58 1.50 0.97, 2.31 642 53 1.34 0.89, 2.03 0.02 0.07 0.17 0.72 2,089 902 57 19 97 1.00 1,036 37 1.00 1,053 60 1.00 85 1.63 1.20, 2.22 462 50 2.52 1.63, 3.91 440 35 1.10 0.71, 1.69 8 1.92 0.89, 4.13 24 5 4.21 1.59, 11.1 33 3 0.91 0.28, 3.03 7 6.23 2.63, 14.8 12 6 13.0 4.90, 34.7 7 1 1.59 0.21, 12.1 <0.0001 <0.0001 0.70 0.0002 868 38 1.00 436 16 1.00 432 22 1.00 587 35 0.96 0.59, 1.54 281 16 1.02 0.50, 2.08 306 19 0.89 0.48, 1.68 475 49 1.78 1.11,2.84 237 26 2.23 1.14, 4.36 238 23 1.53 0.82, 2.86 748 45 1.61 0.93, 2.80 377 23 1.80 0.87, 3.74 371 22 1.46 0.73, 2.92 389 30 1.78 1.00, 3.20 203 17 2.52 1.16, 5.46 186 13 1.31 0.61,2.80 0.01 0.008 0.20 0.23 * Excludes 993 participants: 830 who were eligible on the basis of radiographic findings only, 133 never smokers, 20 who did not meet the asbestos eligibility criteria, eight with missing data on pack-years of smoking, and two who did not have a baseline radiograph. t LC, lung cancer; RR, relative risk; CI, confidence interval. ^ Risk estimate from a Cox regression model with stratification by cohort and study center and inclusion of the following covariates: age (<55, 55-64, or >65 years), years since quitting smoking (current smoking or 0--4, 5-9, 10-14, or >14 years), pack-years of smoking (<40, 40-60, or >60), trade (eight specified high-risk trades and an "other" category), intervention arm (active vitamins or placebo), and the asbestos-related predictors defined above. Risk estimate from a Cox regression model as defined above that also included interaction terms for intervention arm assignment x profusion rating and arm x predictor of interest. ! Test for trend using a grouped linear variable. # Test of interaction for intervention arm assignment x predictor of interest coded as a grouped linear variable. ** Presence of bilateral pleural thickening or plaques upon radiograph, with or without calcification. tt Density of small irregular shadows in the lung fields using the International Labour Organization's 12-point rating scale. exposure-history-eligible participants who had normal radiographs on the ILO scale (ILO grade <0/1) (table 5). In this subcohort, the gradient of increased risk with increasing duration of exposure was stepwise and signifi cant, reaching a fivefold excess in the most heavily exposed groups versus the least heavily exposed after adjustment for the covariates. Pleural change remained predictive; it was associated with approximately a doubling of lung cancer risk after adjustment. Neither of these results seemed to have been modified by the study intervention vitamins; if anything, trends were more conspicuous among the participants receiving placebo. We conducted additional analyses to address two of the more plausible explanations for the exposure-years-response trend among persons with normal lung parenchyma upon radiograph: 1) residual confounding by tobacco--that is, more smoking among men in the groups with longer expo sure, even after adjustment for pack-years--and 2) misclassification of "asbestosis" status--that is, inclusion of more participants with radiographically undetected asbestosis in Am J Epidemiol 2005;161:260-270 Predictors of Lung Cancer in Asbestos-exposed Men 267 TABLE 5. Predictors of lung cancer incidence among work-history-eligible asbestos-exposed participants with an International Labour Organization rating less than 1/0,**^*-C arotene and Retinol Efficacy Trial, 1989-2002 Overall Active Placebo Years in a high-risk trade <10 11-20 21-30 31-40 >40 p for trend p for interaction! Pleural abnormality# Negative Positive p for trend p for interaction! Years since last asbestos exposure 0 1-5 6-10 11-15 >15 p for trend p for interaction!] TMno. 1LC^t +cas;e^s RRt,f 95% CIt 272 6 1.00 570 23 2.07 0.82, 5.19 720 30 2.50 0.97, 6.42 441 26 3.15 1.12, 8.85 86 12 5.17 1.61, 16.6 0.007 1,325 47 1.00 764 50 1.91 1.25, 2.92 0.003 611 18 1.00 370 17 1.07 0.53, 2.16 313 25 2.10 1.07, 4.10 539 21 1.64 0.74, 3.62 256 16 2.79 1.23, 6.34 0.01 no. LC cases RRf 95% CI no. LC cases RRf 95% CI 152 4 1.00 120 2 1.00 285 10 1.59 0.49, 5.17 285 13 3.08 