2.1 Data sources

This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols Statement of 2015.¹⁹ The review protocol was registered with the International Prospective Register of Systematic Reviews (http://www.crd.york.ac.uk/PROSPERO; registration number CRD42022374674). Two authors (TYC, CYL) independently searched the PubMed, Medline, and Cochrane Collaboration Central Register of Controlled Trials databases for studies published between January 2005 and June 2022, and any discrepancies were resolved by a third author (TFC). A combination of keywords or medical terms related to hemodialysis (e.g., dialysis, kidney failure), AF, and anticoagulation (e.g., oral anticoagulation, NOAC, Direct oral thrombin inhibitors, factor Xa inhibitors, dabigatran, rivaroxaban, apixaban, and Edoxaban) was used.

2.2 Selection criteria and data extraction

We included only registered randomized controlled trials (RCTs) comparing the efficacy and safety outcomes of NOACs and VKAs among AF patients with kidney failure receiving hemodialysis. Two authors (TYC, CYL) extracted the data independently, and any discrepancies were resolved via a third author (TFC) (Figure 1).

The principal efficacy and safety outcomes of the three analyzed studies are listed in Table 1. Individual outcomes included ischemic stroke, hemorrhagic stroke, clinically relevant non-major bleeding (CRNB), major bleeding, acute coronary syndrome (ACS), CV mortality, and all-cause mortality.

2.3 Statistical analysis

To explore the associations between NOAC/warfarin and outcomes of interest, we calculated relative risks (RRs) and related 95% confidence intervals (CIs) for binary variables, and mean differences and related 95% CIs for continuous variables. Pooled estimates with 95% CIs were estimated using inverse-variance weighted random-effects models with the DerSimonian and Laird estimators for interstudy variances (τ²). Continuity correction was used to calculate individual study results and conduct a meta-analysis based on the inverse-variance method. To quantify heterogeneity, I² statistics and τ² were assessed. Studies were considered heterogeneous if I² was >50% or p was <0.05 for τ². The results of the meta-analysis are displayed visually using forest plots (Figure 2A,B). All analyses were performed using RStudio (version 1.2.1335).

4.1 Use of VKAs in patients on hemodialysis

VKAs, such as warfarin, are commonly prescribed for patients with AF. However, the use of VKAs in dialysis patients is associated with an increased risk of bleeding. Previous studies investigating the effects of VKAs in this population have yielded conflicting results.^{10, 14, 17, 23-28} Some studies suggested a reduced risk of stroke¹⁴ or a survival benefit^{23, 29, 30} with VKA use compared to no oral anticoagulant therapy, while other studies found no protective effect of VKAs on stroke risk or all-cause mortality.^{17, 25-28, 31}

A meta-analysis involving more than 9000 dialysis patients with AF found that VKA treatment was associated with a 1.2-fold increased risk of stroke (95% CI, 0.8–1.9).²⁴ Additionally, several studies have reported an increased risk of bleeding with VKA use.^{28, 32} Contributing factors such as pre-existing platelet dysfunction, routine heparinization during hemodialysis, and suboptimal time in therapeutic range (TTR) may exacerbate this bleeding risk.^{33, 34} In the current RCTs included in our analysis, the mean TTR in the VKA groups was consistently low, ranging from 44% to 50.7% (AXADIA-AFNET 8 study, RENAL-AF trial, and Valkyrie study), with patients being three times more likely to be subtherapeutic (international normalized ratio [INR] <2.0) than supratherapeutic (INR >3.0). The efficacy of warfarin may be compromised by a low TTR, and bleeding risks could increase if conventional TTR targets were achieved. Nonetheless, the low TTR observed in these trials also underscores the potential advantage of NOACs in this population.

4.2 Use of NOACs in patients on dialysis

In a large retrospective cohort study that included over 25,000 patients from the United States Renal Data System, standard-dose apixaban (5 mg twice daily) was associated with significantly lower rates of stroke, systemic embolism, and mortality compared to both warfarin and low-dose apixaban (2.5 mg twice daily). Furthermore, both dosing regimens of apixaban (5 mg or 2.5 mg twice daily) demonstrated lower rates of major bleeding compared to warfarin, with the standard-dose apixaban (5 mg twice daily) also being associated with reduced rates of thromboembolic events and mortality.

A recently published meta-analysis, which included data from three RCTs and three observational studies, reported comparable outcomes between patients treated with NOACs and those treated with VKAs. The rates of ischemic stroke or systemic embolism (relative risk [RR]: 0.65, 95% confidence interval [CI]: 0.38–1.10), major bleeding events (RR: 0.79, 95% CI: 0.49–1.28), and all-cause mortality (RR: 0.79, 95% CI: 0.56–1.12) were similar between the two groups. As compared with our study, in addition to ischemic stroke and bleeding events, we further reported the principal efficacy and safety endpoints, cardiovascular events including acute coronary syndrome and cardiovascular death. Moreover, we demonstrated the high number of deaths and bleeding events relative to ischemic stroke events in the result of pooled NOAC and VKA study groups, which raises the need for more holistic and integrated AF management in hemodialysis patients, such as shared decision making in the prescription of anticoagulants.

