Endovascular Thrombectomy for Acute Stroke with a Large Ischemic Core: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Our systematic review and meta-analysis adhered to the guidelines provided by the PRISMA statement [22] and the Cochrane handbook for systematic reviews and meta-analyses [23]. The protocol for this review has been registered and published in PROSPERO with the following ID: CRD42023407277.

(B.A. and M.A.) performed a thorough electronic search for relevant literature by utilizing several databases, including PubMed (MEDLINE), Web of Science, SCOPUS, and the Cochrane Central Register of Controlled Trials (CENTRAL) until February 18th, 2023. They did not use any limitations on their search. Further details about the search strategy, including the keywords and search terms, as well as the results of the search, can be found in (Table S1).

A PICO criterion was used to include RCTs: population (P): patients with AIS with a large infarct size defined as large vessel occlusion with ASPECTS score 3 to 5; intervention (I): endovascular thrombectomy plus medical therapy; control (C): medical therapy alone; outcome (O): primary outcomes of this review are the efficacy outcomes: early neurological improvement assessed by ≥ 4 points reduction in the National Institutes of Health Stroke Scale (NIHSS), excellent neurological recovery (modified Rankin Scale (mRS) 0–1), functional Independence (mRS 0–2), and independent ambulation (mRS 0–3). The secondary outcomes included safety outcomes: (any-cause mortality at 90 days, poor neurological recovery (mRS 4–6), any intracerebral hemorrhage, symptomatic intracerebral hemorrhage, and decompressive craniectomy).

We did not consider a range of research designs in our analysis. Specifically, we excluded non-human studies, preliminary reports, various forms of observational studies, single-arm clinical trials, in vitro experiments conducted on tissues and cultures, book chapters, editorial, and press articles, publications that only contain abstracts or posters, unpublished study protocols, and studies that were conducted in languages other than English.

The review process was carried out using the Covidence online tool. (H.A.S. and A.S.) reviewed the retrieved records independently after eliminating any duplicated records. The full-texts of the records that met the initial eligibility criteria were examined through full-text screening. Any discrepancy was resolved by consensual discussion and agreement.

Two reviewers (H.A.S. and A.S.) independently extracted all data using a standardized electronic spreadsheet: study characteristics (country, study design, total participants, main inclusion criteria, intervention, and comparison methods, ASPECTS, timing after symptoms onset (time window), and baseline imaging); baseline characteristics (age, sex, number of patients in each group, ASPECTS score, NIHSS, infarct core volume, and occlusion location); efficacy outcomes data (early neurological improvement, excellent neurological recovery (mRS 0–1), functional Independence (mRS 0–2), and independent ambulation (mRS 0–3)), and safety outcomes (any-cause mortality at 90 days, poor neurological recovery (mRS 4–6), any intracerebral hemorrhage, symptomatic intracerebral hemorrhage, and decompressive craniectomy). Any discrepancy was resolved by consensual discussion and agreement.

Two reviewers (H.A.S. and A.S.) assessed the quality of the studies included in the research independently using the Cochrane RoB2 tool [24]. The domains that were evaluated included the risk of bias resulting from the randomization process, the risk of bias due to deviation from the intended intervention, the risk of bias due to missing outcome data, the risk of bias in measuring the outcome, and the risk of bias in selecting the reported results. In the event of any disagreements, the reviewers discussed and resolved them through consensus.

To appraise the quality of evidence, two reviewers (M.A. and B.A.) utilized the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) guidelines [25, 26]. The evaluation was carried out for each outcome, and the decisions were justified and documented. Any discrepancy was resolved by consensual discussion and agreement.

The RevMan v5.3 software [27] was used for statistical analysis. To combine the outcomes for dichotomous outcomes, the risk ratio was used, while the mean difference (MD) was used for continuous outcomes. Both were calculated with a 95% confidence interval (CI) using the fixed-effects model. However, the random-effects model was used in case of significant heterogeneity. The presence and extent of heterogeneity were evaluated using the Chi-square and I‑square tests, respectively. Following the Cochrane Handbook (chapter nine) [28], heterogeneity was considered significant if the alpha level for the Chi-square test was below 0.1, while the I‑square test results were interpreted as follows: not significant for 0–40%, moderate heterogeneity for 30–60%, and substantial heterogeneity for 50–90%. On significant heterogeneity, a leave-one-out sensitivity analysis was conducted.

A total of 1995 studies were identified and evaluated for screening based on their titles and abstracts. After removing 969 duplicates and 1012 studies that did not match the inclusion criteria, fourteen full-text articles were assessed. Out of these, eleven were found to be irrelevant and excluded, leaving three RCTs to be included in the qualitative and quantitative analysis (Fig. 1).

We included a total of three RCTs [19‐21]. Detailed summary characteristics of the included studies are outlined in (Table 1). A total of 1010 patients were included, of which 509 were allocated to the ET group and 501 patients to the medical management group. Most patients were men, including 297 (58.3%) in the ET group and 302 (60.3%) in the medical management group. Further baseline characteristics are highlighted in (Table 2).

Using the Cochrane RoB2 tool’s five domains, we evaluated each outcome included in the quantitative synthesis’s risk of bias (Fig. 2). All of the included RCTs showed an overall high risk of bias, mainly attributed to the performance bias due to the lack of blinding.

