2.1 Eligibility Criteria

Studies were included in this systematic review and meta-analysis if they met the following criteria: (1) RCTs or observational studies, (2) comparing continuation versus interruption of OAC (including VKA and DOAC) during TAVI, (3) involving patients with an indication for OAC undergoing TAVI, and (4) reporting outcomes of interest such as all-cause mortality, stroke, any bleeding, and major bleeding. Accordingly, studies were excluded if they (1) lacked a control group for comparison, (2) did not provide clear stratified data, (3) were case reports, letters, or comments, (4) were written in languages other than English, or (5) involved overlapping patient populations, identified by studies conducted over the same period with common authors or study centers. In such cases, only the study with the larger sample size was included. No restrictions were applied regarding the publication date or follow-up duration.

2.2 Search Strategy and Data Extraction

We performed a structured search of PubMed (MEDLINE), Embase, and Cochrane Central Register of Controlled Trials databases from inception until September 12th 2024, using the following search terms: (“transcatheter aortic valve replacement” OR “transcatheter aortic-valve replacement” OR TAVR OR “transcatheter aortic valve implantation” OR “transcatheter aortic-valve implantation” OR TAVI) AND (“oral anticoagulation” OR “vitamin K antagonist” OR “vitamin K antagonists” OR warfarin OR “direct oral anticoagulant” OR “direct oral anticoagulants” OR apixaban OR dabigatran OR edoxaban OR rivaroxaban). The search was adapted for each database as necessary. Additionally, reference lists of all included studies and relevant systematic reviews were manually searched for any additional eligible studies. Data extraction was conducted independently by two authors (A.L.C.V and J.M.B.M), each following the predefined criteria. Any disagreements between the authors were resolved through discussion, and reasons for study exclusion were documented. This meta-analysis protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) under protocol CRD42024588855.

2.3 Outcomes

The primary outcome of interest was (1) all-cause mortality. Secondary outcomes included: (2) stroke; (3) any bleeding; and (4) major bleeding. These outcomes were defined according to the study authors' definitions, following the Valve Academic Research Consortium-2 (VARC-2) criteria and the Valve Academic Research Consortium-3 (VARC-3) guidelines, and they all occurred within 30 days post-TAVI [8, 9].

2.4 Quality Assessment

The risk of bias and quality assessment in randomized studies were evaluated with version 2 of the Cochrane Risk of Bias assessment tool (RoB 2) [10]. We collected data on outcomes based on intention-to-treat analysis. Non-randomized studies were assessed with the Risk of Bias in Non-randomized Studies of Interventions tool (ROBINS-I) [11]. Two independent authors (C.K.F and P.L.F.V) completed the risk of bias assessment. Disagreements were resolved through a consensus after discussing the reasons for the discrepancy (J.M.B.M.).

Two independent authors (P.A.M.G.S and R.S.B) assessed the level of certainty of the evidence in this meta-analysis using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) handbook guidelines, with categorizations ranging from very low to high [12]. Any disagreements were resolved through discussion with a third author (J.M.B.M).

2.5 Statistical Analysis

This systematic review and meta-analysis were conducted in accordance with the methodological guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions and in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [13, 14]. All analyzed outcomes were dichotomous, and risk ratios (RR) with 95% CI were used to compare treatment effects. Heterogeneity was assessed using the I² statistic and Cochran's Q test. Outcomes were considered to have low heterogeneity if p > 0.10 and I² < 25%, moderate heterogeneity if I² was between 25% and 75%, and high heterogeneity if I² > 75%. Statistical significance was defined as p < 0.05.

We used the DerSimonian and Laird random-effects model to account for the expected variability between studies, particularly given the inclusion of observational studies [15]. However, to ensure the robustness of our findings and address any uncertainty in the effect estimates, we performed an additional sensitivity analysis using the Hartung-Knapp-Sidik-Jonkman (HKSJ) method, as recommended in the Cochrane Handbook (Section 10.10.4.4) [13]. This method is recommended for meta-analyses with a small number of studies or in cases where there is low apparent heterogeneity but potential uncertainty due to study sizes. The HKSJ analysis was conducted to confirm the robustness of our results by adjusting for potential imprecision. All statistical analyses were performed using R statistical software version 4.2.2 (The R Foundation for Statistical Computing), utilizing the “meta” and “metafor” packages, along with RStudio version 2024.04.2 + 764 (RStudio Team) [16-18].

3.1 Study Selection and Baseline Characteristics

As shown in Figure 1, the search strategy identified 1949 studies. After removing duplicate records and excluding studies based on the title/abstract triage, eight studies remained for full-text assessment. Ultimately, three studies were included in our analysis: one randomized controlled trial and two observational studies, comprising a total of 2773 patients. Of these, 1314 (47.4%) continued OAC during TAVI, while 1459 (52.6%) interrupted OAC. The mean age of the patients ranged from 80 to 82 years, and 46.9% (n = 1301) were female. The follow-up period for all included studies was 30 days. The main characteristics of the study populations are presented in Table 1. The studies excluded from the meta-analysis and the reasons for their exclusion are detailed in Supporting Information S1: Table 2.

