This retrospective, single-center study included all consecutive patients who were admitted for ischemic stroke due to acute BAO at our comprehensive stroke center between March 2008 and June 2021 (n = 199). During the recruitment period, approximately 2800 patients received endovascular treatment for any type of ischemic stroke at our center. Patients with angiographically verified thromboembolic material in the BA who were referred for endovascular treatment were included in this study. Part of the patient cohort has been previously described [6, 8, 9]. For 144 out of the 199 patients, post-stroke magnetic resonance imaging (MRI) was available for ischemia grading. For all analyses not requiring MRI (grading of occlusion location and clinical and baseline clinical and outcome variables), we included all consecutive 199 patients to increase the generalizability of our findings. See Fig. S1, a flowchart of subject inclusion.

Approval was obtained by the local ethics board in accordance with regional law under reference number 274/21 S-SR. Patient informed consent was waived by the ethics committee due to the retrospective nature of the study.

Clinical and imaging data were acquired retrospectively. Clinical, demographic, outcome, and procedural data of patients were gathered. National Institutes of Health Stroke Scale (NIHSS)-certified neurologists assessed the NIHSS score at the time of admission and discharge as part of the clinical routine. Substantial neurological improvement was defined as the difference between admission and discharge NIHSS score of ≤ 8 or discharge NIHSS score of ≤ 1 [6, 8, 10, 11]. NHISS improvement was calculated by subtracting the NHISS score at discharge from the score at admission. The modified Rankin Scale (mRS) score was used to measure disability at admission, at discharge (referred to as post-treatment in the text below), and after 3 months. An early good clinical outcome (EGCO) was defined as mRS score ≤ 3 at discharge. Due to missing follow-up patient data at 3 months (valid cases n = 83), discharge mRS was used as the primary endpoint.

The stroke etiology was classified according to the Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria [12]. Reperfusion success of endovascular therapy was quantified based on the modified Thrombolysis in Cerebral Infarction (mTICI) scale [13] by two independent neuroradiologists.

All image-based grading was performed with the consensus of two experienced neuroradiologists.

After dividing the BA into thirds using digital subtraction angiography (DSA), the location of contrast agent discontinuation was used for grading: (1) proximal third beginning at the confluence, and including occlusions in the V4 segment with contralateral vertebral artery aplasia as previously described [6, 8]; (2) middle third, usually distal to the anterior inferior cerebellar artery; (3) distal third, the “basilar head”, see Fig. S2. Patients with V4 occlusions but without thrombotic material in the BA and without contralateral vertebral artery aplasia were not included in this study. Patients with proximal BAO may have had simultaneous V4 occlusions and patients with middle and distal BAO may have simultaneous posterior cerebral artery (PCA) occlusions.

BAOS was classified on peri-interventional DSA by detecting an underlying BS with vessel wall abnormalities and a trend to repeated thrombosis, excluding other reasons such as dissection. In three cases, classification was not possible due to failure of vessel reperfusion.

Infarct location was graded semiquantitatively on diffusion-weighted imaging (DWI). Infarct grading was performed for the following regions of interest (ROI), exactly matching the anatomical structure: the medulla oblongata, the inferior pons, the superior pons, the mesencephalon, the diencephalon, the cerebellum, and the perfusion territory of the PCA in the telencephalon, especially the occipital lobes. The inferior brain stem was defined as medulla oblongata and pons. The superior brain stem refers to the mesencephalon.

For each ROI, the following scores were distributed visually as described below based on the DWI-positive infarct volume as a percentage of the total ROI volume. The scoring was performed as a consensus reading of two expert raters (MRHP, 3 years of experience, and MB, 6 years of experience). An exemplary scoring in the superior pons can be found in Fig. S3.

0: 0% DWI volume, no infarct

1: 1–20% DWI volume, punctiform infarcts

2: 21–60% DWI volume, larger infarcts

3: 61–100% DWI volume, subtotal to total infarction

Variables with metric and ordinal data were described as the median and interquartile range (IQR). Categorical variables were described using absolute and relative frequencies.

Metric and ordinal data were compared using the Mann–Whitney-U-test when comparing two groups. Metric and ordinal data were compared using the Kruskal–Wallis test when comparing three or more groups. Pearson’s Chi-Square test was used to compare categorical variables. If categorical data was distributed on a 2 × 2 crosstab, Fisher’s exact test was used. p-values ≤ 0.05 were considered statistically significant.

To measure the impact of infarct extension in the ROIs on EGCO, a multivariate logistic regression model was calculated including the semiquantitative ischemia scores of all ROIs. To test the validity of this model to predict good clinical outcomes, ROC analyses were performed.

Univariate associations between baseline variables and EGCO were calculated. A multivariate logistic regression model using a stepwise forward variable selection method was calculated to predict EGCO of all variables that were significantly associated with ECOG in univariate analysis. ROC analyses were performed for this regression model.

Mediation models were calculated using the commonly available PROCESS extension for SPSS using a nonparametric bootstrap approach with 5000 replication samples to obtain a 95% CI [14, 15].