Direct oral anticoagulants versus vitamin K antagonists for postoperative atrial fibrillation after cardiac surgery: a systematic review and meta-analysis of randomized controlled trials

Abstract

Background

Postoperative atrial fibrillation (POAF) is a common complication after cardiac surgery. Although direct oral anticoagulants (DOACs) have revolutionized stroke prevention in non-surgical atrial fibrillation, their role relative to vitamin K antagonists (VKAs) specifically in POAF is unclear.

Methods

A systematic review and meta-analysis were performed from randomized controlled trials comparing DOACs (apixaban, rivaroxaban, or dabigatran) vs. VKAs in adult patients with POAF following cardiac surgery. The primary outcomes included stroke/systemic embolism as an efficacy endpoint and major bleeding as a safety endpoint. Secondary outcomes comprised any bleeding and cost-effectiveness. We searched the following databases: PubMed, Scopus, Cochrane Central, and Google Scholar from inception through November 2025. The study protocol was registered with PROSPERO (Registration number: CRD420251249764).

Results

Four randomized controlled trials (RCTs) were included (n = 202 patients). Stroke/systemic embolism occurred in 1.0% (1/103) of DOAC patients versus 1.0% (1/99) of VKA patients (RR 0.80, 95% CI 0.06–11.5; P = 0.87). Major bleeding occurred in 1.9% (2/103) versus 4.0% (4/99) (RR 0.55, 95% CI 0.11–2.64; P = 0.46). Any bleeding occurred in 10.7% (11/103) versus 11.1% (11/99) (RR 1.01, 95% CI 0.31–3.30; P = 0.99). Two trials reported lower costs with DOACs in their local settings, though these findings are exploratory and not generalizable.

Conclusion

Among patients with POAF following cardiac surgery, DOACs appear to have efficacy and safety profiles similar to those of VKAs, with non-significant trends toward lower major bleeding rates. Reported cost advantages are preliminary and context-specific. These findings are based on a small number of trials with limited sample sizes and very low event rates, and do not demonstrate equivalence; they warrant confirmation in larger, adequately powered studies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-05571-9.

Introduction

Postoperative atrial fibrillation (POAF) is one of the most frequent complications after cardiac surgery, affecting 20–40% of patients receiving coronary artery bypass grafting (CABG) or valve surgery [1]. Usually transient, POAF is associated with significant morbidity, including increased risk of stroke and systemic embolism, prolonged hospital stay, and higher healthcare expenditure [2]. The exact pathophysiology consists of inflammation due to surgery, imbalance in autonomic tone, and, finally, an atrial remodelling that creates a prothrombotic state, which may persist beyond the early postoperative period [3].

Current guidelines support the consideration of anticoagulation for POAF, founded on stroke risk stratification using the CHA₂DS₂-VASc score, in a manner similar to that for non-surgical atrial fibrillation [4]. Moreover, contemporary atrial fibrillation guidelines emphasize early rhythm control in select patients, which may influence both the duration of anticoagulation and the choice of antithrombotic therapy in the postoperative period [5]. However, the best strategy for anticoagulation in this unique population remains a matter of debate. Vitamin K antagonists (VKAs), mainly warfarin, have traditionally been used but pose practical challenges in the postoperative setting, including requirements for frequent monitoring of INR, dietary constraints, and bridging with heparin, all factors that may delay discharge and increase costs [6].

Direct oral anticoagulants have transformed stroke prevention in non-valvular atrial fibrillation by proving themselves to be non-inferior to warfarin with improved safety profiles and greater convenience in landmark trials [7–9]. Their predictable pharmacokinetics, fixed dosing without routine monitoring requirement, and rapid onset of action represent a theoretical advantage in postoperative cardiac surgical patients. However, the bleeding risk in freshly operated patients, potential drug interactions with common postoperative medications, and limited evidence in POAF have acted as a deterrent to wide usage of these drugs [10].

