Gene signature predictor of dose-response to prostate radiation: Validation of PORTOS in phase III trials.

Abstract

308 Background: NRG/RTOG 0126 and SAKK 09/10 were phase III randomized trials examining whether higher dose resulted in better response/outcomes in PCa patients following definitive and post-operative RT, respectively. RTOG 0126 showed a benefit for RT dose escalation (DE) from 70.2Gy to 79.2Gy though SAKK 09/10 did not show a benefit from 64Gy to 70Gy. We hypothesized that a previously developed 24-gene prostate cancer RT gene expression score (PORTOS) could distinguish patients who benefited from RT DE in both trials. Methods: PORTOS scores were calculated on biopsy samples in RTOG 0126 and prostatectomy samples in SAKK 09/10 as published. Since the original PORTOS cutoffs were in the post-op setting, we utilized tertile score groups in RTOG 0126, whereas the published PORTOS cutoffs were used for SAKK 09/10. The primary objective was to evaluate PORTOS as a predictive biomarker for the benefit of RT DE on biochemical failure (BF) via the Phoenix criteria in RTOG 0126 (N=215) and clinical progression-free survival (CFPS) in SAKK 09/10 (n=226). In addition, we also investigated clinical and molecular correlates of PORTOS in large real-world datasets of 31,107 prostate biopsy samples and 42,407 radical prostatectomy samples. Results: In RTOG 0126, in patients with lower tertile PORTOS scores, there was no difference in Phoenix BF (sHR 1.14 [0.54-2.40], P=0.73). However, for patients in the middle and higher tertile PORTOS score range, there was a significant benefit for RT DE for Phoenix BF (middle PORTOS: sHR 0.45 [0.22-0.90], P=0.02; higher PORTOS: sHR 0.30 [0.12-0.75], P=0.009). An interaction test indicated a significant difference in benefit for DE between higher and lower PORTOS groups (P=0.048). Similarly in the post-op SAKK 09/10 trial, only patients in the higher PORTOS score group benefited from RT DE (CPFS HR 0.19 [0.05-0.70]; P=0.01), with a significant biomarker-treatment interaction between lower vs. higher PORTOS and treatment arm (P=0.003). Interestingly, PORTOS was not consistently associated with clinicopathologic variables in either trial or in the large real-world biopsy or prostatectomy datasets. Biologically, in the real-world datasets, PORTOS was modestly associated with hypoxia signatures consistent with its role in radio-resistance, and strongly associated with immune signatures and molecular subtypes. Conclusions: In two phase III randomized trials, we have validated that PORTOS can identify patients who benefit as well as those that do not benefit from RT DE for localized PCa and provides the first randomized evidence for any biomarker to be able to predict RT dose response. PORTOS could be used to personalize radiation dose for patients with prostate cancer clinically, and allow selection of patients most likely to benefit from RT DE while sparing others from the potential increased risk of toxicity.