0.68, 13.9 344 9 1.36 0.40, 4.71 376 21 4.55 1.02, 20.2 212 10 2.19 0.61,7.94 229 16 4.98 1.05, 23.8 43 4 2.29 0.48, 10.9 43 8 10.92 2.07, 57.7 0.23 0.003 0.23 654 19 1.00 671 382 18 1.91 0.98, 3.73 382 0.06 0.99 28 1.00 32 1.91 1.12, 3.27 0.02 310 8 1.00 301 174 4 0.55 0.16, 1.88 196 153 11 1.93 0.73, 5.06 160 264 8 1.54 0.52, 4.55 275 135 6 2.48 0.76, 8.04 121 0.05 0.82 10 1.00 13 1.49 0.64, 3.50 14 2.18 0.91,5.22 13 1.74 0.66, 4.58 10 3.08 1.14, 8.30 0.04 * Excludes 1,971 participants: 978 with a profusion rating greater than 0/1,830 who were eligible on the basis of radiographic findings only, 133 never smokers, 20 who did not meet the asbestos eligibility criteria, eight with missing data on pack-years of smoking, and two who did not have a baseline radiograph. t LC, lung cancer; RR, relative risk; CI, confidence interval. f Risk estimate from a Cox regression model with stratification by cohort and study center and inclusion of the following covariates: age (<55, 55-64, or >65 years), years since quitting smoking (current smoking or 0--4, 5-9, 10-14, or >14 years), pack-years of smoking (<40, 40-60, or >60), trade (eight specified high-risk trades and an "other" category), intervention arm (active vitamins or placebo), and the asbestos-related predictors defined above. Test for trend using a grouped linear variable. ! Test of interaction for intervention arm assignment x predictor of interest coded as a grouped linear variable. # Presence of bilateral pleural thickening or plaques upon radiograph, with or without calcification. the groups with longer exposure. Neither of these explana tions appeared highly likely. The Pearson correlation between pack-years and asbestos-years, measured continu ously, was only 0.09. Moreover, the mean values and stan dard errors for all measures of lung function (data not shown) were almost identical among the groups. While low- normal mean values suggest that there may have been some cases of subclinical asbestosis within each subgroup, the absence of any gradient by exposure length renders residual confounding by subradiographic asbestosis unlikely. Finally, we evaluated the role of baseline lung function in subsequent lung cancer risk in this work-eligible, ILO <1/0 subcohort. As is shown in table 6, decreasing values of FEVj, FVC, and (especially) FEVpFVC ratio were strongly associated with future lung cancer risk, suggesting an effect of preexisting obstructive lung disease. The risk gradients associated with these functional changes in men without asbestosis appeared to be independent of intervention assignment, with the possible exception of the small group with FVC values less than 60 percent predicted. Am J Epidemiol 2005;161:260-270 DISCUSSION Observational analysis of data from this large, random ized, controlled chemoprevention trial appears to confirm excessive rates of lung cancer among men with asbestosis and demonstrates for the first time synergy between the lung parenchymal changes and the intervention vitamins. Among participants with normal lung parenchyma on chest radio graph, there was also evidence of exposure-related risk; men with more than 40 years of exposure in high-risk trades had a risk fivefold higher than men with 5-10 years of exposure after adjustment for smoking and other risk factors. The presence of pleural plaques and obstructive lung disease at baseline was also associated with significantly increased risk of lung cancer. In the subcohort of men who were free of radiographic asbestosis, no interaction with study vitamins was evident. Although this population was recruited for a randomized trial and, as such, differs from typical populations identified 268 Cullen et al. TABLE 6. Lung-function predictors of lung cancer incidence among work-history-eligible asbestos-exposed participants with an International Labour Organization rating less than 1/0 and valid baseline spirometry measures available,* p-Carotene and Retinol Efficacy Trial, 1989-2002 % predicted FEV1t >80 70-79 60-69 <60 p for trend p for interaction! % predicted FVCt >80 70-79 60-69 <60 