Dosimetric predictors of symptomatic radiation necrosis after five-fraction radiosurgery for brain metastases

Background To identify factors predictive of developing symptomatic radiation necrosis (sRN) among patients with either intact or resected brain metastases undergoing five-fraction stereotactic radiosurgery (5fSRS). Methods Multi-institutional retrospective review of 117 brain metastases from 83 patients treated with 5fSRS. The cumulative incidence of sRN and predictors of sRN were calculated using Gray's competing risks and Cox regression. Results The median dose of 5fSRS was 30 Gy (range: 25–40), and 21 lesions (18%) had prior SRS. After a median follow-up of 10.3 months (range: 3–52), the cumulative sRN incidence was 15%, with a median time to sRN of 6.9 months (range: 1.8–31.7). sRN incidence was significantly higher among the lesions treated with prior SRS: hazard ratio (HR): 7.48 [95% confidence interval: 2.57–21.8]. Among lesions without prior SRS, higher volume of uninvolved brain receiving 25 Gy (BrainV25; HR: 1.07 [1.02–1.12]) and 30 Gy (BrainV30; HR: 1.07 [1.01–1.33]) were the most significant factors associated with sRN. Similar results were also observed among the patients with prior SRS. For lesions without prior SRS, BrainV25 > 16 cm3 (HR: 11.7 [1.47–93.3]) and BrainV30 > 10 cm3 (HR: 7.08 [1.52–33.0]) were associated with significantly higher risk of sRN. At two years, the sRN incidence was 21% if violating either dosimetric threshold and 2% if violating neither (p = .007). Conclusion BrainV25 and BrainV30 are significant dosimetric predictors of sRN of brain metastases treated with 5fSRS. In the absence of prior SRS, maintaining BrainV25Gy < 16 cm3 and BrainV30Gy < 10 cm3 may minimize sRN risk.