This retrospective observational cohort study was conducted at Taizhou Hospital of Zhejiang Province, a tertiary care center serving the Taizhou region of Zhejiang Province, China. We retrospectively analyzed data from consecutive patients with acute ischemic stroke who underwent endovascular thrombectomy between January 2021 and December 2024. The study protocol was approved by the institutional review board of Taizhou Hospital of Zhejiang Province (Registration number: K20181204), and the requirement for informed consent was waived due to the retrospective nature of the study.

The selection of SII as our primary inflammatory biomarker was based on its comprehensive integration of multiple inflammatory pathways and demonstrated prognostic value across various pathological conditions. Rather than conducting comparative analyses with other inflammatory indices, which would substantially expand the study scope, we focused on thoroughly investigating SII's prognostic value through comprehensive dose-response relationships, subgroup analyses, and safety endpoint assessments.

All clinical data were systematically collected and maintained in a standardized stroke database, including baseline demographics, medical history, laboratory parameters, imaging findings, treatment variables, and functional outcomes. Data quality was ensured through regular monitoring and validation processes.

Inclusion Criteria: Patients were included if they met the following criteria: (1) age ≥ 18 years; (2) clinical diagnosis of acute anterior circulation ischemic stroke confirmed by neuroimaging; (3) underwent endovascular thrombectomy; and (4) availability of complete laboratory data required for SII calculation within 24 h after the procedure. Exclusion Criteria: Exclusion criteria were: (1) unsuccessful vascular recanalization or pre-procedure recanalization (defined as failure to achieve mTICI grade 2b or 3 flow) (9); (2) missing 3-month functional outcome assessment; (3) patients with multiple endovascular thrombectomy procedures within 1 year; (4) incomplete laboratory data essential for SII calculation, including neutrophil count, lymphocyte count, or platelet count; and (5) posterior circulation stroke. The exclusion of posterior circulation strokes was based on well-established anatomical and pathophysiological differences between vascular territories. The posterior circulation demonstrates distinct ischemic susceptibility patterns, with brain stem and cerebellar regions showing greater resistance to ischemic injury compared to anterior circulation regions (24, 25). Clinical presentations of posterior circulation strokes are often atypical and variable, leading to diagnostic challenges and increased heterogeneity in patient selection and outcome assessment (25, 26). Additionally, hemorrhagic complications after revascularization therapies differ significantly between circulations, with notably lower rates in posterior circulation strokes (25, 27). Given these fundamental differences, we focused on anterior circulation strokes to enhance internal validity and reduce treatment heterogeneity. Anterior circulation acute ischemic stroke was defined as involving the internal carotid artery (ICA) system, including the middle cerebral artery (MCA) and its branches, the anterior cerebral artery (ACA) and its branches, ICA territory infarctions, and clinically significant anterior choroidal artery infarctions.

The primary endpoint was poor functional outcome at 3 months, defined as modified Rankin Scale (mRS) (11) score of 3–6. The mRS assessment was performed by trained neurologists or stroke nurses through structured telephone interviews or clinic visits at 3 months post-procedure. Outcome assessors were blinded to the SII values and other study variables when feasible.

The safety endpoint was symptomatic intracranial hemorrhage (sICH), defined according to international standards as any hemorrhagic transformation combined with either a ≥4-point increase in NIHSS score compared to the immediate pre-deterioration neurological status or requirement for neurosurgical intervention (28). sICH was selected as the primary safety endpoint due to its greater clinical relevance and standardized definition across major EVT trials.

The SII was calculated using the formula: SII = platelet count ( × 10⁹/L) × neutrophil count ( × 10⁹/L)/lymphocyte count ( × 10⁹/L). Blood samples were obtained within 24 h after endovascular thrombectomy, following standardized laboratory protocols. All laboratory measurements were performed using automated analyzers with established quality control procedures.

Baseline demographic and clinical characteristics were systematically recorded, including age, sex, medical comorbidities (hypertension, diabetes mellitus, atrial fibrillation, and previous stroke), smoking status, and laboratory parameters (complete blood count with differential, hemoglobin, serum creatinine, blood glucose, lipid profile, and D-dimer). Blood samples for SII calculation were obtained within 24 h after endovascular thrombectomy. Stroke severity was assessed using NIHSS at admission. Imaging characteristics included ASPECTS and occlusion site. Treatment variables encompassed puncture-to-reperfusion time, use of intravenous thrombolysis, and stroke etiology (TOAST classification).

