Sequential NIHSS–NLR monitoring after endovascular thrombectomy: malignant cerebral oedema and inflammatory futility in a two-phase derivation study

Introduction: Successful reperfusion after endovascular thrombectomy (EVT) does not guarantee recovery. Malignant cerebral oedema (MCE) complicates approximately one in eight patients, and static pre-procedural models do not capture evolving post-reperfusion risk. We evaluated a two-phase sequential clinical-inflammatory monitoring framework. Methods: We conducted a retrospective consecutive-patient cohort study at Xiangya Hospital (2021-2025) including 306 successfully reperfused EVT patients. Phase 1 combined the 24-h NIHSS trajectory with neutrophil-to-lymphocyte ratio (NLR) direction for clinically ascertained MCE risk. Ridge-penalised logistic models were internally validated using Harrell bootstrap correction. Exploratory Phase 2 combined 72-h NLR with CT-perfusion anatomy for futile recanalisation (90-day mRS ≥3). NLR direction provided minimal incremental discrimination (DeltaAUC +0.008, 95% CI -0.007 to +0.020) but separated complication profiles: concurrent NIHSS worsening and NLR rise identified an oedema-enriched subgroup (MCE 26.9%; NNM 3.7), whereas non-rising NLR among worsening patients identified a haemorrhage-enriched subgroup (sICH 37.7%). At 72 h, favourable anatomy with elevated NLR defined a candidate inflammatory futility phenotype with lower functional independence (57.3% vs 84.2%; corrected AUC 0.631, upper bound by construction). Discussion: Because MCE ascertainment was clinically triggered and adjudication was unblinded to NIHSS trajectory, Phase 1 identifies a clinical-inflammatory risk pattern rather than an independent radiological predictor. NLR direction functions primarily as a complication-profile stratifier. Prospective multicentre validation with blinded radiological adjudication is required before clinical or trial-design application.