The current meta-analysis demonstrated that NOACs provide efficacy and safety outcomes comparable to warfarin. However, interstudy heterogeneity was observed in the analysis of trial-defined efficacy endpoints, largely driven by the reduced efficacy outcomes reported in the Valkyrie study, which included fatal cardiovascular disease, non-fatal stroke, cardiac events, and other vascular events as principal efficacy endpoints. In contrast, the AXADIA-AFNET 8 study defined efficacy endpoints as thromboembolic events, including myocardial infarction, ischemic stroke, all-cause mortality, and deep vein thrombosis and/or pulmonary embolism. Since the RENAL-AF trial was not designed to evaluate efficacy, the current analysis used all-cause mortality as a surrogate endpoint.

4.3 Recommendations of guidelines about oral anticoagulants in AF patients on hemodialysis

The JCS/JHRS 2024 Guideline Focused Update on Management of Cardiac Arrhythmias stated that warfarin is not recommended for patients on dialysis (class III recommendation).³⁵ Also, the prior JCS/JHRS 2020 Guideline on Pharmacotherapy of Cardiac Arrhythmias stated that NOACs are contraindicated in patients on dialysis, and so is warfarin except in some cases, such as the perioperative period of AF ablation, mechanical valves, and for secondary stroke prevention.³⁶ The “weak” recommendations of routine use of OACs (either warfarin or NOACs) in AF patients on dialysis are mainly due to the lack of high-quality trials demonstrating the positive net clinical benefits of OACs compared to non-anticoagulation in this population at a high bleeding risk. In 2023 ACC/AHA/ACCP/HRS Guideline for the Diagnosis and Management of Atrial Fibrillation, the use of warfarin or apixaban in AF patients on dialysis was a class IIb recommendation.³⁷ Apixaban was listed based on the results of some retrospective analyses of databases and the trials of apixaban compared to warfarin, rather than any randomized trials of apixaban compared to non-anticoagulation. This issue would remain controversial until the randomized trials are available which directly compare NOACs to warfarin in AF patients on hemodialysis. Based on current guideline recommendations and the findings of the present study, NOACs may be considered only for secondary stroke prevention in AF patients on hemodialysis (as suggested by the 2020 JCS/JHRS guidelines), with apixaban and rivaroxaban being the preferred choices among NOACs—and likely preferred over warfarin—once the decision to initiate anticoagulation is made.

4.4 Holistic management of AF patients with kidney failure receiving hemodialysis

Among patients across three RCTs, 144 individuals (37.6%) died, 61 (15.9%) experienced a major bleeding event, 88 (23%) suffered a CRNB event, and 17 patients (4.4%) experienced an ischemic stroke (Figure 4). A recently published meta-analysis further corroborated these findings, highlighting a high incidence of both minor and major bleeding events as well as all-cause mortality.³⁸ The substantial burden of deaths and bleeding complications compared to ischemic stroke events underscores the necessity of a more comprehensive and integrated approach to AF management, particularly for patients with AF and kidney failure undergoing hemodialysis. In this context, the ABC (Atrial fibrillation Better Care) pathway has been proposed as an integrated management strategy aiming at reducing AF-related mortality, morbidity, and hospitalizations.³⁹

4.5 Study limitation

We included three RCTs in our analysis: the VALKYRIE study, the RENAL-AF trial, and the AXADIA-AFNET 8 study. There were several limitations of the present study. First, the limited number of patients, which may lead to an underpowered analysis, is a major limitation of our study. However, only these three RCTs are currently available in the literature for a meta-analysis assessing the efficacy and safety of NOACs versus VKAs in AF patients on hemodialysis. Therefore, despite the small sample size, we believe the findings remain important, as this is a clinical issue we frequently encounter in daily practice. At the very least, the available data can help inform our discussions with patients and support shared decision making. Second, apixaban and rivaroxaban are distinct drugs with differing pharmacokinetics, administration methods, and clinical indications, so these differences might have influenced the pooled results. However, separate analysis would be difficult to provide more information due to the limited enrolled trials and patient number (1 rivaroxaban trial and 2 apixaban trials). Furthermore, the consistently low time in the therapeutic range observed in these trials reflects the real-world quality of VKA therapy in this patient population. To address these limitations, a meta-analysis was conducted to integrate findings across the studies, offering a more comprehensive perspective that may inform future clinical care.