Volumetric and actuarial analysis of brain necrosis in proton therapy using a novel mixture cure model

High-dose fractionated radiotherapy is often necessary to achieve long-term tumor control in several types of tumors involving or within close proximity to the brain. There is limited data to guide on optimal constraints to the adjacent nontarget brain. This investigation explored the significance of the three-dimensional (3D) dose distribution of passive scattering proton therapy to the brain with other clinicopathological factors on the development of symptomatic radiation necrosis.

All patients with head and neck, skull base, or intracranial tumors who underwent proton therapy (minimum prescription dose of 59.4 Gy(RBE)) with collateral moderate to high dose radiation exposure to the nontarget brain were retrospectively reviewed. A mixture cure model with respect to necrosis-free survival was used to derive estimates for the normal tissue complication probability (NTCP) model while adjusting for potential confounding factors.

Of 179 identified patients, 83 patients had intracranial tumors and 96 patients had primary extracranial tumors. The optimal dose measure obtained to describe the occurrence of radiation necrosis was the equivalent uniform dose (EUD) with parameter a = 9. The best-fit parameters of logistic NTCP models revealed D₅₀ = 57.7 Gy for intracranial tumors, D₅₀ = 39.5 Gy for extracranial tumors, and γ₅₀ = 2.5 for both tumor locations. Multivariable analysis revealed EUD and primary tumor location to be the strongest predictors of brain radiation necrosis.

In the current clinical volumetric data analyses with multivariable modelling, EUD was identified as an independent and strong predictor for brain radiation necrosis from proton therapy.