MMDA-20 PREDICTING SURGICAL BENEFIT PRIOR TO STEREOTACTIC RADIOSURGERY FOR BRAIN METASTASES: DEVELOPMENT AND VALIDATION OF PREDICTIVE AND PROGNOSTIC NOMOGRAMS

Abstract

Treatment of brain metastases is increasingly individualized, and despite broad guidelines favoring resection of large, symptomatic, or solitary lesions, the decision to pursue upfront surgery prior to stereotactic radiosurgery (SRS) remains challenging. Individualized tools to predict which patients benefit from surgical resection are lacking. We analyzed a multi-institutional cohort of 1,172 patients to develop and validate nomograms estimating the likelihood of surgical benefit and post-resection survival. Multivariable logistic regression identified four predictors of surgical resection prior to SRS: single brain metastasis (OR = 3.38; p = 0.0004), lesion diameter ≥2.5 cm (OR = 33.3; p < 0.0001), higher KPS (80–100 vs. 40–70; OR = 2.16; p = 0.048), and neurologic deficit on presentation (OR = 2.38; p = 0.016), with strong discriminative accuracy (Somers’ D = 0.77 training; 0.861 internal validation; 0.802 external validation). A significant interaction between number of metastases and resection status was observed on overall survival (p = 0.003), with resected patients harboring a solitary metastasis demonstrating a median OS of 2.93 years (95% CI 1.72–5.27) compared to 0.69 years (95% CI: 0.54-1.35) for those with multiple metastases. Using propensity score matching, patients were classified as surgery-beneficial or surgery-non-beneficial based on whether their survival exceeded the median OS of matched non-surgical controls. A predictive nomogram was developed using logistic regression to estimate the individualized probability of surgical benefit (C-index 0.705 training, 0.639 validation). A separate prognostic nomogram estimated 1-, 3-, and 5-year OS following resection using Cox regression (C-index at 3 years: 0.734 training, 0.756 validation). Calibration was confirmed by expected-to-observed (E/O) ratios near 1.0 and integrated calibration index (ICI) values of 0.052–0.062. Decision curve analysis demonstrated net clinical benefit across low-to-moderate risk thresholds, suggesting the nomogram is most useful for identifying patients unlikely to benefit from resection. These nomograms provide a quantitative, individualized framework to complement surgical decision-making prior to SRS for brain metastases.