Introduction

Atrial fibrillation (AF) is regarded as the most common arrhythmia with an increasing risk of death and morbidity [1]. Heart failure (HF) is a highly complex clinical syndrome with a prevalence that increases with age. They often coexist in clinical practice, and HF is an independent risk factor of thromboembolic complications among AF patients. As such, HF is incorporated into the CHA2DS2-VASc score (congestive heart failure, hypertension, age 65–74 years, diabetes mellitus, vascular disease (prior myocardial infarction, peripheral artery disease, or aortic plaque), female (1 point each), age ≥ 75 years, and prior stroke/transient ischemic attack/thromboembolism (doubled)) for stroke prediction in AF [2]. More recently, several studies have found that HF patients are at increased risks of thromboembolic events regardless of the presence of AF [3, 4]. Prior randomized clinical trials (RCTs) have found that non-vitamin K antagonist oral anticoagulants (NOACs) are at least as effective as warfarin and could have a better safety profile for stroke prediction in patients with AF [5,6,7,8]. Current guidelines consistently recommend NOACs as the first choice of drugs in AF patients [1, 9, 10]. A number of studies have investigated the efficacy and safety outcomes of NOACs versus warfarin in patients with AF and HF [11,12,13,14].

Two previous systemic reviews including four RCTs have proposed that the NOACs is at least not associated with increased risks of stroke or systemic embolism, major bleeding, and intracranial hemorrhage in HF patients with AF [15, 16]. Nevertheless, the populations in RCTs are generally selected with strict selection criteria, which are sometimes different from the real-world practice [17]. In recent years, several observational studies have investigated the role of NOACs in patients with AF and HF. Therefore, we conducted a meta-analysis to re-assess the efficacy and safety outcomes of NOACs versus warfarin in patients with AF and HF.

Methods

Study search

We systematically searched the PubMed database from inception to December 2019 for studies that reported the efficacy and safety outcomes of any NOAC with warfarin in patients with AF and HF. The following search terms (and their similar terms) were used: atrial fibrillation, heart failure, non-vitamin K antagonist oral anticoagulants, direct oral anticoagulants, dabigatran, rivaroxaban, apixaban, edoxaban, vitamin K antagonists, and warfarin. We also screened the reference lists of the retrieved studies in order to find the additional studies. No language restrictions were applied during the search in this meta-analysis.

Eligibility criteria

Studies could be included if they met the criteria: (1) RCTs or observational studies that compared the effect of NOACs versus warfarin among patients with AF and HF; (2) any NOAC (dabigatran, rivaroxaban, apixaban, or edoxaban; any dose) versus warfarin; (3) studies reporting at least one of the efficacy or safety outcomes. Efficacy outcomes included stroke or systemic embolism, ischemic stroke, hemorrhagic stroke, and all-cause death, while safety outcomes included major bleeding, intracranial hemorrhage and gastrointestinal bleeding; and (4) effect estimates of the study was adjusted risk ratios (RRs) and its 95% confidence intervals (CIs). If the substantial overlap was found among the different studies, we only included the study with the longest follow-up or largest sample size.

Study selection and data abstraction

We first read the titles and abstracts of the retrieved studies to screen out the available studies. The full texts of these eligible studies were then reviewed in more detail. Two independent reviewers screened all of the retrieved studies in the search. Discrepancies were resolved by consensus. In addition, we abstracted the baseline characteristics of the included studies, such as the first author and publication year, type of study, definitions of AF or HF, number of NOACs or warfarin users, type or dose of NOACs, follow-up duration, and outcomes.

Risk of bias assessment

The methodological quality of the RCTs was evaluated according to the Cochrane risk of bias assessment tool [18]. The quality of the observational studies was evaluated using the Newcastle-Ottawa score, which involved the selection of cohorts, comparability of cohorts, and assessment of the outcome [19]. A Newcastle-Ottawa score of < 6 points indicated a low quality [20].

Statistical method

All statistical analyses were performed using Review Manager version 5.30 software (the Cochrane Collaboration, Copenhagen, Denmark). We applied the Cochrane Q test combined with the I [2] values to assess the heterogeneity across the included studies. P < 0.1 and I2 > 50% indicated a significant heterogeneity, respectively. For each study, we calculated the natural logarithm of the RR (Ln[RR]) and its corresponding standard error (SELn[RR]). And then, Ln[RR] and SELn[RR] were pooled using a random-effects model weighted by the inverse-variance method. In addition, we re-analyzed these analyses with a fixed-effects model in the sensitivity analysis. The subgroup analysis was not performed based on the design of study (RCTs versus observational studies). The possible presence of publication bias was checked by observing the symmetry characteristics of the funnel plots.

