The prevalence of anemia among patients with chronic heart failure (CHF) is significant, estimated to be present in >20% of CHF patients.^{[1] and [2]} Anemia among patients with CHF likely has a multifactorial etiology,^{[3] and [4]} and along with renal dysfunction, comprises the triad of the cardiorenal syndrome.^{[5] and [6]} Recently, several clinical trials have identified anemia as a strong prognostic factor associated with poor outcomes among patients with CHF,^{[7], [8], [9] and [10]} with those with anemia having a nearly 2-fold increase in mortality compared with their nonanemic counterparts.¹¹ It has been suggested that treatments for anemia, including erythropoiesis-stimulating agents (ESAs) such as recombinant erythropoietin and its analogue darbepoetin, may decrease this increased mortality rate and improve quality of life among patients with CHF and comorbid anemia.^{[12], [13], [14] and [15]} Several trials have examined the effects of ESAs in this patient population. However, the trials conducted to date have involved limited sample sizes and have produced varying estimates of the effect of these agents. Furthermore, numerous clinical trials examining the role of ESAs in other patient populations have yielded important safety concerns.^{[16], [17] and [18]} We therefore performed a systematic review and meta-analysis of the available randomized controlled trials (RCTs) to evaluate the efficacy and safety associated with the use of ESAs for correcting anemia in patients with CHF.

Methods

We carried out this systematic review and meta-analysis in accordance with the standards described in the Quality of Reporting of Meta-Analyses of Randomized Controlled Trials (ie, QUOROM) guidelines.¹⁹

Search Strategy

We systematically searched MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Clinical Trials to identify RCTs studying the use of erythropoietin or darbepoetin as a treatment for anemia among patients with CHF. This search (Appendix 1) involved the following search terms: “erythropoiesis-stimulating agents,” “hematinics,” “erythropoietin,” “darbepoetin,” “anemia,” and “heart failure.” The search was limited to RCTs published in English before June 15, 2009. We also searched ClinicalTrials.gov as well as bibliographies of retrieved publications and used PubMed's related articles feature to identify studies not captured by our primary literature search.

Inclusion Criteria

The inclusion criteria for our study were: (1) documented CHF among study participants (left ventricular ejection fraction [LVEF] <40% or symptomatic heart failure); (2) comorbid noncritical anemia among participants (hemoglobin between 9.0 and 12.5 g/dL) at the time of randomization; (3) randomized allocation of patients to ESA intervention or control; (4) follow-up duration ≥12 weeks; and (5) assurance that patients were receiving stable, optimal medical management for CHF at the time of randomization.

Data Extraction

Data were independently extracted in duplicate by 2 reviewers using a standardized protocol and reporting form. Disagreements were resolved by consensus or, when necessary, by a third reviewer. Information regarding study design, publication date, country of origin, sample size, and intervention characteristics was recorded. Data were also recorded regarding patient baseline demographic and clinical characteristics, including age, sex, LVEF, levels of hemoglobin, hematocrit, and creatinine, exercise tolerance, severity of CHF symptoms (New York Heart Association class), quality of life, and brain-natriuretic peptide (BNP) level. We also extracted data regarding the following outcomes: all-cause mortality, hospitalization resulting from CHF, change in New York Heart Association symptom severity, exercise tolerance, hemoglobin, LVEF, BNP, quality of life, and adverse events, including venous thrombosis, arterial hypertension, cerebrovascular events, and myocardial infarction (MI). Follow-up durations were noted. We also recorded extended morbidity and mortality follow-up periods, during which studies were unblinded but patients were maintained on their initial randomized treatment (though placebo injections were discontinued). Data were extracted for the initial, blinded treatment period except for mortality, CHF-related hospitalizations, and adverse events, which were based on events observed during both the initial and extended follow-up periods. We contacted the authors of several included RCTs to obtain necessary count data not provided in the manuscript. Where possible, data that were not obtainable through correspondence with the authors were taken from other published, peer-reviewed sources, including a previous meta-analysis.²⁰

Quality Assessment

Quality assessment of all included RCTs was conducted using the Jadad scale.²¹ Briefly, RCTs were awarded one point for each positive response to the following 5 questions: (1) Was the treatment allocation described as random? (2) Was the randomization scheme described and appropriate? (3) Was the study described as double-blind? (4) Were both the patient and the assessor appropriately blinded? (5) Was there a description of dropouts and withdrawals? Trials scoring from 0 to 2 points were classified as “low” quality and those scoring 3 to 5 points were classified as “high” quality.

