Erythropoiesis‐stimulating agents for anaemia in adults with chronic kidney disease: a network meta‐analysis

Suetonia C Palmer; Valeria Saglimbene; Dimitris Mavridis; Georgia Salanti; Jonathan C Craig; Marcello Tonelli; Natasha Wiebe; Giovanni FM Strippoli

doi:10.1002/14651858.CD010590.pub2

Erythropoiesis‐stimulating agents for anaemia in adults with chronic kidney disease: a network meta‐analysis

Authors' declarations of interest

Version published: 08 December 2014 Version history

https://doi.org/10.1002/14651858.CD010590.pub2

Collapse all Expand all

Abstract

available in

Background

Several erythropoiesis‐stimulating agents (ESAs) are available for treating anaemia in people with chronic kidney disease (CKD). Their relative efficacy (preventing blood transfusions and reducing fatigue and breathlessness) and safety (mortality and cardiovascular events) are unclear due to the limited power of head‐to‐head studies.

Objectives

To compare the efficacy and safety of ESAs (epoetin alfa, epoetin beta, darbepoetin alfa, or methoxy polyethylene glycol‐epoetin beta, and biosimilar ESAs, against each other, placebo, or no treatment) to treat anaemia in adults with CKD.

Search methods

We searched the Cochrane Renal Group's Specialised Register to 11 February 2014 through contact with the Trials' Search Co‐ordinator using search terms relevant to this review.

Selection criteria

Randomised controlled trials (RCTs) that included a comparison of an ESA (epoetin alfa, epoetin beta, darbepoetin alfa, methoxy polyethylene glycol‐epoetin beta, or biosimilar ESA) with another ESA, placebo or no treatment in adults with CKD and that reported prespecified patient‐relevant outcomes were considered for inclusion.

Data collection and analysis

Two independent authors screened the search results and extracted data. Data synthesis was performed by random‐effects pairwise meta‐analysis and network meta‐analysis. We assessed for heterogeneity and inconsistency within meta‐analyses using standard techniques and planned subgroup and meta‐regression to explore for sources of heterogeneity or inconsistency. We assessed our confidence in treatment estimates for the primary outcomes within network meta‐analysis (preventing blood transfusions and all‐cause mortality) according to adapted GRADE methodology as very low, low, moderate, or high.

Main results

We identified 56 eligible studies involving 15,596 adults with CKD. Risks of bias in the included studies was generally high or unclear for more than half of studies in all of the risk of bias domains we assessed; no study was low risk for allocation concealment, blinding of outcome assessment and attrition from follow‐up. In network analyses, there was moderate to low confidence that epoetin alfa (OR 0.18, 95% CI 0.05 to 0.59), epoetin beta (OR 0.09, 95% CI 0.02 to 0.38), darbepoetin alfa (OR 0.17, 95% CI 0.05 to 0.57), and methoxy polyethylene glycol‐epoetin beta (OR 0.15, 95% CI 0.03 to 0.70) prevented blood transfusions compared to placebo. In very low quality evidence, biosimilar ESA therapy was possibly no better than placebo for preventing blood transfusions (OR 0.27, 95% CI 0.05 to 1.47) with considerable imprecision in estimated effects. We could not discern whether all ESAs were similar or different in their effects on preventing blood transfusions and our confidence in the comparative effectiveness of different ESAs was generally very low. Similarly, the comparative effects of ESAs compared with another ESA, placebo or no treatment on all‐cause mortality were imprecise.

All proprietary ESAs increased the odds of hypertension compared to placebo (epoetin alfa OR 2.31, 95% CI 1.27 to 4.23; epoetin beta OR 2.57, 95% CI 1.23 to 5.39; darbepoetin alfa OR 1.83, 95% CI 1.05 to 3.21; methoxy polyethylene glycol‐epoetin beta OR 1.96, 95% CI 0.98 to 3.92), while the effect of biosimilar ESAs on developing hypertension was less certain (OR 1.18, 95% CI 0.47 to 2.99). Our confidence in the comparative effects of ESAs on hypertension was low due to considerable imprecision in treatment estimates. The comparative effects of all ESAs on cardiovascular mortality, myocardial infarction (MI), stroke, and vascular access thrombosis were uncertain and network analyses for major cardiovascular events, end‐stage kidney disease (ESKD), fatigue and breathlessness were not possible. Effects of ESAs on fatigue were described heterogeneously in the available studies in ways that were not useable for analyses.

Authors' conclusions

In the CKD setting, there is currently insufficient evidence to suggest the superiority of any ESA formulation based on available safety and efficacy data. Directly comparative data for the effectiveness of different ESA formulations based on patient‐centred outcomes (such as quality of life, fatigue, and functional status) are sparse and poorly reported and current research studies are unable to inform care. All proprietary ESAs (epoetin alfa, epoetin beta, darbepoetin alfa, and methoxy polyethylene glycol‐epoetin beta) prevent blood transfusions but information for biosimilar ESAs is less conclusive. Comparative treatment effects of different ESA formulations on other patient‐important outcomes such as survival, MI, stroke, breathlessness and fatigue are very uncertain.

