Patients were screened from the Nanjing Stroke Registry Program (NSRP, between January 2014 and December 2019) and the Captor trial (a multi-center clinical trial between March 2018 and July 2019, register code: ChiCTR1900025256). Detailed descriptions of NSRP had been published elsewhere (19). The Captor trial was aimed to compare the Captor stent retriever with the Solitaire FR device for rapid flow restoration in AIS-LVO. The non-inferiority of the Captor retrievable stent was proved after enrolling 245 patients from 16 comprehensive stroke centers in China.

The key inclusion criteria were as follows: (1) age ≥18 years; (2) National Institute of Health Stroke Scale (NIHSS) score on admission ≥6; (3) occlusion of the intracranial large vessel (defined as diameter ≥2 mm) proved by CT angiography, magnetic resonance angiography, or digital subtraction angiography; (4) completion of groin-puncture within 8 h since stroke onset.

The key exclusion criteria were as follows: (1) active hemorrhage or hemorrhagic tendency; (2) severe organic diseases such as heart, lung, liver, and kidney failure; (3) occlusion attributed to arterial dissection or arteritis. The Ethics Committee of Jinling Hospital and each participating center approved this study according to the Declaration of Helsinki. The requirement for written informed consent was waived because of its retrospective nature.

Baseline demographic parameters, medical histories, laboratory tests, the NIHSS scores, and radiographic evaluation were obtained. The treatment profiles were also gathered, including intravenous thrombolysis (IVT), number of retrieval attempts, final reperfusion status, and procedural time. IVT was administrated within 4.5 h of stroke onset after excluding contraindications. Retrieval attempts were recommended no more than three times, but it was at the discretion of the interventionalists. Reperfusion was evaluated with the modified Thrombolysis In Cerebral Ischemia scale (mTICI) (20). Successful reperfusion was defined as the mTICI scale score of 2b or 3.

Post-procedural NIHSS scores of all patients were assessed by attending neurologists. Follow-up imaging examinations were performed within 24 ± 6 h after the procedure. Symptomatic intracranial hemorrhage (SICH) was defined as imaging evidence of intracranial hemorrhage (ICH) with an increase of ≥4 points on the NIHSS within 24 h. On day 7 ± 2 or hospital discharge if earlier, assessments of neurological severity were performed on the subjects again.

Post-procedural laboratory tests were carried out routinely in the acute phase (7 ± 2 days after MT) and/or in case of clinical deterioration. Liver function was assessed according to tests of serum aspartate aminotransferase [upper limit of normal (ULN): 40 U/L], alanine aminotransferase (ULN: 40 U/L), total bilirubin (ULN: 21 mmol/L), and direct bilirubin (ULN: 10 mmol/L). ALFT was defined as any elevated value above the ULN (21). Multiple_max was defined as the maximum multiple of ULN in the four variables (aspartate aminotransferase, alanine aminotransferase, total bilirubin, and direct bilirubin) during hospitalization after MT.

Follow-up was performed by outpatient clinical visits or telephone contacts to record the modified Rankin Scale (mRS) score at 90 ± 14 days. A favorable outcome was defined as mRS 0–2 at 90 days, and an excellent outcome was referred to as mRS 0–1 at 90 days.

Median imputation was performed to account for missing values [white blood cell: 6 (1.4%); creatinine: 9 (2.1%); glucose: 11 (2.6%); activated partial thromboplastin time: 17 (4.0%); partial thromboplastin time: 13 (3.1%); international normalized ratio: 13 (3.1%)].

Frequency (percentage) was used for categorical variables. Mean (SD) was adopted for continuous variables with normal distribution and median (interquartile range, IQR) for non-normal variables.

Univariate analysis was conducted with Student t-test, Mann–Whitney U, χ² test, or Fisher exact test as appropriate. Confounding factors and other variables with a statistical trend (p ≤ 0.1) in the univariate analysis were included in multivariate logistic regression analysis (forward selection) to develop a basic model for the favorable prognosis. Composite model was established by factors of basic model and IVT, ASITN/SIR, recanalization, and SICH. Besides, the predictive value of adding preprocedural or post-procedural ALFT into the basic model was estimated by the area under curve (AUC). The improvements on the basic model were assessed by the continuous net reclassification index (NRI) and integrated discrimination improvement (IDI) of which values above 0 were regarded as significant (22). Multiple_max was categorized based on quartiles to evaluate the association between Multiple_max and prognosis. The pattern of the potential dose–response relationship between Multiple_max and prognosis was revealed using a restricted cubic spline with 3 knots after adjusting for confounders. Odds ratio (OR) was reported with a 95% CI.

Subgroup analyses were performed to detect the heterogeneity in the effect of post-procedural ALFT on prognosis. Sensitivity analyses were conducted to explore the effect of ALFTs at different periods on prognosis. A two-tailed p ≤ 0.05 was considered statistically significant. Statistical analyses were performed with the SPSS software package, version 26 (IBM, Armonk, NY) and R statistical software 3.6.3 [R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/].