Data Analysis

Data were aggregated using random-effects meta-analysis models. Meta-analysis models were created for the following outcomes: all-cause mortality, hospitalization for CHF, CHF exacerbations, change in quality of life (based on responses to the Minnesota Living with Heart Failure Questionnaire), and overall adverse events. In sensitivity analyses, we repeated our analyses using a 0.5 continuity correction to examine the effect of zero-event trials. In additional sensitivity analyses, we restricted our analysis of hospitalization for CHF to high-quality RCTs. We also conducted sensitivity analyses in which we restricted analyses to trials that provided iron supplementation to both groups and to trials whose inclusion criteria included the presence of both CHF and low LVEF. Treatment effects are presented as odds ratios (OR) with corresponding 95% confidence intervals (CI) for binary outcomes and as weighted mean differences with corresponding 95% CIs for continuous outcomes. Heterogeneity was estimated using I² statistics. Funnel plots were constructed and visually assessed for the presence of publication bias. All analyses were conducted using MIX version 1.7.²²

Results

Search Results

Study and Patient Characteristics

The characteristics of the 9 identified RCTs are shown in Table 1. RCTs were double-blind, except for 2 single-blind studies^{[24] and [31]} and 1 open-label study.³² All RCTs used either recombinant erythropoietin or its analogue darbepoetin, but varied in their dosing regimens, with 3 studies using hemoglobin-targeted dose titration, 5 using a fixed weight-based dose, and 1 using both (Table 1). All but 2 RCTs administered concurrent iron to both the intervention and control groups; these 2 studies administered supplemental iron only in the treatment group.^{[31] and [32]} Follow-up duration ranged from 3 to 13 months. Four studies included extended morbidity and mortality follow-up periods.^{[23], [24], [25] and [30]}

The total number of patients randomized was 747 (407 in the ESA-treated groups and 340 in the control groups). Most patients enrolled had a baseline LVEF between 20% and 35% and an initial hemoglobin level between 10 and 12 g/dL (Table 2). The etiology of CHF was most commonly attributed to coronary artery disease, hypertension, or idiopathic cardiomyopathy. Study participants were more likely to be male, and the mean age varied from 60 to 75 years. RCTs included patients with mean serum creatinine levels between 1.3 and 1.7 mg/dL, except 1 trial that enrolled patients with a mean serum creatinine of 2.5 mg/dL (standard deviation 0.4) and 2.4 mg/dL (standard deviation 0.6) in the treatment and placebo arms, respectively.³⁰ All 9 RCTs reported that patients were receiving stable, optimal medical management for CHF at the time of enrollment.

Outcomes

In all RCTs, there was a larger increase in hemoglobin in the ESA treatment group relative to the control group (Table 2). Five of the 9 RCTs found a greater decrease in BNP levels among patients randomized to the ESA (range of mean change: −436 to −90 pg/mL) compared with those randomized to control (range of mean change: −26.5 to +315 pg/mL), although the standard deviations of these measurements varied considerably. Patients randomized to ESA had improved LVEF during follow-up (range: +5% to 6%), whereas those randomized to control did not (range: −5% to 1%) in the 4 RCTs that examined this outcome.

Patients randomized to ESAs also had greater improvements in exercise capacity and tolerability (Table 3). In the 4 RCTs that examined maximally tolerated exercise duration, 3 using the modified Naughton treadmill protocol^{[25], [28], [30] and [35]} and 1 using stationary bicycle testing,³¹ randomization to ESAs resulted in improved exercise duration during follow-up; this improvement was greater than that observed in the control group. Mean distance walked during 6-minute walk tests also improved among patients randomized to ESAs (+34 to 73 m) relative to those randomized to control (−47 to +37 m) in the 4 trials that evaluated this outcome.

