Objectives To develop a prediction model, or nomogram, that would predict the probability that a man with benign prostatic hyperplasia would experience acute urinary retention (AUR) or require surgical intervention (SI) within 2 years, with or without dutasteride therapy. Methods We modeled 4294 men treated in the Phase III dutasteride benign prostatic hyperplasia trials. These men were characterized at baseline by a number of parameters, including the American Urological Association Symptom Index, Benign Prostatic Hyperplasia Impact Index questionnaire, prior use of selective alpha 1-blockers, prostate volume, prostate-specific antigen level, and maximal flow rate. Cox proportional hazards regression analysis was used to relate these baseline variables to their future probability of AUR/SI within 2 years. The nomogram was internally validated with bootstrapping to assess its discrimination and calibration. Discrimination was quantified as the concordance index. Results The nomogram appeared to be accurately calibrated and discriminating (concordance index 0.71, P <0.001). Conclusions We constructed a nomogram for predicting the probability that a man would experience AUR or require SI within 2 years of benign prostatic hyperplasia diagnosis. At 24 months of follow-up, 7.4% of placebo patients and 3.7% of dutasteride patients had experienced AUR and/or SI, representing a 50% relative risk reduction and a 3.7% absolute risk reduction. For the greatest risk patient randomized to the Phase III dutasteride trial, the nomogram predicted a maximal risk of 27%, significantly greater than the median risk of the placebo-treated patients.
Read full abstract