Results

Study selection

As shown in Supplemental Fig. 1, a total of nine studies (four RCTs [11,12,13,14] and five observational studies [21,22,23,24,25]) were included in the present meta-analysis. Thereinto, four RCTs were from the sub-analyses of the RE-LY (dabigatran), ROCKET AF (rivaroxaban), ARISTOTLE (apixaban), and ENGAGE AF-TIMI 48 (edoxaban) trials, respectively. Baseline characteristics of the included RCTs are shown in Supplemental Table 1. All of the four RCTs had a low risk of bias, whereas all of the five observational studies had moderate-to-high quality.

Efficacy and safety of NOACs versus warfarin

Efficacy

Compared with warfarin use, the use of NOACs was significantly associated with reduced risks of stroke or systemic embolism (RR = 0.82 (95% CI, 0.73–0.92); P = 0.001; I2 = 35%; Fig. 1) and all-cause death (RR = 0.87 (95% CI, 0.80–0.94); P = 0.0007; I2 = 64%; Fig. 2). There was no difference in the risk of ischemic stroke (RR = 0.89 (95% CI, 0.75–1.04); P = 0.15; I2 = 28%; Supplemental Fig. 2) in patients with NOACs versus warfarin.

Fig. 1
figure 1

Random-effects model for comparing the stroke or systemic embolism of NOACs with warfarin in patients with AF and HF. Abbreviations: AF, atrial fibrillation; HF, heart failure; NOACs, non-vitamin K antagonist oral anticoagulants; DA, dabigatran; RIV, rivaroxaban; API, apixaban; EDO, edoxaban; LVSD, left ventricular systolic dysfunction; HFpEF, heart failure patients with preserved ejection fraction; CI, confidence interval; SE, standard error; IV, inverse of the variance

Full size image
Fig. 2
figure 2

Random-effects model for comparing the all-cause death of NOACs with warfarin in patients with AF and HF. Abbreviations: AF, atrial fibrillation; HF, heart failure; NOACs, non-vitamin K antagonist oral anticoagulants; DA, dabigatran; RIV, rivaroxaban; API, apixaban; EDO, edoxaban; LVSD, left ventricular systolic dysfunction; HFpEF, heart failure patients with preserved ejection fraction; CI, confidence interval; SE, standard error; IV, inverse of the variance

Full size image

Safety

Compared with warfarin use, the use of NOACs was significantly associated with decreased risks of major bleeding (RR = 0.84 (95% CI, 0.74–0.97); P = 0.01; I2 = 82%; Fig. 3), intracranial hemorrhage (RR = 0.50 (95% CI, 0.43–0.59); P < 0.00001; I2 = 0%; Fig. 4), and hemorrhagic stroke (RR = 0.49 (95% CI, 0.38–0.63); P < 0.00001; I2 = 0%; Supplemental Fig. 3). We found no difference in the risk of gastrointestinal bleeding (RR = 1.11 (95% CI, 0.79–1.55); P = 0.54; I2 = 89%; Supplemental Fig. 4) in patients with NOACs versus warfarin.

Fig. 3
figure 3

Random-effects model for comparing the major bleeding of NOACs with warfarin in patients with AF and HF. Abbreviations: AF, atrial fibrillation; HF, heart failure; NOACs, non-vitamin K antagonist oral anticoagulants; DA, dabigatran; RIV, rivaroxaban; API, apixaban; EDO, edoxaban; LVSD, left ventricular systolic dysfunction; HFpEF, heart failure patients with preserved ejection fraction; CI, confidence interval; SE, standard error; IV, inverse of the variance

Full size image
Fig. 4
figure 4

Random-effects model for comparing the intracranial hemorrhage of NOACs with warfarin in patients with AF and HF. Abbreviations: AF, atrial fibrillation; HF heart failure; NOACs, non-vitamin K antagonist oral anticoagulants; DA, dabigatran; RIV, rivaroxaban; API, apixaban; EDO, edoxaban; LVSD, left ventricular systolic dysfunction; HFpEF, heart failure patients with preserved ejection fraction; CI, confidence interval; SE, standard error; IV, inverse of the variance

