Eligible patients were prospectively and consecutively enrolled from October 2022 to October 2024 in a single-center. Inclusion criteria were as follows: (1) age ≥ 18 years and sex not restricted; (2) disabling symptoms or National Institutes of Health Stroke Scale (NIHSS) ≥ 6 due to acute anterior circulation large vessel occlusion (defined as intracranial internal carotid artery, M1 or M2 segment of middle cerebral artery, A1 segment of anterior cerebral artery, or tandem occlusion) and (3) non-contrast CT or MR diffusion weighted imaging (DWI) ASPECTS ≥ 4, or infarct volume < 100 ml defined by CT perfusion Tmax > 10 s; (4) Initiation of endovascular treatment within 24 h from the last-known-well time; (5) Mechanical thrombectomy was applied as first recanalization method while following intracranial balloon catheter dilation and stenting or thrombolysis as rescuing treatment were acceptable. Exclusion criteria included: (1) evidence of cerebral hemorrhage before intervention; (2) non-disabling symptoms or NIHSS ≤ 5 or neurological deficits due to posterior circulation territory infarction; (3) non-contrast CT or DWI ASPECTS ≤ 3, or infarct volume ≥ 100 ml defined by CT perfusion Tmax > 10 s or early signs of malignant edema or mid-line shift; (4) missing important clinical data, such as follow-up mRS information or critical baseline data. According to the timing of admission, the patients were divided into training and validation cohorts in a 7:3 ratio to construct and verify a prognostic nomogram.

Endovascular intervention was performed under general anesthesia or sedation. Stent retrievers (Solitaire FR, ev3, Irvine, USA; Trevo ProVue, Stryker Neurovascular, Salt Lake City, USA), aspiration catheters (Penumbra aspiration system, Penumbra Inc., Alameda, USA; React 68/71 catheter, ev3, Irvine, USA), or a combination of both were all acceptable to remove clots. For patients with stenosis or arterial occlusion owing to atherosclerosis, angioplasty with a Gateway (Boston Scientific, Fremont, USA) or Neuro LPS (SinoMed, Tianjin, CHN) intracranial balloon catheter with or without a Wingspan stent (Stryker Neurovascular, Salt Lake City, USA) implantation was permitted as part of the intervention.

Patients presenting within 4.5 h of symptom onset and without any contraindication received intravenous thrombolysis treatment comprising 0.9 mg/kg recombinant tissue plasminogen activator (rt-PA; Boehringer Ingelheim, Biberach, Germany) after providing informed consent, or were directly moved to the catheter room for endovascular evaluation. Heparinisation by intravenous infusion of unfractionated heparin at 50–100 IU/kg was performed at the start of the procedure, with an additional 1,000 IU administered every hour during the intervention, unless the patient had prior use of intravenous thrombolytic agents or known severe thrombocytopenia and coagulation disorder. Antiplatelet agents, such as tirofiban, were used as necessary. The patients were discharged to their homes or rehabilitation facilities. Postoperative management for all patients included control of underlying risk factors, such as hypertension and diabetes. Additionally, in the absence of any contraindications, all patients were prescribed dual-antiplatelet or anti-coagulation therapy and lipid-lowering drugs, which were maintained for 90 days according to the standard protocol. Ninety-day clinical outcomes were reviewed and collected through face-to-face or telephone interviews by a trained neurologist who were blinded to the quantifying conditions of the patients. To assess potential selection bias, we performed a sensitivity analysis comparing the baseline characteristics of patients who completed follow-up with those lost to follow-up.

Demographic data and traditional cerebrovascular risk factors were recorded. Stroke severity was scored using NIHSS, while the ASPECTS was used to assess the infarct core. Arterial occlusion location and American Society of Interventional and Therapeutic Neuroradiology/Society of Interventional Radiology (ASITN/SIR) collateral scores were evaluated using digital subtraction angiography (DSA). An eTICI score of 2b−3 was defined as reperfusion success. For the purpose of multivariable modeling, the degree of reperfusion was categorized into eTICI 0–2a, eTICI 2b, and eTICI 2c−3. The mRS score 90 days after endovascular intervention was used to quantify long-term outcomes. The laboratory tests for baseline blood glucose level, blood cell count, C-reactive protein level, and neutrophil-to-lymphocyte ratio were run right before MT. While fasting total cholesterol (TC), triglyceride (TG), and low-density lipoprotein cholesterol (LDL-C) levels were tested in the next morning. Intracranial hemorrhage (ICH) was evaluated within 72 h after MT, symptomatic ICH was defined as new intracranial hemorrhage confirmed by computed tomography or magnetic resonance imaging with an increase in NIHSS score of ≥ 4 points or ≥ 2 points in one category. The nutritional status was assessed by the controlling nutritional status (CONUT) (17) score that use the TC level, lymphocyte count, and serum albumin level according to the following formula: albumin (g/dl): ≥3.5 (0); 3–3.4 (2); 2.5–2.9 (4); < 2.5 (6); total cholesterol (mg/dl): ≥180 (0), 140–179 (1), 100–139 (2), < 100 (3); Lymphocyte count (109/l): ≥1.6 (0), 1.2–1.59 (1), 0.8–1.19 (2), < 0.8 (3). Nutritional status was classified as none (0–1), mild (2–4), moderate (5–8), or severe (9–12) (17). The prognostic nutritional index (PNI) was used as another nutritional scoring system, (18) calculated using the following formula: 10 × serum albumin (g/dl) + 0.005 × total lymphocyte count (/mm³). Malnutrition severity was defined as none (>38), moderate (35–38), or severe (< 35) (18).

