Anaemia is often found in patients with chronic heart failure (HF), has multifactorial causes, and is associated with considerable risk.1 2 A recent meta-analysis including more than 150 000 patients showed anaemia to be associated with a doubled mortality risk in patients with chronic HF.3 Therapeutic agents to correct anaemia are clinically available, yet their role in the anaemic patient with chronic HF remains unclear. Several small-scale phase II studies assessed the effect of erythropoiesis-stimulating proteins (ESPs) on surrogate end points including exercise capacity, New York Heart Association class, renal function and neurohormonal changes.4^–7 The results of these trials are ambiguous, showing either beneficial effects in small trials or only non-significant trends in improved exercise endurance in somewhat larger studies.

Recently, concerns over the safety of ESP in patients with renal failure and malignancies were raised. Two separate studies of ESP treatment in chronic kidney disease (CKD) showed that anaemic patients targeted to a higher haemoglobin level had an increased incidence of cardiovascular events.8 9 In patients with cancer-related anaemia, ESP treatment was associated with increased risks of mortality and venous thromboembolism.10 The effects of ESP treatment in anaemic patients with chronic HF remain largely unknown. Therefore, our aim was to perform a meta-analysis of all randomised studies to determine the effect of ESP treatment on mortality, hospitalisation and adverse events in patients with chronic HF.

METHODS

Search strategy

We adhered to the QUORUM and Cochrane Collaboration guidelines as previously published.11 To identify relevant randomised clinical trials we performed a literature search in Medline accessed by PubMed (1966 through July 2008). Two of the authors (HFG and PvdM) developed a search strategy. The searches were performed independently and in duplicate by two authors. To identify ESP studies in patients with HF we used both the medical subject heading term (MESH) and text word search including heart failure, congestive heart failure, chronic heart failure, CHF. These terms were combined with MESH and text words for recombinant erythropoietin, darbepoetin, erythropoietin, erythropoiesis, stimulating proteins, therapeutic use, clinical trials and random allocation. In addition, we identified potentially relevant studies using a manual search of references lists from all eligible studies and review articles. We searched for randomised clinical trials on the website http://www.clinicaltrials.gov (accessed 23 May 2009) and the Cochrane Controlled Clinical Trials Register Database. We consulted experts in the field and searched the ISI Web of Sciences for publications that cited key references.

Study selection

All abstracts were scanned independently by two investigators (HFG and PvdM). We considered for inclusion all prospective randomised placebo controlled clinical trials assessing the effect of ESPs in anaemic patients with chronic HF. Titles and abstracts of all articles identified by the search strategy were evaluated and rejected on an initial screen if they (a) were not randomised controlled trials; (b) included subjects other than patients with chronic HF; (c) included patients ⩽18 years old; (d) were published in a language other than English; (e) had a follow-up <3 months (since it may take 6–8 weeks before patients reach their target haemoglobin levels).4 After obtaining full reports of the candidate studies, the same reviewers independently assessed eligibility. Differences in data between the two reviewers were solved by re-reviewing the corresponding articles and the final set was agreed by consensus. For studies in which not all data were published, or data were insufficient, authors were contacted and asked for additional information.

Quality assessment and data abstraction

Each study was evaluated for quality according to the guidelines provided by the United States Preventive Task Force and published recommendations. The following characteristics were assessed: (a) duration of follow-up >6 months; (b) reporting loss of follow-up; (c) definition of anaemia; (d) full specification of outcome; (e) study sample representative for the mentioned population; (f) full specification of clinical and demographic variables; (g) explanation of sample selection; (h) clear inclusion and exclusion criteria. Studies were graded as poor quality if they met <5 criteria, fair if they met 5–6 criteria and good if they met ⩾7 criteria. Outcomes assessed were all-cause mortality, HF hospitalisation, venous thrombosis (deep venous thrombosis and pulmonary embolism) and hypertension.

Statistical analysis

Risk ratios (RRs) with 95% confidence intervals (CIs) of outcomes were constructed from every study. We pooled the results from the individual studies and performed tests of heterogeneity between studies using a standard χ² test and I² statistic. Significance of the pooled studies was tested with the fixed effects models except when heterogeneity was found to be significant, in which case pooled estimates based on random effects models were reported. Bias in published studies was assessed using a funnel plot of study results against study precision. We tested symmetry of the funnel plot by the Egger test. Significance was set at p<0.05 and all statistical analyses were performed using Stata 9.0, College Station, Texas, USA.

RESULTS

Electronic search retrieved 184 eligible studies. On initial screening 168 studies were rejected based on the title. Of 16 screened abstracts, five were rejected and 11 studies were retrieved for detailed evaluation. One of these studies was excluded for a follow-up <3 months12 and three studies were descriptions of the same cohort,13^–15 in which case we included only the first published study. Altogether seven studies, investigating 650 patients were included in the present meta-analysis (fig 1). The number of patients in each study ranged from 23 to 319. All studies were performed in patients with moderate to severe chronic HF. Over 90% of the patients were in New York Heart Association class II or III. All studies required a left ventricular ejection fraction (LVEF) <40%, except for the study of Mancini et al, which did not have a cut-off value for LVEF.16

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flow chart for study selection. CHF, chronic heart failure; ESP, erythropoiesis-stimulating protein; RCT, randomised controlled trial.

