A longitudinal population study of 1462 women, aged 38-60 years, was carried out in Gothenburg, Sweden, in 1968-69. In univariate analysis anthropometric variables indicating centrally localized adipose tissue (waist circumference, the ratio of waist to hip circumference and the subscapular skinfold) showed significant age-standardized positive associations with the occurrence (prevalence + incidence data) of endometrial carcinoma. Incidence data suggested that measurements of centrally localized adipose tissue might be of predictive value for this malignancy as well as for ovarian carcinoma. In contrast, measurements of generalized obesity (body weight or body mass index) or peripherally localized adipose tissue (triceps skinfold) showed no associations to these malignancies. No relationship was observed between the anthropometric variables studied and breast carcinoma. The association observed between endometrial and ovarian carcinomas with central adipose tissue did, however, not remain in multivariate analysis when generalized obesity was taken into account. Centrally localized adipose tissue is known to be associated with endocrine abberations including irregular ovulation and menstruation, re-emphasizing the importance of endocrine factors for the pathogenesis of endometrial and ovarian carcinomas. No positive association was found between development of the carcinomas and initial measurements of blood glucose, serum lipids or blood pressure, found to be elevated in cross-sectional studies. An increase in these variables therefore probably are parallel phenomena rather than predictors. The women with endometrial or breast carcinomas smoked more than the remaining women. Although the number of end-points observed was limited these results suggest that measurements of adipose tissue distribution might be a valuable addition to the predictors of endometrial and ovarian carcinomas.(ABSTRACT TRUNCATED AT 250 WORDS)