Modelling the risk of radiation induced alopecia in brain tumor patients treated with scanned proton beams.

Abstract

To develop normal tissue complication probability (NTCP) models for radiation-induced alopecia (RIA) in brain tumor patients treated with proton therapy (PT). We analyzed 116 brain tumor adult patients undergoing scanning beam PT (median dose 54 GyRBE; range 36-72) for CTCAE v.4 grade 2 (G2) acute (≤90 days), late (>90 days) and permanent (>12 months) RIA. The relative dose-surface histogram (DSH) of the scalp was extracted and used for Lyman-Kutcher-Burman (LKB) modelling. Moreover, DSH metrics (Sx: the surface receiving ≥ X Gy, D_2%: near maximum dose, D_mean: mean dose) and non-dosimetric variables were included in a multivariable logistic regression NTCP model. Model performances were evaluated by the cross-validated area under the receiver operator curve (ROC-AUC). Acute, late and permanent G2-RIA was observed in 52%, 35% and 19% of the patients, respectively. The LKB models showed a weak dose-surface effect (0.09 ≤ n ≤ 0.19) with relative steepness 0.29 ≤ m ≤ 0.56, and increasing tolerance dose values when moving from acute and late (22 and 24 GyRBE) to permanent RIA (44 GyRBE). Multivariable modelling selected S_21Gy for acute and S_25Gy, for late G2-RIA as the most predictive DSH factors. Younger age was selected as risk factor for acute G2-RIA while surgery as risk factor for late G2-RIA. D_2% was the only variable selected for permanent G2-RIA. Both LKB and logistic models exhibited high predictive performances (ROC-AUCs range 0.86-0.90). We derived NTCP models to predict G2-RIA after PT, providing a comprehensive modelling framework for acute, late and permanent occurrences that, once externally validated, could be exploited for individualized scalp sparing treatment planning strategies in brain tumor patients.