CNS Radiation Necrosis in Pediatric Patients Treated With Pencil-Beam Scanning Proton Therapy: Interim Results From a Phase IV Surveillance Trial

Abstract

<h3>Purpose/Objective(s)</h3> An interim analysis of a Phase IV clinical trial was performed to estimate the incidence of CNS radiation necrosis (CNS-RN) and identify predictive parameters. <h3>Materials/Methods</h3> At the time of analysis, the trial had accrued 366 patients with CNS and base of skull tumors treated with pencil beam scanning proton therapy (PBS-PT). Patients with a prior history of radiation therapy (n = 32), treatments with no reported dose to the brain (n = 4), or incomplete dosimetry (n = 39) were excluded. Cox proportional hazards models were used to evaluate the impact of interventions on time to resolution. Events were graded according to CTCAE v4 guidelines. Eloquent cortex was defined as a brain region which, when damaged, lead to severe focal neurologic deficits. Brain and brainstem dose volume histograms were pooled to create complication incidence maps for grade 1+ radiation necrosis. The total probability that the true rate of necrosis exceeded 5% in the brainstem and 10% in the brain was estimated at all dose-volume combinations. Intersections at volumes of 0.1cc and 1cc were assessed at probabilities of 10% and 50%. <h3>Results</h3> Twenty-three (n = 23) CNS-RN events, corresponding to 15 grade 1 and 8 grade 3 events, were observed in 291 patients resulting in a 2-year CI in grade 3 necrosis of 6.02% +/- 0.013 (s.e). CNS-RN (any grade) was more common in astrocytic-oligodendroglial, embryonal, and ependymal tumors. PBS-PT attribution was likely or plausible in all but 2 events. Although CNS-RN could not be attributed solely to end-of-range effects, 11 (48%) cases occurred in the region of the cervico-medullary junction. Complete resolution of CNS necrosis occurred in all 8 grade 3 patients and in 12 of 15 (80%) grade 1 events. Grade 3 or more severe events were treated with bevacizumab, corticosteroids, or hyperbaric oxygen therapy. Symptomatic resolution was comparable between treatments although bevacizumab showed an advantage for radiographic resolution (HR 4.3, 95% CI 1.6-11). Factors predictive of symptom resolution were eloquence of the region of necrosis (HR 0.15, 95% CI 0.035-0.67, <i>P</i> = 0.0003) and any chemotherapy exposure (HR 0.16, 95% CI 0.03-0.88, <i>P</i> = 0.034). The probability of true necrosis rate exceeding 5% in brainstem increased from less than 10% to greater than 50% at 58.5CGE to a volume of 0.1cc and 57.7CGE to a volume of 1cc. The probability that the true necrosis rate exceeded 10% in the brain increased from less than 10% to greater than 50% at a dose of 63.1CGE to a volume of 0.1cc. At a volume of 1cc, sparsity of data limited analysis to the less than 50% to greater than 50% probability level, which occurred at 60.5CGE. <h3>Conclusion</h3> Incidence of radiation necrosis was consistent with photon historical data. These data will aid the development of PBS-PT delivery guidelines aimed at minimizing CNS necrosis risk for children requiring RT which involves or approximates the CNS.