European multi-centric study on variable proton RBE dose calculations for multiple anatomical sites

Abstract

Purpose or Objective An increasing number of publications has reported on a variable relative biological effectiveness (RBE) in proton therapy (PT) challenging the clinically applied constant RBE. Since no standard for clinical variable RBE calculations exists, many PT centres independently account for RBE variability. This multi-centric study compared variable RBE dose calculations from different PT institutions towards developing a consistent clinical RBE implementation. Materials and Methods Each of six European PT institutions generated clinically acceptable robust pencil beam scanning treatment plans for five cases, comprised of the sites: brain, base of skull (BoS), head and neck, pancreas and prostate. For plan optimisation, centres used their beam model, treatment planning software and an RBE of 1.1 (D1.1). All centres used the same patient-specific prescriptions, fractionation, clinical goals and beam directions. Then, each centre recalculated the corresponding variable RBE weighted dose (DvarRBE) according to their local procedure, including choice of variable RBE model and parameters. Here, an analysis of differences in volume histogram parameters in clinical target volumes (CTVs) and organs at risk (OARs) is presented for all five anatomical sites and three of the six centres. Results Each of the three centres applied a different in-vitro data-based linear energy transfer (LET) driven variable RBE model, including tissue-specific radiosensitivity (α/β) as model parameter (McNamara and Wedenberg models) or without α/β dependency (McMahon model), respectively (Fig.1). For the CTVs, volume-histogram parameters of D1.1 and LET were comparable among the centres for each patient. Considering DvarRBE instead of D1.1 increased inter-institutional differences in dosimetric parameters by 1-7 Gy(RBE) in the CTVs (Fig.2). Additionally, mean DvarRBE deviated from prescribed D1.1 by up to 5 Gy(RBE) for each anatomical site. For OARs, inter-institutional dose differences remained comparable when using DvarRBE instead of D1.1 and all variable RBE models consistently predicted an increase of DvarRBE by 3-9 Gy(RBE) in OARs located at the distal beam edge and for treatment plans with small angles between single incident beams (brain, BoS, pancreas). There, DvarRBE also exceeded OAR near-maximum tolerance doses at all centres. The main drivers of inter-centre DvarRBE variance were differences in the considered centre- and treatment site-specific α/β as well as the model-specific handling of α/β. When all centres used the same variable RBE model and α/β value, inter-centre differences in DvarRBE distributions essentially reduced to the order of those for D1.1. Conclusion Clinical DvarRBE calculations are readily available at multiple institutions and already inform clinicians on treatment plan safety. However, to ensure future comparability of results between PT centres and to realize the full potential of PT, it remains a critical task to collectively agree on how to utilize variable RBE models in clinical practise.