Background and objectiveEndovascular thrombectomy (EVT) has become the gold standard in the treatment of acute stroke patients. However, not all patients respond well to this treatment despite successful attempts. In this study, we aimed to identify variables associated with the failure of improvements following EVT.MethodsWe retrospectively analyzed prospectively collected data of 292 ischemic stroke patients with large vessel occlusion who underwent EVT at three academic stroke centers in China from January 2019 to February 2022. All patients were above 18 years old and had symptoms onset ≤6 h. A decrease of more than 4 points on the National Institute of Health Stroke Scale (NIHSS) after 24 h compared with admission or an NIHSS of 0 or 1 after 24 h was defined as early neurological improvement (ENI), whereas a lack of such improvement in the NIHSS was defined as a failure of early neurological improvement (FENI). A favorable outcome was defined as a modified Rankin scale (mRS) score of 0–2 after 90 days.ResultsA total of 183 patients were included in the final analyses, 126 of whom had FENI, while 57 had ENI. Favorable outcomes occurred in 80.7% of patients in the ENI group, in contrast to only 22.2% in the FENI group (p < 0.001). Mortality was 7.0% in the ENI group in comparison to 42.1% in the FENI group (p < 0.001). The multiple logistic regression model showed that diabetes mellitus [OR (95% CI), 2.985 (1.070–8.324), p = 0.037], pre-stroke mRS [OR (95% CI), 6.221 (1.421–27.248), p = 0.015], last known well to puncture time [OR (95% CI), 1.010 (1.003–1.016), p = 0.002], modified thrombolysis in cerebral infarction = 3 [OR (95% CI), 0.291 (0.122–0.692), p = 0.005], and number of mechanical thrombectomy passes [OR (95% CI), 1.582 (1.087–2.302), p = 0.017] were the predictors of FENI.ConclusionDiabetes mellitus history, pre-stroke mRS, longer last known well-to-puncture time, lack of modified thrombolysis in cerebral infarction = 3, and the number of mechanical thrombectomy passes are the predictors of FENI. Future large-scale studies are required to validate these findings.