Effect of digoxin on all-cause and cardiovascular mortality in patients with atrial fibrillation with and without heart failure: an umbrella review of systematic reviews and 12 meta-analyses

Purpose To perform a systematic umbrella review with meta-analysis to evaluate the certainty of evidence on mortality risk associated with digoxin use in patients with atrial fibrillation (AF) with or without heart failure (HF). Methods We systematically searched MEDLINE, Embase, and Web of Science databases from inception to 19 October 2021. We included systematic reviews and meta-analyses of observational studies investigating digoxin effects on mortality of adult patients with AF and/or HF. The primary outcome was all-cause mortality; secondary outcome was cardiovascular mortality. Certainty of evidence was evaluated by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) tool and the quality of systematic reviews/meta-analyses by the A MeaSurement Tool to Assess systematic Reviews 2 (AMSTAR2) tool. Results Eleven studies accounting for 12 meta-analyses were included with a total of 4,586,515 patients. AMSTAR2 analysis showed a high quality in 1, moderate in 5, low in 2, and critically low in 3 studies. Digoxin was associated with an increased all-cause mortality (hazard ratio [HR] 1.19, 95% confidence interval [95%CI] 1.14–1.25) with moderate certainty of evidence and with an increased cardiovascular mortality (HR 1.19, 95%CI 1.06–1.33) with moderate certainty of evidence. Subgroup analysis showed that digoxin was associated with all-cause mortality both in patients with AF alone (HR 1.23, 95%CI 1.19–1.28) and in those with AF and HF (HR 1.14, 95%CI 1.12–1.16). Conclusion Data from this umbrella review suggests that digoxin use is associated with a moderate increased risk of all-cause and cardiovascular mortality in AF patients regardless of the presence of HF. Trial registration This review was registered in PROSPERO (CRD42022325321). Supplementary Information The online version contains supplementary material available at 10.1007/s00228-023-03470-y.


Introduction
The management of patients suffering from atrial fibrillation (AF) is multifactorial including thromboprophylaxis for cardioembolic stroke by anticoagulant treatment, symptoms management, and rate and rhythm control by anti-arrhythmic drugs [1,2]. Indeed, beyond anticoagulation therapy, rhythm and rate control strategies are cornerstone for the acute and chronic management of patients with AF [3]. Among anti-arrhythmic drugs, digoxin is a still widely used drug to control heart rate in AF patients. The 2020 guidelines from the European Society of Cardiology (ESC) recommend beta-blockers and/or digoxin to control heart rate in AF patients with left ventricular ejection fraction < 40% (class I level of evidence B) [3]. In addition, the ESC guidelines recommend the long-term use of digoxin in patients in whom an adequate rate control cannot be achieved by beta blockers at maximum tolerated dose or when beta-blockers are contraindicated or not tolerated with low class of evidence (IIa) [3].
Digoxin is also recommended by the "2021 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure" for the treatment of patients with heart failure (HF) and sinus rhythm to reduce the risk of hospitalization and symptoms burden (class IIb level B) [4]. A similar recommendation is provided by the 2022 AHA/ACC/ HFSA Guideline for the Management of Heart Failure [5].
During the last decades, concerns regarding the safety of digoxin have been raised. In particular, some studies showed an increased risk of death in patients with AF treated with digoxin, especially when supratherapeutic blood concentrations are reached [6,7]. Given the possibility of the presence of an indication bias (i.e., administration of digoxin to sicker patients), also propensity-matched studies have been performed providing divergent conclusions [8,9].
However, several systematic reviews and meta-analyses investigating the association of digoxin with all-cause and cardiovascular mortality reported conflicting evidence [10,11].
Given the impossibility of obtaining data from a randomized trial testing the safety and efficacy of digoxin in addition to standard therapy, we performed an umbrella review of systematic reviews and meta-analyses, focusing on patients with AF with and without HF, in whom the use of digoxin has recommendation by international guidelines.

