Anticoagulation versus placebo for heart failure in sinus rhythm
Abstract
Background
People with chronic heart failure (HF) are at risk of thromboembolic events, including stroke, pulmonary embolism, and peripheral arterial embolism; coronary ischaemic events also contribute to the progression of HF. The use of long‐term oral anticoagulation is established in certain populations, including people with HF and atrial fibrillation (AF), but there is wide variation in the indications and use of oral anticoagulation in the broader HF population.
Objectives
To determine whether long‐term oral anticoagulation reduces total deaths and stroke in people with heart failure in sinus rhythm.
Search methods
We updated the searches in CENTRAL, MEDLINE, and Embase in March 2020. We screened reference lists of papers and abstracts from national and international cardiovascular meetings to identify unpublished studies. We contacted relevant authors to obtain further data. We did not apply any language restrictions.
Selection criteria
Randomised controlled trials (RCT) comparing oral anticoagulants with placebo or no treatment in adults with HF, with treatment duration of at least one month. We made inclusion decisions in duplicate, and resolved any disagreements between review authors by discussion, or a third party.
Data collection and analysis
Two review authors independently assessed trials for inclusion, and assessed the risks and benefits of antithrombotic therapy by calculating odds ratio (OR), accompanied by the 95% confidence intervals (CI).
Main results
We identified three RCTs (5498 participants). One RCT compared warfarin, aspirin, and no antithrombotic therapy, the second compared warfarin with placebo in participants with idiopathic dilated cardiomyopathy, and the third compared rivaroxaban with placebo in participants with HF and coronary artery disease.
We pooled data from the studies that compared warfarin with a placebo or no treatment. We are uncertain if there is an effect on all‐cause death (OR 0.66, 95% CI 0.36 to 1.18; 2 studies, 324 participants; low‐certainty evidence); warfarin may increase the risk of major bleeding events (OR 5.98, 95% CI 1.71 to 20.93; number needed to treat for an additional harmful outcome (NNTH) 17; 2 studies, 324 participants; low‐certainty evidence). None of the studies reported stroke as an individual outcome.
Rivaroxaban makes little to no difference to all‐cause death compared with placebo (OR 0.99, 95% CI 0.87 to 1.13; 1 study, 5022 participants; high‐certainty evidence). Rivaroxaban probably reduces the risk of stroke compared to placebo (OR 0.67, 95% CI 0.47 to 0.95; number needed to treat for an additional beneficial outcome (NNTB) 101; 1 study, 5022 participants; moderate‐certainty evidence), and probably increases the risk of major bleeding events (OR 1.65, 95% CI 1.17 to 2.33; NNTH 79; 1 study, 5008 participants; moderate‐certainty evidence).
Authors' conclusions
Based on the three RCTs, there is no evidence that oral anticoagulant therapy modifies mortality in people with HF in sinus rhythm. The evidence is uncertain if warfarin has any effect on all‐cause death compared to placebo or no treatment, but it may increase the risk of major bleeding events. There is no evidence of a difference in the effect of rivaroxaban on all‐cause death compared to placebo. It probably reduces the risk of stroke, but probably increases the risk of major bleedings.
The available evidence does not support the routine use of anticoagulation in people with HF who remain in sinus rhythm.
PICOs
Plain language summary
What is the evidence for the benefit and harms of anticoagulants for heart failure?
Background
When the heart’s ability to pump blood is reduced, it is called heart failure (HF). When this happens, some people will develop severe clotting problems (called thromboembolism) in the lungs, legs, and brain. This happens because the blood is flowing more slowly, inflammation is increasing, and there is an overproduction of clotting molecules. These clots can cause stroke, lung or leg damage, or even death.
Why is the question important?
There are strong blood‐thinning medications, called anticoagulants, which are successfully used in people with different clotting problems, such as those with HF who also have irregular heart beats (an arrhythmia called atrial fibrillation). At present, there are no data to recommend the use of anticoagulants to avoid clotting problems in people with HF who have regular heart beats (know as sinus rhythm). In this analysis, we assessed studies that tested anticoagulants in these people to avoid death, death from heart problems, and other severe clotting problems.
How did we identify and evaluate the evidence?
We searched electronic databases of studies (CENTRAL, MEDLINE, and Embase). We looked for randomised studies (like tossing a coin) that compared anticoagulant tablets with no treatment or a placebo (tablet with no drug inside) in people with HF who were in sinus rhythm.
What did we find?
We found three randomised trials, with 5498 participants, which we used for analysis.
What does this mean?
Based on the three randomised trials, we are uncertain about the risk of dying between people who took warfarin and those who did not. However, people who took warfarin may have been more likely to have episodes of major bleeding. The evidence suggested that there was no difference in the risk of dying for people who took rivaroxaban compared with those who did not. Rivaroxaban probably reduced the risk of stroke, but the risk of episodes of major bleeding was higher than in those who did not take rivaroxaban.
Our analysis does not support the routine use of anticoagulation in people with HF who remain in sinus rhythm.
How up‐to‐date is this review?
The literature is current to March 2020.
