We studied histological samples and clinical data from 111 breast carcinoma patients originally treated in 1975-1977 who did not have distant metastases at the time of diagnosis. A multivariate survival analysis using the Cox model selected the studied variables in the following order according to their prognostic association (breast cancer deaths, or deaths from any cause): axillary lymph node status, tumor size, mitotic index, and DNA ploidy status. The association of the different variables to the prognosis in respect to breast cancer deaths was evaluated 1-10 years after treatment by stepwise logistic regression. Lymph node status, tumor size, mitotic index, and DNA ploidy all showed significant relation to the prognosis but this association varied considerably with time of observation.