For consumers, clinicians and funders, considerations such as drug cost and availability and preferences for dosing frequency might be considered as the basis for individualising anaemia care due to lack of data for comparative differences in clinical benefits and harms.

PICOs

Population

Intervention

Comparison

Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Plain language summary

available in

The relative safety and effectiveness of different epoetin drugs for treating anaemia in people with chronic kidney disease

Several drugs are available to treat anaemia for people who have kidney disease but whether these drugs are similar or different in their ability to improve symptoms of anaemia, such as tiredness and breathlessness, and whether they are equally safe based on their risks of causing a stroke or a heart attack, is not clear. This is because research studies that compare the effects of one drug directly with another are not common. We have found 56 studies that measure the safety and how these drugs help to improve how patients who have kidney disease feel, function and survive that have involved 15,596 people. Our last search of the literature was in February 2014.

We are somewhat confident that four of the drugs (epoetin alfa, epoetin beta, darbepoetin beta and methoxy polyethylene glycol‐epoetin beta) are better than a placebo injection to prevent patients needing to have a blood transfusion. We are less certain that biosimilar drugs are better than placebo to help patients avoid a blood transfusion.

All erythropoiesis‐stimulating agents cause high blood pressure, but we cannot be very sure if biosimilar products have effects on blood pressure. We cannot be confident in the other important effects of these drugs ‐ we are not sure whether the drugs are similar or different in their effects on the chances of death, a heart attack or stroke; the risk of having a clot in a fistula or vascular catheter needed for dialysis; or the chances of needing dialysis for people who have milder kidney disease. We are unsure whether the different drugs are better at improving symptoms such as tiredness or breathlessness than others as the available research studies generally do not measure these aspects of treatment very well.

Overall, whether different drugs are safer or better at treating symptoms of anaemia for people with kidney disease is poorly known. It is likely that most if not all the drugs prevent the need for a patient to require a blood transfusion. The choice of which drug to use to treat anaemia when a patient has kidney disease can be decided between patients and health professionals based on shared preferences for how frequently the drug is given and considering drug costs and availability.

Authors' conclusions

available in

Implications for practice

Our review includes direct and indirect comparisons of ESAs for anaemia in CKD and is currently the best available evidence for consumers and health professionals on the relative safety and efficacy of differing ESA prescribing patterns. On the basis of moderate to low quality evidence, epoetin alfa, epoetin beta, darbepoetin alfa and methoxy polyethylene glycol‐epoetin beta are all superior to placebo for preventing blood transfusions. It is unclear whether ESA formulations have similar or different efficacy for patient‐centred benefits including blood transfusions, fatigue and breathlessness in very low quality data.

There are presently insufficient high quality data for a definitive statement on whether differing ESAs differ from placebo or each other for their effects on mortality and cardiovascular outcomes including stroke, myocardial infarction or death due to a cardiovascular cause.

In general, data for biosimilar ESA formulations are sparse and very low quality, and are not suitable to inform patients and health providers about the balance of their benefits and risks.

As seen in earlier reviews, reporting of treatment effects of ESAs on potentially patient‐important outcomes is infrequent and heterogeneous, precluding a robust understanding of the effects of ESA therapy on the way patients feel and function. Given the inconclusive effects of the differing ESAs on quality of life and survival, decisions about different agents in clinical practice and policy might be based on drug cost and availability and patient preferences for treatment frequency until additional data become available.

Implications for research

We believe there is a key need that the research agenda should address. Large RCTs of ESAs on patients‐centred outcomes that are considered most relevant to patients and health services should be undertaken using consistent methods for reporting of outcomes and consideration of clinically important benefits for these drugs. Currently, given the lack of evidence for treatment benefits from ESA therapy (Clement 2009; Phrommintikul 2007) studies of treatment effects on health‐related quality of life and key symptoms of anaemia and advanced kidney disease are required before widespread ongoing use of these agents can be justified. Additional work to identify core research outcomes that are priorities for consumers and health providers would inform the design of future studies for treatment of anaemia of CKD to increase their research relevance to health and clinical practice.

Summary of findings

Open in table viewer

Summary of findings for the main comparison. Erythropoiesis‐stimulating agents (ESAs) for anaemia in adults with chronic kidney disease (CKD)

