ESTRO 2022 - Abstract Book

S1548

Abstract book

ESTRO 2022

The relation between RBE-LET d is often proposed to be nonlinear, but linearity is often implied in normal tissue complication probability (NTCP) modelling and LET d optimization. In this study we aim to investigate the linearity of the RBE-LET relation using two commonly used variable RBE models. Materials and Methods This study was conducted in three sets of 20 consecutive head & neck, breast and brain cancer patients resulting in a total sample of 60 patients. The RBE weighted dose (D RBE ) was calculated for the Wedenberg (WED) and McNamara (MCN) RBE models using an α / β value between 1 and 10 Gy. A linear fit was made for each RBE model based on the average D RBE and LET d to the target and all relevant organs-at-risk (OARs). The D RBE for clinically relevant dosimetric parameters was then calculated and compared using the WED and MCN models as well as the linear fits of the RBE models. Results The average D RBE parameters models plotted against the D ⋅ LETd are shown in figure 1 for selected α / β values. For the average D RBE of OARs, the coefficient of determination (R 2 ) was between 0.94 and 1.00 and between 0.87 and 0.99 for the MCN and WED models respectively, with lower R 2 values occurring for low α / β values (figure 2). The R2 for an α / β of 2 Gy or higher was higher than 0.92 for both models. For clinically relevant dosimetric parameters, the mean absolute error (MAE) between the WED and MCN model decreased with increasing α / β . The MAE between the MCN model with an α / β of 2 Gy and its linear estimate was 1.57, 0.44 and 0.64 Gy for HNC, breast and brain patients, respectively. The MAE between the MCN and WED model with an α / β of 2 Gy was 0.73, 0.22 and 0.32 Gy for HNC, breast and brain patients respectively. The MAE between an RBE model and its linear fit was of a similar order of magnitude as the MAE between two RBE models. Conclusion Both considered RBE models are more linear for high α / β values which are typically observed for acutely responding tissues and the MCN model was more linear than the WED model. The linear fits to estimate mean D RBE parameter generated good fits with high (>0.90) R 2 values for α / β of 2 Gy and very high (>0.95) for α / β of 3 Gy or higher. Using a linear RBE-LET d relation does not clinically meaningfully contribute to the uncertainty of the D RBE when considering clinically relevant dosimetric parameters.