Safety and efficacy of idarucizumab-assisted thrombolysis in dabigatran-related acute ischemic stroke: a systematic review and single-arm meta-analysis

Figure 1. PRISMA 2020 flow diagram of study selection. Records identified from databases and clinical trial registers were deduplicated. After title/abstract screening, full-text articles were assessed for eligibility. Studies were excluded for prespecified reasons including small sample size, conference-only publication, overlapping cohorts, or lack of relevant outcomes. Twelve studies were ultimately included: 11 for single-arm meta-analysis and 1 as an external historical reference cohort for descriptive comparison.

Figure 2. Risk of bias assessment using the JBI Critical Appraisal Checklist for Case Series. Each row represents an included study. Green circles (+) indicate low risk (Yes), yellow circles (−) indicate unclear risk, and red circles (×) indicate high risk (No). Overall judgment was determined by the predominant domain rating; studies with ≥3 high/unclear domains were classified as unclear risk. D1–D10 correspond to: D1 Clear aims, D2 consecutive enrollment, D3 participants clearly described, D4 demographics reported, D5 clinical characteristics reported, D6 outcomes/follow-up clearly reported, D7 outcome data provided, D8 statistical analysis included, D9 adverse events reported, D10 long-term outcomes reported. Detailed domain judgments and full checklist questions are provided in Supplementary material S9.

Figure 3. Forest plot summarizing the pooled proportion of symptomatic intracranial hemorrhage (sICH) across 11 single-arm studies. Individual study estimates are represented by black squares (size proportional to the inverse-variance weight), with horizontal lines denoting 95% confidence intervals (CIs). The vertical dotted line indicates the null value (0%). The gray diamond with a black border at the bottom represents the pooled estimate derived from a random-effects generalized linear mixed model (GLMM) with a logit link, along with its 95% CI. The pooled incidence of sICH was 5.4% (95% CI: 2.6–10.6%). Between-study heterogeneity was estimated at τ2 = 0.362 (Cochran’s Q = 17.20, df = 10, p = 0.120). The 95% prediction interval (PI) was 1.4–18.5%, indicating the expected range of true sICH rates in future clinical settings with similar patient characteristics and treatment protocols. Studies reporting zero events are appropriately accounted for within the GLMM framework without requiring continuity corrections. SICH, symptomatic intracranial hemorrhage; CI, confidence interval; PI, prediction interval; GLMM, generalized linear mixed model.

Figure 4. Forest plot of the pooled proportion of good functional outcome (modified Rankin Scale score 0–2 at 90 days) across the six single-arm studies. Proportions were pooled using a random-effects generalized linear mixed model (GLMM) with a logit link. The size of each square is proportional to the study weight. Horizontal lines represent 95% confidence intervals (CIs). The gray diamond with a black border indicates the pooled proportion with its 95% CI. Between-study variance was τ2 = 0.116 (95% CI 0.000–0.542). The 95% prediction interval (PI) for the true good functional outcome rate in a future setting with similar patient characteristics and treatment conditions was 46.1–82.3%.

Figure 5. Forest plot of the pooled proportion of any intracranial hemorrhage (any ICH) across the 11 single-arm studies. Proportions were pooled using a random-effects generalized linear mixed model (GLMM) with a logit link. The size of each square is proportional to the study weight. Horizontal lines represent 95% confidence intervals (CIs). The gray diamond with a black border indicates the pooled proportion with its 95% CI. Heterogeneity was low. Most hemorrhagic transformations were asymptomatic; only one event was symptomatic (already counted in sICH).

Figure 6. Doi plots for publication bias assessment. (A) Symptomatic intracranial hemorrhage (sICH). The LFK index was −1.93 (|LFK| between 1 and 2 indicates moderate asymmetry). (B) 90-day good functional outcome (mRS 0–2). The LFK index was −0.5 (|LFK| < 1 indicates minor or no asymmetry). Given the limited number of studies (k = 6 for the 90-day outcome), these results are exploratory. The performance of publication bias assessment in single-arm proportion meta-analyses is limited; therefore, these findings should be interpreted with caution and in conjunction with other sources of heterogeneity.