RADT-68. PREDICTING SURVIVAL OUTCOMES OF PATIENTS RECEIVING CAVITY BOOST RADIOTHERAPY FOLLOWING SURGICAL RESECTION OF BRAIN METASTASES: A 5-YEAR RETROSPECTIVE ANALYSIS

Abstract INTRODUCTION Due to inconclusive evidence, the role and timing of cavity boost radiotherapy (CBRT) following resection of metastases varies in practice globally. Better patient stratification is urgently needed to identify those that might benefit, but limited predictive models exist. This study aims to explore factors associated with improved outcomes in this patient cohort, to inform a clinically useful predictive algorithm. METHODS Patients (n=163, M:64, F:99): histologically confirmed brain metastases undergoing resection at a UK Brain Tumour Centre of Excellence from 2016-21. Data collected: primary site, staging, total intracranial metastases, steroid treatment, KPS, surgery, overall survival (OS), local recurrence. Statistical analysis: Restricted Mean Survival Times (RMST), following Inverse Probability of Treatment Weightage (IPTW) confounder adjustment; Univariate and Multivariate Piecewise Cox Proportional Hazards models to identify significant risk factors, with Leave-One-Out Cross-Validation (LOOCV), Likelihood Ratio Test (LRT) and time-dependent Area Under the Curve (AUC) performed as validation. RESULTS 50 (31%) patients received CBRT (fractionated=31, single-fraction-SRS=19). CBRT demonstrated significant cumulative OS gains from 1 year [RMST: 2.31, 95%CI:-3.30 - -1.40, p=0.002] to 4.6 years [RMST: 6.33, 95%CI:-13.33 - -0.38, p=0.049]. CBRT had the most significant protective effect on OS in the high-risk mortality group (OS<7months) [HR: 0.14, Bootstrapped 95%CI: 0.04-0.39, p=0.0003] compared to moderate-risk (7months<OS<45months) and low-risk (OS>45months) groups. Per 10-point increase in postoperative KPS, mortality risk decreased by 19% [HR: 0.98, Bootstrapped 95%CI: 0.97-0.99, p=0.009]. Multivariate modelling was statistically significant overall (LRT χ²=95.5, p<0.0001), with good predictive power (LOOCV C-index: 0.70, 95%CI: 0.64–0.75) across 1-5 years respectively (time-dependent AUCs: 0.711, 0.715, 0.710, 0.638, 0.611). CONCLUSION Based on a large, multivariate dataset, with long-term follow-up, we identified postoperative KPS with early, upfront CBRT as being key to conferring significant and sustained benefits to patients. These data informed a risk-adapted predictive model that can be used in the management of this patient cohort.