ABSTRACT NUMBER: ESOC2026A218 STRUCTURAL CONFOUNDING EXPLAINS MOST HETEROGENEITY IN LATE-WINDOW REPERFUSION TRIALS: A NETWORK META-EPIDEMIOLOGIC ANALYSIS OF 14 RANDOMIZED STUDIES

Abstract Background and aims Late-window reperfusion trials (>4.5 hours) show substantial variation in imaging criteria, infarct severity, and workflow, complicating indirect comparisons across thrombectomy, thrombolysis, and best medical therapy. We aim to quantify structural sources of heterogeneity to interpret treatment contrast betwwen trials. Methods Using a PROSPERO-registered protocol, we identified 14 randomized trials (n = 4,384) enrolling adults 4.5–24 hours from last-known-well using imaging-based selection. Random-effects pairwise and network meta-analyses estimated odds ratios (ORs) with between-study τ2 and I2. Results EVT significantly improved functional independence versus BMT (OR 3.14, 95% CI 1.82–5.41; τ2 = 0.366; I2 = 76.9%). Mortality was likewise reduced (OR 0.82, 95% CI 0.67–1.00). Meta-regression identified baseline ASPECTS as the dominant ecological modifier, explaining 75.5% of EVT-BMT heterogeneity (R2 = 0.755). Each +1-point increase in ASPECTS corresponded to a 14-18% relative attenuation of the EVT effect (slope range –0.15 to –0.20; p values 0.008–0.035). Bridging-thrombolysis proportion varied markedly across regions (~5% in North America vs 20-30% in Europe/China; χ2 = 35.0; p < 0.0001). Bridging was strongly correlated with faster workflow (β = –20.2 minutes per 1.0 unit increase; R2 = 0.95; p = 0.0002), explained only 11% of between-trial heterogeneity in EVT effect size (β = –1.90; p = 0.31). Conclusions In late-window stroke, EVT consistently confers benefit, but cross-trial variation primarily reflects structural confounding rather than biological treatment differences. Harmonizing imaging thresholds and workflow metrics is essential for generating reliable comparative evidence for late-window reperfusion strategies. Conflict of interest Hung Phan Huu, MS-5: Nothing to disclose. Han Hong Huynh, MD: Nothing to disclose. Dang Nguyen, MD: Nothing to disclose. Nghia Nguyen Phu, MD: Nothing to disclose. Linh Nguyen Ngoc Yen, MSc-2: Nothing to disclose. Khoa Ngoc-Dang Tran, MD: Nothing to disclose. Uyen Dang, MD: Nothing to disclose. Tam Quoc Minh Tran, MD: Nothing to disclose. Chau Dang, MD: Nothing to disclose. Wei Jun Lee, MD: Nothing to disclose. Tuan Vinh, BA: Nothing to disclose. Thu Huynh Minh Le, PharmD: Nothing to disclose. Nhi Huu Hanh Le, MD: Nothing to disclose. Hoai Huu Le, MD: Nothing to disclose. Huyen Le, MD: Nothing to disclose. Lam Duc Chau, BA: Nothing to disclose. Uyen Do, MD: Nothing to disclose. Minh Le, MD, PhD: Nothing to disclose. Thach Nguyen, MD: Nothing to disclose.