ESAs for anaemia in adults with CKD
Intervention	Comparison/intervention	Nature of the evidence	Confidence in the evidence	Reasons for downgrading our confidence in the evidence*	Network treatment estimate OR (95% CI)
Preventing blood transfusion
Epoetin alfa	Placebo	Mixed	Low	Study limitations (‐1) Inconsistency (‐1)	0.18 (0.05 to 0.59)
Epoetin beta	Placebo	Mixed	Low	Study limitations (‐1) Inconsistency (‐1)	0.09 (0.02 to 0.38)
Darbepoetin alfa	Placebo	Mixed	Moderate	Inconsistency (‐1)	0.17 (0.05 to 0.57)
Methoxy polyethylene glycol‐epoetin beta	Placebo	Indirect	Low	Study limitations (‐1) Inconsistency (‐1)	0.15 (0.03 to 0.70)
Biosimilar ESA	Placebo	Indirect	Very low	Study limitations (‐1) Imprecision (‐1) Inconsistency (‐1)	0.27 (0.05 to 1.47)
Epoetin alfa	Epoetin beta	Indirect	Very low	Study limitations (‐1) Imprecision (‐1) Imprecision (‐1)	2.04 (0.38 to 11.0)
Epoetin alfa	Darbepoetin alfa	Mixed	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	1.06 (0.35 to 3.29)
Epoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Indirect	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	1.14 (0.27 to 4.97)
Epoetin alfa	Biosimilar ESA	Mixed	Very low	Study limitations (‐1) Imprecision (‐1) Imprecision (‐1)	0.66 (0.19 to 2.28)
Epoetin beta	Darbepoetin alfa	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.52 (0.10 to 2.67)
Epoetin beta	Methoxy polyethylene glycol‐epoetin beta	Mixed	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.56 (0.11 to 3.00)
Epoetin beta	Biosimilar ESA	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.33 (0.04 to 2.60)
Darbepoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Mixed	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	1.08 (0.38 to 3.04)
Darbepoetin alfa	Biosimilar ESA	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.62 (0.12 to 3.30)
Methoxy polyethylene glycol‐epoetin beta	Biosimilar ESA	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.58 (0.09 to 3.92)
All‐cause mortality
Epoetin alfa	Placebo	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	1.25 (0.71 to 2.21)
Epoetin beta	Placebo	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.82 (0.45 to 1.48)
Darbepoetin alfa	Placebo	Mixed	Moderate	Imprecision (‐1)	1.06 (0.91 to 1.24)
Methoxy polyethylene glycol‐epoetin beta	Placebo	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	1.16 (0.74 to 1.82)
Biosimilar ESA	Placebo	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	1.31 (0.65 to 2.62)
Epoetin alfa	Epoetin beta	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	1.53 (077 to 3.03)
Epoetin alfa	Darbepoetin alfa	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	1.17 (0.68 to 2.05)
Epoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	1.08 (0.54 to 2.15)
Epoetin alfa	Biosimilar ESA	Mixed	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	0.95 (0.62 to 1.44)
Epoetin beta	Darbepoetin alfa	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.77 (0.43 to 1.38)
Epoetin beta	Methoxy polyethylene glycol‐epoetin beta	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.71 (0.35 to 1.42)
Epoetin beta	Biosimilar ESA	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.62 (0.29 to 1.37)
Darbepoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Mixed	Very low	Study limitations (‐2) Imprecision (‐1)	0.91 (0.60 to 1.40)
Darbepoetin alfa	Biosimilar ESA	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	0.81 (0.41 to 1.61)
Methoxy polyethylene glycol‐epoetin beta	Biosimilar ESA	Indirect	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	0.88 (0.40 to 1.97)
CI: Confidence interval; OR: Odds Ratio *There was moderate heterogeneity in the network for preventing blood transfusion (τ = 0.89 which was between the 50th and 75th quartile of empirical distributions of heterogeneity variances specific to the type of outcome and types of treatments being compared) (Turner 2012) We did not downgrade for reasons of indirectness or publication bias as insufficient studies contributed to network treatment estimates to draw meaningful conclusions. We downgraded for inconsistency when the network did not include a closed loop of evidence for the comparison and accordingly the presence of inconsistency could not be excluded.
GRADE Working Group grades of evidence (GRADE: Rating the quality of evidence 2011) High quality: We are very confidence that the true effect lies close to that of the estimate of effect Moderate quality: We are moderately confident in the estimate of effect: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of effect Very low quality: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect

Background

available in

Description of the condition

Anaemia, literally meaning lack of blood, is defined as "a condition in which the number of red blood cells or their oxygen‐carrying capacity is insufficient to meet physiological needs" (http://www.who.int/topics/anaemia/en/). Circulating red blood cells transport oxygen to tissues bound to iron ions within the metalloprotein, haemoglobin. In anaemia, insufficient numbers of circulating red blood cells or inadequate quantities of iron or functional haemoglobin are available to transport and release oxygen to tissues, which is essential for aerobic (oxygen‐dependent) metabolism. Anaemia, defined by the World Health Organization as a haemoglobin level below 130 g/L in men and below 120 g/L in women, affects approximately a quarter of the world's population, particularly children and pregnant women (WHO 2008). Anaemia is common in the expanding global populations of chronic disease including people affected by solid malignancies (50%), blood cancers (60% to 70%) (Ludwig 2004), human immunodeficiency virus (HIV) causing acquired immunodeficiency syndrome (AIDS; 40%) (Shah 2007), chronic heart failure (20%) (Ezekowitz 2003) and nearly all individuals who have advanced chronic kidney disease (CKD). Symptoms caused by insufficient oxygen delivery to tissues in anaemia include weakness and fatigue, breathlessness, light‐headedness, and palpitations. Observational cohort data show that anaemia in people who have chronic disease is also consistently associated with negative effects on quality of life (Lefebvre 2006), role function (Ludwig 2004; Semba 2005), and survival (Caro 2001; Groenveld 2008; Locatelli 2004; Melekhin 2012).

