Local Failure and Gleason Score 9-10 Disease Independently Predict for Survival Outcomes: A Meta-Analysis of Six Randomized Trials

Emerging data suggest that androgen deprivation therapy (ADT) can act as a radiosensitizer, and the benefit of upfront ADT in high Gleason score (GS) tumors--which has been validated in multiple randomized trials--may be ascribed to improvements in both local and systemic control. Patients with high GS cancers are known to have inferior survival outcomes, which has primarily been attributed to a higher risk of developing distant metastases (DMs) or having micrometastatic disease at presentation. The impact of LF in patients with high GS cancers is unknown. The purpose of the present meta-analysis was to explore whether having GS 9-10 disease and having an LF event were independent predictors of survival outcomes. Individual patient-level data were obtained for patients with GS 8-10 PCa enrolled on three RTOG (8531, 8610, and 9202) and three EORTC trials (22863, 22961, and 22991). An LF event could occur in 4 ways: palpable tumor (all 6 trials), PSA>4 (1 trial), initiation of ADT without DM (1 trial), or urethral obstruction (2 trials). For the purposes of this analysis, LF was only counted if it was identified before a DM was identified. Multivariable Cox proportional hazard models were developed to obtain hazard ratio (HR) estimates of the effect of GS 9-10 vs. GS 8 disease and LF vs. no LF on overall survival (OS) and cancer specific-survival (CSS), after adjustment for ADT duration, age, and T-stage. All analyses were repeated in the subgroups of patients who developed DMs within 3 years ("early DM") or after 3 years ("late DM"). Nine hundred ninety-two patients (593 GS 8 and 399 GS 9-10) were included, with a median follow-up of 7.1 years for surviving patients. Overall, 22.9% of GS 8 and 26.3% of GS 9 patients experienced LF (HR for GS 9-10 vs. 8 of 1.39, p<0.05). LF and GS 9-10 disease were significant predictors of both CSS (HRs 2.23 and 1.69, respectively; p<0.05) and OS (HRs 1.32 and 1.31; p<0.05). On subgroup analysis, the LF effect for CSS and OS was not significant in patients with early or late DMs, while the GS 9-10 effects for both CSS and OS were significant only in patients who developed early DMs (HRs 1.93 and 2.00; p<0.05). Having GS 9-10 PCa and experiencing a LF are independent predictors of inferior CSS and OS. These results underscore the dual importance of local and systemic control as part of the upfront management of high GS PCa. Notably, the GS 9-10 PCa effect is most pronounced in patients who develop metastases within 3 years of treatment, suggesting that micrometastatic disease at presentation may be a significant concern. Limitations are that assessment for LF was not uniformly pursued and that the doses of radiation used in the majority of these trials would be considered suboptimal today, such that fewer LF outcomes would be expected in a modern cohort.