Peripheral immune-inflammatory markers associated with intracranial progression-free survival after Gamma Knife radiosurgery in patients with non-small cell lung cancer brain metastases: a retrospective cohort study

BackgroundIntracranial outcomes after Gamma Knife radiosurgery (GKRS) for non-small cell lung cancer (NSCLC) brain metastases vary substantially. Peripheral immune-inflammatory markers may provide additional prognostic information, but their associations with composite intracranial progression-free survival (iPFS) and with documented intracranial progression remain unclear. We therefore evaluated these two outcomes separately.MethodsWe retrospectively analyzed 104 patients with NSCLC brain metastases treated with GKRS between July 2019 and June 2021. Composite iPFS was defined from GKRS to intracranial progression or death from any cause. Documented intracranial progression was additionally analyzed with death before progression treated as a competing event. Six covariates, predefined before fitting the primary multivariable models on clinical grounds, were entered directly into a composite-iPFS Cox model, a Fine-Gray model, and a cause-specific Cox model. Continuous predictors were modeled continuously, restricted cubic splines assessed non-linearity, and time-varying coefficients addressed non-proportionality. A reduced static prediction model was evaluated as a secondary exploratory analysis with 1,000-resample bootstrap optimism correction.ResultsIntracranial progression preceded death in 79 patients, whereas 25 died without previously documented progression. Median overall survival was 14.65 months (95% CI, 11.3–16.2), and median composite iPFS was 6.50 months (95% CI, 6.0–8.5). At 12 months, the cumulative incidences of intracranial progression and competing death were 63.5% and 17.3%, respectively. In complete-case analyses (n = 89), each doubling of neutrophil-to-lymphocyte ratio (NLR) was associated with higher hazards in the composite Cox model (HR, 1.47; 95% CI, 1.10–1.96), time-varying Fine-Gray model (sHR, 1.44; 95% CI, 1.03–2.01), and cause-specific Cox model (HR, 1.58; 95% CI, 1.14–2.19). CD4/CD8 ratio was associated with composite and cause-specific hazards but not with the Fine-Gray subdistribution hazard. The reduced static model had an optimism-corrected C-index of 0.703 and calibration slope of 0.838; calibration weakened at later horizons.ConclusionHigher baseline NLR showed the most consistent association with both general prognosis and documented intracranial progression after GKRS. The endpoint-dependent CD4/CD8 findings and the internally validated prediction model remain exploratory and require external multicenter validation before clinical use.