3.2 Primary Outcome

At 30 days post-TAVI, all-cause mortality occurred in 2.13% (28/1314) of patients in the continued OAC group, compared with 2.95% (43/1459) in the interrupted OAC group. There was no significant difference in all-cause mortality between the groups (RR 0.74; 95% CI 0.45–1.20; p = 0.22; I² = 5%; Figure 2).

3.3 Secondary Outcomes

The risk of stroke in the continued OAC group compared to the interrupted OAC group did not reach significance in the primary analysis (RR 0.65; 95% CI 0.42–1.01; p = 0.053; I² = 0%; Figure 3).

There were no significant differences between the groups in the incidence of any bleeding (RR 1.08; 95% CI 0.81–1.43; p = 0.60; I² = 81%; Figure 4A) or major bleeding (RR 0.90; 95% CI 0.67–1.21; p = 0.48; I² = 48%; Figure 4B). However, subgroup differences in major bleeding became evident when results were pooled by bleeding definition (VARC-2 vs. VARC-3).

3.4 Sensitivity Analyses

Concerning the risk of stroke, the sensitivity analysis showed a significant reduction, indicating that the continued OAC group had a lower incidence of stroke compared with the interrupted group (RR 0.65; 95% CI 0.47−0.90; p < 0.03; Figure 5). In terms of all-cause mortality, the results remained non-significant (RR 0.72; 95% CI 0.23−2.27; p = 0.34; I² = 5%; Figure 6a). Similarly, there was no significant difference in the occurrence of any bleeding (RR 1.08; 95% CI 0.81−1.44; p = 0.61; I² = 81%; Figure 6b), or major bleeding (RR 0.90; 95% CI 0.66−1.23; p = 0.52; I² = 48%; Figure 6c). However, subgroup differences in major bleeding were observed depending on the bleeding definition used (VARC-2 vs. VARC-3).

3.5 Risk of Bias Assessments

The ROBINS-I tool was applied to evaluate Mangner et al. and Brinkert et al., both of which showed an overall moderate risk of bias [6, 7]. Mangner et al. [6] had moderate risk due to confounding factors and deviations from the intended interventions, while Brinkert et al. [7] demonstrated moderate bias only in deviations from the intended interventions. Both studies exhibited a low risk of bias across all other domains, reflecting a generally sound methodology, despite limitations related to confounding and intervention deviations (Figure 7A).

The ROB-2 assessment of Van Ginkel et al. [2] shows a low risk of bias across all domains, including bias arising from the randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of reported results. Overall, the study was determined to have a low risk of bias [2] (Figure 7B).

According to the GRADE evaluation, the risk of bias was considered serious for all outcomes due to the inclusion of observational data in the overall analysis, as well as imprecision related to wide CI and a limited number of events. The certainty of evidence for hemorrhagic outcomes was very low, with moderate and high heterogeneity impacting confidence in these results (I² = 48% for major bleeding and I² = 81% for any bleeding, respectively). For stroke prevention in the continued OAC group, the evidence was classified as low certainty, downgraded due to the limited number of studies and the lack of statistical significance in the primary analysis (Supporting Information S1: Table 1).

4.1 Limitations

Therefore, this systematic review and meta-analysis suggest a similar clinical profile for both continuation and interruption of OAC in patients undergoing TAVI, with continuation potentially offering a significant reduction in stroke risk, consistent with the published literature. However, our findings have certain limitations. First, the limited number of studies restricts the depth of subgroup analysis and the ability to analyze the underlying indication for anticoagulation (e.g., atrial fibrillation with varying CHA2DS2-VASc scores) and bleeding risk (e.g., the presence of peripheral arterial disease), which could affect the risk/benefit of continuing or interrupting anticoagulation during the TAVI, yet randomization assecured the balanced presence of atrial fibrillation, hypertension, and diabetes in the RCT. Additionally, the type of OAC used in the studies represents another important limitation. In observational studies, many patients on continued anticoagulation were treated with DOACs, while in the RCT, the use of VKAs was minimally represented, which may influence the applicability of the findings. Concerning the observational studies, only previous myocardial infarction history at baseline was different between groups, and only one study had a significant baseline difference for the presence of diabetes mellitus [6]. CHA2DS2-VASc scores were paired between all studies.

Second, although the number of studies available for inclusion was small, the patient sample of around 3000 was substantial for most outcomes. Third, variations in patient populations across individual studies may have contributed to the high heterogeneity in any or major bleeding. Additionally, non-standardized approaches in clinical practice have made it difficult to categorize patients for research purposes uniformly. Fourth, all studies reported only 30-day outcomes, leaving uncertainty about whether the benefits and risks of continued anticoagulation persist beyond this period. The lack of long-term follow-up limits our ability to assess sustained effects in TAVI patients. Fifth, although adjustments for multiple comparisons were not applied, the possibility of type I error should be acknowledged. Although these limitations may restrict the robustness of the conclusions, this meta-analysis provides important insights into the current evidence on OAC continuation versus interruption in patients undergoing TAVI. Further RCTs are needed to better understand the implications in specific subgroups, such as those at high risk of bleeding or stroke, and to eliminate confounding effects.