UCSF

Purpose

Two methods widely used to predict the risk of treatment failure after radical prostatectomy for localized prostate cancer are the 3 level D'Amico risk classification and the Kattan nomogram. Although they have been previously validated, to our knowledge they have not been compared in a community based cohort. We tested the 2 instruments in the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) database, a national registry of patients with prostate cancer, to assess their accuracy in a community based cohort.

Materials and methods

Men were invited to join CaPSURE from 33 American urology practices, of which 30 were community based. A total of 1,701 men with localized prostate cancer (T1-3a) were treated with radical prostatectomy between 1989 and 2000. Patients who received neoadjuvant or adjuvant therapy were excluded. Recurrence was defined as 2 or more consecutive prostate specific antigen measurements of 0.2 ng/ml or greater, or a second treatment greater than 6 months after surgery. Freedom from progression (FFP) was based on life table estimates and Kaplan-Meier curves. Risk groups were compared using a Cox proportional hazards model and ANOVA.

Results

Based on the D'Amico classification 671 cases (39%) were classified as low risk, 446 (26%) were intermediate risk and 584 (34%) were high risk. Five-year FFP was 78%, 63% and 60% in the low, intermediate and high risk groups (HR 1.00, 1.87 and 2.32 respectively, p <0.0001). Mean 5-year FFP predicted by the Kattan nomogram in the same risk groups was 91%, 74% and 69%, respectively. Outcomes in the low risk group were tightly grouped about the mean but there was considerable dispersion of outcomes in the intermediate (30% to 98% FFP) and high (17% to 98%) risk groups.

Conclusions

Stratifying patients in CaPSURE into low, intermediate and high risk categories for disease as described by D'Amico or applying the Kattan nomogram resulted in statistically significant differences in predicted 5-year FFP. However, there was considerable overlap of outcomes between the intermediate and high risk groups. This analysis suggests that simply estimating disease recurrence by stratifying patients into low, intermediate and high risk groups may not provide sufficient information for predicting outcomes among individuals.