[Prognostic factor and prediction tables for clinically localized prostate cancer].

INTRODUCTION AND OBJECTIVES To identify the independent prognostic factors of prostate cancer and to develop a table for predicting the probabilities of not-localised prostate cancer occurrence, thus permitting to restrict the radical prostatectomy indication to those patients who have greater probabilities of being cured by the procedure. METHODS 1293 patients with clinically localised prostate cancer, with histories of neither hormone therapy nor prostate radiation therapy, in whom radical prostatectomy was performed during the period 1990-2003, were retrospectively evaluated. In order to analyse the prognostic factors, logistic regression was carried out by studying all the potential confusion and interaction factors, and by introducing the independent variables in a forward fashion with the following criteria: BCON(0.0001) LCON(0.00001) ITER(50) POUT(0.1). The prognostic variables were categorised, and the prediction table of the not-localised prostate cancer probability was developed from them. RESULTS Clinical stage, prostatic specific antigen and Gleason's "grade" were identified as prognostic factors, taking into account that the higher they are, the higher the probability of not-localised prostate cancer occurrence. Logistic regression enabled us to develop a table to predict the probability of not-localised prostate cancer in which, taking a 50% probability as the cutoff point, a 26.13% sensitivity and a 94.65% specificity are obtained. The pathological examination of the prostate and the lymphadenectomy showed that, globally, 70% of the tumours were localised. CONCLUSIONS Application of tables to predicting the probability of having not-localised prostate cancer offers the possibility of improving the prognostic accuracy of the so-called "risk groups", and enables to issue a therapeutic indication better adjusted to the actual status of the disease.