Authors' conclusions
Summary of findings
| Warfarin compared to placebo or no treatment in people with chronic heart failure in sinus rhythm | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with heart failure in sinus rhythm | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with placebo or no treatment | Risk with warfarin | |||||
| All‐cause death | 203 per 1000 | 144 per 1000 | OR 0.66 | 324 | Low certaintya | |
| Stroke | Stroke was not reported as an individual outcome. | |||||
| Major bleeding eventsc | 0/143 | 11/181 | OR 5.98 | 324 | Low certaintyb | 0 per 1000 in the control arm for major bleeding events. It is not based on the estimated relative effect. |
| *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded twice – once for imprecision, and once for inconsistency due to high heterogeneity cA major bleeding event is defined by the International Society on Thrombosis and Haemostasis as overt bleeding associated with a decrease in haemoglobin level of at least 2 g/dL, transfusion of two or more units of packed red cells or whole blood, bleeding in a critical site (intracranial, intraspinal, intraocular, pericardial, intraarticular, intramuscular with compartment syndrome, or retroperitoneal), or a fatal outcome. | ||||||
| Rivaroxaban compared to placebo in people with heart failure in sinus rhythm | ||||||
| Patient or population: people with heart failure in sinus rhythm | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with placebo | Risk with rivaroxaban | |||||
| All‐cause death | 227 per 1000 | 225 per 1000 | OR 0.99 | 5022 | High certainty |
|
| Stroke | 30 per 1000 | 20 per 1000 | OR 0.67 | 5022 | Moderate certaintya | |
| Major bleeding eventsb | 20 per 1000 | 32 per 1000 | OR 1.65 | 5008 | Moderate certaintya | |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded for imprecision, due to very low event rates. | ||||||
Background
Description of the condition
Chronic heart failure (HF) is a growing clinical and social problem. HF has an estimated prevalence of approximately 1% to 2% of the adults in developed countries, reaching ≥ 10% among people over 70 years of age (Ponikowski 2016). It is associated with high morbidity and annual mortality, with rates higher than 30% in people with severe symptoms (CONSENSUS 1987).
HF predisposes a person to stroke and thromboembolism, including pulmonary embolism and peripheral arterial embolism. These thromboembolic events contribute to the high morbidity in HF (Fuster 1981; Kyrle 1985). Ischaemic and thromboembolic events, particularly stroke, myocardial ischaemia, and myocardial infarction (MI), also contribute to the high hospital admission rates for people with HF (Brown 1998). The outcomes for people with HF remain unsatisfactory. The healthcare and mortality burden of HF are substantial. In the USA, the comorbid HF event rate increased between 2006 and 2014, which was also associated with rising healthcare costs (Jackson 2018).
According to the European Society of Cardiology (ESC) guidelines, HF comprises people with normal left ventricular ejection fraction (HF with preserved EF (HFpEF), when EF ≥ 50%; HF with mid‐range EF (HFmrEF), when EF = 40% to 49%; and HF with reduced EF (HFrEF), when EF < 40% (Ponikowski 2016)). The previous ESC guidelines divided people with HF into those with reduced EF (EF ≤ 35%), and preserved EF (EF ≥ 50% (McMurray 2012)). As discussed previously, HF predisposes people to thromboembolic events.
People hospitalised due to HF have an increased risk of venous thromboembolism (VTE) because of vascular abnormalities, increased coagulability, and impaired blood flow (Haas 2003). In cases of no thromboprophylaxis, the incidence of VTE ranges from 4% to 26% in people with decompensated HF. In one study, participants with more severe HF (defined by high N‐terminal pro‐B type natriuretic peptide plasma concentration) were at elevated risk of VTE (Mebazaa 2014). In people with new onset HF, the cumulative incidence of VTE was 1.4% at 30 days, 2.5% at 1 year, and 10.5% at 5 years (Smilowitz 2019).
Description of the intervention
Oral anticoagulants include vitamin K antagonists (VKA) and non‐vitamin K antagonist oral anticoagulants (NOAC). VKA inhibit the synthesis of the active forms of vitamin K‐dependent procoagulant proteins: prothrombin, factor VII, IX, and X, and protein C and S. Its narrow therapeutic interval, fluctuations of dose‐response relationship, and interactions with food and drugs for concomitant diseases, limit the use of VKA. Warfarin blocks vitamin K. NOACs include: dabigatran, rivaroxaban, apixaban, and edoxaban. Dabigatran is a direct thrombin inhibitor. Rivaroxaban, apixaban, and edoxaban act as factor Xa inhibitors. NOACs differ from VKA: they have rapid onset of action, fewer drug interactions, and their activity is not influenced by dietary vitamin K intake. They also have a predictable anticoagulant effect, and thus, they do not need routine coagulation monitoring.
HF has not been well defined in NOAC trials. There were numerous definitions based on reported symptoms, and occasionally, on left ventricular EF. People with HF may be prone to VKA instability and drug interactions, which happen to a lesser extent with NOACs. NOACs are an efficient and safe alternative to VKA in people with HF. The available data provide assurance that the relative risks and benefits of NOACs versus VKAs likely apply to people with HF. However, a more homogeneous definition of HF should be implemented in future trials. In people with HF and previous thromboembolism, newly diagnosed intracardiac thrombus, and right HF with concomitant pulmonary hypertension, anticoagulation may potentially be considered, but evidence is limited. The question whether to use oral anticoagulants (OAC) in people with HF and extensive infarct scar remains open, and needs further research.