Methods
This review was registered in PROSPERO (CRD42022325321). This umbrella review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [12]. (Supplementary Fig. 1).

Data sources and searches
A systematic literature search was conducted by two independent authors on MEDLINE, Embase, and Web of Science databases from inception to 19 October 2021. Keywords used to perform the search were "digoxin", "atrial fibrillation", "mortality", and "meta-analysis" combined with Boolean operators were used to find articles. Search strategy was adapted for each database; a complete list of search strings is available in Supplementary Material 1.

Study selection
Criteria of inclusion were defined as follows: Meta-analysis of studies investigating digoxin effects compared to standard of care on mortality of patients with AF; patients included in studies must be at least 18 years old; effect sizes must be provided as relative risk (RR) or hazard ratio (HR) with 95% confidence intervals (95%CI). Only articles with full-text available were considered. No restrictions were placed on language or publication date. Meta-analyses which did not report data concerning mortality were excluded. Multiple meta-analyses reported in a single paper (e.g., multiple outcomes or based on different types of studies) were included separately.
Study selection was performed by two independent authors, and disagreements were resolved through discussion with the senior author. Titles and abstracts of each article were screened to remove duplicates, and full texts of promising articles were read to assess eligibility. Reference lists of eligible articles were hand-searched to identify additional relevant meta-analyses.

Data extraction
Two investigators independently extracted the following data from each eligible study: name of first author; year of publication; outcomes; databases whose searches were based on; period of time searched; number and type of included studies; followup period; digoxin indication; number of patients with AF; overall mortality; cardiovascular mortality; mortality in patients with only AF; and mortality in patients with AF and HF.

Outcomes
All-cause mortality was the primary outcome. Secondary endpoint was cardiovascular mortality. A subgroup analysis in patients with AF alone or AF and HF was performed.

Quality evaluation and risk of bias assessment
Quality of each included meta-analysis was evaluated with the assessment of multiple systematic reviews (AMSTAR) 2 tool [13]. This tool aims at evaluating systematic reviews quality by answering "no", "partial yes", or "yes" to 16 different items. Items 4,9,11,12, and 15 are considered critical domains. The quality of studies was defined as follows: high (no or 1 non-critical weakness), moderate (more than 1 noncritical weakness), low (1 critical flaw with or without noncritical weaknesses), or critically low (more than one critical flaw with or without non-critical weaknesses) ( Table 1).

Data synthesis and analysis
Meta-analyses for each endpoint separately were performed based on random effects, using the logarithm of hazard ratios (HRs) associated as outcome. Inverse variance weights were used in all cases. Pooled effects were obtained through maximum likelihood.
Heterogeneity was evaluated by calculating the I 2 index. According to arbitrary cut-offs, low, moderate, and high heterogeneity was defined as an I 2 of < 25%, 25-75%, and > 75%, respectively.
Publication bias was assessed for studies reporting outcomes according to digoxin use, with the use of funnel plots. Egger's test was also performed.
Analyses were performed using the R (R Development Core Team) software version 3.6.1; statistical significance level was set at 0.05, and all p values were two-tailed.

Certainty of evidence
Two authors independently evaluated certainty of evidence using GRADEproGDT according to the GRADE Handbook [14]. GRADE categorizes certainty of evidence into very low, low, moderate, or high; the higher the category, the greater the confidence that the true effect is close to the reported findings. The following characteristics are considered in order to assess the right category: design of study (observational, randomized clinical trial), inconsistency across studies (I 2 statistics), imprecision of the findings, indirectness (e.g., due to mixed outcome), publication bias, size effect, and presence of dose-response gradient.

Ethics approval
This is an umbrella review with meta-analysis. No ethical approval is required.

