We examined data on body weight and height from 55 male and 26 female lung cancer cases and up to 10 sex-ethnicity-age matched controls per case from a large prospective cohort. All four body mass indices (W/H, W/H2, W/Hp) were highly intercorrelated. Conditional logistic regression, using each index as the exposure variable, yielded odds ratios for lung cancer with magnitude and dose-response gradient that were somewhat different among the four indices. These results suggest that the body mass indices are not necessarily interchangeable in measuring obesity-disease associations.