The recent meta-analysis by Benedetto et al. focused on anticoagulation for POAF and did not directly compare DOACs versus VKAs [11]. This important knowledge gap has recently been informed by several pilot randomized controlled trials, but results remain fragmented and individually underpowered to conclusively inform practice. To date, no systematic review or meta-analysis has synthesized this emerging evidence.

This systematic review and meta-analysis thus seek to compare the efficacy and safety of DOACs versus VKAs specifically in patients with POAF after cardiac surgery and to evaluate their comparative cost-effectiveness to provide much-needed evidence for guiding clinical decision-making in this complex population.

Methods

Study protocol and registration

The design for this systematic review and meta-analysis was specified in a study protocol registered at the International Prospective Register of Systematic Reviews, PROSPERO (Registration number: CRD420251249764). All procedures were performed following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement.

Search strategy

A systematic literature search was performed following PRISMA 2020 guidelines. We searched the following electronic databases from inception to November 23, 2025: PubMed, Scopus, and the Cochrane Central Register of Controlled Trials (CENTRAL). We also searched ClinicalTrials.gov for ongoing or completed but unpublished trials relevant to our review. A supplementary search was conducted in Google Scholar. Only English-language publications were considered. The reference lists of all included studies and relevant review articles were manually screened. For this review, we developed a search strategy, including a combination of keywords and controlled vocabulary terms related to the concepts “postoperative atrial fibrillation,” “cardiac surgery,” “direct oral anticoagulants,” “DOAC,” “vitamin K antagonists,” and “warfarin.” Known relevant RCTs’ names were also included in the search.

All identified studies were published and reported the outcomes of interest in full; therefore, no authors were contacted for additional outcome data. However, one eligible trial (Moonsamy et al. 2025, the NEW-AF Trial) was excluded because the full text could not be accessed despite attempts to contact the corresponding author, and the abstract did not report all prespecified outcomes.

Eligibility criteria

We included RCTs enrolling adult patients ≥ 18 years who developed new-onset or pre-existing POAF after any major cardiac surgery, including coronary artery bypass grafting (CABG), bioprosthetic valve replacement or repair, or combined procedures. We excluded patients with mechanical heart valves, moderate to severe mitral stenosis, or any other valvular disease that would contraindicate the use of DOACs according to current guidelines (e.g., significant mitral stenosis, prosthetic valves). This ensured that the included population aligned with the approved indications for DOAC use in atrial fibrillation.We restricted inclusion to randomized controlled trials to ensure the highest level of evidence for causal inference regarding treatment effects. Although observational studies may provide additional data, they are inherently subject to greater confounding and selection bias, which could compromise the validity of comparative effectiveness conclusions.

The intervention of interest was any DOAC, either apixaban, rivaroxaban, or dabigatran, at doses approved for stroke prevention in AF. The comparator was VKAs, mainly warfarin, with or without bridging therapy.

The primary efficacy outcome was stroke (ischemic or hemorrhagic) or systemic thromboembolism. The primary safety outcome was major bleeding, defined according to the ISTH criteria or the study authors. Secondary outcomes of interest included any bleeding (major or minor) and cost-effectiveness outcomes where reported. Nonrandomized studies, observational cohorts, case reports, and studies that did not report on the outcomes of interest were excluded.

Study selection and data extraction

Study selection was done by two independent reviewers. First, titles and abstracts of all records retrieved were checked against the eligibility criteria, followed by full-text review of potentially relevant articles to reach a final decision. Discrepancies in decisions taken by the reviewers were resolved through consensus or consultation with a third reviewer.

Two investigators independently extracted data from each eligible study using a standard data extraction form. The following data were extracted: study identification (first author and year); details of the intervention (type of DOAC, use of warfarin, sample sizes in each group, and duration of follow-up); and raw event data for all pre-specified outcomes, including stroke, systemic embolism, major bleeding, and any bleeding events for both treatment arms. Cost-effectiveness results were extracted separately for narrative synthesis.

