SU-D-137-07: Verification of the Method Used in Clinical Estimation of Alpha/beta Ratio, Using a Model Simulation Including Proliferation and Hypoxic Effects

Abstract

Purpose: We hypothesize that the interplay between cell kill, proliferation, and hypoxia make the traditional interpretation of clinically‐derived α/β ratios problematic. In this study, we simulate resulting α/β ratios given a recently developed tumor‐response simulation model that includes the interplay between hypoxia and proliferation, as well as the resulting tumor reoxygenation during a course of radiotherapy. Methods: Using a state‐driven tumor response model, various typical fractionation schemes were simulated until the same level of tumor control probability was achieved (50%). Depending on the inclusion of effects, four different cases were evaluated: (1) neither proliferation nor hypoxia; (2) proliferation only; (3) hypoxia only; and (4) both proliferation and hypoxia. We judge the full model to be most realistic. The resulting α/β ratio was estimated based on the Withers isoeffect equation (as a negative slope of linear regression line in the plot of TD5 0 vs. TD5 0 xd) and compared with the ratio in model input, α/β(model) (=6.63). Results: Without any other effects (case 1), the model estimated an α/β ratio, α/β(est), that equals to the model assigned value. Including proliferation (case 2), smaller fractional doses become less effective due to increased repopulation, resulting in a reduced α/β(est). For case 3, including hypoxia only, longer schedules with smaller fraction sizes become more effective due to reoxygenation and resulted in unrealistic negative α/β(est). Interestingly, in case 4, the two effects, proliferation and hypoxia, approximately cancel each other and unmask the underlying cell‐kill characteristics, resulting in an α/β(est) comparable to the model input value α/β(model). Conclusion: Simulations show that the estimated α/β ratio is dependent on underlying radiobiological effects (including proliferation and hypoxia), and the resulting model estimation may not be an accurate reflection of cell‐kill sensitivity alone, which implies the clinical estimation might be biased if other effects are not properly controlled.