Body size and breast cancer prognosis: a statistical explanation of the discrepancies.

A historical cohort of 68 female breast cancer patients from one institution who were enrolled in a multicenter randomized controlled trial between 1971 and 1973 were followed up to the beginning of 1986. Weight and height at the time of mastectomy were transformed into two indices of body size, namely the Quetelet Index and a weight to "ideal weight" ratio. These two indices were analyzed for their relation with overall and disease-free survival, while controlling for the effect of several potential confounding variables. While neither index was linearly related to the hazard of death or recurrence, a significant quadratic (curvilinear) relation was found for both indices and both hazards. In all cases the hazard function was concave up, indicating that not only overweight but also underweight status is predictive of an unfavorable prognosis of breast cancer. This finding offers a possible explanation for the discrepancies among previous studies on this topic.