Results
Three hundred ninety-seven articles were obtained from the initial search. After duplicates removal, 350 papers were evaluated. After a first screening, 39 studies were eligible for detailed analysis, but 5 had no full-text available. Finally, 11 studies met the eligibility criteria and were included in the umbrella review. The strategy search is summarized in PRISMA flow diagram ( Supplementary Fig. 1). Table 1 shows AMSTAR 2 items evaluation for every included meta-analysis. Overall, out of the 11 papers, 1 had a high, 5 moderate, 2 low, and 3 critically low quality. In particular, as far as it regards critical domains, each paper reported a proper use of comprehensive literature search queries and a significant publication bias assessment (Q4, 11/11, and Q15, 11/11, respectively). Almost every paper used an appropriate method for statistical combination of the results (Q11, 10/11), while some critical issues were found in the evaluation of technique for assessing risk of bias and in determining its implication on the results of the metaanalysis (Q9 8/11 and Q12 6/11, respectively). Table 1 AMSTAR2 tool evaluation of the quality of the included systematic reviews/meta-analyses  Table 2. Table 3 summarizes meta-analysis characteristics. Each study reported synthesis of results expressed as HR or RR. One paper (Sethi et al.) included only RCTs, while the others included observational studies and data from registries as well. A total of 4,586,515 patients were included. The length of follow-up ranged from 0.4 to 4.7 years. Notably, Ziff and colleagues [11] reported data on all-cause mortality and cardiovascular mortality separately for observational and interventional studies; given the different populations in the two analyses, we considered them as two different studies.
Funnel plots reported in the Supplementary Fig. 2 did not show significant publication bias.

All-cause and cardiovascular mortality
All the included studies reported data on the primary outcome, the all-cause mortality. Figure 1 shows the results of our umbrella review concerning mortality. Digoxin was associated with an increased mortality in the overall population (HR 1.19, 95%CI 1.14-1.25, panel A) with moderate certainty of evidence according to the GRADE (Table 4) and moderate-high heterogeneity (I 2 75.8%).
Data on cardiovascular mortality were provided by 5 studies (Fig. 2). Overall, the evidence suggests that digoxin might result in an increase in cardiovascular mortality (HR 1.19, 95%CI 1.06-1.33) with moderate certainty of evidence according to the GRADE (Table 4) and moderate heterogeneity (I 2 70.5%).

Subgroup analysis according to the HF
As far as it concerns mortality in AF-only population (Fig. 1B), only 8 papers provided data concerning this outcome. Our analysis shows that digoxin may result in an increase in mortality in this group of patients (HR 1.23, 95%CI 1.19-1.28) with low certainty of evidence according to the GRADE (Table 4) and moderate heterogeneity (I 2 70.8%).
Mortality in patients affected by AF and HF outcome (Fig. 1C) was explored by 8 papers and provided similar results. Even in this population, digoxin was associated with an increase in mortality (HR 1.14, 95%CI 1.12-1.16) with moderate certainty of evidence according to the GRADE (Table 4) and no heterogeneity (I 2 0%).