Stroke severity was assessed using the National Institutes of Health Stroke Scale (NIHSS) at admission. Imaging characteristics included Alberta Stroke Program Early CT Score (ASPECTS) and occlusion site. Treatment variables encompassed puncture-to-reperfusion time (PRT), use of intravenous thrombolysis, reperfusion outcomes (modified Thrombolysis in Cerebral Infarction [mTICI] (9) scale), and stroke etiology according to Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification.

All procedures were performed by experienced neurointerventionalists according to institutional protocols. Successful recanalization was defined as achieving modified Thrombolysis in Cerebral Infarction (mTICI) (9) grade 2b or 3 flow.

Baseline characteristics were presented according to SII tertiles. Continuous variables were expressed as mean ± standard deviation for normally distributed data or median (interquartile range [IQR]) for non-normally distributed data. Categorical variables were presented as counts (percentages). Differences across tertiles were compared using one-way ANOVA or Kruskal-Wallis test for continuous variables and chi-squared test or Fisher's exact test for categorical variables, as appropriate.

Due to the skewed distribution of SII values, logarithmic transformation (log_SII) was applied when SII was analyzed as a continuous variable. Associations between SII and outcomes were examined using logistic regression models with three levels of adjustment: Model 1 (unadjusted), Model 2 (adjusted for age and sex), and Model 3 (fully adjusted).

The fully adjusted model (Model 3) included age, gender, smoking, hypertension, diabetes, atrial fibrillation, thrombolysis, ASPECTS, and puncture-to-reperfusion time (PRT). Variable selection was guided by directed acyclic graph (DAG) principles (29, 30) to distinguish between confounders and potential mediators. Baseline NIHSS was conceptualized as a mediator on the causal pathway (patient characteristics → stroke severity → inflammatory response → functional outcomes) rather than a confounder. Adjusting for mediators can introduce over adjustment bias by blocking causal pathways and underestimating total effects (31). Therefore, baseline NIHSS was excluded from the primary model to estimate the total effect of SII. In contrast, ASPECTS was retained as it represents baseline anatomical burden (early ischemic changes) that is a common cause of both SII levels and functional outcomes, fulfilling the definition of a confounder. To ensure transparency and assess robustness, sensitivity analyses including baseline NIHSS were performed (Model 4, Supplementary Table S1).

In addition to continuous analysis, SII was also analyzed as categorical tertiles based on the original values, with the lowest tertile (T1) serving as the reference group. P for trend across tertiles was calculated by modeling tertiles as a continuous variable.

Subgroup analyses examined the association between log-transformed SII and poor functional outcome across patient characteristics using fully adjusted logistic regression models (Model 3 variables) with interaction terms to assess effect modification. For each subgroup, we: (1) ran separate fully adjusted models to obtain subgroup-specific odds ratios and P-values; (2) tested interaction terms to assess whether the SII-outcome association differed significantly across subgroup categories. Both individual subgroup P-values and interaction P-values are reported, with the latter being the primary test for effect modification.

Restricted cubic spline analyses were performed to assess the shape of the association between log (SII) and outcomes, using three knots at the 10th, 50th, and 90th percentiles. The reference value was set at the 10th percentile (log_SII = 6.224).

No formal a priori sample size calculation was performed for this retrospective study. All consecutive eligible patients during the study period were included to maximize statistical power and minimize selection bias. The final sample of 741 patients demonstrated adequate statistical power, as evidenced by the significant findings with narrow confidence intervals.

All analyses were performed using R version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value < 0.05 was considered statistically significant.

Among 918 patients with acute ischemic stroke who underwent endovascular thrombectomy between January 2021 and December 2024, 741 patients met the inclusion criteria and were included in the final analysis (Figure 1). A total of 177 patients were excluded: 56 due to unsuccessful vascular recanalization (TICI < 2b) or pre-procedure recanalization, 84 with posterior circulation stroke, 26 lacking 3-month follow-up data, 10 with multiple endovascular thrombectomy procedures within 1 year, and 1 lacking relevant blood test data for SII calculation.