There is evidence of benefit from long‐term oral anticoagulation in certain groups of people with HF. For instance, oral anticoagulation is extremely effective in reducing stroke and other embolic events in people with HF in atrial fibrillation (BAATAF 1990; Petersen 1989; SPAF 1991). The role of anticoagulation in the broader HF population is less well established. There is wide variation in the use of oral anticoagulants in people with HF (Edep 1997).
How the intervention might work
Oral anticoagulation has been associated with a reduction in the number of thromboembolic events in various cardiovascular disease states, but the potential risks of bleeding must also be considered. Importantly, the control of anticoagulation is reported to be more difficult, and bleeding complications more frequent in HF, as a result of hepatic congestion and potential drug interactions (Davis 1977; Husted 1976; Landefeld 1989).
In clinical practice, people with HF in sinus rhythm have a high risk of thromboembolism and bleeding. Because there are no clear guideline recommendations for people with HF in sinus rhythm, the therapeutic decisions tend to be based on a case‐by‐case assessment, during which the physician takes into account factors including age, aetiology of HF, risk of bleeding, severity of functional impairment, cardiac remodelling, and the risk of drug non‐adherence (Pozzoli 2020). Recent data indicate that CHA₂DS₂‐VASc and HAS‐BLED scores can be useful to predict thromboembolic and haemorrhagic events in those with HF in sinus rhythm (Melgaard 2015; Ye 2015).
Stroke risk is linked with bleeding risk. There is evidence that thromboembolic factors, like older age, hypertension, or history of stroke are also bleeding risk factors. The benefits of OAC in the elderly are evident, despite their co‐morbidities, frailty, and increased risk of falls (Annoni 2016; Deandrea 2013).
Before initiation of OAC therapy, risk of bleeding should be evaluated. Potentially modifiable and non‐modifiable risk factors should be identified. High risk of bleeding should not justify withholding OAC. Modifiable bleeding risk factors should be managed, and systematically reassessed during regular and frequent visits. It should be highlighted that the risk of bleeding is dynamic, and should be re‐evaluated systematically (Hindricks 2021). It should be emphasised that absolute contraindications to OAC are rare.
Why it is important to do this review
Various small and large studies have addressed the incidence of ischaemic and thromboembolic events, and the risk factors associated with a high thromboembolic risk, although the reported incidence of these events appears to vary between studies, depending on the study methodologies and populations. Nevertheless, as an example, mild to moderate HF was associated with an annual stroke risk of approximately 1.5% compared with an annual stroke risk in the general population of less than 0.5%, whilst the annual risk of stroke increases to almost 4% in people with severe HF (CONSENSUS 1987; PROMISE 1993; SOLVD 1998; V‐HeFT 1993).
The debate over the potential benefit of using oral anticoagulants in people with HF in sinus rhythm in clinical practice continues. The COMMANDER‐HF study added new evidence to this debate (COMMANDER HF 2018). Thrombin plays a role in the adverse interplay between endothelial dysfunction, stasis, and inflammation, which are present in HF. The COMMANDER‐HF trial examins an intervention to modulate thrombin‐mediated 'crosstalk', a potential trigger of numerous negative feedback cycles, involving inflammation, endothelial dysfunction, and thrombosis in people with HFrEF and coronary artery disease (Ferreira 2016).
Objectives
To determine whether long‐term oral anticoagulation reduces total deaths and stroke in people with heart failure in sinus rhythm.
Methods
Criteria for considering studies for this review
Types of studies
We included parallel‐arm randomised controlled trials (RCTs), which were reported in full text.
We did not include cluster‐randomised studies, as the analysis aimed to assess the ability of the treatment to improve individual outcomes. We did not include cross‐over studies, to enable collection of the outcomes after completion of the trial treatment or withdrawal from it. Given the chronic nature of heart failure (HF), we also excluded studies with short treatment duration (i.e. less than one month).
Types of participants
We included adults (≥ 18 years of age) with a diagnosis of HF. Participants with HF were defined clinically, and if possible, by a more objective assessment of the left ventricular systolic function (for example, echocardiography, radionuclide ventriculography, cardiac MRI).
We excluded trials with participants with the following co‐morbidities and characteristics: diagnoses other than HF as inclusion criteria; trials with a mixed population if the subgroup of participants with HF was not analysed and reported separately (to analyse treatment effects specific to HF); atrial fibrillation (AF) at randomisation (as AF is an established indication for oral anticoagulation in people with or without HF).
Types of interventions
We included trials comparing oral anticoagulation with placebo or no treatment. The oral anticoagulants could include vitamin K antagonists (VKA; for example, warfarin) and non‐vitamin K antagonist oral anticoagulants (NOAC; for example, dabigatran, apixaban, rivaroxaban, or edoxaban). We excluded studies with co‐interventions (for example, beta‐blockers, ACE inhibitors).
Types of outcome measures
The reporting of one or more of our outcomes of interest was not an inclusion criterion for the review. When a published report did not appear to report one of these outcomes, we accessed the trial protocol and contacted the trial authors to ascertain whether the outcomes were measured but not reported. We included relevant trials, which measured these outcomes but did not report the data at all, or not in a usable format, in the review as part of the narrative.
Primary outcomes
-
All‐cause death
-
Stroke
Secondary outcomes
-
Cardiovascular death (including sudden death)
-
Myocardial infarction
-
Pulmonary embolism
-
Peripheral arterial embolism
-
Major bleeding events (major bleeding was defined as fatal bleeding; symptomatic bleeding in a critical area or organ, such as intracranial, intraspinal, intraocular resulting in vision changes, retroperitoneal, intraarticular, pericardial, or intramuscular with compartment syndrome; bleeding causing a fall in haemoglobin level of 2 g/dL or more; bleeding leading to transfusion of two or more units of whole blood or red cells, or both (Schulman 2005)).
