Obesity has been associated with poor prognosis of breast cancer, but studies investigating indicators of adiposity, independent of height, such as skinfold thickness and weight/height ratio, have been limited. From a cohort of 89,835 women in the National Breast Screening Study, data on histologically confirmed, incident cases of invasive carcinoma of breast, diagnosed between January 1981 to June 1992, were analyzed. Body measurements were taken at enrollment and other risk-factor information from a self-administered questionnaire completed prior to diagnosis of the breast cancer. Mortality data were obtained by linkage to Statistics Canada, provincial cancer registries, and annual follow-up of cases through physicians. Of the 5 measures of body size studied among cases, poor prognosis was associated with higher triceps skinfold thickness only. With every 5 mm increase in skinfold, the risk of dying increased approximately 12% (relative risk, 1.12; 95% confidence intervals, 1.01 to 1.24) in the 1033 cases examined and adjusted for age at diagnosis, number of positive axillary nodes involved, and body weight. Compared with women in the lowest quartile of skinfold thickness, women in the fourth quartile had a relative risk (RR) of 1.64 (95% CI, 0.99 to 2.73). None of the other measures--weight, height, body mass index or weight/height ratio--reached any statistical significance. Since the reliability and validity of skinfold measures used in our study were never established, and no data were available for changes in body size after diagnosis, the results of this study on skinfold thickness should be used only as indicative of a need for further studies.