Abstract
Several large randomized prospective studies have demonstrated a survival benefit with the addition of long-term androgen deprivation to definitive radiotherapy for patients with Gleason score 8–10 or T3-T4 prostate cancer. However, these studies were performed before the routine use of prostate-specific antigen (PSA) measurement. The purpose of this study was to determine what pretreatment (initial) PSA (iPSA) level, if any, warrants the addition of long-term androgen deprivation in the PSA era.
The data set evaluated consisted of 1003 prostate cancer patients treated definitively with three-dimensional conformal radiotherapy between May 1, 1989 and November 30, 1999 (median follow-up, 61 months). Specifically excluded were patients with T3-T4 disease or Gleason score greater than 7 or those who had undergone androgen deprivation as a part of their initial therapy. The median radiation dose was 76 Gy. Patients were randomly split into two data sets, with the first (
n = 487) used to evaluate the optimal iPSA cutpoint for which a statistically significant difference in outcome was noted. The second data set (
n = 516) served as a validation data set for the initial modeling. The analysis of the optimal iPSA cutpoint was based on a recursive partitioning approach for censored data using the log–rank test for nodal separation of freedom from biochemical failure (FFBF) as defined by the American Society for Therapeutic Radiology and Oncology definition. Cox multivariate regression analysis was used to confirm independent predictors of outcome among the clinical and treatment-related factors: iPSA (grouped as defined by the recursive partitioning analysis), Gleason score (2–6 vs. 7), T stage (T1c-T2a vs. T2b-T2c), and total radiation dose (continuous).
The recursive partitioning analysis data set resulted in an optimal iPSA cutpoint of 35 ng/mL, such that the 5-year Kaplan-Meier estimate of FFBF was 80%, 69%, and 19% for iPSA groups of 0–9.9, 10–35, and >35 ng/mL, respectively. The validation data set demonstrated the optimal iPSA cutpoint to be 30 ng/mL. Conservatively choosing 30 ng/mL as the optimal cutpoint, the 5-year FFBF estimate for the entire 1003 patients was 82%, 69%, and 20% for iPSA groups 0–9.9 (
n = 630), 10–30 (
n = 329), and >30 (
n = 44) ng/mL, respectively. On multivariate regression analysis, with the iPSA grouped as above, the Gleason score and radiation dose were independent predictors of outcome in this patient group (all
p < 0.001). On univariate analysis, a higher radiation dose improved FFBF when the iPSA level was between 10 and 30 ng/mL (
p = 0.001) but not when the iPSA level was >30 or <10 ng/mL.
Recursive partitioning techniques defined an iPSA cutpoint of 30 ng/mL for delineating intermediate vs. high risk. Patients with a PSA level >30 ng/mL in the absence of Gleason score >7 or T3 disease do poorly when treated with radiotherapy alone and should be considered for long-term androgen deprivation or other aggressive systemic therapy.