Risk of bias assessment

The methodological quality of the included randomized controlled trials was independently assessed by two reviewers using the Cochrane Risk of Bias (RoB 1) tool described in the Cochrane Handbook. This included seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias.

Data synthesis and analysis

We used Review Manager (RevMan) software, version 5.4, for all statistical analyses. For dichotomous outcomes, the pooled RR with 95% CIs was estimated according to the Mantel-Haenszel method under a random-effects model. In addition, we quantified heterogeneity of the included studies using the I² statistic, where < 25%, 25–50%, and > 75% were interpreted as low, moderate, and substantial heterogeneity, respectively. The summary effect was considered statistically significant when the p-value was less than 0.05. Two studies included cost-effectiveness analyses alongside clinical outcomes. Due to methodological heterogeneity in cost reporting (different healthcare systems, currencies, and time horizons), these findings were synthesized narratively.

Assessment of certainty of evidence

The certainty of evidence for the primary outcomes (stroke/systemic embolism and major bleeding) was evaluated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. According to GRADE methodology, evidence from randomized controlled trials begins as high certainty but may be downgraded based on five domains: (1) risk of bias, (2) inconsistency, (3) indirectness, (4) imprecision, and (5) publication bias. Two review authors independently assessed each domain, with disagreements resolved through consensus or consultation with a third reviewer. We documented reasons for downgrading or upgrading the certainty of evidence for each outcome.

Results

Selection of studies

The systematic search of electronic databases yielded 290 records from PubMed, Scopus, and Cochrane Central. Duplicate records were removed separately for database-derived citations and grey literature/registry searches. Removing duplicates resulted in 200 records being screened by title and abstract. From these records, 195 were excluded, and 5 full-text articles were sought for retrieval. Of the 5 articles, 1 was not retrievable, and thus, 4 articles were assessed for eligibility. All 4 studies met the inclusion criteria. Separately, in Google Scholar, 50 records were identified and screened; 20 were duplicates, 10 were not retrievable, and the remaining 20 were excluded as they were editorials, reviews, or non-RCTs. Four randomized controlled trials were thus included for qualitative and quantitative synthesis. The selection of studies for inclusion is summarized in the PRISMA flow diagram below (Fig. 1).

Fig. 1.

PRISMA flow diagram. Duplicates were removed in two stages: 90 duplicates from 290 database records, and 20 duplicates from 50 grey literature/registry records. *Consider, if feasible to do so, reporting the number of records identified from each database or register searched (rather than the total number across all databases/registers).**If automation tools were used, indicate how many records were excluded by a human and how many were excluded by automation tools. Source: Page MJ, et al. BMJ 2021;372:n71. 10.1136/bmj.n71. This work is licensed under CC BY 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/

Study characteristics

We included a total of four RCTs [12–15] that comprised 202 participants. The following RCTs were included: Chapin 2020 [12], Ghavami 2024 [13], Duraes 2016 [14], and Pereira 2023 [15]. Participants in all trials included adult patients who developed new-onset postoperative atrial fibrillation (POAF) after cardiac surgery. In three of the trials, participants had undergone CABG (Chapin, Ghavami, Pereira), and in one trial, patients had undergone bioprosthetic valve replacement. Three trials were small sample-sized pilot trials (Chapin, Ghavami, Duraes), whereas Pereira 2023 was a full-scale randomized trial that also included an economic evaluation. Table 1 describes the baseline characteristics of all the trials included.

Table 1.

Characteristics of included randomized controlled trials

Study (Year)	Country	Design	Population	Surgery Type	DOAC (n)	Warfarin (n)	DOAC Type & Dose	INR Target	Follow-up	Primary Outcomes
Chapin 2020	USA	Pilot RCT	POAF after CABG	CABG	28	28	Apixaban 5 mg BID	2.0–3.0	30 days	Cost, Safety
Ghavami 2024	Iran	Pilot RCT	POAF after CABG	CABG	34	32	Rivaroxaban 15/20 mg OD	2.0–3.0	30 days	Major bleeding
Duraes 2016	Brazil	Pilot RCT	POAF after valve	Bioprosthetic valve	15	12	Dabigatran 110 mg BID	2.0–3.0	90 days	Intracardiac thrombus
Pereira 2023	Brazil	RCT	POAF after CABG	CABG	26	27	Rivaroxaban 15/20 mg OD	2.0–3.0	30 days	Cost-effectiveness

