Objective: Our objective was to develop a nomogram that predicts the probability of cancer-specific survival in men with untreated androgen-independent prostate cancer (AIPC).
Methods: AIPC was diagnosed in 129 consecutive patients between 1989 and 2002. No patient received cytotoxic chemotherapy. Univariate and multivariate Cox regression models were used to test the association between prostate-specific antigen (PSA) level at initiation of androgen deprivation, PSA doubling time (PSADT), PSA nadir on androgen deprivation therapy (ADT), time from ADT to AIPC, and AIPC-specific mortality. Multivariate regression coefficients were then used to develop a nomogram predicting AIPC-specific survival at 12-60 mo after AIPC diagnosis. Two-hundred bootstrap resamples were used to internally validate the nomogram.
Results: AIPC-specific mortality was recorded in 74 of 129 patients (57.4%). Other-cause mortality was recorded in 7 men (5.4%). Median overall survival was 52.0 mo (mean, 36.0 mo) and median AIPC-specific survival was 54.0 mo (mean, 35.0 mo). In univariate regression models, all variables were significant predictors of AIPC-specific survival (p < or = 0.02). In multivariate models, PSADT and time from androgen deprivation to AIPC remained statistically significant (p < or = 0.004). Bootstrap-corrected predictive accuracy of the nomogram was 80.9% versus 74.9% for our previous model.
Conclusions: A nomogram predicting AIPC-specific survival is between 13% and 14% more accurate than previous nomograms and 6% more accurate than tree regression-based predictions obtained from the same data. Moreover, a nomogram approach combines several advantages, such as user-friendly interface and precise estimation of individual recurrence probability at several time points after AIPC diagnosis, which all patients deserve to know and all treating physicians need to know.