Search methods for identification of studies
Electronic searches
We identified trials through systematic searches of the following bibliographic databases:
-
Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 3) in the Cochrane Library (searched 20 March 2020);
-
Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, MEDLINE Daily, and MEDLINE Ovid (1946 to 19 March 2020);
-
Embase Ovid (1980 to week 11 2020).
We updated the 2005 searches (Appendix 1), in February 2010 (Appendix 2), June 2013 (Appendix 3), and March 2020 (Appendix 4).
We applied the Cochrane sensitivity‐maximising RCT filter to MEDLINE Ovid, and adapted it to use for Embase Ovid (Lefebvre 2019).
We searched all databases from their inception to the present, and imposed no restriction on language of publication or publication status. We translated relevant foreign language papers.
As records from Clinicaltrials.gov (www.clinicaltrials.gov) and the World Health Organization (WHO) International Clinical Trials Registry platform (ICTRP; apps.who.int/trialsearch/) are included in CENTRAL, we did not search these clinical trials registers separately.
We only considered adverse effects described in included studies.
Searching other resources
We checked reference lists of all included studies and relevant systematic reviews identified as part of the search, for additional references to trials. We also examined relevant retraction statements and errata for included studies.
Data collection and analysis
Updating this review
Over the course of the original review and updates, six authors (GYHL, IC, BJW, RP, ES, and MK) reviewed the inclusion criteria, the search strategies, the methodology criteria, and methods for pooling the data for this review.
Contacting authors
In the rare instances where the authors disagreed over the grading and inclusion of studies, recourse was made to an additional author. When resolving the disagreement was not possible, we added the article to those ‘awaiting assessment’, and contacted the trial authors for clarification. However, there was no disagreement during the current update.
Selection of studies
Two review authors (ES, MK) independently screened titles and abstracts to include all the potential studies we identified as a result of the search, and coded them as 'retrieve' (eligible or potentially eligible or unclear), or 'do not retrieve'. If there were any disagreements, they asked the third review author (GL) to arbitrate. We retrieved the full‐text study reports or publication, and two review authors independently screened the full text to identify studies for inclusion, and identify and record reasons for exclusion of the ineligible studies. We resolved any disagreement through discussion, or if required, we consulted a third person (GL). We identified and excluded duplicates, and collated multiple reports of the same study, so that each study, rather than each report, was the unit of interest in the review. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram and 'Characteristics of excluded studies' table (Liberati 2009).
Data extraction and management
We used a data collection form for study characteristics and outcome data, which was piloted on at least one study in the review. One review author (MK) extracted the following study characteristics from included studies.
-
Methods: study design, total duration of study, details of any 'run‐in' period, number of study centres and location, study setting, and date of study
-
Participants: N randomised, N lost to follow‐up or withdrawn, N analysed, mean age, age range, gender, aetiology of HF, left ventricular ejection fraction (LVEF), inclusion criteria, and exclusion criteria
-
Interventions: intervention, comparison, concomitant medications, and excluded medications
-
Outcomes: primary and secondary outcomes specified and collected, and time points reported
-
Notes: funding for trial, and notable conflicts of interest of trial authors
Two review authors (ES, MK) independently extracted outcome data from the included studies. We resolved disagreements by consensus, with a provision of involving a third person (GL), would it be required. One review author (MK) transferred data into the Review Manager 5 file (Review Manager 2020). We double‐checked that data were entered correctly by comparing the data presented in the review with the data extraction form. A second review author (ES) spot‐checked study characteristics for accuracy against the trial report.
Assessment of risk of bias in included studies
Two review authors (ES, MK) independently assessed risk of bias for each study, using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved any disagreements by discussion or by involving another author (GL). We assessed the risk of bias according to the following domains.
-
Random sequence generation
-
Allocation concealment
-
Blinding of participants and personnel
-
Blinding of outcome assessment
-
Incomplete outcome data
-
Selective outcome reporting
-
Other bias
We classified each potential source of bias as high, low, or unclear, and provided a quote from the study report, together with a justification for our judgment in the risk of bias table. We summarised the risk of bias judgements across different studies for each of the domains listed. When information on risk of bias was related to unpublished data or correspondence with a trialist, we noted this in the risk of bias table.
When considering treatment effects, we took into account the risk of bias for the studies that contributed to that outcome.
Measures of treatment effect
We analysed dichotomous data as Peto odds ratio. We did not include any continuous outcomes in the review.
We narratively described skewed data reported as medians and interquartile ranges.
Unit of analysis issues
The reviews included RCTs with parallel design. We did not include trials with multiple time points, or trials with more than two arms.
Dealing with missing data
We contacted investigators or study sponsors, to verify key study characteristics and obtain missing numerical outcome data, where possible (e.g. when a study was identified as an abstract only). When this was not possible, and we thought the missing data might introduce serious bias, we explored the impact of including such studies in the overall assessment of results with a sensitivity analysis.