Full size image

Sensitivity analysis and subgroup analysis

For the efficacy and safety outcomes of NOACs versus warfarin, a fixed-effects model analysis produced the similar results with the aforementioned analyses. In subgroup analysis on the design of study, we found no significant interaction between data of RCTs versus observational studies for the outcomes of stroke or systemic embolism, ischemic stroke, hemorrhagic stroke, major bleeding, intracranial hemorrhage and gastrointestinal bleeding (all Pinteraction > 0.05). However, the use of NOACs versus warfarin decreased the risk of all-cause death in observational studies (RR = 0.80 (95% CI, 0.72–0.88); P < 0.0001; I2 = 67%), but not in RCTs (RR = 0.95 (95% CI, 0.88–1.03); P = 0.19; I2 = 0%) (Pinteraction = 0.009; Fig. 2).

Publication Bias

The publication biases assessed by the funnel plots of the reported efficacy and safety outcomes are shown in Supplemental Figs. 5, 6, 7, 8, 9, 10, and 11.

Discussion

In the present analysis, we combined the data of RCTs and observational studies to compare the efficacy and safety outcomes of NOACs with warfarin in patients with AF and HF. Our pooled data indicated that [1] compared with warfarin use, the use of NOACs was associated with the reduced risks of stroke or systemic embolism, all-cause death, major bleeding, intracranial hemorrhage, and hemorrhagic stroke [2], there were no differences in the risks of ischemic stroke and gastrointestinal bleeding in patients with NOACs versus warfarin. Re-analyses with a fixed-effects model produced the similar results with the main analyses.

The populations in RCTs are generally selected with strict selection criteria, which are sometimes different from the real-world settings. Our current meta-analysis included the real-world data adding to an already existing meta-analysis including four RCTs. We found no significant interactions between data of RCTs versus observational studies for the efficacy and safety outcomes except all-cause death. As such, our findings suggest that NOACs had similar or lower risks of thromboembolic and bleeding events compared with warfarin in patients with AF and HF. NOACs might be reasonable alternatives to warfarin in patients with AF and HF. The sub-type of heart failure is important and should be included in the further study.

Previously, data from the Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) trial [26] has indicated that in patients with stable coronary artery disease and sinus rhythm, the use of low-dose rivaroxaban (2.5 mg twice daily) plus aspirin might decrease the risk of major vascular events compared with the use of aspirin alone. In addition, data from the Cardiovascular Outcome Modification, Measurement AND Evaluation of Rivaroxaban in patients with Heart Failure (COMMANDER-HF) trial [27] has found that in HF patients without AF, low-dose rivaroxaban (2.5 mg twice daily) is associated with a reduced risk of stroke. A prior meta-analysis including 9490 patients with HF and sinus rhythm has suggested that anticoagulant therapy could decrease the stroke risk but increase the risk of major bleeding [28]. More recently, data from the observational studies suggest that anticoagulant therapy could help decrease the risk of stroke induced by HF [29]. On the other hand, HF patients with the anticoagulant therapy might be also at a high risk of bleeding [30]. However, no studies have been performed to observe the effect of NOACs versus warfarin in patients with with HF and sinus rhythm. HF patients without AF are at increased risks of stroke or death [4, 31]. Based on our current data, the use of NOACs was at least non-inferior to warfarin for stroke prevention in patients with AF and HF.

Limitations

The following limitations should be acknowledged in our current meta-analysis. First, the subgroup analyses based on the type or dosage of NOACs were not performed due to the limiting data. Second, we did not take the time in the therapeutic range of warfarin users into consideration. Third, the residual confounders (e.g., age, baseline chronic conditions) in the real-world studies might exist. Fourth, the different definitions of the studied outcomes might affect our pooled data. Finally, we included both HF patients with reduced ejection fraction and those with preserved ejection fraction for analysis. Further study should include a sub-analysis between preserved and reduced ejection fraction as they represent totally different populations.

Conclusions

Compared with warfarin use, the use of NOACs had similar or lower risks of thromboembolic and bleeding events in patients with AF and HF. The use of NOACs was at least non-inferior to warfarin for stroke prevention in patients with AF and HF.