Anonymized data not published within this article will be made available by request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

This retrospective study utilized data from the Specialized Multi-Center Attributed Registry of Stroke-CLOT (SMART-CLOT), a prospective nationwide multicenter registry that includes patients with acute consecutive ischemic stroke from 13 centers in South Korea who underwent EVT due to large vessel occlusion [8]. Each participating center maintains a site-specific de-identified unique patient identifier, which enabled within-center linkage of encounters from the same individual. When more than one EVT with thrombus retrieval occurred for a patient at the same center during the study period, only the first (index) EVT thrombus and its corresponding clinical data were retained; subsequent EVTs were not analyzed as separate index entries. To assess the possibility of cross-center re-registration, we reviewed baseline histories of previous stroke and queried the registry for evidence of prior EVT with thrombus retrieval. Among 34 patients with a previous stroke history, none had a prior EVT, confirming that all analyzed thrombi correspond to index EVTs of unique patients. Reperfusion therapy was performed using protocols based on the guidelines. In addition, during hospitalization at the study hospitals, all patients underwent a comprehensive assessment that included medical history, blood tests, computed tomography, magnetic resonance imaging, carotid ultrasonography, transcranial Doppler, 12-lead electrocardiography (EKG), echocardiography, Holter monitoring, and continuous EKG monitoring.

Written informed consent was obtained from prospectively enrolled patients or their legal representatives. The Institutional Review Board approved the registry of each participating hospital. For this study, we included patients from 5 of the 13 participating centers (Seoul Medical Center, Yongin Severance Hospital, Ewha Woman’s University Seoul Hospital, National Health Insurance Service Ilsan Hospital, and Severance Hospital), as these centers currently have data on long-term outcomes. This study was approved by the Institutional Review Board of Yonsei University College of Medicine (approval number: 4-2024-1662).

Thrombi retrieved from EVT were promptly placed in tubes containing 4% paraformaldehyde and fixed overnight at 4°C. Subsequently, all thrombi were transported immediately to the central laboratory at Yonsei University College of Medicine. They were processed into paraffin blocks and stored until analysis. Using these paraffinized blocks, the thrombi composition was subsequently analyzed using immunohistochemistry.

The paraffin-embedded thrombi were sectioned into 3-μmthick slices. Following deparaffinization and blocking with 1% horse serum and 5% non-fat milk in Tris-buffered saline, the sections were incubated overnight at 4°C with primary antibodies. For immunohistochemistry, the following primary antibodies were used: monoclonal anti-CD42b (ab134087, 1:100; Abcam, Cambridge, UK) for platelets, rabbit polyclonal anti-fibrinogen (ab34269, 1:200; Abcam) for fibrin/fibrinogen, and rabbit monoclonal anti-glycophorin A (ab129024, 1:400; Abcam) for red blood cells (RBCs). Antigen retrieval was performed using the IHC-Tek epitope retrieval solution and a steamer (IHC World, Woodstock, MD, USA), except for anti-CD42b. Secondary antibody binding was performed using the avidin-biotin-horseradish peroxidase complex (Vector Laboratories, Peterborough, UK), followed by color development with 3,3’-diaminobenzidine. Hematoxylin counterstaining was conducted, and the slides were mounted with Permount Mounting Medium (Fisher Scientific, Fair Lawn, NJ, USA).

In this study, thrombi from individual patients were analyzed based on various characteristics, including the proportion of primary thrombus components such as RBCs, fibrin, and platelets. Whole-slide images of stained thrombi were acquired using a whole-slide scanner (MoticEasyScan Pro 6; Motic Hong Kong Limited, Hong Kong, China) at 40× (0.26 μm/pixel). The scanned images were analyzed using the Automated Region-of-Interest-Based Image Analysis, open-source software for automated composition analysis [9]. A fixed threshold was applied, and the proportions of RBCs, platelets, and fibrin/fibrinogen were calculated as the percentage of pixel density relative to the total thrombus area.

The thrombus distribution patterns were subsequently analyzed. The extracted thrombi were categorized into four groups based on their overall distribution patterns of common components of thrombus, especially platelets and RBCs, as described in previous reports (Figure 1) [10,11]. The platelet distribution pattern was the most distinctive. First, a layered pattern was defined when platelets were linearly deposited in layers or clustered in spots within the thrombus [10,11]. Second, some thrombi exhibited an erythrocytic pattern, where platelets were predominantly located at the periphery, with densely packed RBCs in the core [10,11]. Third, a diffuse platelet pattern was defined as a uniform distribution of platelets throughout the thrombus without any specific organization. A single dominant pattern was assigned when that pattern occupied ≥70% of the thrombus in the reviewed section. Specimens that did not meet this rule were classified as mixed pattern. Two investigators (H.L. and Y.D.K.) independently assessed the thrombus pattern without any clinical information and achieved good agreement (κ_un-weighted, 0.88; 95% confidence interval [CI], 0.84-0.91). Any disagreements were resolved by consensus.

