There is as yet no generally accepted explanation for the common finding that low body mass index (BMI) is associated with an increased risk of lung cancer. We investigated this association in a Canadian population-based case-control study (1996)(1997)(1998)(1999)(2000)(2001)(2002) with a particular view to assessing the hypothesis that the observed association was due to residual confounding by smoking. Analyses were based on 1,076 cases and 1,439 controls who provided their height at enrollment and their weight at two points in time, at age 20 and 2 years before enrollment. BMI, in kg/m 2 , was classified into underweight (<18.5), normal (18.5-24.9), overweight (25.0-29.9), and obese (30). Smoking history was synthesized into a comprehensive smoking index (CSI) that integrated duration, intensity and time since quitting. Odds ratios (ORs) and 95% confidence intervals (CIs) for BMI-lung cancer associations were estimated, adjusting for CSI as well as several sociodemographic, lifestyle and occupational factors. The normal BMI category was used as the reference. Among those who were underweight at age 20, there was a lower risk of lung cancer (OR 5 0.69, 95% CI: 0.50-0.95). Conversely, lung cancer risk was increased among those who were underweight 2 years before enrollment (OR 5 2.30, 95% CI: 1.30-4.10). The results were almost identical when stratifying analyses based on smoking history into never/lighter and heavier smokers. The inverse association between recent BMI and lung cancer is unlikely to be largely attributable to residual confounding by smoking. Reverse causality or a true relationship between BMI and lung cancer remain plausible.Elevated body mass index (BMI) has been associated with an increased risk of several cancers, including esophageal, pancreatic, colorectal, endometrial, postmenopausal breast, and prostate cancers.1 By contrast, BMI appears to be inversely related to lung cancer risk.2-16 However, the robustness and meaning of this ostensible inverse association has been challenged in some studies by the finding that such an inverse association was not present among nonsmokers. 4,9,10,12 Smoking is both a powerful risk factor for lung cancer, 17 and is also inversely associated with body weight. 18,19 Several explanations have been hypothesized regarding the observed inverse BMI-lung cancer association. First, a methodological argument, according to which residual confounding by smoking accounted for observing an inverse, or disguising a true positive, BMI-lung cancer relation. Second, reverse causality due to pre-clinical effects of lung cancer, especially for studies in which weight was documented shortly before cancer diagnosis. Third, an etiologic explanation, namely through carcinogenic DNA adducts 6 or effects of estrogens. 16 In any case, the interplay between BMI, smoking, and lung cancer remains uncertain.In the context of a large population-based case-control study on the environmental etiology of lung cancer, data on sociodemographic, anthropometric, and lifestyle factors were c...