p for trend p for interaction! FEV1:FVC ratio >80 70-79 60-69 <60 p for trend p for interaction!] Total no. 1,064 391 219 209 1,364 317 124 78 363 928 420 172 Overall No. of LCt cases RRt,f 95% CIt 23 1.00 14 1.42 0.72, 2.79 15 2.42 1.22, 4.80 31 5.17 2.81,9.50 <0.0001 41 1.00 23 1.85 1.09, 3.16 12 2.47 1.24, 4.93 7 2.35 0.99, 5.56 0.003 4 1.00 25 2.20 0.76, 6.41 28 4.32 1.46, 12.8 26 9.87 3.23, 30.1 <0.0001 Active Total no. No. of LC cases RRJ 95% CI 515 9 1.00 213 7 1.80 0.66, 4.91 105 5 2.10 0.68, 6.47 106 12 4.70 1.88, 11.7 0.001 0.62 684 18 1.00 153 7 1.58 0.64, 3.93 70 7 2.94 1.20, 7.23 32 1 0.81 0.10, 6.34 0.14 0.64 173 1 1.00 484 11 2.99 0.38, 23.4 199 12 6.47 0.82, 51.4 83 9 9.98 1.22, 81.8 0.001 0.54 Placebo Total no. No. of LC cases RRJ 95% CI 549 14 1.00 178 7 1.17 0.47, 2.94 114 10 2.63 1.12, 6.14 103 19 5.55 2.60, 11.8 <0.0001 680 23 1.00 164 16 2.01 1.03, 3.92 54 5 1.96 0.70, 5.51 46 6 3.50 1.31,9.39 0.01 190 3 1.00 444 14 1.90 0.54, 6.72 221 16 3.56 1.01, 12.6 89 17 10.49 2.90, 37.9 <0.0001 * Excludes 2,177 participants: 206 who were missing valid data on baseline spirometry measures, 978 with a profusion rating greater than 0/1,830 who were eligible on the basis of radiographic findings only, 133 never smokers, 20 who did not meet the asbestos eligibility criteria, eight with missing data on pack-years of smoking, and two who did not have a baseline radiograph. t LC, lung cancer; RR, relative risk; CI, confidence interval; FEV,, forced expiratory volume in 1 second; FVC, forced vital capacity. ^ Risk estimate from a Cox regression model with stratification by cohort and study center and inclusion of the following covariates: age (<55, 55-64, or >65 years), years since quitting smoking (current smoking or 0--4, 5-9, 10-14, or >14 years), pack-years of smoking (<40, 40-60, or >60), trade (eight specified high-risk trades and an "other" category), intervention arm (active vitamins or placebo), years of working in a high-risk trade (<10, 11-20, 21-30, 32-40, or >40), presence of pleural abnormality (yes/no), and years since last asbestos exposure (0, 1-5, 6-10, 11-15, or >15). Test for trend using a grouped linear variable. ! Test of interaction for intervention arm assignment x predictor of interest coded as a grouped linear variable. in situ for epidemiologic research, we believe that the breadth of the sample and the completeness of its character ization justify these analyses, which were envisioned at the study's inception, as long as comparisons with any external population are avoided. Still, our findings must be viewed in light of the strengths and limitations of our study design. A strength is that the participants were identified and enrolled in the study prior to development of lung cancer. Although they are not necessarily representative of all occupationally exposed men, the asbestos-exposed cohort in CARET is very diverse and typical of many persons who seek clinical evalu ation because of their exposure. Clinical characterization at baseline was standardized, including consistent interpreta tions of occupational and smoking histories, radiographs, and lung function. Although almost 15 percent of those enrolled in the study discontinued active vitamin use during the course of the study, very few subjects refused to partici pate in the aggressive follow-up effort; thus, ascertainment of lung cancer in this cohort was almost complete. Further more, the rules governing endpoint assessment assured a very high degree of precision regarding diagnosis (39, 40) in comparison with death certificates or the case ascertainment methods used in previous cohort studies (1-12). However, these data were not without limitations. Almost 80 percent of the participants in this study had been exposed to asbestos in shipbuilding or other construction-related trades; manufacturing and other exposure settings were proportionally underrepresented. The diversity of exposures and workplaces limited our ability to quantify exposure beyond the crude surrogate measure used in the analysis-- years of employment in a high-risk