Description of the intervention

Recombinant erythropoietin and its synthetic derivatives (epoetin alfa, epoetin beta, darbepoetin alfa, methoxy polyethylene glycol‐epoetin beta; collectively known as erythropoiesis‐stimulating agents (ESAs)), are widely used to treat anaemia. Erythropoietin is a glycoprotein made by peritubular cells in the kidney (with an additional smaller contribution from liver cells (15% total)) and is released in response to low tissue oxygen levels (hypoxia) through the actions of hypoxia‐inducible factor to stimulate the formation and viability of red blood cells in the bone marrow (erythropoiesis). The average red blood cell survives in the circulation for 120 days although red cell survival is reduced by chronic disease. Causes of anaemia are numerous and include: reduced production of erythropoietin in response to hypoxia (CKD; chronic inflammatory conditions); abnormal bone marrow responses to the actions of erythropoietin (chronic inflammatory conditions, bone marrow failure due to infiltration or drug‐related therapy); insufficient iron stores; abnormal production or function of haemoglobin (thalassaemia or haemoglobinopathies); excessive red blood cell losses (destruction within the circulation or haemorrhage); or reduced red blood cell survival (Figure 1).

Figure 1

Overview of anaemia in chronic disease

Before the development of recombinant human erythropoietin (rHuEPO) in the late 1980s (Eschbach 1987), blood transfusions and iron supplementation (both oral and intravenous (IV)) were the mainstays of treatment for anaemia in populations with severe CKD, in which haemoglobin levels were commonly in the range of 70 to 80 g/L. Androgen treatment for anaemia was also used in CKD but provided small and unsustained responses in haemoglobin levels and was poorly tolerated (Neff 1981). In the pre‐recombinant erythropoietin era, blood transfusions effectively increased haemoglobin levels to provide acute symptom relief but were associated with hospitalisation, iron overload, antibody formation against blood cell antigens, sensitisation to transplant antigens, and transfusion‐related infections, particularly viral hepatitis. Technological advances and successful cloning of the erythropoietin gene enabled large‐scale production of rHuEPO which effectively and rapidly increases haemoglobin levels when administered IV or subcutaneously (SC). The United States Food and Drug Administration (FDA) approved rHuEPO for the treatment of anaemia in people with CKD on dialysis in 1989 and broadened approval to include people with CKD without dialysis, and in patients with HIV and anaemia on zidovudine (AZT) in 1990.

Clinical guidelines published soon after initial drug approval suggested that patients with CKD and haemoglobin concentrations below 80 g/L who were symptomatic should receive ESA treatment in conjunction with sufficient iron supplementation once other causes of anaemia were excluded (Macdougall 1990). However, rapid widespread uptake of ESAs occurred in numerous clinical settings, and by 2007, clinical practice guidelines recommended the use of ESAs to achieve target haemoglobin levels of 110 to 120 g/L in people with CKD (KDOQI 2007). ESA prescription also subsequently expanded to treat anaemia in cancer and heart failure populations, as well as for people undergoing surgery likely to require blood transfusion who could not undergo pre‐operative blood collection. Presently, epoetin alfa is approved by the FDA for treatment of anaemia due to CKD, zidovudine in HIV‐infected patients, effects of concomitant myelosuppressive chemotherapy and to reduce red blood cell transfusions in patients undergoing elective, noncardiac, nonvascular surgery. Darbepoetin alfa is currently approved by the FDA for the treatment of anaemia resulting from CKD or myelosuppressive chemotherapy (FDA website).

How the intervention might work

Despite an association between low haemoglobin levels and higher mortality in uncontrolled studies, prompting speculation that correcting anaemia with ESA therapy might lower cardiovascular events and mortality, the opposite was observed in subsequent meta‐analyses of RCTs (Bohlius 2009; Palmer 2010; Phrommintikul 2007; Strippoli 2006). Correction of anaemia and maintenance of haemoglobin levels to near normal levels with ESAs reduced the need for red blood cell transfusions, but increased mortality, cardiovascular events and cancer progression, without consistently improving quality of life. The precise mechanisms for treatment‐related harm are not understood, but observational studies suggest that impaired haemoglobin responses to erythropoietin treatment, together with higher erythropoietin doses are associated with increased treatment‐related toxicity (Kilpatrick 2008; Szczech 2008).

Treatment guidelines for ESAs to treat anaemia have become more conservative over the last decade and FDA labelling now suggests that ESA treatment should be considered in people with CKD when the haemoglobin level is less than 100 g/L, and treatment objectives are to increase haemoglobin levels sufficient to reduce the need for red cell transfusions (FDA website). Clinical practice guidelines have also responded to increasing evidence of harm when higher haemoglobin levels are targeted by ESA treatment (Bohlius 2009; Palmer 2010; Phrommintikul 2007). Recent clinical practice guidelines for the use of ESAs to treat anaemia in CKD suggest the potential benefits of reducing blood transfusions and anaemia‐related symptoms should be balanced against the risks of harm (e.g. stroke, vascular access thrombosis and hypertension) for individual patients. Currently, guidelines do not suggest specific haemoglobin targets for patients not treated with dialysis, while for dialysis patients, the recommended approach is to use ESA therapy to avoid a haemoglobin level below 9.0 g/dL (KDIGO 2010).