Table 1 summarises the included studies. Six of the seven studies excluded patients with significant renal disease; serum creatinine >2.5 mg/dl (221 μmol/l),16 17 creatinine >3.0 mg/dl (265 μmol/l),4 5 18 serum creatinine >5 mg/dl (442 μmol/l),6 except for the study of Silverberg et al7 which did not have renal failure as an exclusion criterion. The average creatinine values ranged from 1.3 mg/dl (115 μmol/l) to 2.5 mg/dl (221 μmol/l). The duration of follow-up ranged from 3 months to 1 year. Average baseline haemoglobin levels ranged from 10.3 g/dl (6.4 mmol/l) to 11.8 g/dl (7.4 mmol/l). Table 2 shows the baseline, target and achieved haemoglobin levels of the different studies. Three studies with a total of 516 patients specifically excluded patients with malignancies or patients receiving chemotherapy and/or radiation therapy within 12 weeks of randomisation.4 5 17 No studies were terminated prematurely.

In total 650 patients were included, of these, 363 patients (55.8%) were treated with ESPs and 287 were randomised to the placebo group. In the study by Van Veldhuisen et al two different treatment regimens were used, 56 patients in the weight-based dose (0.75 μg/kg darbepoetin alfa starting dose) and 54 patients in the fixed dose group (50 μg darbepoetin alfa starting dose). Since no differences in the haemoglobin values reached were seen between the two different treatment regimens, both treatment arms from this study were combined by the authors.5 Other studies included used one treatment regimen. Table 1 shows the study characteristics. Of the 363 patients receiving ESPs, the vast majority of patients (n = 310) were treated with darbepoetin-alfa. Of the remaining patients, 33 were treated with erythropoietin alfa and 20 with erythropoietin beta). We found that in most studies both groups (placebo and ESP) were treated with iron. In the studies of Mancini et al16 and Silverberg et al7 only patients randomised to ESPs were treated with iron. These two studies comprised 55 patients (8.5% of the total included patients).

All studies reported data on all-cause mortality. In the ESP treatment group 20 patients died (5.5%), whereas in the placebo group 25 patients (8.7%) died (RR = 0.69; 95% CI 0.39 to 1.23, p = 0.21) (fig 2). Hospitalisation data were available in all seven studies. Of the 363 patients in the ESP treatment group, 37 patients were admitted to hospital (10.2%), while 56/287 placebo-treated patients were hospitalised (19.5%). This resulted in a significantly reduced hospitalisation risk for patients treated with ESPs (RR = 0.59, CI 0.41 to 0.86; p = 0.006) (fig 3). For the outcomes, no significant heterogeneity was found between the included studies for mortality (I² = 0.0%, p = 0.45) and HF hospitalisation (I² = 4.3%, p = 0.39), and neither showed a clear indication for publication bias (Egger test, p = 0.60 and p = 0.06, respectively).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Risk ratio for mortality in erythropoiesis-stimulating protein treatment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Risk ratio for hospitalisation for heart failure in erythropoiesis-stimulating protein treatment.

A risk of poorly controlled blood pressure was seen in 4.7% of the ESP-treated patients and in 4.2% of the placebo treated patients RR = 1.21 (95% CI 0.61 to 2.39, p = 0.59). Venous thrombosis (two patients with deep venous thrombosis and one patient with pulmonary embolism) was found in only three patients, all randomised to the placebo group. No data were published on the incidence of newly diagnosed malignancies in any of the included studies.

We were unable to do a formal meta-analysis for the effect of ESPs on left ventricular function because of differences in presentation and assessment of left ventricular function. Three studies assessed LVEF by echo, whereas one study used a multiple gated acquisition scan. Three of the four individual studies showed a significantly improved LVEF6 7 17 but the largest study assessing LVEF showed no improvement.5

DISCUSSION

Recent trials and meta-analyses have raised concerns about the safety of ESPs in patients with malignancies and renal failure. The most important finding of our current meta-analysis is that in anaemic patients with chronic HF ESP treatment appears to be safe. Moreover, ESP treatment was associated with a significant 47% reduction in the risk of HF hospitalisations. Further, we found no increase in the occurrence of other adverse events, including poorly controlled blood pressure and venous thrombosis.

