E. Scott1, M. U. Zahid2, J. J. Caudell1, J. F. Torres-Roca1, and H. Enderling2; 1H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL, 2MD Anderson Cancer Center, Department of Radiation Oncology, Houston, TX
Purpose/Objective(s):This research pioneers a personalized approach to radiation therapy, emphasizing tumor volume dynamics as a surrogate for survival. Acknowledging existing fractionation schemes, such as acceleration/boost techniques, our work focuses on real-time adaptation of RT fractionation based on individual tumor behavior during therapy. By using our model to explore alternative options beyond conventional methods, we aim for maximal therapeutic impact. This initiative represents a shift towards leveraging dynamic tumor response for precision in radiation medicine, offering a nuanced understanding of treatment personalization. Materials/
Methods: In our study, we developed a mathematical model for tumor dynamics during radiotherapy (RT), with growth modeled as logistic growth, described by the equation dV/dt = ?(1-V/K), where V is tumor volume, ? the growth rate, and K the carrying capacity, or the putative maximum tumor size supportable by local tissue. The RT effect is seen as an immediate decrease in carrying capacity per radiation fraction. We modeled locoregional recurrence by estimating a residual tumor volume (RTV) post-RT; if RTV exceeded a threshold for minimum viable volume, recurrence time was predicted by exponential tumor regrowth to a detectable size. Parameters (radiosensitivity, minimum viable size) were calibrated with Kaplan-Meier curves abstracted from RTOG 9003, which compared multiple fractionation schedules and assessed their efficacy. Using this calibrated model, we introduce a methodology for selecting adaptive fractionation schedules tailored to individual patient responses. We examined the differential impacts of standard vs. hyperfractionation on tumor volume reduction to construct maps showing when standard fractionation is favored over hyperfractionation for the remaining RT weeks. Results: A complete search of the parameter space using our calibrated model yielded two scenarios yielding optimal patient outcomes: (1) a fixed schedule of standard fractionation, where no fractionation adaptation is necessary and (2) upfront hyperfractionation followed by a switch to standard fractionation. Conclusion: Our study presents a calibrated mathematical model integrating logistic growth modeling, instantaneous RT effects, and decision-making maps, offering a new approach to optimizing RT with personalized, adaptive fractionation schemes. Its important to underscore that while our model is calibrated against population-level data from a substantial retrospective cohort, we are exploring various adaptive fractionation schedules as potential options rather than making definitive recommendations. These preliminary findings emphasize the need for further prospective calibration and validation of our model before it can be applied clinically. This approach underscores the importance of continued research in this area to refine and validate predictive models for RT personalization.