Why it is important to do this review

Darbepoetin alfa and methoxy polyethylene glycol‐epoetin beta (a continuous erythropoietin‐receptor activator (CERA)) are newer synthetic forms of naturally‐occurring erythropoietin that have a longer duration of action (Macdougall 2008). These agents have similar effects on haemoglobin concentrations as epoetin alfa and beta and require less frequent administration (Levin 2007; Macdougall 2001). Darbepoetin alfa treatment in people with earlier stages of CKD and diabetes mellitus has been shown to nearly halve the risk of blood transfusion but has no beneficial effects on survival and increases the risk of stroke and death related to cancer recurrence (TREAT Study 2005).

The apparent narrow therapeutic balance between potential treatment benefits (avoidance of red blood cell transfusions and improving symptoms of anaemia) and hazards (cardiovascular events and mortality) together with the availability of several agents in this drug class (epoetin alfa, epoetin beta, darbepoetin alfa, methoxy polyethylene glycol‐epoetin beta and biosimilar epoetins) to treat anaemia builds the case for a comprehensive and systematic head‐to‐head comparison of the available ESAs to treat anaemia. However, large‐scale studies directly comparing different epoetins have been relatively uncommon, and the comparative efficacy and safety of each agent relative to each other is poorly understood.

In addition, the expiration of several epoetin patents has prompted companies to produce similar biological medicinal products that are second versions of biological medicines that depend on the same mechanism of action and are intended to be used for the same therapeutic indication as the earlier product, known as "biosimilars" or "follow‐on biologicals". Global clinical guidelines assume that available epoetins are all equally safe and effective, including true biosimilar products (KDIGO 2010), although the drug formulations differ widely in molecular structure, cost, availability and duration of action.

While patient and policy decisions about anaemia management of CKD are highly dependent on the comparative effectiveness of ESAs, existing studies have focused mainly on the evaluation of targeting differing haemoglobin levels with treatment. Head‐to‐head studies of ESAs in CKD are lacking. To overcome the known limitations of single randomised studies, we have conducted a systematic review of the literature and a network meta‐analysis to estimate the comparative efficacy and safety of ESAs for treating anaemia in people with CKD.

Objectives

available in

Methods

available in

Criteria for considering studies for this review

Types of studies

We included all RCTs comparing ESA versus ESA, placebo or no treatment to treat anaemia in people with CKD. We did not restrict inclusion based on language of publication. We did not include quasi‐RCTs (studies in which treatment allocation was by date of birth, alternation, or similar predictable method). We included studies in which allocation to treatment was not adequately concealed but considered study methodological quality in our analyses and discussion.

Types of participants

Inclusion criteria

Studies in adults aged 18 years or older with anaemia due to CKD were included. CKD was characterised by clinically relevant proteinuria, haematuria, and/or structural kidney disease with or without estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m², recipients of a kidney transplant, and people with Stage 5 CKD treated with dialysis (KDIGO 2013).

Exclusion criteria

As network meta‐analysis requires reasonable homogeneity in study design and populations, we excluded data in children and from studies in which follow‐up was less than three months.

Types of interventions

We included studies of ESAs (epoetin alfa, epoetin beta, darbepoetin beta, methoxy polyethylene glycol‐epoetin beta, biosimilar) to treat or prevent anaemia in CKD administered via any route (IV or SC), compared with each other, placebo or no treatment. Dose adaptation of ESAs and non‐randomised iron supplementation depending on haematological response were allowed. We included studies in which iron was administered as a randomised intervention in all arms of the study.

We coded the comparisons within a study where iron was a randomised co‐intervention in all study arms as follows.

ESA1 plus iron (any route) versus ESA2 plus iron (any route) = ESA1 versus ESA2
ESA plus oral iron versus oral iron = ESA versus no treatment
ESA plus oral iron versus oral iron plus placebo injection = ESA versus placebo
ESA plus intravenous iron versus intravenous iron plus placebo injection = ESA versus placebo
ESA plus intravenous iron versus intravenous iron = ESA versus no treatment.

We excluded studies in which iron therapy was a randomised co‐intervention combined with an ESA in a single arm of the study (e.g. ESA plus iron versus ESA alone, ESA plus iron versus placebo). Studies of hypoxia‐inducible factor stabilisers and peginesatide were excluded.

Types of outcome measures

We evaluated the following outcomes occurring at any time during study follow‐up.

Primary outcomes

We estimated the comparative effects of the competing interventions according to the following outcomes:

Response to treatment

Preventing blood transfusion

Safety

All‐cause mortality.

Secondary outcomes

Response to treatment

Fatigue (as defined by study authors)
Dyspnoea (as defined by study authors)

Safety

Cardiovascular mortality
Fatal or nonfatal MI
Fatal or nonfatal stroke
Vascular access thrombosis
Major adverse cardiovascular event (as adjudicated by investigators)
End‐stage kidney disease (ESKD).

Search methods for identification of studies

Electronic searches

We searched the Cochrane Renal Group's Specialised Register to 11 February 2014 through contact with the Trials' Search Co‐ordinator using search terms relevant to this review. The Cochrane Renal Group’s Specialised Register contains studies identified from several sources:

Monthly searches of the Cochrane Central Register of Controlled Trials (CENTRAL)
Weekly searches of MEDLINE OVID SP
Handsearching of renal‐related journals and the proceedings of major renal conferences
Searching of the current year of EMBASE OVID SP
Weekly current awareness alerts for selected renal journals
Searches of the International Clinical Trials Register (ICTRP) Search Portal and ClinicalTrials.gov.

