2231 - Clinical Outcomes of Gamma Knife Lightning for Brain Metastases: Consistent Local Control but Size-Dependent Radionecrosis, Influenced by Dosimetric Shifts and Clinician Planning Priorities

Purpose/Objective(s): Stereotactic radiosurgery treatment planning has evolved with the introduction of inverse planning platforms such as Gamma Knife Lightning. Prior dosimetric analysis at our institution demonstrated that inverse-planned cases exhibit increased selectivity but worse gradient index for small targets compared to forward-planned cases. Whether these dosimetric shifts translate into meaningful changes in local control (LC) or radionecrosis (RN) remains unknown.

Materials/Methods: Retrospective single-institution cohort analyzing consecutive intact brain metastases from lung or breast primaries treated with Gamma Knife radiosurgery from 2020–2024. Patients with less than nine months follow-up were excluded. Lesions were categorized as pre-inverse (before January 1, 2022) or inverse era (on or after January 1, 2022) and stratified by volume: small (<1cc) and large (=1cc). Primary endpoints were LC and RN. RN rates were compared using chi-square testing. Cox models assessed LC and RN over time with era and lesion size as covariates. Kaplan-Meier curves were generated stratified by era and size.

Results: The cohort included 198 brain metastases (pre-inverse:106; inverse:92) with comparable tumor distributions (breast: pre 20, inverse 7; lung: pre 86, inverse 85) and size distributions (pre: 74 small, 32 large; inverse: 67 small, 25 large). LC was excellent across both eras. Pooled 12-month LC was 100.0% for small and 91.2% for large lesions. Cox analysis showed no association between era and LC (HR=0.746, p=0.679), whereas lesion size was significant (small vs large: HR=5.792, p=0.013). Overall RN rates were low and not significantly different between eras (p=0.299), though crude proportions suggested higher RN post-implementation (pre 8.5% vs inverse 13.0%). Twelve-month RN-free rates were 98.0% pre-inverse and 86.6% post-inverse. Lesion size was significantly associated with RN (HR=4.033, p=0.003). Stratified by size, RN showed a non-significant trend toward higher incidence for small lesions in the inverse era (HR=4.615, p=0.062) but no difference for large lesions (HR=1.232, p=0.731). Joint stratification showed 12-month RN-free rates of 100.0% (small) and 93.1% (large) pre-inverse versus 88.9% (small) and 80.0% (large) in the inverse era.

Conclusion: Inverse planning adoption did not compromise LC, which remained excellent across eras, despite dosimetric shifts that could theoretically increase marginal failure risk. Overall RN rates were very low, likely limiting statistical power. However, a concerning non-significant trend toward increased RN was observed for small lesions in the inverse era, aligning with our finding of worse gradient index for small targets. These findings may inform clinician plan-selection priorities, suggesting that attention to gradient index for small lesions may help mitigate RN risk without compromising LC. Larger prospective studies with standardized dosimetric reporting are warranted.