BACKGROUND Studies have suggested that leanness in adulthood may be a risk factor for lung cancer; however, there is justifiable concern that the observed association may be due to residual confounding by smoking, preclinical weight loss, competing causes of death, or some combination of these. METHODS To examine this association we used data from the Canadian National Breast Screening Study, which included 89,835 women ages 40-59 years at recruitment between 1980 and 1985. During a mean of 16 years of follow-up, we observed 750 incident lung cancer cases. We used Cox proportional hazards models to estimate hazard ratios and 95% confidence intervals for the association between body mass index (BMI) and lung cancer. RESULTS After adjustment for pack-years of smoking and other covariates, there was some evidence for inverse associations in current smokers (hazard ratio for highest BMI quintile relative to the lowest = 0.63; 95% confidence interval = 0.48-0.83) and in former smokers (0.69; 0.39-1.23), whereas in never-smokers, BMI was positively associated with lung cancer (2.19; 1.00-4.80). The results for current and former smokers were not altered by exclusion of cases diagnosed within the first 5 years of follow-up; however, in never-smokers the strength of the association was reduced. CONCLUSIONS The present study contributes to the aggregate evidence suggesting that there may be an inverse association between BMI and lung cancer among smokers. However, the contrasting pattern of associations between BMI and lung cancer seen in ever-smokers and never-smokers in this study requires explanation.