Abbreviations: BID twice daily, CABG coronary artery bypass grafting, DOAC direct oral anticoagulant, INR international normalized ratio, OD once daily, POAF postoperative atrial fibrillation, RCT randomized controlled trial

Quantitative analysis

Primary efficacy outcome: stroke or systemic embolism

The meta-analyzed data from four trials, involving 202 total patients, showed no statistically significant difference, though the confidence intervals were extremely wide in the rate of stroke or systemic embolism between the DOAC and warfarin groups. Stroke or reversible ischemic neurological deficit occurred in 1.0% (1/103) of patients receiving DOACs compared to 1.0% (1/99) receiving warfarin (Risk Ratio [RR] 0.80, 95% Confidence Interval [CI] 0.06 to 11.5; P = 0.87). Systemic embolism (including intracardiac thrombus) occurred in 0% (0/103) of DOAC patients compared to 1.0% (1/99) of warfarin patients (RR 0.27, 95% CI 0.01–6.11; P = 0.41). Heterogeneity was not applicable for these outcomes as only one trial contributed events (Figs. 2 and 3).

Fig. 2.

Forest plot for stroke or reversible ischemic neurological deficit comparing DOACs versus warfarin

Fig. 3.

Forest plot for systemic embolism comparing DOACs versus warfarin

Primary safety outcome: major bleeding

Major bleeding occurred in 1.9% (2/103) of patients receiving DOACs compared to 4.0% (4/99) receiving warfarin (RR 0.55, 95% CI 0.11–2.64; P = 0.46) with no heterogeneity (I² = 0%) (Fig. 4). The clinical interpretation of this finding is limited by the very small number of events.

Fig. 4.

Forest plot for major bleeding comparing DOACs versus warfar

Secondary outcome: any bleeding

Any major or minor bleeding occurred in 10.7% of patients receiving DOACs (11/103) compared to 11.1% receiving warfarin (11/99) (RR 1.01, 95% CI 0.31–3.30; P = 0.99) with moderate heterogeneity (I² = 36%) (Fig. 5).

Fig. 5.

Forest plot for any bleeding comparing DOACs versus warfarin

Exploratory cost analysis

Two trials reported exploratory cost analyses in their respective settings. The study by Pereira 2023 found significantly lower total costs were associated with rivaroxaban compared to warfarin (median $586.80 vs. $1,423.20; P = 0.002), yielding a cost savings of 52.4%, mainly resulting from shorter hospital lengths of stay (2 vs. 5 days; P = 0.01), and due to avoidance of bridging therapy. Another study, Chapin 2020, similarly showed lower total costs with apixaban ($522.50 vs. $778.22; P = 0.003), which was apparently driven by the elimination of INR monitoring and laboratory expenses, and travel related to warfarin therapy. Both these studies reported similar quality-adjusted life years between treatment groups. These economic findings should be interpreted as exploratory and are not generalizable beyond the local contexts and healthcare systems in which the studies were conducted.

Risk of bias assessment

The methodological quality of the four included trials was assessed using the Cochrane Risk of Bias tool (RoB 1). Regarding selection bias, three studies were rated as having low risk for random sequence generation, while one study had an unclear risk. Allocation concealment was rated as low in one study, using sealed opaque envelopes, and as unclear in three studies. In terms of performance bias, all four studies were rated as at high risk due to open-label designs; blinding of participants and personnel was not possible due to different monitoring requirements between DOACs and warfarin. For detection bias, the risk was rated as unclear across all studies since blinding of outcome assessors was not explicitly reported. Attrition and reporting biases were rated as low in all trials. Other sources of bias were rated as low across all studies. A detailed study-level traffic light summary of methodological quality and bias risk for each included trial is shown in Fig. 6.