Endovascular thrombectomy significantly increased the rates of functional independence (mRS ≤ 2) (RR: 2.54 with 95% CI [1.85, 3.48], P = 0.00001) (low-quality evidence) (Fig. 3a; Table 3), independent ambulation (mRS ≤ 3) (low-quality evidence) (RR: 1.78 with 95% CI [1.28, 2.48], P = 0.0006) (low-quality evidence) (Fig. 3b; Table 3), and early neurological improvement (RR: 2.46 with 95% CI [1.60, 3.79], P = 0.0001) (low-quality evidence) (Fig. 3c; Table 3). However, there was no difference between endovascular thrombectomy and medical care in excellent neurological recovery (mRS ≤ 1) (RR: 1.35 with 95% CI [0.88, 2.08], P = 0.17) (low-quality evidence) (Fig. 3d; Table 3).

Pooled studies were homogenous in functional independence (mRS ≤ 2) (P = 0.64, I² = 0%), early neurological improvement (P = 0.20, I² = 38%), and excellent neurological recovery (mRS ≤ 1) (P = 0.19, I² = 40%). However, studies were heterogenous in independent ambulation (mRS ≤ 3) (P = 0.09, I² = 58%). Therefore, we performed a sensitivity analysis, and heterogeneity was best resolved by excluding ANGEL-ASPECT RCT (P = 0.60, I² = 0%) (Table S2).

Endovascular thrombectomy significantly reduced the rate of poor neurological recovery (mRS 4–6) (RR: 0.79 with 95% CI [0.72, 0.86], P = 0.00001) (low-quality evidence) (Fig. 4a; Table 3), with no difference regarding all-cause mortality (RR: 0.95 with 95% CI [0.78, 1.16], P = 0.61) (low-quality evidence) (Fig. 4b; Table 3), symptomatic intracranial hemorrhage (RR: 1.83 with 95% CI [0.95, 3.55], P = 0.07) (low-quality evidence) (Fig. 4c; Table 3), and decompressive craniectomy (RR: 1.22 with 95% CI [0.43, 3.41], P = 0.71) (very low-quality evidence) (Fig. 4d; Table 3). However, endovascular thrombectomy was associated with more incidence of any intracranial hemorrhage (RR: 2.30 with 95% CI [1.50, 3.51], P = 0.0001) (very low-quality evidence) (Fig. 4e; Table 3).

Pooled studies were homogenous in poor neurological recovery (mRS 4–6) (P = 0.96, I² = 0%), all-cause mortality (P = 0.55, I² = 0%), and symptomatic intracranial hemorrhage (P = 0.50, I² = 0%). However, studies were heterogenous in decompressive craniectomy (P = 0.07, I² = 70%) and any intracranial hemorrhage (P = 0.06, I² = 71%), and sensitivity analysis was not applicable.

Our meta-analysis is based on data from the three most recent RCTs, with minimal statistical heterogeneity (any heterogeneity encountered was resolved by sensitivity analysis) among outcomes guaranteeing the relevance and reliability of our conclusions to be the gold standard evidence regarding this matter. Nevertheless, there are however certain limitations due to the inherent characteristics of the included studies: first, our results might not be generalizable to all patient populations due to two of the included studies being based in Asian geography (Japan & China) [19, 20]. Second, there is also variability in the time window the patients were enrolled, which can influence patient outcomes. As in ANGEL-ASPECT [20], 63.3% of the patients were enrolled in the 6‑to-24-hour time window, whereas in RESCUE-Japan LIMIT [19], 28.6% of the patients were enrolled in this late window. Third, due to the lack of agreement on the management between different centers, several confounding variables may have gone unreported & therefore impacted our results. Similarly, there may be institutional variability in assessing AIS depending on whether CT or diffusion-weighted MRI was used to calculate ASPECTS values. Fourth, another factor that can introduce bias and hence influence the validity is the difference in group sizes notably, the number of patients receiving IV thrombolysis remained smaller than ET, possibly due to eligibility limitations. Finally, the GRADE assessment yielded low to very low-quality evidence, limiting the clinical endorsement of our findings.

To address these limitations, First, large-scale RCTs should be conducted, including patients with diverse characteristics: demographic, comorbidities, and stroke risk factors. Studies should be designed to explore patient populations with low ASPECT scores < 5, baseline NIHSS score, large ischemic core, and optimal time window for ET. Further imaging protocols should be standardized by using a specific imaging modality, such as CT or DW-MRI, for stroke assessment. In this regard, results from ongoing RCTs are anticipated to provide promising results. In this regard, ongoing RCTs are committed to exploring the horizons of ET based on varying imaging modalities and inclusion criteria. Results from the European (TENSION, NCT03094715) trial investigating ASPECTS 3–5 at baseline in the extended time window of up to 12 h, the French (IN-EXTREMIS-LASTE, NCT03811769) trial exploring ET in a seven-hour time window, ASPECTS 0–5 on DWI-MRI or non-contrast CT, and the North American (TESLA, NCT03805308) assessing moderately large infarct volume NCCT ASPECTS 2–5 are likely to provide conclusive evidence.

Second, a patient-level & subgroup analysis is warranted for an in-depth exploration of patient factors to determine if patients with certain characteristics are more likely to benefit from ET. Additionally, a long-term follow-up duration, i.e., 12 months in contrast to the typical three months, should be conducted to better assess the impact of ET on patients’ quality of life vs. medical therapy alone. Finally, for ET to be adopted in widespread clinical practice, a cost-effectiveness analysis should be conducted between the two interventions. This is especially important in low and middle-income countries (LMIC) settings where the best evidence is required to convince policymakers and stakeholders to sponsor treatments.