Abstract DP016: Comparative Efficacy of Interventions for Supratentorial Stroke: A Bayesian NMA with Meta-Regression

Methods: We synthesized evidence from 43 randomized trials (87 arms; 3,399 patients) of supratentorial stroke—predominantly intracerebral haemorrhage—comparing surgical (craniotomy, minimally invasive surgery [MIS] including endoscopic evacuation, decompressive craniectomy) and pharmacologic strategies against standard care. Primary outcomes were 6-month all-cause mortality and functional status (mRS). A Bayesian random-effects network meta-analysis (HalfNormal(0,0.5) prior on τ) used standard care as the reference; network meta-regression (NMR) examined prespecified covariates (age, ICH volume, GCS, NIHSS, timing, etc.). Consistency was assessed by node-splitting and a design-by-treatment test; small-study effects by comparison-adjusted funnel and Egger regression. Results: The network was well connected (27 interventions across 36 trials). Surgery outperformed medical management. Endoscopic MIS reduced mortality (RR≈0.66) and increased functional independence (RR≈1.31 for good outcome) versus conservative care; craniotomy also lowered mortality (RR≈0.75). Pharmacologic options (e.g., tranexamic acid, atorvastatin, cerebrolysin) showed no significant benefit for mortality or function compared with standard care. Ranking favoured endoscopic MIS (SUCRA ≈85%) and stereotactic MIS (≈80%); standard care ranked lowest. Between-study heterogeneity was low (τ≈0.10; I^2≈14%). There was no global inconsistency (design-by-treatment P=0.32). Funnel plots and Egger tests suggested no important small-study bias (P>0.25). NMR identified earlier treatment (