The primary outcome measures were the rate of mRS score of 0–3 during the 90-day follow-up period meaning the capability to ambulate, and the performance of the prognostic nomogram. Secondary outcome measures included recanalization status, symptomatic hemorrhage and all-cause death.

Continuous variables were tested for normality distribution using the Kolmogorov–Smirnov test. Normally distributed variables were presented as the mean ± standard deviation, and means were compared using Student's t-test. Non-normally distributed variables were presented as median (interquartile range) or counts and percentages, and were compared using the U test. The χ² or Fisher exact tests were used to compare categorical variables. Two-tailed p values of < 0.05 were considered statistically significant. Individualized prognostic models for adverse outcomes were constructed using a multivariate logistic regression analysis. Backward elimination based on the Akaike Information Criterion was applied for variable selection. Only participants with complete data on all variables were included in the final multivariable model analysis. A nomogram was subsequently developed and internal validation was conducted by bootstrapping with 1,000 bootstrap samples. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUROC), and goodness of fit was assessed using the Hosmer–Lemeshow test. Model consistency and accuracy were further examined using calibration curves. Decision curve analysis (DCA) was applied to quantify the net benefit and determine the clinical utility of the nomogram. All tests were two-sided, and statistical significance was defined as p < 0.05. All statistical analyses were conducted using the SPSS software version 26 (IBM, New York, NY, USA) and R software (version 4.4.2).

The 90 day mRS assessment identified 374 (65.7%) patients with mRS 0–3 and 195 (34.3%) patients with mRS 4–6; Univariate analysis showed multiple discrepant characteristics: patients in the mRS 0–3 group had a higher proportion of men, with younger age and lower prevalence of traditional vascular risk factors. Higher ASPECT score and lower NIHSS before and after MT were seen in mRS 0–3 group. Local anesthesia or intravenous sedation, better DSA ASITN/SIR collateral, and shorter puncture-to-recanalization time were related to 90 day mRS 0–3, whereas patients with mRS 4–6 showed higher rate of postoperative hemorrhage, eTICI 0–2a and distal embolizaiton (Table 1).

Significant differences were found in inflammation and nutritional indexes between 90 day mRS 0–3 and 4–6 groups (Table 1); Patients with 90 day mRS 4–6 showed higher white blood cell counts, neutrophil counts, neutrophil/lymphocyte ratios, and C-reactive protein levels, while lymphocyte count and albumin were lower (p < 0.05). Patients with a CONUT evaluation of malnutrition as “none” had a higher probability of achieving an mRS score of 0–3 compared with patients with malnutrition levels of “mild”, “moderate”, or “severe”, which is consistent with the results found using the PNI scoring system. These differences indicate that patients with severe inflammation and malnutrition stress are more likely to be disabled after MT.

Multivariate logistic regression analysis identified female sex, baseline NIHSS score, C-reactive protein, puncture to recanalization time, history of cerebrovascular disease, ICH after MT and insufficient re-perfusion (eTICI 0–2a) as adverse factors for mRS 0–3 (OR < 1, p < 0.05); Compared with “moderate to severe” malnutrition patients based on the CONUT criteria, patients with “none” or “mild” malnutrition had a higher chance of achieving mRS 0–3 (none: OR: 4.813 (2.119–10.935), p < 0.001, mild: OR: 2.651 (1.340–5.242), p = 0.002) (Table 2). Table 3 showed efficacy of prognostic models based on multivariate logistic regression analysis. Supplementary Figure S1 showed that CRP and malnutrition status contributes to better performance of the model: AUC: 0.845 (0.812–0.877) vs. 0.808 (0.773, 0.845), p = 0.0002.

The overall population was divided into the training (n = 398) and validation (n = 171) cohorts in a 7:3 ratio according to the chronological order of enrollment. Characteristics of the two cohorts were shown in Table 4. The two cohorts were well-balanced, with no statistically significant differences observed in demographic features, stroke severity, laboratory parameters (including inflammation and nutrition markers), comorbidities, or treatment modalities (all p > 0.05).

Significant variables identified using multivariate logistic regression were applied in the training cohort to construct a nomogram (Figure 2) which was then tested for effectiveness in the validation cohort.

The AUROC values of this nomogram in the training and validation cohorts were 0.865 (95% CI: 0.826–0.905) (Figure 3A) and 0.861 (95% CI: 0.800–0.922) (Figure 3D), respectively, indicating that the model had good discrimination. The calibration curve indicated that the model had reliable predictive accuracy. The p values for Spiegelhalter's Z-test were 0.745 and 0.303, respectively, indicating no significant departure from a perfect fit. The Brier values were 0.142 and 0.154, indicating that the calibration curves were close to the ideal curve (Figures 3B, E). Also, the DCA curve indicated a relatively favorable net clinical benefit (Figures 3C, F).