Distance to thrombus, ischemic lesion volume and clinical outcome after thrombectomy for M1 middle cerebral artery occlusion

For this study all patients admitted for acute ischemic stroke due to LVO at two comprehensive stroke centers from 1 January 2018 to 31 December 2020 coming within 24 h since last-seen-well and without contraindications to mechanical thrombectomy were eligible. Specifically, we excluded patients with a modified Rankin scale (mRS) greater than 3 prior to presentation. Treatment with intravenous alteplase was allowed. For definitive study inclusion, performance of neurovascular imaging by either CT or MR angiography (CTA or MRA) followed by EVT due to an occlusion of the M1 segment of the MCA (defined as thrombus material situated between the T‑bifurcation of the ICA and the MCA bifurcation) was required. Additionally, individual comorbidities such a history of prior stroke, atrial fibrillation and peripheral arterial occlusive disease in each patient, together with prestroke medication (e.g., intake of anticoagulants) were documented. Devices used for the EVT technique (i.e., thrombus aspiration alone vs. pass with stent retriever alone vs. combination of both), procedural times, complications, and the type of anesthesia was noted. At admission, stroke severity, classified by the National Institutes of Health stroke scale (NIHSS), was evaluated by a certified stroke physician. Due to the study’s retrospective nature, the ethics committee’s approval was waived. The measurements were performed by a neurologist with > 20 years experience in vascular neurology (SP) and by a second experienced neurologist (> 10 years experience) to perform interrater variability.

The CTA or MRA performed in each patient needed to include the depiction of the arterial circle of Willis together with adequate native image resolution for evaluation of the ASPECTS score and leptomeningeal collaterals (LC) [10]. The presence and the quality of the observed LC were then classified as being either “good” or “poor,” as described previously [11]. In this sense, a situation of “good” collaterals was present if the collateral filling was equal as opposed to “poor” collaterals if the filling was less than the other hemisphere or no filling [11]. The DT was then measured manually with the Deep Unity® curve tool (Dedalus HealthCare, Bonn, Germany) (in mm) from the middle of the T‑bifurcation of the internal carotid artery to the cessation of contrast medium as the proximal end of the thrombus as suggested by Friedrich et al. [6] (see Fig. 1). The DT was additionally divided into equally numbered groups with progressive distance and designated “proximal” (0.0–7.2 mm), “middle” (7.3–12.1 mm) and “distal” (12.2–27.8 mm).

The outcome of the EVT was categorized according to the modified thrombolysis in cerebral infarction (mTICI) score ranging from 0 to 3. These were grouped as insufficient reperfusion (0–2a) or sufficient reperfusion (2b, 3). The volume of the ischemic lesion volume (i.e., infarction) in each patient was calculated 24 h after performing EVT and according to the ABC/2 formula [12] and documented (in ml). Virtually all patients had CT scans as control imaging. Edema and/or hemorrhagic transformation were included in the ILV measurement. The signs of evidently old infarctions in the ipsilateral hemisphere were neglected and best efforts were made to exclude possible gliosis associated with old infarcts in the vicinity of a new infarction.

Clinical outcome was measured by a modified Rankin scale at 3 months. The patients were examined at 3 months by a clinician or interviewed via telephone (the caregiver would be questioned when the patient could not provide information). In the event of death, the data were sampled from the relevant municipal registries.

Categorical variables were prepared as absolute numbers and percentages, whereas continuous variables were tested (Kolmogorov-Smirnov test) for normal distribution. Non-normally distributed data were presented as the median and interquartile range (IQR, 25–75 percentile). As data were non-normally distributed, non-parametric tests of associations (Wilcoxon Rank-Sum test, Spearman Rho for correlation between continuous variables) were performed. Exploration analysis was performed with ln-transformed ILV, and ILV was categorized into four categories (0–15 ml, 15.1–70 ml, 70.1–200 ml, and > 200 ml) as the dependent variable [9]. First, we performed univariate non-parametric analysis (Kruskal-Wallis test for binary-continuous, and Fisher’s test with simulated p values, when needed, for categorical variables) for possible associations. The ILV and ln-transformed ILV was also inspected as a continuous predictor. All predictors having p < 0.1 entered the multivariate analysis. In multivariate ordinal regression analysis, where ILV was grouped into the previously described categories, the following adjusting variables were accounted for: (1) continuous predictors: age in years at symptom onset, total thrombectomy steps performed, distance to thrombus in mm, (2) categorical variables: the technique used for thrombus removal (aspiration + stent retriever vs. aspiration only, vs. stent retriever only) as a binary variable state of the leptomeningeal collaterals as seen on first imaging grouped into good/equal vs. bad; ASPECT score greater than 6, TICI outcome 2b–3 vs. 0–2a and intravenous thrombolysis. Missing data were managed with multivariate imputation via chained equation (MICE) package in R. We also performed univariate mediation analysis, adjusted for age (in years), TICI outcome (0-2a vs. 2b–3, as binary variable), ASPECTS score greater than 6 and thrombolysis (binary), where DT was considered a predictor, ln-transformed ILV as a mediator, and mRS at 3 months (0–2 vs. 3–6) as outcome. The interrater reliability was determined through an intraclass correlation calculation on 19 randomly selected patients. The level of statistical significance was set at P < 0.05. For statistical analyses, we used R Software v. 4.2.0 [13].

