Clinical correlates, procedural benchmarks, and reperfusion-outcome dissociation after mechanical thrombectomy for...

Endovascular thrombectomy (EVT) is an established treatment for selected patients with acute ischemic stroke caused by large-vessel occlusion (LVO). Single-center registries remain useful for reviewing outcomes, safety events, and workflow in routine care provided that their analytic limits are clearly stated.

We performed a retrospective analysis of a single-center mechanical thrombectomy registry covering January 2023 to December 2025. The final cohort included 212 patients; 90-day modified Rankin Scale (mRS) data were available in 206 (97.2%). The primary outcome was functional independence at 90 days (mRS 0-2). Procedural and safety outcomes were summarized for the full cohort. Two pre-specified logistic regression models were used to explore clinical and procedural correlates and were interpreted with explicit caution regarding sample size and events-per-variable limitations. Robustness was assessed using (i) Firth penalized maximum-likelihood estimation, (ii) an anterior-circulation-only sensitivity Model 1, (iii) E-value computation and inverse-probability-of-treatment weighting for the intravenous thrombolysis (IVT) association, (iv) four-scenario handling of missing 90-day mRS including multiple imputation by chained equations, (v) optimism-corrected area under the curve (AUC) with 2,000-iteration bootstrap, and (vi) a post-hoc Model 1 incorporating collateral circulation score.

Among patients with available 90-day follow-up, 111 of 206 (53.9%) achieved functional independence; across alternative handling of the 6 patients with missing follow-up, the proportion ranged from 52.4% (worst-case) to 55.2% (best-case), with multiple imputation yielding 53.3%. Baseline NIHSS score showed the most consistent adjusted association with lower odds of functional independence [Model 1: adjusted odds ratio (aOR) 0.95, 95% CI 0.91-0.99; Firth aOR 0.95, 95% CI 0.91-0.99]. Intravenous thrombolysis before thrombectomy showed a numerical adjusted association with functional independence (Model 1 aOR 2.14, 95% CI 1.11-4.13; Firth aOR 2.04, 95% CI 1.09-3.91). This magnitude is much larger than randomized-trial pooled estimates (relative risk approximately 1.05-1.15) and is best explained by confounding by indication. The E-value of 2.10 for the point estimate (1.29 for the lower confidence bound) indicates that a moderately strong unmeasured confounder could explain the observed association. In an anterior-circulation-only sensitivity analysis (n = 171), the IVT association no longer reached statistical significance (aOR 2.02, 95% CI 0.98-4.17), reflecting both reduced statistical power and territory-dependent case mix. Overlap weighting, which achieved exact balance on all measured propensity-score covariates whereas standard inverse-probability-of-treatment weighting did not (residual onset-to-door SMD = 0.74), produced an attenuated IVT aOR of 1.88 (95% CI 0.58-6.09)-no longer statistically significant and close to the pooled randomized-trial effect size. No workflow or procedural variable reached statistical significance after adjustment. Apparent discrimination was moderate (C-statistic: Model 1, 0.679; Model 2, 0.703); optimism-corrected AUC was 0.60 for both models, the expected pattern of overfitting in low-EPV settings. Despite a TICI 2b-3 reperfusion rate of 99.5%, 45.9% of reperfused patients failed to achieve functional independence-a reperfusion-outcome gap that mirrors the increasingly recognized phenomenon of clinically ineffective reperfusion.

In this single-center retrospective registry, baseline stroke severity was the most consistent clinical correlate of 90-day functional independence after mechanical thrombectomy, holding up across Firth-penalized, anterior-circulation-only, and collateral-adjusted sensitivity analyses. The IVT association lost statistical significance under overlap weighting (aOR 1.88, 95% CI 0.58-6.09), which we read as confounded rather than causal and which brings the present registry in line with the small randomized-trial effect size. Procedural success was exceptionally high yet decoupled from functional recovery in nearly half of patients. The data are best viewed as a local EVT benchmark and as a basis for quality review, including continued attention to door-to-EVT workflow and to mechanisms of futile reperfusion.