Studies contained in the Specialised Register are identified through search strategies for CENTRAL, MEDLINE, and EMBASE based on the scope of the Cochrane Renal Group. Details of these strategies, as well as a list of handsearched journals, conference proceedings and current awareness alerts, are available in the Specialised Register section of information about the Cochrane Renal Group.

See Appendix 1 for search terms used in strategies for this review.

Searching other resources

Database of Abstracts of Reviews of Effects (DARE) (The Cochrane Library)
Health Technology Assessment (HTA) database (The Cochrane Library)
NHS Economic Evaluation Database (The Cochrane Library)
Reference lists of review articles, relevant studies and clinical practice guidelines
Letters seeking information about unpublished or incomplete studies to investigators known to be involved in previous studies.

Data collection and analysis

Selection of studies

The search strategy described was used to obtain titles and abstracts of studies that were relevant to the review. The titles and abstracts were screened independently by two authors, who discarded studies that were not applicable; however studies and reviews that might have included relevant data or information on studies were retained initially. Two authors independently assessed retrieved abstracts, and if necessary the full text, of these studies to determine which studies satisfied the inclusion criteria. Systematic reviews were screened to identify any studies not retrieved by the electronic database search.

Data extraction and management

Data extraction was carried out independently by two authors using standard data extraction forms. Data were cross checked between authors and discussed. Studies reported in non‐English language journals were translated electronically before assessment. Where more than one publication of one study existed, reports were grouped together and the publication with the most complete data was used in the analyses. Where relevant outcomes were only published in earlier versions, these data were used. Any disagreements in data extraction were discussed with a third author.

Any further information required from the original authors or sponsors of studies included in the review was requested by written correspondence (e.g. emailing or writing to corresponding author/s) and any relevant information obtained in this manner was included in the review. Data requested included numbers of events and numbers of participants at risk for important dichotomous clinical outcomes (blood transfusions, all‐cause mortality, cardiovascular mortality, fatal or nonfatal stroke, fatal or nonfatal MI, vascular access thrombosis, ESKD, major adverse cardiovascular events, fatigue, breathlessness). We also requested additional information on the use of iron supplementation in treatment arms where this was not clear from reading the study report.

Assessment of risk of bias in included studies

The following items were independently assessed by two authors using the risk of bias assessment tool (Higgins 2011) (see Appendix 2).

Was there adequate sequence generation (selection bias)?
Was allocation adequately concealed (selection bias)?
Was knowledge of the allocated interventions adequately prevented during the study (detection bias)?
- Participants and personnel
- Outcome assessors
Were incomplete outcome data adequately addressed (attrition bias)?
Are reports of the study free of suggestion of selective outcome reporting (reporting bias)?
Was the study apparently free of other problems that could put it at a risk of bias (imbalance in interventions, publication only as abstract or letter, premature termination of study and industry sponsor involvement in authorship or data management and analysis)?

Measures of treatment effect

Relative treatment effects

We calculated comparative effect sizes for pairwise and network meta‐analysis as odds ratios (ORs) with their 95% confidence intervals (CIs).

Relative treatment rankings

To rank the treatments available according to safety or efficacy, we planned to use the surface under the cumulative ranking (SUCRA) probabilities which express as percentages each intervention to an imaginary intervention that is always the best without uncertainty (Salanti 2011). For example, a SUCRA of 80% means that the drug achieved 80% of the effectiveness of this imaginary drug, and accordingly, larger SUCRAs denote greater efficacy. However, the large uncertainty in the resulting estimates rendered ranking of the competing treatments imprecise.

Assessment of clinical and methodological heterogeneity within treatment comparisons

To evaluate the presence of clinical heterogeneity, we generated descriptive statistics for the population characteristics across all eligible studies that compared each pair of interventions. We assessed the presence of clinical heterogeneity within pairwise comparisons by comparing these characteristics.

Assessment of transitivity across treatment comparisons

The assumption of transitivity ‐ that one can learn about treatment A versus treatment B via treatment C (e.g. learning about epoetin alfa versus darbepoetin alfa via placebo) ‐ underlies network meta‐analysis (Salanti 2012). Evaluation of the assumption is important and its plausibility determines the validity of the network meta‐analysis results. We inferred about the assumption of transitivity:

We assessed whether the included interventions were similar when they were evaluated in studies with different designs, for example, whether ESAs are administered the same way in studies comparing ESAs to placebo and in those comparing ESAs to other ESAs
We compared the distribution of the potential effect modifiers (age, stage of CKD, duration of treatment) across the different pairwise comparisons.

Data synthesis

Methods for direct treatment comparisons

First, we conducted pairwise meta‐analyses by synthesising studies that compared the same interventions using a random‐effects model (DerSimonian 1986) that contained two or more studies. We compared treatments that used the same haemoglobin target (e.g. epoetin high target versus darbepoetin high target). For dichotomous outcomes (avoiding red blood cell transfusions; all‐cause and cardiovascular mortality; major cardiovascular event; fatal or nonfatal myocardial infarction; fatal or nonfatal stroke; vascular access thrombosis; ESKD; fatigue; breathlessness) results were expressed as an OR with 95% CI.