Assessment of heterogeneity
We inspected forest plots visually to consider the direction and magnitude of effects, and the degree of overlap between confidence intervals. We used the I² statistic to measure heterogeneity among the trials in each analysis, but acknowledged that there was substantial uncertainty in the value of I² when there was only a small number of studies. We also considered the P value from the Chi² test. If we identified substantial heterogeneity, we have reported it. However, we did not explored possible causes by prespecified subgroup analysis due to the small amount of studies. We defined substantial heterogeneity using a threshold I² of 50% or above.
Assessment of reporting biases
If we were able to pool more than 10 trials, we would have created and examined a funnel plot to explore possible small study biases for the primary outcomes.
Data synthesis
We undertook meta‐analyses only when this was meaningful, i.e. if the treatments, participants, and the underlying clinical question were similar enough for pooling to make sense.
We used a fixed‐effect model, as the previous updates of the analysis indicated a relatively small number of eligible trials, and we assumed the same intervention effect.
Subgroup analysis and investigation of heterogeneity
We planned to carry out the following subgroup analyses:
-
Ischaemic versus non‐ischaemic aetiology of HF
-
Preserved versus reduced EF (e.g. left ventricular EF ≥ 50% versus < 40%)
for the following outcomes:
-
All‐cause death
-
Stroke
Due to the small number of included studies, we did not conduct these planned subgroup analyses.
Sensitivity analysis
We planned to carry out the following sensitivity analyses, to test whether key methodological factors or decisions affected the main result.
-
Only including studies with a low risk of bias (e.g. low risk in at least four domains)
-
If we found substantial heterogeneity, comparing results from a random‐effects model with those from the planned fixed‐effect model
Due to the small number of included studies, we did not conduct these planned sensitivity analyses.
Summary of findings and assessment of the certainty of the evidence
We created a summary of findings table using the following outcomes: all‐cause death, stroke, and major bleeding events. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias) to assess the quality of a body of evidence as it related to the studies that contributed data to the meta‐analyses for each important outcome. We used methods and recommendations described in Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions and GRADEpro software (GRADEpro GDT; Schünemann 2020). We created two summary of findings tables for the different comparisons: warfarin versus control, and rivaroxaban versus placebo. We justified all decisions to downgrade the quality of studies in footnotes, and made comments to aid the reader's understanding of the review, where necessary.
Two review authors (ES, MK) independently made judgements about the quality of the evidence. They resolved disagreements through discussion, or by involving a third review author (GL). They justified and documented the judgements, incorporating them into the reports of the results for each outcome.
We extracted study data, formatted our comparisons in data tables, and prepared a summary of findings table before writing the results and conclusions of our review.
Results
Description of studies
Results of the search
The previous searches in 2001 and 2005 identified 1100 records, four studies (four records) of which met the inclusion criteria.
The updated search in 2010 identified 1219 records, one of which was previously identified as an ongoing study and was included (HELAS 2006); and another of which was previously identified as an ongoing study and was excluded (WATCH 2009). One study remained ongoing (WARCEF 2012).
The updated search in 2013 identified 1460 new records, none of which we included.
The updated search in 2020 identified 2767 records. We removed 489 duplicates, and screened 2278 records. We assessed 49 records for eligibility, and excluded 47 records with reasons: 31 studies included the wrong population, and 16 studies were not randomised. We described only the studies that mostly narrowly missed inclusion. We included one new RCT (with two references (COMMANDER HF 2018); see Figure 1.
PRISMA flowchart
We identified three small prospective controlled studies of warfarin in HF in the previous search, but they were over 50 years old, with methods not considered reliable by modern standards, so we excluded them (Anderson 1950; Griffith 1952; Harvey 1950).
In total, this review includes three studies (see Characteristics of included studies).
Included studies
We included three studies, with a total of 5498 participants (COMMANDER HF 2018; HELAS 2006; WASH 2004).
WASH 2004 was a pilot study of 279 participants with heart failure (HF) randomised to anticoagulation with warfarin (89 participants; target international normalized ratio (INR) 2.5); aspirin (300 mg; 91 participants); or no antithrombotic therapy (99 participants). This was an open label trial design and therefore, performance bias could not be excluded (see risk of bias table).
HELAS 2006 was a study of 115 participants with ischaemic heart disease randomised to anticoagulation (target INR 2.0 to 3.0) or aspirin, and a group of 82 participants randomised to anticoagulation (target INR 2.0 to 3.0; 38 participants) or placebo (44 participants). This was a double‐blind trial with an independent data and safety monitoring committee.
COMMANDER HF 2018 was a study of 5022 participants with HF and coronary artery disease randomised to rivaroxaban (2507 participants) or placebo (2515 participants). This was a double‐blind placebo‐controlled trial. Participants were randomised using an interactive web‐response system and permuted blocks of four. An independent data and safety monitoring committee had access to unblinded data during the trial and was responsible for the safety of the enrolled individuals.
Two studies exclusively or predominantly recruited from European countries (HELAS 2006; WASH 2004). One study was conducted in Europe, Asia, North and Latin America, and Africa (COMMANDER HF 2018). One trial reported receiving relevant industry sponsorship (COMMANDER HF 2018).
Excluded studies
We excluded 16 studies, four of which were observational studies (EPICAL 2002; Fuster 1981; Kyrle 1985; Natterson 1993). We provided the adverse event information in the discussion. See Characteristics of excluded studies for details.