ESAs also reduced symptom severity and improved quality of life (Table 4). Mean New York Heart Association symptom class decreased more among patients receiving ESAs than among control patients. In addition, patients' responses to quality of life questionnaires consistently reflected improved quality of life among patients randomized to ESAs. In 3 RCTs, ESAs were associated with a greater decrease than control in mean scores on the Minnesota Living with Heart Failure Questionnaire (MLHFQ),³⁶ corresponding to improved self-rated health-related quality of life.^{[25], [29] and [31]} In contrast, randomization to control resulted in a larger decrease in MLHFQ score relative to ESA treatment in 1 RCT.²⁸ Summary scores on the Kansas City Cardiomyopathy Questionnaire³⁷ in 3 studies also showed a consistent trend toward improved quality of life in the ESA-treated group compared with control, though there was large variability across studies.

Pooled Results

There was a significant reduction in hospitalization due to CHF among patients receiving ESAs relative to control (OR 0.41; 95% CI 0.24-0.69; I² = 19%; Fig. 2). This was based on 46 events in the ESA group compared with 76 in the control group. The effect of ESAs on mortality was inconclusive (OR 0.60; 95% CI 0.32-1.11; I² = 0%; Fig. 3). Sensitivity analyses that included a 0.5 continuity correction to account for zero-event trials did not appreciably alter these results. Additionally, restriction to high-quality studies in further sensitivity analyses did not substantially affect our results. We conducted additional sensitivity analyses in which we restricted inclusion to those that provided iron supplementation to both treatment arms (CHF hospitalization: OR 0.54, 95% CI 0.35-0.84; I² = 0%; all-cause mortality: OR 0.69, 95% CI 0.36-1.31; I² = 0%) as well as those that only included subjects with both CHF and low LVEF (CHF hospitalization: OR 0.50, 95% CI 0.32-0.77; I² = 2%; mortality: OR 0.63, 95% CI 0.33-1.18; I² = 0%).

When quality of life data were pooled from the 3 trials reporting MLHFQ,^{[25], [28] and [29]} comprising 525 total patients, the weighted mean difference in MLHFQ scores was −2.29 (95% CI -2.64, -1.94; I² = 0%).

Adverse Events

There was no evidence of an increase in the overall incidence of severe adverse events among patients randomized to ESAs (OR 0.86; 95% CI 0.51-1.42; I² = 28%). The incidence of specific adverse events was insufficient to permit pooled-analysis, but was noted among reporting trials: venous thrombosis (0/342 vs. 3/277 events/patient treatment vs. control group, respectively), arterial hypertension (18/342 vs. 12/277), MI (6/342 vs. 6/277), CHF exacerbations (52/342 vs. 60/277), transient ischemic attack or cerebrovascular accident (10/327 vs. 5/269), and seizure (1/342 vs. 2/277). Meta-analysis of the CHF exacerbation adverse event data provided inconclusive results (OR 0.64; 95% CI 0.37-1.10; I² = 14%).

Publication Bias

Visual inspection of funnel plots provided no evidence to suggest that publication bias was present (data not shown).

Discussion

Our study was designed to examine the therapeutic effects of ESAs, including recombinant erythropoietin and darbepoetin, in patients with CHF and comorbid anemia. Our meta-analysis suggests that the use of ESAs can significantly decrease the number of hospitalizations from CHF (OR 0.41; 95% CI 0.24-0.69). Our meta-analysis of their effect on all-cause mortality was inconclusive, likely because of the small number of patients enrolled in included RCTs, but suggests that ESAs may have beneficial mortality effects. We also found that the use of ESAs improved quality of life, as measured by the MLHFQ, compared with control. Although there were insufficient data to meta-analyze heart failure symptoms and Kansas City Cardiomyopathy Questionnaire quality of life data, available data suggest that treatment with ESAs is associated with improvements in these outcomes compared with control. Patients treated with ESAs also generally experienced improved exercise tolerance and capacity, decreased BNP levels, and improved LVEF compared with those randomized to control. Finally, there was no detectable difference in the overall rate of serious adverse events between treatment and control groups (OR 0.86; 95% CI 0.51-1.42), and only minor variation in the observed frequency of individual adverse events, including cerebrovascular accident/transient ischemic attack, MI, arterial hypertension, and venous thrombosis.

