Abstract
BACKGROUND AND OBJECTIVE: The 24-h National Institutes of Health Stroke Scale (NIHSS) is a promising early surrogate for 90-day functional outcome after mechanical thrombectomy (MT). However, its predictive performance in posterior circulation infarctions (PCI) is not well established. In addition, whether this association is consistent across anterior (ACI) and PCI territories, and whether it follows a linear or non-linear pattern, remains unclear. METHODS: This prospective cohort included patients from two centers (2015-2022). The primary outcome was a favorable 90-day functional outcome (The modified Rankin Scale [mRS] score ≤2). The primary analysis used multivariable logistic regression to evaluate the predictive value of the 24-h NIHSS, with results expressed as the area under the curve (AUC). Secondary analyses included testing for interaction between the 24-h NIHSS and infarct location (ACI vs. PCI) and modeling non-linear relationships using restricted cubic splines. RESULTS: A total of 334 patients were included (ACI: 282 [84.4%], PCI: 52 [15.6%]). Median baseline and 24-h NIHSS were 16 (IQR: 11-19) and 14 (6-20), respectively. The 24-h NIHSS demonstrated the highest predictive accuracy (AUC: 0.849, 95% CI: 0.805-0.893) compared to other metrics. An optimal cutoff of ≤6 yielded a sensitivity of 94.9% and specificity of 65.8%. Critically, no significant interaction between infarct location and NIHSS metrics was found (p for interaction > 0.05), indicating consistent predictive performance across ACI and PCI. Furthermore, the 24-h NIHSS exhibited a non-linear association with functional independence (p for non-linearity = 0.024), suggesting that its predictive value does not increase uniformly across the entire score range. CONCLUSION: The 24-h NIHSS is a reliable early surrogate for 90-day functional recovery post-MT in both ACI and PCI, outperforming both baseline NIHSS and Δ-NIHSS. The non-linear relationship between 24-h NIHSS and functional independence highlights that treating it as a continuous variable has limitations, strongly supporting the use of this pragmatic cutoff rather than treating the scale as a simple continuous linear predictor.