Personalizing Radiotherapy Prescription Dose Using Genomic Markers of Radiosensitivity

Purpose/Objective(s): Although radiotherapy (RT) remains a critical curative agent for cancer, it has yet to adapt a biological basis in the clinic. We previously proposed the gene expression-based radiosensitivity index (RSI) and the genomic-adjusted radiation dose (GARD), as a clinically-feasible approach to biology-based RT. We hypothesize that empiric one-size fits all RT dose is biologically imprecise and negatively impacts clinical outcome. Materials/Methods: We utilize the RSI/GARD formalism to calculate a personalized RT prescription dose that biologically optimizes prescription dose for each patient (RxRSI). We integrated RxRSI into a commercial treatment planning system to create the first approach to personalized biologically-optimized RT plans. We compare RxRSI with standard of care empiric radiotherapy dose in 2 cohorts of NSCLC patients (n=60 and n=1,747) from an IRB-approved de-identified institutional biorepository. To quantify the opportunity inherent in personalized RT prescription, we develop a first-in kind precision RxRSI model that estimates local control (LC) based on whether an optimal RT dose (RxRSI) is delivered and penalizes outcome based on exposure to excess normal tissue complication risk (penalized local control, pLC). To validate the precision RxRSI model, we ran an in silico clinical trial similar to RTOG 0617 in the modeling cohort. Results: We demonstrate that up to 75% of patients receive non-optimal radiotherapy doses, when treated with a one-size fits all empiric approach. Conversely, we show that personalized, radiotherapy prescription dose delivers optimal doses to up to 75% of the patients even when restricted to doses within the standard of care. These differences result in quantifiable excess doses to normal tissue that negatively impact clinical outcome. We validate the precision RxRSI model by demonstrating that it predicts that empiric radiotherapy dose escalation to 74 Gy results in no improvement in radiotherapy-associated outcome (pLC) as demonstrated in RTOG 0617. The precision RxRSI model demonstrates that only an additional 16.2% of lung cancer patients achieve RxRSI by receiving 74 Gy (beyond 60Gy). A genomic-based strategy to deliver 74 Gy to only those patients could have produced a 6.3% absolute improvement in clinical outcome (pLC) for the whole population (p<0.05). Conclusion: The current empiric approach to radiotherapy dose prescription is biologically imprecise and over or under-treats up to 75% of lung cancer patients. These inaccuracies negatively impact patient outcome by either an over-exposure to normal tissue complications or by achieving a sub-optimal tumor control. Personalized RT prescription dose may improve radiotherapy-associated outcome in lung cancer by an absolute 6.3% using radiotherapy doses within current standard of care without an expected increase in normal tissue toxicity. Funding Statement: DeBartolo Personalized Medicine Institute. Declaration of Interests: JTR, SAE and JGS report IP (RSI (JTR, SAE), GARD (JTR, JGS) and RxRSI (JTR, JGS). JTR and SAE report stock in Cvergenx. TF, AW report stock in Varian. Ethics Approval Statement: Data was compiled for two cohorts of patients from an IRB-approved de-identified institutional bio-repository.