Fig. 6.

Risk of bias summary for included studies

Sensitivity analysis

A sensitivity analysis excluding Pereira et al. [4] – which focused primarily on cost-effectiveness – was performed to assess the robustness of the findings. Results remained consistent across all outcomes. For stroke or reversible ischemic neurological deficit, the risk ratio was unchanged at 0.80 (95% CI 0.06–11.50; P = 0.87). For systemic embolism, the risk ratio remained 0.27 (95% CI 0.01–6.11; P = 0.41). Major bleeding showed a similar risk ratio of 0.81 (95% CI 0.12–5.34; P = 0.82; I² = 4%). Any bleeding demonstrated a risk ratio of 1.72 (95% CI 0.43–6.91; P = 0.44; I² = 16%), indicating a non-significant trend toward more bleeding with DOACs in this subset, though the overall conclusion remained unchanged (Figs. 7, 8, 9 and 10). The results of this sensitivity analysis are summarized in Table 2.

Fig. 7.

Sensitivity analysis forest plot for stroke comparing DOACs versus warfarin, excluding Pereira et al. [15]

Fig. 8.

Sensitivity analysis forest plot for systemic embolism comparing DOACs versus warfarin, excluding Pereira et al. [15]

Fig. 9.

Sensitivity analysis forest plot for major bleeding comparing DOACs versus warfarin, excluding Pereira et al. [15]

Fig. 10.

Sensitivity analysis forest plot for any bleeding comparing DOACs versus warfarin, excluding Pereira et al. [15]

Table 2.

Sensitivity analysis comparing Meta-Analysis with and without Pereira et al. [15]

Outcome	All Four Studies	Excluding Pereira et al.	Interpretation
Stroke/Reversible Ischemic Neurological Deficit	RR 0.80 (0.06–11.50), P = 0.87	RR 0.80 (0.06–11.50), P = 0.87	No change
Systemic Embolism	RR 0.27 (0.01–6.11), P = 0.41	RR 0.27 (0.01–6.11), P = 0.41	No change
Major Bleeding	RR 0.55 (0.11–2.64), P = 0.46, I² = 0%	RR 0.81 (0.12–5.34), P = 0.82, I² = 4%	Direction and significance unchanged
Any Bleeding	RR 1.01 (0.31–3.30), P = 0.99, I² = 36%	RR 1.72 (0.43–6.91), P = 0.44, I² = 16%	Trend toward more DOAC bleeding in subset; overall conclusion stable

Sensitivity analysis based on risk of bias

To assess whether studies with a higher risk of bias influenced our findings, we conducted a sensitivity analysis excluding the two trials with the highest overall risk of bias [2, 3]. For the outcome of stroke or systemic embolism, this analysis could not be performed because both events occurred in the excluded trials. For bleeding outcomes, the results are presented in Table 3. The risk ratios and confidence intervals remained consistent with the main analysis, suggesting that the overall conclusions are robust despite variations in methodological quality among the included trials.

Table 3.

Sensitivity analysis of bleeding outcomes after exclusion of studies with higher risk of bias (Ghavami et al. [13] and duraes et al. [14])

Outcome	Studies Included	DOAC Events/Total	VKA Events/Total	Risk Ratio (95% CI)	P-value	I²
Major bleeding	Chapin 2020, Pereira 2023	1/54	2/55	0.75 (0.05–10.31)	0.83	31%
Any bleeding	Chapin 2020, Pereira 2023	7/54	8/55	1.12 (0.13–9.62)	0.92	68%
Stroke / Systemic Embolism	Not analyzable*	–	–	–	–	–

Sensitivity analysis for stroke/systemic embolism was not feasible as both events occurred in the excluded studies

Certainty of evidence

The GRADE assessment for primary outcomes is summarized in Table 4. For both stroke/systemic embolism and major bleeding, the certainty of evidence was rated as very low. This rating resulted from downgrading for serious risk of bias (all trials were open-label with high or unclear risk of performance and detection bias) and for very serious imprecision (total sample size of 202 patients, few events [≤ 6 per outcome], and extremely wide confidence intervals that included both clinically significant benefit and harm; e.g., RR 0.06–11.5 for stroke). No concerns were identified regarding inconsistency, indirectness, or publication bias. Very low certainty indicates that the true effect may be substantially different from the estimated effect.