Fuster 1981 was a retrospective study of 104 participants with idiopathic dilated cardiomyopathy followed up for a total of 725 person‐years. Kyrle 1985 was a study of 38 participants with non‐ischaemic cardiomyopathy followed for a total of 72 person‐years. Natterson 1993 was a more recent study of 224 participants awaiting cardiac transplantation. EPICAL 2002 was a study of 417 participants with an average follow‐up period of five years. EPICAL 2002 was eventually excluded as 24% had concomitant AF. Only survival was reported and event rates were not reported separately for the sinus rhythm group or those treated with aspirin, warfarin or no therapy. Aspirin was used in 31% of participants, warfarin in 28% of participants, and warfarin plus aspirin in 2% of participants. Participants given any antithrombotic treatment compared to none had a better survival at five years (40.4% versus 31%, P = 0.01).
One case series, conducted over 50 years ago in an heterogeneous HF population (N = 61), was excluded but it did report that the prevalence of thromboembolism on dicumarol (6.5%) was lower compared to previously published reports (22%) (Wishart 1948).
Five publications from RCTs were excluded because the participants in the analyses were not randomised to anticoagulant or control in the original study (CONSENSUS 1987; PROMISE 1993; SAVE 1997; SOLVD 1998; V‐HeFT 1993). The analyses were post hoc analyses of participants treated with oral anticoagulation at the discretion of the investigators. The V‐HeFT 1993 trials included participants with symptomatic HF with radiological, echocardiographic or radionuclide evidence of left ventricular systolic dysfunction (V‐HeFT 1993). The SOLVD 1998 trials included participants with left ventricular systolic dysfunction, defined as a left ventricular EF of 35% or more, who were symptomatic and enrolled into the treatment trial or asymptomatic and enrolled into the prevention trial (SOLVD 1998). Limited information from a retrospective analysis of one trial has been presented in abstract form only (PROMISE 1993). The SAVE 1997 study included participants post‐MI with a left ventricular EF of 40% or more and no overt HF, that is, asymptomatic participants (SAVE 1997). One further study was a cohort study of people attending an outpatient anticoagulation clinic for a variety of co‐morbidities (Visser 2004). Two studies were excluded as the control group received aspirin or clopidogrel, but not placebo (WARCEF 2012; WATCH 2009).
The earlier prospective controlled studies were performed over 50 years ago in hospitalised people with a high prevalence of rheumatic heart disease and AF (Anderson 1950; Griffith 1952; Harvey 1950). Although described as 'randomised', the trial methodologies in these studies are more properly described as quasi‐randomised and cannot be seen as entirely reliable by modern standards. Moreover, routine standards of care over 50 years ago were different than current standards of care.
Risk of bias in included studies
Risk of bias is shown in Figure 2 and Figure 3.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Allocation
The three studies were at low risk of selection bias, because they were designed as RCTs. In HELAS 2006, computerised randomisation was conducted by fax or by phone to the Coordinating Centre. In WASH 2004, participants were randomised according to prospectively constructed blocks of random numbers. COMMANDER HF 2018 study, participants were randomly assigned to receive rivaroxaban or placebo in a ratio 1:1. Central randomisation was based on a computer‐generated randomisation schedule. Participants were randomised using an interactive Web response system and permuted blocks of four. Randomisation was stratified by country. The investigator was not informed about randomisation codes. The codes were maintained within the system, which had the functionality to allow the investigator to break the blind for an individual subject.
Blinding
We assessed one study at high risk of performance bias (open label trial), but low risk of detection bias (blinded endpoint design (WASH 2004)).
Another study was a double‐blind study, which we judged at unclear risk of performance bias (HELAS 2006). The investigator‐supervisor was blinded to the study and placebo tablets were given daily or according to sham adjustment. The dose of the drug was administered to the participants by second physician. In the placebo group, the supervisor recommended the dose according to sham results. This study was monitored by an independent data and safety monitoring committee. The occurrence of clinical endpoints was recorded (unclear risk of detection bias).
We judged the third study at a low risk of performance bias (double‐blind study) and a low risk of detection bias (outcomes were confirmed by investigators using an extensive, dedicated case‐report form, with source‐data verification by the sponsor's clinical operations team (COMMANDER HF 2018)). However, events were not centrally adjudicated.
Incomplete outcome data
We assessed two studies at unclear risk of attrition bias (COMMANDER HF 2018; WASH 2004). In COMMANDER HF 2018, data on vital status were available for 5013 participants (99.8%). WASH 2004 did not provide data on participants who were lost to follow‐up. We assessed one study at a low risk of attrition bias, since all participants were included in the follow‐up analysis (HELAS 2006). All participants were included in the follow‐up analysis and an independent data committee performed interim analyses to determine whether continuation of any of the treatment arms might be detrimental to the participants.
Selective reporting
We judged all three studies at a low risk of reporting bias, since they all reported on predefined primary and secondary outcomes. The results of the above mentioned studies showed that treatment had no effect on outcome, and therefore were negative studies.
Other potential sources of bias
In one study, cross‐over rates were 25% at 12 months what might have affected the certainty of the results or the risk of bias (WASH 2004).
Effects of interventions
See: Summary of findings 1 Warfarin compared to placebo or no treatment in people with chronic heart failure in sinus rhythm; Summary of findings 2 Rivaroxaban compared to placebo in people with heart failure in sinus rhythm
Warfarin compared to control
Two studies contributed data for this comparison. In the WASH 2004 trial, 99 participants received no treatment compared to 89 participants who received warfarin; they were followed for a mean of 27 months. The HELAS 2006 study did not achieve the target recruitment of 6000 participants. Eighty‐two participants with idiopathic dilated cardiomyopathy were randomised to receive either warfarin or placebo.
