Personalizing Prescription of Radiation Therapy Utilizing Genomic Markers of Radiosensitivity

Abstract

In previous studies, we developed the radio-sensitivity index (RSI) and genomic adjusted radiation dose (GARD) to capture genomic heterogeneity and personalize radiation therapy (RT) prescription. We hypothesize that the current one-size fits-all paradigm to RT prescription is inefficient for individual patients given observed distributions of tumor RSI across large cohorts of patients. We demonstrate a workflow for clinical use of genomic features in RT prescription. We assessed tumor genomics in a cohort of 1229 patients with non-small lung cancer (NSCLC, Cohort 1) using RSI. To model the impact of genomic heterogeneity on RT dose prescription, we utilized a cohort of 60 NSCLC patients treated with surgery and postoperative RT (45 – 70 Gy, median 54 Gy, Cohort 2). For all analyses, expression data (Affymetrix Hu-RSTA-2a520709) were used to calculate RSI. Using RSI, we derive a patient specific alfa to calculate GARD. Previously we demonstrated that lung cancer patients in cohort 2 who achieved a GARD >= 33 had an improved local control (HR=3.4, 1.3, 9.1 p=0.02). We defined the personalized RSI/GARD RT dose prescription (RxRSI) as the physical dose in Gy required to achieve the GARD threshold of 33. We integrated GARD into the planning workflow of a commercial treatment planning system to create personalized genomic radiation treatment plans (pGRT) using standardized plan optimizations for both standard of care (SoC) and RxRSI doses. RxRSI plans were evaluated against standard planning criteria and normal tissue constraints, and compared with SoC prescription plans using biological effective voxel based dose scaling. In cohort 1, RSI demonstrated a bimodal distribution, indicating a uniform treatment dose may be sub-optimal. A similar heterogeneity was observed in cohort 2, (RSI (IQR: 0.25, 0.40), GARD (IQR: 24.7, 38.62)) where we calculated a personalized RxRSI dose for each patient. In 24% of cohort 2, RxRSI dose was lower than the prescribed dose by more than 10%, with median RxRSI at 47Gy. In 60% of the cohort, RxRSI dose was greater than prescribed dose by more than 10%, ranging from 62Gy to 95Gy. pGRT planning achieved both desired target coverage for RxRSI and normal tissue constraints for patients under 74Gy, with esophageal dose falling below clinically acceptable levels for 80% of plans at 88Gy and above. One-size fits-all RT dose prescription may be sub-optimal for some patients given the genomic heterogeneity observed. Using a clinically-validated model (RSI/GARD) we identify that in up to 84% of patients the prescribed dose and personalized RxRSI did not match, suggesting that there is a large opportunity to optimize dose with genomics. A novel pGRT function within a commercial system can effectively optimize plans to present a choice between a personalized RxRSI and SoC treatment plans for consideration. To our knowledge this is the first demonstration of a feasible clinical approach to precision medicine in radiation oncology.