Methods for indirect and mixed comparisons

To determine comparative efficacy and safety, we then conducted network meta‐analysis. Network meta‐analysis is a method of synthesising information from a network of studies addressing the same questions but involving different interventions. Joint analysis of data within a network framework allows novel inferences on treatment comparisons that have not been previously addressed directly in any studies, and it may increase precision for comparisons with few data (Caldwell 2010; Lu 2004; Salanti 2008). For a given comparison, say A versus B, direct evidence is provided by studies that compare these two treatments directly (epoetin alfa versus darbepoetin alfa) as in standard direct comparisons meta‐analysis. In addition, indirect evidence for A versus B can be provided if studies that compare A versus C and B versus C are analysed jointly (e.g. epoetin alfa versus placebo studies and darbepoetin alfa versus placebo studies can allow indirect comparison of epoetin alfa versus darbepoetin alfa via the use of placebo). Network meta‐analysis aims to combine the direct and indirect evidence into a single effect size and thus may help to increase the precision of the comparison, while randomisation is respected. The combination of direct and indirect evidence for any given treatment comparison can be extended when ranking more than three types of treatments according to their effectiveness or safety; every study contributes evidence in the network about a subset of the competing treatments. We performed network meta‐analysis in STATA (www.stata.com) using the 'mvmeta' command (White 2012) and self‐programmed STATA routines described in Chaimani 2013 and available at http://www.mtm.uoi.gr/index.php/stata‐routines‐for‐network‐meta‐analysis.

Assessment of statistical heterogeneity

Assumptions when estimating heterogeneity

In standard pairwise meta‐analyses we estimated different heterogeneity variances for each pairwise comparison. In network meta‐analysis we assumed a common estimate for the heterogeneity variance across the different comparisons.

Measures and tests for heterogeneity

We evaluated for the presence of heterogeneity within meta‐analyses using the Cochran Q test and I² statistic (Higgins 2003) that measures the percentage of variability that cannot be attributed to random error. We considered the I² thresholds to represent heterogeneity that might not be important (0% to 40%), might be moderate heterogeneity (30% to 60%), might be substantial heterogeneity (50% to 90%), and was considerable heterogeneity (75% to 100%) considering also the magnitude and direction of treatment effects and strength of evidence for heterogeneity (P value from the Chi² test) (Higgins 2011). The assessment of statistical heterogeneity in the entire network was based on the magnitude of the heterogeneity variance parameter (τ²) estimated from the network meta‐analysis models. We compared the magnitude of a common heterogeneity variance for the specific network of interest with an empirical distribution of heterogeneity variances specific to the type of outcome and the types of treatments being compared (Turner 2012).

Assessment of statistical inconsistency

Local approaches for evaluating inconsistency

To evaluate the presence of inconsistency locally, we used the loop‐specific approach. A loop of evidence is formed by at least three treatment pairs which have been compared in studies forming a closed path. Indirect evidence can be contrasted to direct evidence and their difference defines their disagreement (inconsistency factor). To infer whether the inconsistency factor is incompatible with zero, we looked at the magnitude of the inconsistency factors and their 95% confidence intervals (Bucher 1997). We extended analysis to all closed triangular and quadratic loops assuming a single loop‐specific heterogeneity and examine the estimates of inconsistency together with 95% confidence intervals for each loop using a graphical representation (Salanti 2009), This approach can be easily applied and indicates loops with large inconsistency, but cannot infer consistency of the entire network or identify the particular comparison that is problematic. It should be noted that in a network of evidence there may be many loops and estimates of inconsistency factors and with multiple testing there is an increased likelihood that we might find an inconsistent loop by chance. Therefore, we were cautious deriving conclusions from this approach.

Global approaches for evaluating inconsistency

To check the assumption of consistency in the entire network, we used the design‐by‐treatment interaction model as fully explained in Higgins 2012 (pp. 102 to 103). This method accounts for different sources of inconsistency that can occur when studies with different designs (two‐arm studies versus three‐arm studies) give different results as well as disagreement between direct and indirect evidence. Using this approach, we inferred about the presence of inconsistency from any source in the entire network based on a Chi² test. The design‐by‐treatment model was performed in STATA using the 'mvmeta' command. Inconsistency and heterogeneity are interwoven: to distinguish between these two sources of variability we employed the I² for inconsistency that measures the percentage of variability that cannot be attributed to random error or heterogeneity (within comparison variability).

It should be noted in general that the power of statistical tests for inconsistency are low, which implies that the absence of statistically significant inconsistency is not evidence of consistency.

Investigation of heterogeneity and inconsistency

We planned to perform meta‐regression or subgroup analyses to explore important heterogeneity and/or inconsistency. When we identified potential evidence of inconsistency and heterogeneity, we first checked for any mistakes and inconsistencies in data extraction and entry. We then evaluated for evidence based on the following effect modifiers as possible sources of inconsistency and/or heterogeneity. However, insufficient data precluded these analyses.