Primary outcomes
All‐cause death
Both studies reported all‐cause death. The evidence is uncertain if warfarin has any effect on all‐cause death compared to no treatment or placebo, because the certainty of the evidence is low, and the 95% confidence interval (CI) is consistent with possible benefit and possible harm (odds ratio (OR) 0.66, 95% CI 0.36 to 1.18; 2 studies, 324 participants; low‐certainty evidence; Analysis 1.1).
Stroke
None of the studies reported stroke as an individual outcome.
Secondary outcomes
Cardiovascular death (including sudden death)
The evidence is uncertain if warfarin has any effect on cardiovascular deaths (OR 0.98, 95% CI 0.58 to 1.65; 2 studies, 324 participants; Analysis 1.2).
Myocardial infarction (MI)
The evidence is uncertain if warfarin has any effect on the risk of MI (OR 0.64, 95% CI 0.20 to 2.08; 2 studies, 324 participants; Analysis 1.3).
Pulmonary embolism
None of the studies reported data on pulmonary embolism.
Peripheral arterial embolism
None of the studies reported data on peripheral arterial embolism.
Major bleeding events
Both studies reported major bleeding events. Warfarin may increase the risk of major bleeding events (OR 5.98, 95% CI 1.71 to 20.93; 2 studies, 324 participants; low‐certainty evidence; NNTH 17, Analysis 1.4).
Rivaroxaban compared to control
In the COMMANDER HF 2018 study, 2507 participants were randomised to receive rivaroxaban, and 2515 participants were randomised to receive placebo; they were followed for a median duration of 21.1 months.
Primary outcomes
All‐cause death
Rivaroxaban makes little to no difference to the risk of all‐cause death (OR 0.99, 95% CI 0.87 to 1.13; 1 study, 5022 participants; high‐certainty evidence; Analysis 2.1).
Stroke
Rivaroxaban probably reduces the risk of stroke (OR 0.67, 95% CI 0.47 to 0.95; NNTB 101, 1 study, 5022 participants; moderate‐certainty evidence; Analysis 2.3).
Secondary outcomes
Cardiovascular death (including sudden death)
Rivaroxaban may make little to no difference to the risk of cardiovascular death (OR 1.00, 95% CI 0.86 to 1.15; 1 study, 5022 participants; Analysis 2.2).
Myocardial infarction (MI)
The evidence is uncertain if rivaroxaban has any effect on the the risk of MI (OR 0.83, 95% CI 0.63 to 1.09; 1 study, 5022 participants; Analysis 2.4).
Pulmonary embolism
The evidence is uncertain if rivaroxaban has any effect on the risk of symptomatic pulmonary embolism (OR 1.23, 95% CI 0.51 to 2.97; 1 study, 5022 participants; Analysis 2.5).
Peripheral arterial embolism
The study did not report data on peripheral arterial embolism.
Major bleeding events
Rivaroxaban probably increases the risk of major bleeding events compared to no treatment (OR 1.65, 95% CI 1.17 to 2.33; 1 study, NNTH 79; 5008 participants; moderate‐certainty evidence; Analysis 2.6).
Discussion
Summary of main results
We included three prospective randomised controlled studies (RCT) of oral anticoagulation in 5498 people with heart failure (HF) in this review (COMMANDER HF 2018; HELAS 2006; WASH 2004).
The evidence is uncertain if warfarin has any effect on all‐cause death compared to placebo or no treatment. Warfarin may increase the risk of major bleeding events.
There is evidence of no difference for rivaroxaban's effect on all‐cause death compared to placebo. Rivaroxaban compared to placebo probably reduces the risk of stroke, but increases the risk of major bleedings events.
Therefore, these three studies do not support the routine use of oral anticoagulation therapy for people with HF in sinus rhythm.
Overall completeness and applicability of evidence
The included studies had some limitations in overall completeness and applicability of evidence. WASH 2004 was not blinded and was underpowered to provide an answer to our review question. Unfortunately, cross‐over rates were 25% at 12 months what might have affected the certainty of the results or the risk of bias. In HELAS 2006, participants were representative of those in clinical population. In COMMANDER HF 2018, the events were not centrally adjudicated, and thus, the risk of misclassification of causes of hospitalisation and death exist. Moreover, the rate of discontinuing the trial regimen was relatively high (16.3 in the rivaroxaban group, and 13.6 per 100 person‐years in the placebo group).
Quality of the evidence
The three studies were RCTs, but contributed to the quality of the evidence with some limitations.
For warfarin versus placebo or no treatment, we downgraded all‐cause death by two levels, once for imprecision, since the 95% confidence interval crossed the line of no effect, and once for inconsistency, due to high heterogeneity (80%) across studies. We downgraded major bleeding events twice for imprecision, due to the small sample size and low event rate.
We downgraded the single trial of rivaroxaban included in the review by one level for imprecision for stroke and major bleeding events, since both of these outcomes have very low event rates (COMMANDER HF 2018).
