Stereotactic Radiosurgery (SRS) for Large Brain Metastases: Dosimetric and Clinical Predictors of Local Progression and Radionecrosis

Abstract

Stereotactic radiosurgery (SRS) provides high rates of local control for small brain metastases with low rates of radionecrosis (RN). Larger targets are associated with increased risk of both local progression (LP) and RN. In this analysis, we hypothesized that dosimetric and clinical parameters predict for risk of LP and RN in SRS targets larger than two centimeters. We retrospectively reviewed patients with one or more targets with either an intact versus post-operative cavity larger than 2.0 cm treated with LINAC-based SRS between 2017 and 2022 at one institution. We assessed for association between patient, treatment, and disease variables with LP and RN. Variables assessed included tumor resection status, PDL1 positivity, target volume, maximum and minimum target dose, EQD2 and BED (a/b = 2 for necrosis and a/b = 10 for tumor control), as well as receipt of steroids, bevacizumab, or systemic therapy before or after SRS. Radionecrosis was determined by characteristic radiographic changes. Analyses were performed for the entire cohort and within subsets including by resection status and dose fractionation. A total of 178 lesions in 143 patients were included. Targets with volume diameters measuring at least 2 cm were used. Median follow-up was 2.3 years. Overall survival at 1 and 2 years was 56% and 32%, respectively. Most lesions (n = 119) were resected and treated with SRS post-operatively. The most common dose and fractionation schemes used were 30 Gy in 5 fractions (n = 89) and 27 Gy in 3 fractions (n = 63). For the entire cohort, the cumulative incidence of LP 1 and 2 years was 26% and 34%, respectively. The cumulative incidence of radiographic radionecrosis at 1 and 2 years was 12% and 17%, respectively. There was no difference in LP or RN between 27 Gy in 3 fractions versus 30 Gy in 5 fractions (p>0.5 for both). Median planning target volume (PTV) size was 18.5 cc for the 27 Gy in 3 fraction group compared to 21.9 cc in the 30 Gy in 5 fraction group. Minimum or maximum dose within the target was not associated with increased risk of LP or RN. Among patients receiving 27 Gy in 3 fractions, patients treated with resection followed by SRS had lower risk of LP compared to those treated with SRS alone (HR: 0.15, 95% CI: 0.03-0.64, p = 0.011). Among patients receiving 30 Gy in 5 fractions, patients who received corticosteroids prior to SRS had a lower risk of RN (HR: 0.14, 95% CI: 0.03-0.66, p = 0.013). For the entire cohort as well as within all subgroups, PD-L1≥1% was associated with increased risk of RN (p<0.001 for all). Selecting the optimal SRS dose fractionation and planning parameters to minimize both LP and RN remains a challenge for large targets. In this analysis, 27 Gy in 3 fractions appeared to provide equivalent LP and RN compared to 30 Gy in 5 fractions, and may be more convenient for patients. Patients with PD-L1≥1% with large brain targets treated with SRS may be at increased risk of RN; corticosteroid prophylaxis may be considered in this population.