trade. Although several attempts have been made in the literature to provide semiquantitative adjustment factors differentiating trades by levels of exposure (43), none was incorporated here, nor was there any information consistently available that would have allowed better exposure classification. Another limitation is that the radiographs were not read by a single panel but rather interpreted by separate B-readers at each study site. Am J Epidemiol 2005;161:260-270 Predictors of Lung Cancer in Asbestos-exposed Men 269 The results of the small study undertaken to assess inter reader agreement (41) revealed some differential classifica tion of the radiographs, though generally by only one minor category. Finally, as with all randomized prospective trials, study participants cannot, as a group, be unmindfully compared with other comparable men at risk. CARET participants were volunteers in a long-term study involving daily ingestion of potentially toxic vitamins, uncertainty regarding treatment arm, and the need to submit to regular examinations and follow-up surveys. By virtue of these selection effects, such volunteers may well be healthier than comparable men in the general population; or, alternatively, they may be more worried because of unmeasured factors. For this reason, all comparisons were internal, that is, made between various subgroups within the study, including the larger heavy-smoker cohort not exposed to asbestos. We made no comparisons with any external population. Like wise, generalization of our results to women, men with lesser amounts of asbestos exposure, or lifetime nonsmokers cannot be inferred from our data. These limitations notwithstanding, certain conclusions can be drawn. We have confirmed previous observations that heavy asbestos exposure causes lung cancer in male smokers with asbestosis and that severity of asbestosis is a strong predictor of risk. The observation of synergy between asbestosis stage and study vitamins is new and provocative but remains as yet unconfirmed and unexplained mechanisti cally. The finding of increased lung cancer risk among men without radiographic evidence of fibrosis is noteworthy, but we cannot resolve the issue of whether this is a primary effect of asbestos or is due to asbestosis developing in the absence of any radiographic findings (44). Hence, we can neither affirm nor refute the contention (33-35) that asbestosis obligatorily mediates the relation between exposure and cancer, although the absence of any interaction with intervention arm in this subcohort may suggest a different carcinogenic pathway. Regardless of whether that speculation withstands further scrutiny, we have provided new evidence that risk of lung cancer is increased in men with ILO-negative radiographs and that the risk rises steeply with years of exposure, contrary to the assertion of Wilkinson et al. (29). Our finding also supports the earlier observation of Hillerdal and Henderson (26) that the presence of pleural plaques on radio graph appears to almost double the risk, contrary to the impression of Partanen et al. (27) and the inference of Edelman (28). We offer no specific hypothesis for this, other than the likelihood that persons with plaques had more expo sure on average than those without them. We also confirmed that obstructive lung dysfunction is a strong independent cancer predictor in this population. Neither residual confounding by tobacco use nor the possibility of differential proportions of subclinical asbestosis in these higher-risk subgroups appears to explain the observations. Impossible to entirely exclude as contributory to these risk factors are the effects of completely unmeasured exposures, such as welding fumes, diesel exhaust, silica, nickel, hexavalent chromium, or other carcinogens to which construction and shipyard workers are exposed, and to whom higher doses might have accrued among those with more years of Am J Epidemiol 2005;161:260-270 working in the asbestos trades (45). However, such expo sures would be unlikely to explain the steep and stepwise dose-response