Potential biases in the review process
The process of updating the review was done independently by two authors (ES and MK). The potential bias in the review process might be associated with selection of outcomes. For example, we did not assess all‐cause hospitalisations or cardiovascular hospitalisations. Moreover, cross‐over rates in the studies seem to be undesirable when testing therapeutic agents.
Agreements and disagreements with other studies or reviews
The available, current evidence does not support the routine use of oral anticoagulation over no therapy or placebo in people with HF and sinus rhythm. Beggs 2019 included five studies in their systematic review and meta‐analysis (COMMANDER HF 2018; HELAS 2006; WARCEF 2012; WASH 2004; WATCH 2009). They found that oral anticoagulant therapy did not reduce all‐cause mortality, hospitalisation due to HF, or non‐fatal myocardial infarction (MI) when compared with control therapy. Oral anticoagulation reduced the rate of non‐fatal stroke and increased the incidence of major haemorrhages. The authors concluded that thrombosis or embolism did not play an important role in the morbidity and mortality associated with heart failure with reduced ejection fraction (HFrEF), with the exception of stroke‐related morbidity.
EPICAL 2002, an observational study, found that of the 417 participants with a mean left ventricular ejection fraction of 22%, New York Heart Association (NYHA) class III or IV, 45% with ischaemic origin demonstrated a significant reduction in thromboembolic events. Therefore, it is possible that the efficacy of oral anticoagulation may differ according to the cause of HF, as people with idiopathic cardiomyopathy may have a greater risk of cardiogenic thromboembolism whilst those with atherosclerosis are also at risk of other vascular events including in situ coronary artery thrombosis. EPICAL 2002 suggested that anticoagulation with warfarin was beneficial in people with severe HF, in particular those with NYHA class III or IV.
In conclusion, the present (limited) data do not support the routine use in people with HF who remain in sinus rhythm.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Comparison 1: Warfarin versus control, Outcome 1: All‐cause death
Comparison 1: Warfarin versus control, Outcome 2: Cardiovascular death (including sudden death)
Comparison 1: Warfarin versus control, Outcome 3: Myocardial infarction
Comparison 1: Warfarin versus control, Outcome 4: Major bleeding events
Comparison 2: Rivaroxaban versus placebo, Outcome 1: All‐cause death
Comparison 2: Rivaroxaban versus placebo, Outcome 2: Cardiovascular death (including sudden death)
Comparison 2: Rivaroxaban versus placebo, Outcome 3: Stroke
Comparison 2: Rivaroxaban versus placebo, Outcome 4: Myocardial infarction
Comparison 2: Rivaroxaban versus placebo, Outcome 5: Pulmonary embolism
Comparison 2: Rivaroxaban versus placebo, Outcome 6: Major bleeding events
| Warfarin compared to placebo or no treatment in people with chronic heart failure in sinus rhythm | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with heart failure in sinus rhythm | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with placebo or no treatment | Risk with warfarin | |||||
| All‐cause death | 203 per 1000 | 144 per 1000 | OR 0.66 | 324 | Low certaintya | |
| Stroke | Stroke was not reported as an individual outcome. | |||||
| Major bleeding eventsc | 0/143 | 11/181 | OR 5.98 | 324 | Low certaintyb | 0 per 1000 in the control arm for major bleeding events. It is not based on the estimated relative effect. |
| *The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded twice – once for imprecision, and once for inconsistency due to high heterogeneity cA major bleeding event is defined by the International Society on Thrombosis and Haemostasis as overt bleeding associated with a decrease in haemoglobin level of at least 2 g/dL, transfusion of two or more units of packed red cells or whole blood, bleeding in a critical site (intracranial, intraspinal, intraocular, pericardial, intraarticular, intramuscular with compartment syndrome, or retroperitoneal), or a fatal outcome. | ||||||
| Rivaroxaban compared to placebo in people with heart failure in sinus rhythm | ||||||
| Patient or population: people with heart failure in sinus rhythm | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with placebo | Risk with rivaroxaban | |||||
| All‐cause death | 227 per 1000 | 225 per 1000 | OR 0.99 | 5022 | High certainty |
|
| Stroke | 30 per 1000 | 20 per 1000 | OR 0.67 | 5022 | Moderate certaintya | |
| Major bleeding eventsb | 20 per 1000 | 32 per 1000 | OR 1.65 | 5008 | Moderate certaintya | |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded for imprecision, due to very low event rates. | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1.1 All‐cause death Show forest plot | 2 | 324 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 0.66 [0.36, 1.18] |
| 1.2 Cardiovascular death (including sudden death) Show forest plot | 2 | 324 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 0.98 [0.58, 1.65] |
| 1.3 Myocardial infarction Show forest plot | 2 | 324 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.64 [0.20, 2.08] |
| 1.4 Major bleeding events Show forest plot | 2 | 324 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 5.98 [1.71, 20.93] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 2.1 All‐cause death Show forest plot | 1 | Peto Odds Ratio (Peto, Fixed, 95% CI) | Totals not selected | |
| 2.2 Cardiovascular death (including sudden death) Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Totals not selected | |
| 2.3 Stroke Show forest plot | 1 | Peto Odds Ratio (Peto, Fixed, 95% CI) | Totals not selected | |
| 2.4 Myocardial infarction Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Totals not selected | |
| 2.5 Pulmonary embolism Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Totals not selected | |
| 2.6 Major bleeding events Show forest plot | 1 | Peto Odds Ratio (Peto, Fixed, 95% CI) | Totals not selected | |
