Investigation of the combined role of dose and dose-averaged LET in the occurrence of brain necrosis after intensity modulated proton therapy for meningioma and solitary fibrous tumours.

Abstract

To investigate the combined effect of RBE-weighted dose (D_RBE) and dose-averaged linear energy transfer (LET_d) on the development of brain radionecrosis (BRN) in patients with intracranial meningioma and solitary fibrous tumors treated with intensity modulated proton therapy (IMPT). The study included 124 patients with a median follow-up of 47.4 months (range 5.8 ÷ 109.6). Dose prescriptions ranged between 54 and 66 Gy(RBE) (RBE = 1.1) in 28 ÷ 33 fractions. Patients reporting a BRN of grade ≥ 1 (CTCAE v5.0) at follow-up constituted the adverse event (AE) group (n = 26). The remaining 98 patients constituted the control group. The BRN volume was contoured on the follow-up MRI images that first revealed the presence of the AE. We extracted both D_RBE and LET_d distributions to build the cumulative dose LET_d volume histogram (DLVH) of the healthy-brain. Univariable and multivariable logistic regressions were fitted to explore the impact of DLVH indexes on BRN (α = 0.01). Voxel-wise analysis was performed on the AE group using greedy random matching (caliper = 0.1) to match healthy-brain with AE voxels with a 5 Gy(RBE) dose-binning. A logistic mixed-effects model (random intercept) was fitted for each bin (α = 0.05). The best fitting DLVH index was selected as the median partial brain volume receiving a D_RBE > 42.9 Gy(RBE) and LET_d > 4.6 keV/µm (8.2 cc for the AE group and 1.1 cc for the control group). The effect of LET_d reached the highest statistically significant impact in the dose bin 40-45 Gy(RBE) (OR: 2.24 [1.62; 3.20], q(False Discovery Rate-adjusted) = 0.01). We identified a DLVH constraint, predictive of BRN, to be implemented in clinical practice aiming at a reduction of treatment toxicity.