Table 4.

GRADE assessment of certainty of evidence for primary outcomes

Outcome	No. of studies (patients)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Overall certainty
Stroke or systemic embolism	4 RCTs (n = 202)	Serious¹	Not serious	Not serious	Very serious²	Undetected	⊕○○○ VERY LOW
Major bleeding	4 RCTs (n = 202)	Serious¹	Not serious	Not serious	Very serious²	Undetected	⊕○○○ VERY LOW

Explanations:

¹All trials were open-label with high/unclear risk of performance and detection bias

²Downgraded two levels due to: (a) extremely wide confidence intervals spanning both appreciable benefit and harm, and (b) optimal information size not met (total sample < 300, events < 10)

Fragility of findings

To assess the robustness of our non-significant findings, we calculated the Fragility Index for primary outcomes. For stroke/systemic embolism, the Fragility Index was 2, meaning that a change in only two events would be required to render the result statistically significant. For major bleeding, the Fragility Index was ≥ 3. These small indices underscore the fragility of our conclusions and highlight the limited statistical power due to low event rates.

Discussion

This systematic review and meta-analysis of four randomized controlled trials represents the first comprehensive comparison of DOACs versus VKAs solely in patients with postoperative atrial fibrillation following cardiac surgery. Our findings demonstrate no statistically significant differences between treatment strategies across all efficacy and safety outcomes, though several important patterns and clinical implications emerge.

The similar rates of stroke and systemic embolism seen in our analysis, 1.0% (1/103) vs. 1.0% (1/99) for DOACs and warfarin, respectively, are consistent with the established non-inferiority of DOACs in general atrial fibrillation populations [7–9]. However, the very low event rates in our included studies reflect both the limited sample sizes and the potentially lower thromboembolic risk of POAF compared to chronic atrial fibrillation [16]. This raises important questions about the necessity and duration of anticoagulation in POAF, particularly given that most episodes are transient and sinus rhythm is often restored within weeks [17].

Concerning safety outcomes, a non-significant trend toward reduced major bleeding with DOACs was observed, 1.9% vs. 4.0%, RR 0.55. This is in keeping with the established safety profile of DOACs in non-surgical populations, where numerous meta-analyses have demonstrated reduced intracranial hemorrhage compared to warfarin [18]. Similar rates of any bleeding, 10.7% vs. 11.1%, suggest that while DOACs may reduce the most severe bleeding events, overall bleeding risk remains similar between strategies in the postoperative setting. This is particularly relevant given the heightened bleeding risk following cardiac surgery, where multiple suture lines, recent heparinization, and frequent antiplatelet use coexist [19].

Exploratory cost analyses from two trials suggested potential economic benefits of DOACs in their respective settings. Both Pereira et al. [15] and Chapin et al. [12] reported lower total costs with DOACs, mainly driven by reduced length of stay and elimination of monitoring. These findings are of interest in contemporary healthcare systems that emphasize value-based care and rapid postoperative recovery pathways [20]. However, they are context-specific and not generalizable, given differences in healthcare systems, currencies, and cost structures. They should be viewed as hypothesis-generating for future health-economic evaluations rather than as definitive evidence of cost-effectiveness.

Interpretation in context of limited evidence

It is crucial to interpret our findings within the context of the limited available evidence. Our meta-analysis included only 202 patients from four trials, three of which were pilot studies. The extremely low event rates, with only two stroke events and six major bleeding events total, render our estimates highly imprecise, as reflected in the very wide confidence intervals (e.g., RR 0.06–11.5 for stroke). While point estimates suggest no significant differences between DOACs and VKAs, the very low certainty of evidence (per GRADE assessment) means we cannot draw definitive conclusions about equivalence or non-inferiority. These results should be viewed as hypothesis-generating rather than practice-changing.

