The European Organization for Research on Treatment of Cancer Genitourinary Group performed a multivariate statistical analysis of prognostic factors based on 436 patients entered between 1976 and 1981 in 2 randomized prospective trials that compared 4 different hormonal treatment regimens. Only previously untreated patients with advanced (stage T3/T4/M0 or M1) prostatic cancer were eligible. After identification of prognostic factors by means of univariate analyses a multivariate analysis using Cox's proportional hazards regression model was done. This test identified performance status (according to the Eastern Cooperative Oncology Group scale) as the most important factor, followed by acid phosphatase (more than 2 times normal) for stage M0 cancer patients, and alkaline phosphatase, T category and the presence or absence of associated chronic disease for stage M1 cancer patients. Based on these 4 variables nonbedridden patients with metastatic disease can be divided into 2 groups: poor and good risk patients, with median survivals of 1 and 3 years, respectively. This study shows that routine clinical and laboratory data already provide an excellent indication as to the prognosis.