2523 - Mathematical Model-Driven Response Stratification and Regimen Adjustment in Personalized Ultra-Fractionated Stereotactic Adaptive Radiotherapy (PULSAR) for Non-Small Cell Lung Cancer

Purpose/Objective(s): Personalized Ultra-Fractionated Stereotactic Adaptive Radiotherapy (PULSAR) is a novel treatment paradigm that delivers high-dose pulses separated by intervals of several weeks, which may facilitate tumor response–guided adaptation. This study aims to develop a patient-specific modeling approach to simulate tumor dynamics during radiotherapy and to evaluate the optimal timing for response assessment and potential adaptive strategies.

Materials/Methods: Thirty-seven patients with non–small cell lung cancer (NSCLC) were analyzed, including 24 treated with PULSAR and 13 with stereotactic body radiotherapy (SBRT). A two-compartment ordinary differential equation (ODE) model was developed to characterize tumor volume dynamics, consisting of active tumor cells (T) and inactive (radiation-damaged) tumor cells (I). Active tumor growth followed logistic kinetics, while radiation-induced damage was modeled using the linear–quadratic formalism, transitioning a fraction of active cells to the inactive compartment. Inactive cells decayed at a clearance rate. The individualized radiosensitivity parameter (a) was estimated by fitting the model to longitudinal tumor volume measurements. Patients were stratified as responders, partial responders, or non-responders based on 3-month tumor volume reduction relative to baseline. To distinguish partial responders from non-responders, we evaluated the timing and magnitude of post-pulse tumor shrinkage. Adaptive strategies (e.g., dose escalation) were simulated in potentially radioresistant patients.

Results: The absolute inter-pulse tumor volume variation was greater in PULSAR than in SBRT (21.58% vs. 15.28%). The model accurately captured tumor dynamics during treatment for both regimens, with normalized root mean square errors (NRMSE) of 0.144 ± 0.085 for PULSAR and 0.120 ± 0.058 for SBRT. Simulations indicated that maximal tumor shrinkage in PULSAR occurred at 24 ± 10 days after each pulse, consistent with the current 3-week clinical interval. In partial responders (20–49% tumor reduction), simulated 20% dose escalation delayed tumor regrowth and improved predicted tumor control. When fitted using early treatment data, the model demonstrated strong predictive performance for PULSAR, achieving an R² of 0.87 in forecasting subsequent tumor dynamics.

Conclusion: We developed a patient-specific modeling framework to characterize tumor volume dynamics during PULSAR, potentially supporting adaptive response assessment and personalized treatment adjustment.