Overall, 353 patients underwent endovascular thrombectomy in the anterior circulation for LVO at both participating centers, and 346 fulfilled all criteria for inclusion in this study. The median age was 76 years (IQR 64–83 years), and the proportion of women was 58%. The median NIHSS at admission was 16 (IQR 12–19), and the proportion of wake-up stroke was 20%. A known history of previous stroke or TIA was present in 37 patients (11%). Intravenous thrombolysis with recombinant tissue plasminogen activator (rt-PA) (IVT) was administered in 187 (54%) patients. Median time from end of EVT and first control CT imaging was 1 day (IQR 1–1 day). Detailed demographic characteristics are listed in Table 1 and information about missing data is given in supplemental Table 1.

The median DT was 9mm (IQR 6–14mm) and was progressively smaller across larger ILV, except for the largest ILV, where no difference between other ILV groups was found (Fig. 2). A larger DT was associated with smaller ILV (ln-transformed), although the effect was small, r (344) = −0.15, p = 0.005 (Supplemental Fig. 1). Additionally, after splitting DT into three equal groups (range of values in mm: 0–7, > 7–12, > 12–28 for proximal, middle, and distal groups, respectively), the association with continuous ILV was significant at p = 0.003. Similarly, continuous DT was significantly associated with the ILV groups 0–15 ml vs. 70.1–200 ml and 15.1–70 ml vs. 70.1–200 ml, at p = 0.047. The measurements made by the two raters, who both have over 5 years of experience in the neurovascular field, demonstrated a high level of agreement in determining the DT on CT/MR images with an intraclass correlation coefficient of 0.87 (95% CI 0.69–0.95) for DT on CTA/MRA images.

The median ILV was 13.9 ml (IQR 2.2–53.1 ml). Age was significantly, albeit complexly, associated with ILV. The group with the largest ILV was the youngest; generally, the age had a negative correlation with ILV, r (345) = −0.13, p = 0.013. Patients waking up with stroke symptoms had significantly larger ILV (40.8 ml vs. 11.1 ml, p < 0.001). A lower ASPECTS score was predictive of a larger ILV (51.0 ml vs. 11.0 ml, p < 0.001). Good LC was protective for larger ILV (equal group having 5.8 ml vs. absent group having 17.3 ml, p = 0.021). Procedure time was the shortest in the smallest ILV group (35 min vs. 67 min. for 0–15 ml vs. 70.1–200 ml group, respectively, p < 0.001). Both devices used were associated with the largest ILV (13.9 ml vs. 8.7 ml for aspiration + stent retriever vs. aspiration only, p = 0.021). Total EVT passes performed were also positively correlated with ILV, 3 (IQR 2-4) steps in the largest ILV group, p < 0.001. Unsuccessful EVT measured by mTICI was associated with larger ILV (96.2 ml vs. 11.5 ml for 0–2a vs. 2b–3 mTICI outcome, respectively, p < 0.001) (Table 2). With an intraclass correlation coefficient 0.78 (CI 0.54–0.91) for ischemic infarct volume, good agreement in determining the ischemic lesion volume was also shown.

There were 64 (18%) missing mRS data. Therefore, the logistic regression analysis was performed on 282 patients. Good outcome (0–2 at 3 months) was recorded in 139 (49%) patients. Age (71 vs. 79 years, OR 0.96, CI 0.96–0.98, p < 0.001), not taking vitamin K oral anticoagulant (OAC) (1% vs. 7%, OR 0.20, CI 0.03–0.77, p = 0.038), thrombolysis (61% vs. 47%, OR 1.68, CI 1.05–2.70, p = 0.030), ASPECTS > 6 (93% vs. 83%, OR 2.58, CI 1.22–5.87, p = 0.017), less total thrombectomy steps performed (1 vs. 2, OR 0.82, 0.70–0.95, p = 0.009), successful first pass (59% vs. 52%, OR 2.44, CI 1.50–3.98, p < 0.001), 2b and 3 vs. 0–2a TICI outcome (94% vs. 86%, OR 0.34, CI 0.14–0.77, p = 0.013), lower ln-ILV (3.9 vs. 35.9, OR 0.61, CI 0.52–0.71, p < 0.001) were all associated with good outcome (Supplemental Table 2). The above predictors entered multivariate logistic regression analysis.

The mediation model in which DT was considered as a predictor, ILV a mediator, and good clinical outcome at 3 months as an outcome, indicated that the effect of distance to a thrombus (i.e., longer DT) on the clinical outcome at 3 months (mRS 0–2 vs. 3–6) was fully mediated via the ln-transformed infarct volume. The indirect effect of distance to thrombus in the mediation model was small: 0.005; however, it was statistically significant (p = 0.018). After adjusting for age, thrombectomy outcome, and ASPECTS (> 6 vs. ≤ 6), the mediation stayed statistically significant (p = 0.016); however, the effect was small at 0.004.

Although our study was conducted in the contemporary 3‑year period in 2 large comprehensive stroke centers with significant stroke patient turn-around, some limitations prevent the generalization of our conclusions. The study’s retrospective design most probably introduced biases we could not account for. We did not include all patients with unknown time of symptom onset, only those having symptoms upon awakening. Only M1 occlusions were included in this analysis and the observed results cannot be extrapolated for ICA occlusions. Also, this analysis did not assess the exact position of the thrombus in relation to the perforating arteries or their patency. We do not have the status of infarct core volume at first imaging as we did not routinely perform perfusion imaging or MRI-DWI in every patient. This reflects the real-life setting of the study. Using CT can be less sensitive for measuring ILV compared to MRI; however, at least one study showed no significant difference between both techniques [14]. Furthermore, ABC/2 volume determination has some limitations, and results could be different when automatic ILV measurement has been used. Additionally, the results could have been different if we had had MRI as control imaging. There is a lack of universal mRS documentation for our patients (18% missing); however, as already mentioned, ILV could serve as a proxy outcome measure. As we included only MCA occlusions our results are not immediately generalizable to other types of LVO.