Several important limitations constrain the interpretation of our findings. First, our analysis is substantially underpowered, combining only 202 patients from four trials, three of which were explicitly designed as pilot studies. The resultant extremely low event rates (e.g., only 2 stroke events) render any statistical inference highly uncertain, as demonstrated by the exceptionally wide confidence intervals. A formal GRADE assessment rated the certainty of evidence as very low for all key outcomes. Second, we restricted inclusion to randomized controlled trials to prioritize internal validity and causal inference; however, this also limited the available data. In a research area with few RCTs, the exclusion of observational studies may have omitted potentially informative data from high-quality cohort studies.Third, the open-label design of all studies introduces the potential for performance and detection bias, although this is largely unavoidable given the distinct monitoring requirements of each drug class. Fourth, there is clinical heterogeneity in surgical populations (CABG vs. valve surgery), DOAC types, and follow-up durations. Fifth, one eligible RCT (Moonsamy et al. 2025, the NEW-AF Trial) was identified but could not be included in the meta-analysis due to the inability to access the full text. This may have introduced selection bias and limited the completeness of our evidence base. Sixth, the exclusion of patients with mechanical valves or severe renal impairment limits generalizability to all cardiac surgery patients. Seventh, we did not perform subgroup analyses by type of surgery (CABG vs. valve) or DOAC agent due to the limited number of trials and patients, which limits our ability to assess whether treatment effects vary across these clinically relevant subgroups.

Despite these limitations, our findings offer preliminary insights. For clinicians who manage POAF, the available data suggest DOACs may be considered as a potential alternative to warfarin, though this should be done cautiously given the very low certainty of evidence. Any perceived benefits in convenience or cost are based on exploratory data and require further validation. As such, the choice of strategy may rest on patient-specific factors, including likelihood of medication adherence, access to INR monitoring, and issues related to insurance coverage.

Future studies should therefore focus on adequately powered, randomized trials with longer follow-up to better define the balance between risk and benefit in this population. Furthermore, investigations that consider the optimal duration of anticoagulation for POAF and the role of risk prediction tools to identify patients who would most benefit from anticoagulation are urgently needed [21].

Conclusion

This meta-analysis found no statistically significant differences between DOACs and VKAs for stroke prevention or major bleeding in POAF after cardiac surgery, but the evidence is of very low certainty, and equivalence cannot be concluded. Non-significant trends toward reduced major bleeding and exploratory cost findings with DOACs warrant further investigation but must not be overinterpreted. Our findings are strictly preliminary and hypothesis-generating; they highlight an urgent need for larger, adequately powered randomized trials before any definitive conclusions can be drawn regarding the use of DOACs as an alternative to VKAs in this population.

Abbreviations

AF

Atrial Fibrillation

CABG

Coronary Artery Bypass Grafting

CI

Confidence Interval

DOAC

Direct Oral Anticoagulant

INR

International Normalized Ratio

ISTH

International Society on Thrombosis and Haemostasis

LOS

Length of Stay

POAF

Postoperative Atrial Fibrillation

RCT

Randomized Controlled Trial

RR

Risk Ratio

SE

Systemic Embolism

VKA

Vitamin K Antagonist

Authors’ contributions

OAOA conceptualized the research idea. OAOA and AAOAmanaged and planned the research process. MA, EY, MEEE, and RaAOA undertook database searches and article screening. MAI, MEMA, MB, ReAOA, and MAA undertook risk of bias assessment. HAS and CG extracted and summarized data and critically reviewed the manuscript for important intellectual content. AAOA analyzed data. OAOA, MB, and AAOAdrafted the manuscript. ASE provided supervision. All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. The complete data extraction sheet is provided as Additional File 1. The full electronic search strategies for all databases are provided as Additional File 2.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.