The identification of surrogate endpoints for survival would accelerate the conduct of prostate cancer clinical trials. We evaluated the incidence of distant metastasis (DM) and general clinical failure (GCF) as potential surrogate markers for cause-specific survival (CSS) using data from RTOG 9202.
Materials/Methods
Men with locally-advanced prostate cancer were treated with 4 months of neoadjuvant and concurrent androgen deprivation (AD) with external beam radiation therapy (RT) followed by no additional therapy (standard arm) or 24 additional months of AD (experimental arm). General clinical failure (GCF) was defined as local progression, regional/distant metastasis, initiation of hormonal therapy, or PSA ≥ 25 ng/mL after completion of RT (Taylor et al, IJROBP 50:1212, 2001). DM and GCF with 5 years follow up were formally tested for surrogacy for CSS at 10 years. Surrogacy testing utilized Prentice's criteria: 1) that the treatment is prognostic for the true endpoint; 2) that the treatment is prognostic for the surrogate endpoint; 3) that the surrogate endpoint is prognostic for the true endpoint; and 4) that the full effect of treatment on the true endpoint is explained by the surrogate. Landmark analyses at 3 and 5 years were performed to address potential follow up length bias.
Results
Among 1521 eligible, randomized patients, there were 218 cause-specific deaths, 291 DMs and 726 GCFs. The 10-year CSS was 85% in the standard arm, compared with 89% in the experimental arm (p = 0.0089). The 5-year DM and GCF rates were 14% and 43% in the standard arm, compared with 9% and 33% in the experimental arm (HR = 1.7 [1.2–2.4] and 1.6 [1.3–1.9], respectively). Both 5-year DM and 5-year GCF were significantly associated with 10-year CSS (p < 0.0001 and p = 0.0005, respectively). 10-year CSS was independent of treatment among patients who experienced DM (p = 0.63) or GCF (p = 0.23), and 10-year CSS was independent of treatment among patients who did not experience DM (p = 0.60) or GCF (p = 0.69), indicating that the 5-year surrogates both capture the full effect of treatment on 10-year CSS. Therefore, without adjusting for length bias, 5-year DM and GCF both satisfied all of Prentice's criteria for 10-year CSS. The analysis using only the 1364 patients alive at landmark time 3 years, showed that both 3-year DM and GCF satisfied all of Prentice's criteria for 10-year CSS. In the 1178 patients alive at a landmark time of 5 years, there was no statistical difference in CSS between the treatment arms (p = 0.08). Nonetheless, 5-year DM and GCF met the most important of the remaining Prentice's criteria for 10-year CSS.
Conclusions
This analysis of patient data from RTOG 9202 demonstrates that DM and GCF at 5 years satisfy Prentice's surrogacy criteria for CSS at 10 years. These endpoints should be validated in future randomized clinical trials for localized prostate cancer.
Supported by NCI grants U10 CA21661, U10 CA37422, U10 CA32115.
FULL TEXT