Many studies have examined the associations of body mass index (weight (kg)/height (m)(2)) with risk of various cancers. However, optimal scaling of weight for height may depend on the population studied. The authors used data from a large cohort study of women (Canadian National Breast Cancer Screening Study, 1980-2000; n = 89,835) to examine how the scaling of weight for height (W/H(x)) influenced the association with risk of 19 different cancers. Cox proportional hazards models were used to estimate the hazard ratio for each cancer site with W/H(x), with x increasing from 0 to 3.0 by increments of 0.1. The correlation between weight and W/H(x) decreased monotonically with increasing x, whereas W/H(x) was minimally correlated with height when x = 1.4. W/H(x) showed significant positive associations with postmenopausal breast cancer, endometrial cancer, kidney cancer, and lung cancer in never smokers. W/H(x) was inversely associated with lung cancer in ever smokers. The value of x for which W/H(x) produced the largest statistically significant hazard ratio ranged from 0.8 (endometrial cancer) to 1.7 (postmenopausal breast cancer). For lung cancer in ever smokers, the inverse association was statistically significant for all values of x. These findings suggest that the scaling of weight for height may vary depending on the cancer site and that optimal scaling may be considerably different from W/H(2) or, alternatively, that a range of scaling should be considered when examining the association of body weight with risk of disease.