The flowchart of this study is illustrated in Supplementary Figure 1. The demographic characteristics of all participants are displayed in Table 1. Of the 420 patients enrolled in this study, men accounted for 60.7% and the median age was 68 years. In total, 234 (55.7%) subjects presented as post-procedural ALFT, and 186 (44.3%) had normal liver function after MT.

Patients were divided into two groups (the ALFT group and the non-ALFT group) according to the post-procedural ALFT. Medical histories were balanced between the two groups. The ALFT group had a significantly larger proportion of preprocedural ALFT (40.2 vs. 16.1%, p < 0.001). Higher NIHSS score on admission (median, 18 vs. 15, p < 0.001), higher white blood cell count (p = 0.019), higher glucose level (p = 0.035), more pneumonia (65.4 vs. 38.2%, p < 0.001), and more SICH (12.0 vs. 3.8%, p < 0.001) were detected in the ALFT group than the non-ALFT group. The ratio of IVT pretreatment (34.2 vs. 28.5%, p = 0.213) was similar between patients with normal and abnormal liver function. The comparisons between the two groups on median NIHSS scores at 24 h (19 in the ALFT group vs. 11 in the non-ALFT group, p < 0.001) and 7 days (14 in the ALFT group vs. 6 in the non-ALFT group, p < 0.001) showed similar trends with that of the admission NIHSS score (Table 1).

The mortality rate at 7 days was comparable between the two groups (p = 0.190). More deaths occurred in the ALFT group at 90 days (29.1 vs. 15.6%, p = 0.001, Table 1).

In total, 168 (40.0%) patients ranked 0–2 on mRS score at 90 days after MT, and the proportion of excellent prognosis was 26.9%. Compared with the non-ALFT subjects, the ALFT group had a lower percentage of favorable prognosis (26.5 vs. 57.0%, p < 0.001) and excellent prognosis (15.4 vs. 41.4%, p < 0.001) at 90 days (Figure 1). The density plot of log-transformed Multiple_max distribution for dichotomous prognosis is shown in Supplementary Figure 2. Smaller Multiple_max was detected in patients achieving better prognosis.

The basic model for predicting favorable prognosis was developed after adjusting for confounding factors and variables with a statistical trend (p ≤ 0.1) in the univariate analysis. Finally, age [OR (95% CI): 0.96 (0.94–0.98), p < 0.001], the NIHSS score at 24 h [OR (95% CI): 0.85 (0.82–0.88), p < 0.001], pneumonia [OR (95% CI): 0.43 (0.25–0.72), p = 0.002], and glucose level [OR (95% CI): 0.81 (0.71–0.92), p = 0.001] on admission were included in the basic model (Table 2).

An improved model was established after adding ALFT into the basic model. As shown in the improved model, post-procedural ALFT had a significant correlation [OR (95% CI): 0.48 (0.28–0.81), p = 0.006] with mRS 0–2 at 90 days after adjusting for the factors involved in the basic model (Table 2). Compared with the basic model, the predictive value of the improved model on favorable prognosis showed a significant improvement when assessed by AUC (basic model: 0.875; improved model: 0.881, p = 0.007), continuous NRI (value: 0.401, p < 0.001), and IDI (value: 0.013, p < 0.001), as shown in Table 3. In addition, preprocedural ALFT was not a statistically significant predictor of clinical prognosis (all p-values > 0.05).

The role of Multiple_max in predicting favorable prognosis was also assessed in multivariable analysis. After adjusting for age, pneumonia, glucose level, and median NIHSS score at 24 h, Multiple_max remained an independent predictive factor of favorable prognosis (p = 0.019, Table 4). To explore potential dose–response association, the restricted cubic spline was drawn with three knots after adjusting for covariates mentioned previously. Interestingly, increasing Multiple_max was no longer lower odds of favorable prognosis (Figure 2) further when Multiple_max was more than 2.

A dominant tendency toward a lower proportion of favorable prognosis in the ALFT group was elucidated in the subgroup analysis (all p-values < 0.01, except for subjects who had NIHSS score ≥15) (Figure 3). Of note, no significant interaction was detected between post-procedural ALFT and these clinical features on the favorable prognosis.

In sensitivity analyses, patients with post-procedural ALFTs alone had a lower proportion of favorable prognosis [OR (95% CI): 0.23 (0.14–0.38), p < 0.001] than patients with preprocedural ALFTs alone [OR (95% CI): 1.94 (0.84–4.52), p = 0.120] (Figure 4). They even had less favorable prognoses than patients with both pre- and post-procedural ALFTs [OR (95% CI): 0.23 (0.14–0.38) vs. 0.43 (0.25–0.73)].