Dose-Size Response Relationship For Melanoma Brain Metastases Treated With Stereotactic Radiosurgery In The Era Of Immunotherapy

Abstract

To determine whether the relationship between the stereotactic radiosurgery (SRS) planning parameter peripheral dose per lesion diameter (Gy/mm) and lesion control differs for intact melanoma brain metastases (MBM) treated with single-fraction SRS with versus without immunotherapy (IO). We retrospectively evaluated an institutional database of consecutive patients with intact MBM treated with single-fraction linear accelerator-based SRS from 2008 to 2019. Response was assessed using RECIST 1.1. Kaplan-Meier methods were used to estimate survival outcomes including lesion control (LC) with a frailty term to account for multiple lesions within a single patient. Cox frailty models were fit to assess the impact of IO. A multivariate Cox frailty model was constructed to include confounders selected by penalized Cox frailty regression models with a LASSO selector. Interaction effect testing was used to determine whether a significant effect between IO and peripheral dose per lesion size could be demonstrated with respect to LC. The study cohort comprised 70 patients with 244 MBMs treated with SRS, including 30 patients with 122 lesions treated with both SRS and IO. Immunotherapy included anti-PD-1 (n = 41 lesions), anti-CTLA-4 (n = 52 lesions), and combination anti-PD-1/anti-CTLA-4 (n = 29 lesions) therapy administered within 12 weeks of SRS. The median lesion diameter was similar for the SRS with (7.9mm, IQR: 5.0 – 12.8 mm) versus without (7.4mm, IQR: 5.9- 9.5 mm) IO subgroups (p = 0.811). LC at 12 months was 83.9% (95% CI 64.4 - 97.1%) versus 75.3 (95% CI 59.1 - 87.8%) for the SRS with versus without IO subgroups, respectively (marginal HR = 0.30, p = 0.04). The logarithm of peripheral dose per lesion diameter was selected as a predictor of LC by the LASSO operator (HR 0.307, 95% CI 0.098 – 0.441), adjusting for IO receipt (HR 0.363, 95% CI 0.108 -1.224) with prior IO receipt, peripheral SRS dose, and BRAF mutation status excluded. Interaction effect testing demonstrated a differential effect of peripheral dose per lesion diameter by IO receipt, with respect to LC (HR 0.634, 95% CI 0.168 -2.399 for SRS and IO; HR 0.117, 95% CI 0.044 – 0.316 for SRS without IO; p = 0.048). The estimated 12-month LC rates for a 7.5 mm lesion receiving a peripheral SRS dose of 18 Gy with versus without IO were 83.8% (95% CI 54.7 – 97.9%) versus 73.6% (95% CI 38.4 – 92.7%), respectively. Peripheral dose per lesion size was predictive of LC in patients with small MBMs treated with single-fraction SRS. We found a differential effect of dose per lesion size and LC by immunotherapy receipt. Future studies are needed to determine whether lower doses of single-fraction SRS may result in similarly effective LC for patients with small MBM receiving immunotherapy.