A cell‐type resolved proton RBE model for potential clinical application

Background Despite decades of effort, researchers have failed to develop a clinically implementable Relative Biological Effectiveness (RBE) model that can improve dose accuracy in proton therapy. A key challenge is that RBE's dependence on cell‐type is not well understood. This study aimed to overcome this challenge. Methods A total of 139 eligible experimental datasets covering 33 cell lines from the Particle Irradiation Data Ensemble (PIDE) database were used in this study. A Poisson‐based cell survival model, which has two new parameters α and β conceptually different from that in the conventional LQ model, was used to derive a proton RBE model by globally fitting the proton experimental data. The two parameters’ fitting values for a cell's photon data, α X and β X , were used to characterize the cell type. RBE's dependence on cell type was investigated by studying RBE variation on α X and β X . Results The Poisson‐based model showed better global fitting than the LQ model (RMSE w = 6.4% vs. 7.8%, p = 0.00003). A comprehensive RBE model expressed as RBE (LET, D P , α X , β X ) was consequently derived. Interestingly, RBE appears to vary minimally within 95% CI of α X variation, while vary greatly for β X variation. Cells were divided into two groups according to β X , and the comprehensive model was simplified into a cell‐type resolved model by establishing two cell‐type‐independent models for the two groups, respectively. This cell‐type resolved model shows better agreement with experimental data than the comprehensive model derived from the LQ model. Conclusion The variation of β X , which represents a cell's repair capability for critical damages, mainly contributes to RBE's cell‐type dependence. A cell‐type resolved model agreeing well with experimental data has been established.