Our findings are somewhat in contrast to the observations in patients with severe renal failure. Recently, two large trials in patients with kidney failure showed that correction of haemoglobin levels to near normal values resulted in an increased morbidity and mortality. In the first trial, CHOIR, patients were included with baseline haemoglobin levels of 10.1 g/dl. and an estimated glomerular filtration rate of 27.2 ml/min/1.73 m², both well below the values seen in the seven trials of chronic HF included.8 Half of the patients in the CHOIR study were randomly assigned to receive a dose of epoetin alfa targeted to achieve a haemoglobin level of 13.5 g/dl, whereas the other half were randomised to achieve a level of 11.3 g/dl. During a median follow-up of 16 months, the group with the higher haemoglobin target had an increased incidence of the composite end point (death, myocardial infarction, HF hospitalisation and stroke) with a hazard ratio of 1.34 (p = 0.03). There was a significant increase in HF hospitalisation in patients targeted to a higher haemoglobin level (64 vs 47 patients). In CREATE, the 603 patients studied had a baseline haemoglobin of 11.6 g/dl and estimated glomerular filtration rate of 24.5 ml/min/1.73 m².9 During a mean follow-up of 3 years, correction of haemoglobin with epoetin beta to a target level of 13.0–15.0 g/dl, compared with partial correction (haemoglobin of 10.5 g/dl), was not associated with an increase in the incidence of cardiovascular events in contrast to the results of CHOIR.

To further examine the safety of complete versus partial correction of haemoglobin values in patients with kidney disease, Krum and colleagues recently published a meta-analysis in which they included nine studies with similar conditions to those of CREATE and CHOIR. They found that randomisation to higher haemoglobin values was associated with an increased risk of mortality, venous thrombosis and poorly controlled blood pressure.19 Interestingly, no differences in the risk of myocardial infarction were seen.

Since less than one-third of the patients included in the CKD trials are diagnosed with chronic HF, it is difficult to extrapolate these findings to patients with chronic HF. The main difference between the studies in CKD and chronic HF is that in all chronic HF trials a placebo group was included, whereas in all trials with CKD both arms included active treatment with ESPs. Also, average renal function is, as expected, considerably higher in the patients with chronic HF included in the evaluated trials compared with the CKD trials and may influence the effects of ESP treatment. The aetiology of anaemia in patients with CKD and chronic HF may be different. Relatively low endogenous erythropoietin levels are seen in the majority of patients with CKD.20 However, in chronic HF most patients have normal or raised erythropoietin levels.21 22 This may be related to higher cytokine levels, resulting in resistance of the bone marrow to endogenous erythropoietin levels.23 In addition, iron deficiency and expanded plasma volume may play a major role in the anaemia seen in patients with chronic HF.24 25 Investigating the aetiology of anaemia in chronic HF is important for providing optimal treatment.

Although anaemia is a prime risk predictor in HF, current guidelines of chronic HF management do not recommend the use of ESPs. However, these agents are at present widely used in anaemic patients with chronic HF, which may in part be related to the treatment regimen of the comorbid renal failure so commonly found in these patients.26 For patients with anaemia and severe renal failure there are guidelines for initiation of ESP treatment, regardless of the presence or absence of chronic HF. Our results are reassuring in that they show no increased incidence of adverse events from ESP treatment in HF, with a suggestion of a beneficial effect on HF hospitalisation, a risk factor which has been clearly linked to future events and mortality.27

Although our data are potentially interesting, it goes without saying that a meta-analysis is not a substitute for an adequately powered randomised trial. The Red-HF study, presently enrolling patients with haemoglobin levels between 9 and 12 g/dl, will examine the value of ESPs in this patient population.28 An estimated 3400 patients with haemoglobin levels between 9 and 12 g/dl (5.6 and 7.5 mmol/l) will be randomised to either placebo or darbepoetin alfa with a target haemoglobin level of 13 g/dl. In the current guidelines the haemoglobin target for patients with severe renal failure is 12 g/dl. Currently, it is unknown what the optimal haemoglobin level is for patients with chronic HF, which might differ from that for patients with severe renal failure. The Red-HF morbidity/mortality trial specifically designed for anaemic patients with chronic HF will further explore the effects of anaemia correction with ESPs in chronic HF.

There are several limitations some of them are inherent to meta-analyses in general. There might have been a bias towards reporting of positive trials. To minimise publication bias we performed a rigorous literature search, in accordance with the QUOROM guidelines. The numbers of events in the reported studies are low. In addition, heterogeneity should be kept in mind when interpreting the results of meta-analyses. However, in the current report, there was very little heterogeneity in the studied outcomes. Furthermore, not all patients were treated with iron. In two studies only patients randomised to ESP received iron.7 16 Since iron might have a beneficial effect in anaemic patients with chronic HF this may lead to bias. On the other hand, these two studies comprised only 55 patients; which represents 8.5% of the total number of patients included in our meta-analysis. HF hospitalisation is a “soft” end point; however, most studies had a double-blind, placebo-controlled design (n = 565 patients). Only the two largest studies (n = 484 patients) had an adjudication committee which reviewed the cause of death and hospitalisation. The other studies based their results on the individual doctor’s judgment.

In summary, among patients with chronic HF, treatment with ESPs is not associated with a higher mortality rate or more adverse events, whereas a beneficial effect on chronic HF hospitalisation was seen. These outcomes are in contrast with studies in cancer and kidney disease and support a large phase III morbidity and mortality trial in anaemic patients with chronic HF.