We analyzed data on demographics, vascular risk factors, comorbidities, and laboratory results. Active cancer was defined as cancer diagnosis, recurrence, ongoing treatment within the past 6 months, inoperable cancer, or metastasis. Premorbid disability was defined as a modified Rankin Scale score of 2-5 [12]. Stroke subtypes were classified according to the criteria of the Trial of Org 10172 in Acute Stroke Treatment [13]. Stroke severity was assessed using the National Institutes of Health Stroke Scale (NIHSS). Immediate modified Thrombolysis in Cerebral Infarction (mTICI) was defined as the angiographic reperfusion grade recorded on the final digital subtraction angiography run at the end of the EVT. Successful recanalization was defined as achieving the mTICI score of 2b or 3.

The primary outcome measured in this study was major adverse cardiovascular events (MACE), defined as a composite of allcause mortality, ischemic heart disease (IHD), and stroke (ischemic or hemorrhagic). Each event within the composite outcome was separately analyzed to compare individual incidences. The IHD events included acute myocardial infarction, unstable angina, percutaneous transluminal coronary angioplasty/stent placement, and coronary artery bypass grafting. Data on vascular events and death were collected through face-to-face or telephone interviews conducted by stroke specialists or stroke research nurses at 3, 6, 9, and 12 months after the index event. Medical chart reviews were conducted when necessary to supplement the interviews and, beyond 12 months, served as the primary source of longitudinal follow-up to capture MACE and its components. The censoring date was October 31, 2024. Patients with at least one post-discharge follow-up were included and, if event-free, censored at their last contact; individuals with no post-discharge information contributed no at-risk time and were excluded. Because deaths within 4 weeks of the index EVT are predominantly attributable to the acute index stroke and periprocedural course rather than longer-term thrombus biology, patients who died before day 28 were excluded by design [14,15].

Data were presented as means±standard deviations, medians (interquartile ranges), or numbers (%), as appropriate. Continuous variables were compared between groups using the t-test, Kruskal-Wallis test, and Mann-Whitney U test. Associations between categorical variables were analyzed using the chi-squared or Fisher’s exact test, as appropriate. MACE incidence rates were calculated as events per 100 person-years. Kaplan-Meier survival curves were used to estimate cumulative incidence, and the log-rank test was used to evaluate differences in thrombus patterns. Cox proportional hazards regression models assessed the association between thrombus characteristics and the risk of clinical events, including MACE. Hazard ratios (HRs) with 95% CI were reported.

Variables for the multivariable analyses were selected from significant univariable analyses (P<0.05). We assessed linearity of the log-hazard for each continuous covariate using Martingale residuals and restricted cubic splines (four knots). Based on Wald tests, age was modeled with restricted cubic splines (nonlinearity P=0.010); C-reactive protein (CRP)/D-dimer/glucose were natural log-transformed and entered linearly, and NIHSS and onset-to-recanalization were retained as linear terms. For continuous variables entered linearly, HRs represent the relative hazard associated with a one-unit increase in the covariate. For biomarkers that were natural log-transformed, HRs correspond to a one-unit increase in the natural log-transformed value. Because age was modeled using restricted cubic splines, HRs for age are presented as contrasts from the spline model, as noted in the table footnotes. For low-event endpoints (stroke and IHD), we conducted sensitivity analyses using Firth penalized Cox regression. To preserve degrees of freedom, adjustment was parsimonious: the stroke model included thrombus pattern, age (per 10 yr, linear), sex, active cancer, and initial NIHSS; the IHD model included thrombus pattern, age (per 10 yr, linear), and sex. The predictive performance of the multivariable model was assessed at 1, 3, and 5 years using the receiver operating characteristic (ROC) curve, and the area under the curve (AUC) was calculated to quantify the model’s discriminative ability. Subgroup analyses were conducted to examine MACE risk based on age (≥65 vs. <65 yr), atrial fibrillation, cancer status, and discharge medications. Interaction terms were included to evaluate effect modification. A two-sided P<0.05 was considered statistically significant for all tests. All statistical analyses were conducted using the R software (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria).