This retrospective analysis was based on data from a prospective nationwide multi-center registry, the Specialized Multi-center Attributed Registry of Stroke-CLOT (SMART-CLOT). The SMART-CLOT is a prospective registry that includes data of patients from 13 centers in South Korea who experienced acute consecutive ischemic strokes and underwent EVT for LVO. During hospitalization at each study hospital, all patients underwent a comprehensive assessment that included a review of their medical history, blood tests, CT, magnetic resonance imaging, carotid ultrasound, transcranial Doppler, 12-lead electrocardiography, echocardiography, and Holter monitoring or continuous electrocardiogram monitoring. Stroke severity was also evaluated using the National Institutes of Health Stroke Scale (NIHSS). Reperfusion therapy was performed according to guideline-based protocols and the attending physician’s discretion, depending on the patient’s clinical status. Written informed consent was obtained from prospectively enrolled patients or their next of kin. The institutional review board approved the registry of each participating hospital.

This study included consecutive patients who underwent EVT and whose thrombi were obtained between July 2017 and July 2023. In addition, to reliably measure the volume of the culprit thrombus, only 235 patients who underwent thin-section (1 or 1.25 mm) NCCT using the same protocols before EVT at the Severance Stroke Center were included. This study was approved by the Institutional Review Board of the Yonsei University College of Medicine (approval number: 2023-2440-001).

For patients with intracranial LVO, thrombus volume and density (Hounsfield units) were measured using baseline thin-section NCCT with a semi-automatic three-dimensional software (Xelis; Infinitt, Seoul, Korea), as described previously (14). Pixel segmentation suggested that the thrombus area should be blue at the threshold between 50 and 100 Hounsfield units. The region of interest for the thrombus was determined by clicking on any portion of the target area. Thereafter, automatic pixel dilation and growth of a region to the margin at a threshold between 40 and 100 Hounsfield units were performed by simply clicking the “dilate” icon. Subsequently, the thrombus volume was automatically measured and presented on a screen. Two stroke neurologists (Y.D.K. and J.Y.) who were not aware of any clinical information measured the thrombus volume. Given the excellent inter-rater agreement (intraclass correlation coefficient [ICC], 0.982; 95% confidence interval [CI], 0.970–0.989; p < 0.001), the measurements obtained by J.Y. were used for the analysis.

Thrombi retrieved from the EVT were promptly preserved in 4% paraformaldehyde, embedded in paraffin blocks, and stored. The fixed specimens were sectioned into 3 μm-thick slices, deparaffinized using xylene, and rehydrated using graded ethanol. 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 RBCs. Antigen retrieval was performed using the IHC-Tek epitope retrieval solution and a steamer (IHC World, Woodstock, MD, USA) for all antibodies except for anti-CD42b.

The 3 μm-thick tissue sections were incubated overnight at 4°C. Secondary antibody reactions were performed using an avidin/biotin/horseradish peroxidase complex (Vector Laboratories Ltd., Peterborough, Cambridgeshire, UK). Positive signals were visualized using 3,3′-diaminobenzidine.

After counterstaining with hematoxylin, the slides were sealed using Permount mounting medium (Fisher Scientific, Fair Lawn, NJ, USA). One slide was prepared per stain for each patient. Images of stained thrombi were obtained using a digital scanner (Aperio AT2; Leica Biosystems, Wetzlar, Germany). The scanned images were analyzed using Automated Region-of-interest based Image Analysis (ARIA) for automated composition analysis (15). The fractions of each component, including RBCs, platelets, and fibrin, were calculated as the percentage pixel density of the total thrombus area.

The extracted thrombi were characterized into four groups based on their overall distribution patterns of common components (especially platelets and RBCs) based on previous reports (Figure 1) (9, 16). The distribution pattern of the platelets and RBCs was distinctive: first, the platelets and RBCs were deposited alternatively deposited in a linear, layered pattern or clustered in spot-like formations within the thrombus (9, 16). In some thrombi, the platelets were located predominantly at the periphery, filled with densely packed RBCs, which represented the erythrocytic pattern. They were similar the erythrocytic-type thrombi (red clot) reported in previous studies (9, 16). Specimens showing both peripherally located platelets and with packed RBCs, along with and internally alternating deposited platelets and RBC layers, were categorized under mixed pattern. Finally, thrombi, which is filled with platelets without any specific pattern were classified under diffuse platelet pattern.

Stroke neurologists and research nurses regularly contacted the patients or their caregivers during follow-up sessions via regular face-to-face visits or telephone interviews, with or without a medical chart review, to investigate the clinical outcomes, including the modified Rankin score (mRS). The radiological outcomes, including the device pass number, modified Thrombolysis in Cerebral Infarction (TICI) grade, symptomatic intracranial hemorrhage (ICH), and procedure time, were also assessed. The first-pass effect was defined as near-complete or complete recanalization (TICI 2c or TICI 3) after the first pass of the device. The procedure time was defined as the interval (min) from the femoral puncture to the first achievement of a TICI of 2b or 3. In addition, we determined the number of fragmented thrombi during the procedure, which was defined as the number of newly apparent occlusions in the downstream vessel after recanalization of the culprit lesion.

Data are presented as the mean ± standard deviation (SD), median (interquartile range), or percentage (%), as appropriate. Continuous variables were compared using Student’s t-tests or analysis of variance, while categorical variables were compared using the Chi-square or Fisher’s exact tests, as appropriate. Correlations between thrombus volume and proportion of thrombus components were investigated using Pearson’s correlation analysis. Thrombus volume tertiles were used when comparing the distribution patterns according to thrombus volume. To identify independent factors associated with thrombus characteristics or outcomes, multiple linear or ordinal regression analyses were performed after adjusting for age, sex, and other variables with a p-value of < 0.1 in the univariable analyses. p < 0.05 was considered statistically significant. Statistical analyses were conducted using the R software package version 4.3.1¹ or SPSS for Windows (version 27; SPSS, Chicago, IL, USA).

Among 235 patients, we excluded the thrombi of 8 patients that were not detected on thin-section NCCT and those of 17 patients with no available immunohistochemical data. Finally, 210 patients were included in the study (Supplementary Figure 1).

The mean age of the 210 patients (± SD) was 73.0 ± 12.9 years, and 103 (49.0%) patients were men. The mean interval from stroke onset to baseline thin-section NCCT was 381.7 ± 480.2 min. Among the stroke mechanisms, cardioembolism (60.0%) was more common than large artery atherothrombosis (11.0%). Other baseline characteristics, including risk factors, medication history, laboratory variables, and treatment outcomes, are shown in Table 1.

The most common occlusion site was the middle cerebral artery (n = 151 [71.9%]), followed by the distal internal carotid artery (ICA; terminal occlusion, n = 19 [9%]; cavernous occlusion, n = 11 [5.2%]), basilar artery (n = 27 [12.9%]), and posterior cerebral artery (n = 2 [1.0%]). The median (IQR) volume of the thrombus was 43.7 (23.5, 74.5) mm³ (Table 1). The median (IQR) thrombus volume based on the occlusion site was middle cerebral artery, 36.0 (22.0, 66.4) mm³; ICA-terminus occlusion, 77.6 (51.1, 118.4) mm³; ICA-cavernous occlusion, 66.9 (29.6, 122.8) mm³; basilar artery, 62.9 (43.6, 91.0) mm³; and posterior cerebral artery, 17.2 (14.6, 19.7) mm³ (Table 1).

The mean proportion (±SD) of the common components of the thrombus was 34.1 ± 15.5% for RBCs, 13.7 ± 15.5% for platelets, and 36.9 ± 17.9% for fibrinogen. The proportion of RBCs (anti-glycophorin A) increased (r = 0.359, p < 0.001), whereas that of platelets (anti-CD42b) decreased (r = −0.194, p = 0.005) with an increase in thrombus volume. However, fibrin (anti-fibrinogen) levels did not correlate with thrombus volume (r = 0.65, p = 0.347) (Figure 2). A positive association between the proportion of RBC and thrombus volume was observed, regardless of the stroke mechanism, when we determined the association between thrombus volume and composition across stroke mechanisms (Supplementary Figure 2).

The mixed pattern was the most common (n = 81 [38.6%]), followed by layered (n = 57 [27.1%]), erythrocytic (n = 47 [22.4%]), and diffuse platelet patterns (n = 25 [11.9%]). The mixed pattern was likely to survive the longest from stroke onset to CT (427.3 ± 503.8 min) and had the highest median NIHSS (15.0, interquartile range 9–20). The layered pattern was associated with coronary artery disease, atrial fibrillation, and cardioembolism, whereas the erythrocytic pattern was related to large artery atherothrombosis. More than half of the patients with diffuse platelet patterns had active cancer (n = 16 [64%]). The proportion of thrombus components differed according to the distribution patterns: the proportion of platelets was the highest in the diffuse platelet pattern and that of RBCs was the highest in the mixed pattern (all p < 0.05). However, the proportion of fibrinogen did not differ according to this pattern (Supplementary Table 1).

We found that the distribution pattern differed according to thrombus volume. The mixed pattern had the largest volume (median [IQR]) of thrombus (71.2 [41.2, 122.4] mm³), followed by layered (36.4 [23.5, 62.9] mm³), erythrocytic (29.3 [19.3, 48.7] mm³), and diffuse platelet (24.1 [20.0, 35.5] mm³) patterns (Supplementary Table 1). As the tertiles of thrombus volume increased, a mixed pattern was commonly observed, whereas an erythrocytic or diffuse platelet pattern was noted infrequently (p < 0.001) (Figure 3). When we divided the study groups based on the median volume (43.7 mm³) of the thrombus (smaller vs. larger thrombus), the mixed pattern was the most common in larger thrombi, regardless of the stroke mechanism (Supplementary Figure 3).

Additionally, thrombus volume was associated with serum hemoglobin level, initial NIHSS score, history of dyslipidemia, previous stroke or transient ischemic attack, and presence of active cancer (all p < 0.05), whereas it was not correlated with the interval from stroke onset to CT, use of tissue plasminogen activator or prior use of antithrombotic agents, and stroke mechanism. Multivariable analysis adjusting for age, sex, and significant variables in the univariable analysis revealed an independent and significant positive association between thrombus volume and the mixed pattern (β 21.04, SE 10.10, p = 0.038) compared to the layered pattern or proportion of RBCs (β 1.00, SE 0.27, p < 0.001) and an inverse association with the erythrocytic pattern (β − 29.78, SE 11.54, p = 0.011) compared to the layered pattern (Table 2).

The number of fragmented thrombi, procedure time, and frequency of symptomatic ICH increased with thrombus volume (all p < 0.005). Poor functional outcome (indicating mRS ≥ 3) or mortality after 3 months from index stroke was not related to thrombus volume (Table 2). Univariable and multivariable ordinal regression analyses showed that the number of fragmented thrombi was independently associated with the proportion of RBCs and thrombus volume. However, the distribution pattern was not associated with the number of fragmented thrombi (Table 3; Supplementary Table 2). Other radiological, clinical outcomes were described at Supplementary Table 3.