Abstract
BACKGROUND: Mechanical thrombectomy (MT) is an established reperfusion therapy for acute ischemic stroke due to large vessel occlusion (LVO-AIS) and has been proven to significantly improve 90-day functional outcomes. However, some patients still experience early neurological deterioration (END) despite successful recanalization. This study aimed to systematically identify independent risk factors for END after MT via retrospective cohort analysis and construct a nomogram by integrating laboratory and clinical characteristics. METHODS: A total of 486 LVO-AIS patients with successful recanalization (eTICI≥2b) were first categorized as END (ΔNIHSS≥4) or non-END (ΔNIHSS<4) according to the change in the NIHSS score from baseline to 24 h post-procedure. The entire patient group was then randomly divided to obtain training (70%, n = 341) and validation (30%, n = 145) cohorts. A nomogram was constructed and subsequently validated. RESULTS: We conducted a LASSO regression analysis on the clinical data of the patients in the modelling cohort and identified 6 disease characteristic variables. These variables were subsequently entered into a multivariate logistic regression model, which ultimately retained 5 independent predictors: smoking status (OR = 3.90, 95% CI 1.760-9.681, p = 0.002), platelet count (OR = 1.54, 95% CI 1.129-2.130, p = 0.007), systolic blood pressure (SBP) (OR = 1.80, 95% CI 1.300-2.543, p = 0.001), puncture-to-recanalization time (PRT) (OR = 1.57, 95% CI 1.084-2.346, p = 0.024), and the neutrophil-to-lymphocyte ratio (NLR) (OR = 1.86, 95% CI 1.311-2.703, p = 0.001). The prediction model showed good discriminative performance, with a C-index of 0.803 (95% CI: 0.738-0.868; p < 0.001). The area under the ROC curve (AUC) was 0.803 (95% CI 0.738-0.868) in the training cohort and 0.799 (95% CI 0.693-0.905; p < 0.001) in the validation cohort. The model also exhibited good calibration. The Hosmer-Lemeshow goodness-of-fit test showed no significant difference between the predicted and observed results (χ(2) = 3.607, p = 0.891), and the mean absolute error of the calibration curve in the training cohort was 0.169. CONCLUSION: The constructed prediction model accurately estimated the risk of END in LVO-AIS patients who underwent MT with successful recanalization and may help optimize patient selection for endovascular therapy and provide reliable prognostic information.