Factors associated with prognosis in human breast cancer. II. Multivariate stratification analysis

Prediction of the individual fate of breast cancer patient is one of the most important problems for establishment of therapeutic plans. We have analyzed the relationships of clinical factors to 3-year and 8-year survivals in a series of 400 patients treated for breast carcinoma. Multivariate stratification analysis for obtaining subgroups with different survivals was performed with the aid of Bintree, a computer program which produces binary splits of a population according to the criterion of maximal reduction of variance, and generates a binary stratification tree. Lymph node status was the major predictor for both short and long term prognosis but had no important role in the survival of patients who relapsed. Antecedent abortions were related to less relapses. Age, previous benign breast diseases, nulliparity and lactation conditioned better prognosis without association with relapse probability, indicating an association with the speed of evolution of the disseminated disease. In general clinical data were insufficient for establishment of good and bad prognosis polar groups of patients, especially in patients in which no clinical lymph node involvement was recorded.