relations observed, since the associations between these exposures and lung cancer have been of generally low relative risk and exposures within our popula tion diverged from trade to trade and work setting to work setting (e.g., general construction vs. shipbuilding). In conclusion, we found that among current and former smokers exposed occupationally to asbestos, risk of lung cancer increases with increased exposure duration, even in persons without clinical evidence of asbestosis. Men with radiographically apparent pleural plaques and preexisting obstructive lung disease are at higher risk than men without them. Whether it is related to the apparently vitamin-sensi tive effect observed in subjects with asbestosis or an inde pendent effect of asbestos, intense occupational exposure, even in the absence of asbestosis, confers significant lung cancer risk in this population. ACKNOWLEDGMENTS This work was supported by National Cancer Institute grant U01 CA63673. REFERENCES 1. Greenberg M. A study of lung cancer mortality in asbestos workers: Doll, 1955. Am J Ind Med 1999;36:331-47. 2. Goodman M, Morgan RW, Ray R, et al. Cancer in asbestosexposed occupational cohorts: a meta-analysis. Cancer Causes Control 1999;10:453-65. 3. Liddell FD, McDonald AD, McDonald JC. Dust exposure and lung cancer in Quebec chrysotile miners and millers. Ann Occup Hyg 1998;42:7-20. 4. Stayner L, Smith R, Bailer J, et al. Exposure-response analysis of respiratory disease associated with occupational exposure to chrysotile asbestos. Occup Environ Med 1997;54:646-52. 5. Dement J, Brown DP, Okun A. Follow-up study of chrysotile asbestos workers: cohort mortality and case-control analyses. Am J Ind Med 1994;26:431-47. 6. Berry G. Mortality and cancer incidence of workers exposed to chrysotile asbestos in the friction-products industry. Ann Occup Hyg 1994;38:539-46. 7. Liddell D. Cancer mortality in chrysotile mining and milling: exposure-response. Ann Occup Hyg 1994;38:519-23. 8. Neuberger M, Kundi M. Individual asbestos exposure: smoking and mortality--a cohort study in the asbestos cement industry. J Occup Med 1990;32:499-507. 9. Hughes JM, Weill H, Hammad YY. Mortality of workers employed in two asbestos cement manufacturing plants. Br J Ind Med 1987;44:161-74. 10. 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. 11. McDonald AD, Fry JS, Woolley AJ, et al. Dust exposure and mortality in an American chrysotile friction products plant. Br J Ind Med 1984;41:151-7. 12. Dement JM, Harris RL Jr, Symons MJ, et al. Exposures and mortality among chrysotile asbestos workers. Part II: mortality. Am J Ind Med 1983;4:421-33. 270 Cullen et al. 13. Cullen MR. Controversies in asbestos-related lung cancer. Occup Med State Art Rev 1987;2:259-72. 14. Landrigan PJ, Nicholson WJ, Suzuki Y, et al. The hazards of chrysotile asbestos: a critical review. Ind Health 1999;37:27180. 15. Stayner LT, Dankovic DA, Lemen RA. Occupational exposure to chrysotile asbestos and cancer risk: a review of the amphibole hypothesis. Am J Public Health 1996;86:179-86. 16. McDonald AD, Fry JS, Woolley AJ, et al. Dust exposure and mortality in an American chrysotile textile plant. Br J Ind Med 1983;40:361-7. 17. Lee PN. Relation between exposure to asbestos and smoking jointly and the risk of lung cancer. Occup Environ Med 2001; 58:145-53. 18. Gustavsson P, Nyberg F, Pershagen G, et al. Low-dose expo sure to asbestos and lung cancer: dose-response relations and interaction with smoking in a population-based case-referent study in Stockholm, Sweden. Am J Epidemiol 2002;155:101622. 19. Liddell FD, Armstrong BG. The combination of effects on lung cancer of cigarette smoking and exposure in Quebec chrysotile miners and millers. Ann Occup Hyg 2002;46:5-13. 20. Vainio H, Boffetta P. Mechanisms of combined effect of asbes tos and smoking in the etiology of lung cancer. Scand J Work Environ Health 1994;20:235-42. 21. Saracci R. Asbestos and lung cancer: an analysis of the epide miologic evidence on the asbestos-smoking interaction. Int J Cancer 1977;20:323-31. 