Our findings suggest potential benefit for the use of ESAs in reducing CHF-related hospitalizations. These agents also result in improvements in quality of life, exercise tolerance, and other measures of functional status. At the present time, however, given the small overall sample size (n = 747), there is not sufficient evidence to recommend their routine use in CHF patients with anemia. These data do, however, present compelling support for the continuation of a large RCT powered to conclusively examine clinical events—the Reduction of Events with Darbepoetin alfa in Heart Failure trial^{[38] and [39]}—which will further evaluate the effects of ESAs in heart failure. This trial is still in the enrollment phase.

Our analysis of adverse events has important implications for ongoing large clinical trials evaluating ESAs in CHF. In a recent RCT studying the role of darbepoetin in patients with renal failure and type 2 diabetes—the Trial to Reduce Cardiovascular Events With Aranesp Therapy (TREAT)—the investigators reported a 2-fold increase in the occurrence of cerebrovascular accidents in patients on treatment, and there was no overall treatment benefit.¹⁶ Furthermore, results of trials looking at the use of ESAs in patients with renal failure^{[17] and [40]} and cancer^{[18] and [41]} have raised concerns regarding the safety of these agents. A meta-analysis of RCTs of ESAs in patients with renal failure, including the Correction of Hemoglobin and Outcomes in Renal Insufficiency (CHOIR) trial⁴² and the Cardiovascular Risk Reduction by Early Anemia Treatment with Epoetin Beta (CREATE)⁴³ trial, demonstrated that ESAs resulted in an increase in all-cause mortality in renal failure patients.¹⁷ These effects may be mediated through nonhemantic, pleiotropic effects of ESAs, including the stimulation of angiogenesis and inhibition of apoptosis. These effects may be desirable in certain patient populations; for example, plausibly in patients with ischemic cardiomyopathy, and detrimental in others. These trials have also emphasized that optimum hematocrit targets are far from clear. And in addition to the hemoglobin target itself, the rate of hemoglobin increase has long been a concern at the Food and Drug Administration, and original labeling included a warning regarding the risk of side effects associated with the rate of correction.⁴⁴ Unger and others have contended that the instability in hemoglobin concentrations (rapid and overcorrections) could translate into increased cardiovascular risk by altering hemodynamic or rheologic homeostasis. Although inconclusive, our mortality analysis suggests that ESAs may have favorable effects overall in patients with CHF and anemia. In addition, our analysis of available data did not show evidence of an increased risk of overall adverse events, nor was a difference apparent in risk of specific adverse events, including venous thrombosis, arterial hypertension, MI, and cerebrovascular accident/transient ischemic attack, although the total patient population was small—perhaps too small to show such a difference. A recent patient-level analysis of the safety of darbepoetin alpha supports our findings.⁴⁵ However, it must be emphasized that, although these preliminary safety data are reassuring, a larger number of patients will be needed to draw definitive safety conclusions in this population.

Anemia is a common comorbidity among patients with CHF and represents a significant challenge faced by clinicians. Given its strong association with poorer prognosis in CHF, strategies designed to correct anemia are intuitively appealing. Recent studies have found that patients with anemia and CHF have an increased risk of mortality compared to nonanemic CHF patients (OR 1.96, 95% CI 1.74-2.21).¹¹ Various mechanisms for anemia in CHF have been postulated, including anemia of chronic disease resulting from neurohormonal and proinflammatory cytokine activation, defective iron utilization, inappropriate erythropoietin production, and depressed bone marrow function.³ However, it is presently not known whether this interaction is causal or associative.

