Purpose: We evaluated whether posttreatment prostatic-specific antigen doubling time (PSADT) was predictive of prostate cancer mortality by testing the Prentice requirements for a surrogate endpoint.
Methods and materials: We analyzed posttreatment PSA measurements in a cohort of 1,514 men with localized prostate cancer (T2c-4 and PSA level <150 ng/mL), treated and monitored prospectively on Radiation Therapy Oncology Group Protocol 92-02. From June 1992 to April 1995, men were randomized to neoadjuvant androgen deprivation and 65-70 Gy of radiation therapy (n = 761), or in combination with 24 months of adjuvant androgen deprivation (n = 753). Using an adjusted Cox proportional hazards model, we tested if PSADT was prognostic and independent of randomized treatment in this cohort. The endpoints were time to PSADT (assuming first-order kinetics for a minimum of 3 rising PSA measurements) and cancer-specific survival (CSS).
Results: After a median follow-up time of 5.9 years, randomized treatment was a significant predictor for CSS (p(Cox) = 0.002), PSADT <6 months (p(Cox) < 0.001), PSADT <9 months (p(Cox) < 0.001), and PSADT <12 months (p(Cox) < 0.001) but not for PSADT <3 (p(Cox) = 0.4). The significant posttreatment PSADTs were also significant predictors of CSS (p(Cox)< 0.001). After adjusting for T stage, Gleason score and PSA, all of Prentice's requirements were not met, indicating that the effect of PSADT on CSS was not independent of the randomized treatment.
Conclusions: Prostatic specific antigen doubling time is significantly associated with CSS, but did not meet all of Prentice's requirements for a surrogate endpoint of CSS. Thus, the risk of dying of prostate cancer is not fully explained by PSADT.