The median SII was 1,247.9 [IQR: 804.9–2,127.1]. Patients were stratified into tertiles: T1 ( ≤ 928.25), T2 (928.26–1,809), and T3 (>1,809), each n = 247.

Higher SII tertiles showed progressively higher neutrophil counts (5.15 vs. 7.04 vs. 9.69 × 10⁹/L) and platelet counts (167.01 vs. 195.45 vs. 229.98 × 10⁹/L), but lower lymphocyte counts (1.49 vs. 1.10 vs. 0.79 × 10⁹/L, all P < 0.001). Clinical presentation severity increased with SII tertiles, as evidenced by higher admission NIHSS scores (mean: 12.98 vs. 13.72 vs. 15.34, P < 0.001) and a greater proportion of patients with NIHSS ≥15 (36.8% vs. 38.9% vs. 53.4%, P < 0.001). Other baseline characteristics including age, gender distribution, vascular risk factors, and procedural variables were generally balanced across tertiles, except for serum creatinine (P < 0.001), total cholesterol (P = 0.02), triglycerides (P = 0.02), and D-dimer levels (P < 0.001), which showed significant differences. Details are in Table 1.

Poor functional outcome (mRS 3–6) at 3 months occurred in 317 patients (42.8%) overall, with rates increasing across SII tertiles: 37.2% (92/247) in T1, 42.5% (105/247) in T2, and 48.6% (120/247) in T3 (P = 0.011; Figure 2).

In multivariable analysis (Table 2), log-transformed SII as a continuous variable was significantly associated with poor functional outcome. The association remained significant across all models: unadjusted (OR 1.38, 95% CI 1.13–1.68, P = 0.002), age and sex-adjusted (OR 1.45, 95% CI 1.18–1.78, P < 0.001), and fully adjusted (OR 1.428, 95% CI 1.159–1.759, P < 0.001).

When analyzed by tertiles, compared to the lowest tertile (T1), patients in T3 had significantly higher odds of poor functional outcome in all models. In the fully adjusted model, the OR for T3 was 1.73 (95% CI 1.18–2.52, P = 0.005), while T2 showed no significant difference (OR 1.25, 95% CI 0.86–1.84, P = 0.243). The P for trend across tertiles was significant (P = 0.005), confirming a dose-response relationship.

Sensitivity analyses including baseline NIHSS (Model 4) demonstrated directionally consistent associations with attenuated effect estimates, as theoretically predicted when adjusting for a mediator (Supplementary Table S1). For continuous log-transformed SII, the adjusted OR was 1.29 (95% CI 1.04–1.60, P = 0.023). When analyzed by tertiles, compared to T1, the adjusted OR for T3 was 1.39 (95% CI 0.93–2.06, P = 0.107), though the trend test across tertiles was no longer statistically significant (P for trend = 0.107). This pattern of effect attenuation while maintaining directional consistency supports our theoretical framework and confirms that our primary analysis (Model 3) better estimates the total effect of SII on functional outcomes.

Subgroup analyses demonstrated generally consistent associations across patient characteristics (Figure 3). Individual subgroup analyses showed varying levels of statistical significance, with some subgroups not reaching significance due to reduced sample sizes after stratification. However, no significant interactions were detected (all P for interaction > 0.05), indicating that the prognostic value of SII was consistent across different patient populations regardless of age, gender, comorbidities, stroke severity, or treatment modalities.

Restricted cubic spline analysis (Figure 4) demonstrated a significant overall association between log (SII) and poor functional outcome (overall P < 0.001), with no evidence of non-linearity (P for non-linearity = 0.173). This indicates a linear relationship between increasing log (SII) values and the odds of poor functional outcome throughout the observed range.

Symptomatic intracranial hemorrhage (sICH) occurred in 39 patients (5.3%) overall, with rates of 5.3% (13/247) in T1, 3.2% (8/247) in T2, and 7.3% (18/247) in T3 (P = 0.148; Table 3).

Neither log (SII) as a continuous variable nor SII tertiles showed significant associations with sICH in any of the regression models. The fully adjusted OR for log (SII) was 1.32 (95% CI 0.84–2.07, P = 0.235). No significant trends were observed across tertiles (P for trend = 0.260).