Clinical trials have sought to reverse the mortality effect by correcting the anemia, some with intravenous iron and others with a combination of ESAs and supplemental enteral iron. Recently, a multicenter RCT—the Ferinject Assessment in Patients with IRon Deficiency and Chronic Heart Failure (FAIR-HF) trial—by Anker and colleagues demonstrated that the administration of intravenous iron (ferric carboxymaltose) to CHF patients with iron-deficiency (with or without anemia) improved symptoms, functional capacity, and quality of life.⁴⁶ This may prove to be another treatment option for patients with CHF, either as an alternative to or in conjunction with ESAs. Additionally, the selection of iron deficiency for inclusion rather than anemia raises questions about what is a more important target in this population. As our study demonstrates, multiple RCTs have looked at the treatment effect of ESAs on exercise capacity and symptomatic benefit, as well as on morbidity and mortality. A prior meta-analysis examining this issue concluded that ESAs were not associated with a higher incidence of hypertension or venous thrombosis and suggested that their use could decrease hospitalization resulting from CHF and possibly all-cause mortality.²⁰ This analysis, however, did not look at the effect of ESAs on exercise capacity, quality of life, CHF symptom control, or hemodynamic and hematologic parameters. In addition, the previous meta-analysis used a fixed-effects analysis model, which assumes no between-study variability. Despite low levels of statistical heterogeneity, we pooled data using the more conservative random-effects models to account for potential between-study variability (eg, variability in inclusion criteria, dosage, duration of follow-up, other study characteristics). Our analysis also differed by the inclusion of 3 studies, 1 on the basis of study design and 2 more recent publications. The inclusion of these data resulted in more precise estimates. Another more recent meta-analysis supports our findings as well.⁴⁷ Our systematic literature review was conducted more recently to capture newer RCTs, and used stricter inclusion criteria, namely a minimum follow-up time of 12 weeks. We feel that allowing adequate follow-up for the physiologic effects of these agents to translate into clinically significant effects is important, and as such our analyses differed by several RCTs. Nonetheless, we reached similar conclusions.

Our study has potential limitations. First, our meta-analysis involved a small number of RCTs with limited sample sizes. We were therefore unable to conclusively examine the effect of ESAs on mortality or specific adverse events. However, this study provides additional information regarding these end points that support the continuation of ongoing RCTs that will definitively address this issue. Second, there was some heterogeneity in study design, patient characteristics, duration of follow-up, and outcomes reported. Although we used a random-effects model to account for between-study variability and our analysis suggests relatively little statistical heterogeneity is present, this heterogeneity remains a potential limitation. In addition, we conducted a number of sensitivity analyses, which provided results that were consistent with our primary analysis. Third, given the limited number of studies performed to date, we could only look at ESAs as a class and could not assess the safety and efficacy of different agents. Given the similar improvements observed in hemoglobin during follow-up (Table 2), we felt that this was a valid approach. The limited number of studies also limited our ability to stratify analyses by study-level covariates. Fourth, we were unable to meta-analyze all outcomes of interest because these outcomes were not reported for all RCTs and, among those that did report these parameters, data were not consistently provided in a poolable format. To allow for more careful examination of these results, we have presented data in tabular form rather than as pooled estimates, and our systematic review of these outcomes suggests that ESAs have favorable effects. Finally, as is true with all meta-analyses, our study may be affected by publication bias. However, we found no evidence of its occurrence in the present study.

Conclusion

In patients with CHF and anemia, treatment with ESAs is associated with a decrease in CHF-related hospitalizations. These medications also show consistent beneficial effects on patient quality of life, exercise tolerance, BNP levels, and LVEF and do not appear to increase the incidence of adverse events. Mortality data are inconclusive but available data suggest that ESAs may have favorable effect on all-cause mortality. However, ongoing large RCTs are needed to confirm this observation before incorporating the use of ESAs in the everyday treatment of patients with CHF and comorbid anemia. Nonetheless, currently available evidence suggests that ESAs have favorable effects in this patient population.

Acknowledgments

We would like to thank Miriam Abouelouafaa for her help with data abstraction.

Disclosures

Dr. Eisenberg is a National Researcher of the Quebec Foundation for Health Research.