Beyond Blind Dose-Escalation: Modeling Precision Genomic-Based Radiation Dose-Response In Rectal Cancer

Abstract

Improved pathologic complete response with preoperative radiation dose escalation has been reported in locally advanced rectal cancer (LARC), but risk stratification to personalize radiation dose prescription has not been explored. We modeled genomic-based radiation dose-response utilizing the previously validated radiosensitivity index (RSI) and the clinically actionable genomic-adjusted radiation dose (GARD). We hypothesized that optimal therapeutic gain could be achieved in LARC stratified by genomic heterogeneity and RSI-based dose prescription (RxRSI). We assessed the patterns of pathologic response in a cohort of LARC patients treated with neoadjuvant chemoradiotherapy (NACRT) (cohort 1). Tumor regression grade (TRG) was analyzed for complete (TRG 0; pCR), favorable (TRG 1, moderate response), and poor (TRG 2 and 3) responders. We characterized the tumor genomics in a separate cohort of 113 rectal cancer tissue samples (cohort 2) to calculate tumor specific RSI, from which a tumor specific α was derived to generate GARD. RxRSI was modeled using the GARD threshold for which the percentile of RSI distribution mirrored the patterns of pathologic response in cohort 1. In cohort 1, the primary tumor stage was T3 (84%). The majority of patients (82%) received an RT dose of 50.4 Gy with concurrent 5-FU or Capecitabine. The rates of complete, favorable and poor responders were 21%, 53%, and 26%, respectively. The RSI of rectal cancer ranged from 0.19 to 0.81 in a bimodal distribution. A pCR rate of 21% would be achieved in tumors with RSI < 0.31 at a minimal GARD of 29.76 when modeling RxRSI to the commonly prescribed physical dose of 50.4 Gy. Modeling RxRSI dose escalation to 55 Gy in tumors with an RSI 0.31-0.34 would lead additional 10% tumors to achieve a pCR, whereas some radioresistant tumors would require much higher genomic-based dose to achieve pCR. For example, in our model, 105 Gy would be required for a tumor with an RSI of 0.57 to achieve a complete response. In regards to modeling RxRSI for a favorable response, tumors with RSI 0.32-0.51 would have TRG 1 in response to 50.4 Gy at a minimal GARD of 16.9. RxRSI dose escalation to 60 Gy could potentially bring 95% tumors (RSI <0.57) to a favorable response. We modeled personalized dose escalation in rectal cancer based on genomic expression profiling. This study could provide a theoretical platform for the development of an RxRSI-based prospective trial in rectal cancer. We anticipate individualizing radiation dose selection for each patient and optimizing patient selection for nonoperative management.