22. Koskinene K, Pukkala E, Martikainen R, et al. Different mea sures of asbestos exposure in estimating risk of lung cancer and mesothelioma among construction workers. J Occup Environ Med 2002;44:1190-6. 23. Pohlabeln H, Wild P, Schill W, et al. Asbestos fibreyears and lung cancer: a two phase case-control study with expert expo sure assessment. Occup Environ Med 2002;59:410-14. 24. Schaeffner ES, Miller DP, Wain JC, et al. Use of an asbestos exposure score and presence of pleural and parenchymal abnor malities in a lung cancer case series. Int J Occup Environ Health 2001;7:14-18. 25. Rogan WJ, Ragan NB, Dinse GE. X-ray evidence of increased asbestos exposure in the US population from NHANES I and NHANES II, 1973-1978. Cancer Causes Control 2000;11: 441-9. 26. Hillerdal G, Henderson DW. Asbestos, asbestosis, pleural plaques and lung cancer. Scand J Work Environ Health 1997; 23:93-103. 27. Partanen T, Nurminem M, Zitting A, et al. Localized pleural plaques and lung cancer. Am J Ind Med 1992;22:185-92. 28. Edelman DA. Asbestos exposure, pleural plaques and the risk of lung cancer. Int Arch Occup Environ Health 1988;60:38993. 29. Wilkinson P, Hansell DM, Janssens J, et al. Is lung cancer asso ciated with asbestos exposure when there are no small opacities on the chest radiograph? Lancet 1995;345:1074-8. 30. Case BW, Dufresne A. Asbestos, asbestosis and lung cancer: observations in Quebec chrysotile workers. Environ Health Perspect 1997;105(suppl):1113-19. 31. Finkelstein MM. Radiographic asbestosis is not a prerequisite for asbestos-associated lung cancer in Ontario asbestos-cement workers. Am J Ind Med 1997;32:341-8. 32. Browne K. Is asbestos or asbestosis the cause of the increased risk of lung cancer in asbestos workers? Br J Ind Med 1986;43: 145-9. 33. Commins BT. Estimations of risk from environmental asbestos in perspective. IARC Sci Publ 1989;90:476-85. 34. Hughes JM. Human evidence: lung cancer mortality risk from chrysotile exposure. Ann Occup Hyg 1994;38:555-60. 35. Churg A. Asbestos-related disease in the workplace and the environment: controversial issues. Monogr Pathol 1993;36:5477. 36. Omenn GS, Goodman GE, Thornquist MD, et al. The Carotene and Retinol Efficacy Trial (CARET) to prevent lung cancer in high-risk populations: pilot study with asbestos-exposed work ers. Cancer Epidemiol Biomarkers Prev 1993;2:389-96. 37. Omenn GS, Goodman GE, Thornquist MD, et al. Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease. N Engl J Med 1996;334:1150-5. 38. Omenn G, Goodman G, Thornquist M, et al. Risk factors for lung cancer and for intervention effects in CARET, the BetaCarotene and Retinol Efficacy Trial. J Natl Cancer Inst 1996; 88:1550-9. 39. Omenn GS, Goodman G, Thornquist M, et al. The beta-Caro tene and Retinol Efficacy Trial (CARET) for chemoprevention of lung cancer in high risk populations: smokers and asbestosexposed workers. Cancer Res 1994;54(suppl):2038s-43s. 40. Thornquist MD, Omenn GS, Goodman GE, et al. Statistical design and monitoring of the Carotene and Retinol Efficacy Trial (CARET). Control Clin Trials 1993;14:308-24. 41. Barnhart S, Keogh J, Cullen MR, et al. The CARET asbestosexposed cohort: baseline characteristics and comparison to other asbestos-exposed cohorts. Am J Ind Med 1997;32:57381. 42. Mannino DM, Aguayo SM, Petty TL, et al. Low lung function and incident lung cancer in the United States: data from the First National Health and Nutrition Examination Survey follow-up. Arch Intern Med 2003;163:1475-80. 43. Nicholson WJ, Perkel G, Selifoff IJ. Occupational exposures to asbestos: population at risk and projected mortality-- 1980-2030. Am J Ind Med 1982;3:259-311. 44. Rosenstock L, Barnhart S, Heyer NJ, et al. The relation among pulmonary function, chest roentgenographic abnormalities, and smoking status in an asbestos-exposed cohort. Am Rev Respir Dis 1988;138:272-7. 45. Welch LS. Construction industry hazards. In: Rosenstock L, Cullen M, Brodkin C, et al, eds. Textbook of clinical occupa tional and environmental medicine. 2nd ed. London, United Kingdom: Elsevier Science, 2005:215-26. Am J Epidemiol 2005;161:260-270