Discussion
Results from this umbrella review of meta-analyses indicate that the use of digoxin may be associated with an increased risk of all-cause and cardiovascular mortality in patients with AF.
The need for this umbrella review and meta-analysis came from literature analysis in which a growing number of observational studies reported a potential harmful effect of digoxin in AF patients [15,16]. However, this evidence became conflicting after the publication of some meta-analyses providing discordant results. For this reason, we adopted the methodology of the umbrella review, which represents one of the highest levels of evidence synthesis currently available [17], to provide more robust data on the association between digoxin and mortality in patients with AF, given the lack of data from a controlled randomized setting. Our analysis indicates an association of digoxin use with all-cause and cardiovascular mortality in patients with AF.
Notably, the association with all-cause mortality persisted in the subgroup of patients with AF and HF, even if with a lower strength of association. The use of digoxin in AF patients with HF is well established in clinical practice, but it should be noted that consolidated evidence showed that the effect of digoxin may not be so evident when a stable haemodynamic has been already reached with other drugs such as diuretics and vasodilators [18], and that digoxin may work less when an activation of sympathetic system is present (e.g., acute decompensated HF) [19]. Thus, HF should not represent per se an indication to the use of digoxin as possible harmful effects are also evident in this subgroup of patients. One of the arguments for the still wide use of digoxin is its ability to reduce the rate of hospitalization and the improvement of symptoms in patients treated with this drug [20]. However, more recently, the TREAT-AF trial included patients with permanent AF and HF randomized to receive digoxin or bisoprolol [21]. This study showed no difference between the two treatments group regarding symptoms after 6 months of therapy [21].
Strengths of the study are that it is the first systematic umbrella review of evidence from meta-analyses including a large sample of patients; even if a number of patients may be counted more than once given the design of umbrella review, it still remains the largest number of subjects considered to our knowledge. Furthermore, we also performed an accurate quality evaluation, certainty of evidence analysis, and risk of bias.
The different results obtained from the present umbrella review and meta-analysis in comparison to other previously published meta-analyses may rely on several reasons including different selection of studies, definition of outcomes variable length of follow-up, and lack of quality evaluation of evidence. In addition, 3 previous meta-analyses had critically low quality, and 2 had low quality at AMSTAR evaluation.
Clinical implications of our results are relevant considering that a high number of patients are currently treated with digoxin worldwide. Clinicians should be aware that digoxin may be harmful in AF patients, especially in some specific settings such as in chronic kidney disease or electrolyte Wang et al. [33] 2021 Pubmed, Embase, Cochrane library September 2020 -FU > 6 months -RCT, observational, retrospective analysis -Digoxin compared with no digoxin or other heart rate control treatment in patients with AF -HRs with 95%CIs for outcomes associated with digoxin treatment were reported -Length of FU ≥ 6 months -All-cause mortality was the endpoint imbalance, both conditions increasing the risk of adverse effects. In addition, patients prescribed on digoxin should be adequately informed about the potential side effects and the need of regular medical and laboratory follow-up while taking this medication.
Our results indicate that digoxin should be considered only in patients who do not achieve an adequate rate control or who experience symptoms with other anti-arrhythmic drugs. In addition, digoxin may be considered in patients with contraindication to the use of beta blockers (e.g., pulmonary disease) or to the use of calcium channel antagonists (such as heart failure). The use of laboratory monitoring and careful electrocardiographic examination may help recognize the early signs of digoxin toxicity, allowing a prompt intervention to reduce the risk of mortality associated with supra-therapeutic values of plasma digoxin. Indeed, values exceeding the therapeutic range may result in an increased risk of pro-thrombotic [22] and pro-arrhythmogenic effect and in an increased endothelial platelet activation [19,22,23], all mechanisms leading to an increased risk of cardiovascular mortality.
There are limitations of this analysis to acknowledge. First, despite the umbrella review approach provides robust evidence regarding the association between digoxin and mortality, the inclusion of observational studies carries some intrinsic limitations, mainly due to the impossibility of eliminating the bias by indication, which implicates that patients prescribed on digoxin may be sicker than those not treated with this drug despite the multivariable adjustment for the most common comorbidities [24]. However, it should be noted that subgroup analysis of propensity-matched populations provided similar results [11]. However, what cannot be deduced from clinical studies is the reason for mortality, so we do not know if toxicity, arrhythmia, and HF were the causes of death. Indeed, data on serum digoxin concentration, renal function, acute coronary syndrome, potassium levels may provide important additional information to understand the association of digoxin with clinical outcomes. Furthermore, we do not know if patients were adequately followed after digoxin prescription.
In conclusion, despite its wide use, the use of digoxin should be considered with caution in patients with AF and should be reserved to those patients in whom an adequate rate control is difficult to achieve with other anti-arrhythmic drugs.
Author contribution All authors contributed to the study conception and design. Material preparation, literature search, and data analysis were performed by Gianluca Gazzaniga, Daniele Pastori, Danilo Menichelli, and Alessio Farcomeni. The first draft of the manuscript was written by Daniele Pastori and Gianluca Gazzaniga, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding Open access funding provided by Università degli Studi di Roma La Sapienza within the CRUI-CARE Agreement.

Competing interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source,