Population: iron status at baseline (iron replete versus iron deficient); stage of CKD (CKD stages 1 to 3, CKD stage 4 to 5, CKD stage 5D, transplantation); baseline haemoglobin (< 10 g/dL, 10 to 12 g/dL, > 12 g/dL); mean age; gender; proportion with diabetes or cardiovascular disease
Intervention: dose, frequency or route; iron supplementation (fixed iron treatment, iron treatment as necessary, or not clear)
Risk of bias: allocation concealment; blinding of outcome assessment; attrition; premature termination of study; publication (full text publication, abstract publication, unpublished data); funding source
Study design: duration of ESA treatment (12 to 16 weeks; 16 to 24 weeks; 24 to 48 weeks; > 48 weeks); duration of follow‐up (≥ 12 months, versus < 12 months); number of participants; date of publication.

Sensitivity analysis

Insufficient data and wide confidence intervals for most treatment estimates precluded additional such analyses.

Summary of findings table

The main results of the review for the primary outcomes (preventing blood transfusion and all‐cause mortality) are presented in a summary of findings table (summary of findings Table for the main comparison). The summary of findings table was provided for the network estimates only and included an overall grading of the evidence for these outcomes.

We used an adapted Grading of Recommendations Assessment, Development approach to grading evidence quality in pairwise meta‐analysis that was developed specifically for network meta‐analysis (Salanti 2014). We considered five components to evidence quality: study limitations, indirectness, inconsistency, imprecision, and publication bias. The interpretations of each of the grades are provided in GRADE: Rating the quality of evidence 2011 and described in the footnote of the summary of findings Table for the main comparison.

For publication bias, due to small numbers of contributing studies (< 10), we considered that funnel plots would have insufficient power and specificity to evaluate for evidence of publication bias and therefore we did not downgrade our confidence in the evidence for reasons of publication bias in this review because of the comprehensive search strategy we followed.

In networks in which there were no closed loops (where three or four treatments were not connected by direct comparisons in individual studies, we couldn't evaluate for consistency between direct information (two drugs compared in a study) and indirect information (two drugs compared via a third treatment strategy using network meta‐analysis). In this situation, we downgraded evidence quality because we could not show absence of inconsistency between these two sources of information.

The adjudication of each component of evidence quality then resulted in maintaining or downgrading evidence quality from a high‐quality rating to moderate, low or very low.

Results

Description of studies

Results of the search

Figure 2 shows the results of the electronic search.

Figure 2

Study flow diagram.

Included studies

The search strategy identified 5223 unique citations. After exclusions during title and abstract screening (4303 citations excluded) and full text analysis (659 citations excluded), 56 studies involving 15,596 participants published between 1989 and 2013 were included in the systematic review and 40 studies involving 12,103 participants could be included within pairwise and network meta‐analyses (Characteristics of included studies). We received unpublished data from investigators of seven studies (Akizawa 2011; CORDATUS Study 2011; EPOCARES Study 2010; Hirakata 2010; Nissenson 2002; Patel 2012; TIVOLI Study 2013).

Median follow‐up was six months (range 3 to 29), with 77% of studies reporting outcomes before 12 months. The average age of participants was highly variable (range 43 to 84 years). Of the 40 studies contributing outcome data, 22 studies included 5583 dialysis patients, two studies provided data for 111 kidney transplant recipients, and 16 studies included in 6409 participants with an estimated GFR between 15 to 90 mL/min/1.73 m². Among studies included in meta‐analyses, seven were placebo controlled (4638 participants) and eight compared ESAs with standard care (787 participants). The remainder were head‐to‐head studies of epoetin alfa versus darbepoetin alfa (8 studies, 1783 participants), epoetin beta versus darbepoetin alfa (1 study, 219 participants), epoetin beta versus methoxy polyethylene glycol‐epoetin beta (2 studies, 462 participants), darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta (5 studies, 1505 participants), epoetin alfa versus biosimilar ESA (8 studies, 2419 participants) and epoetin beta versus biosimilar ESA (1 study, 290 participants).

Other studies

We identified seven ongoing studies (Besarab 2006; NCT00442702; NCT00559273; NCT00717821; NCT00773513; PRIMAVERA Study 2011; STIMULATE Study 2011) and there are five studies which appear to have been completed but as yet there are no results available (Barany 1998; Carrera 2003; Nissenson 2007; Ostrvica 2010; Palazzuoli 2011). These studies will be included in a future update of this review.

Excluded studies

After full‐text review we excluded 29 studies (134 records). Sixteen studies compared the same ESA derivative is the different treatment arms (ACORD Study 2004; Besarab 1998; CHOIR Study 2006; Cianciaruso 2008; CREATE Study 2001; ECAP Study 2006; Eschbach 1989; Foley 2000; Gouva 2004; Johnson 1999; Levin 2005; Linde 2001; Locatelli 2008; Parfrey 2005; Salek 2001; SLIMHEART Study 2004); six studies didn't compare different ESAs (BA16260 Study 2006; BA16285 Study 2007; BA16286 Study 2005; Brier 2010; CAPRIT Study 2012; Macdougall 2007); and one study in which the type of ESA was unknown (Acchiardo 1991a). We also excluded a cross‐over study (Wizemann 2008); four studies of insufficient duration (Kawanishi 2005; Neo‐PDGF Study 2010; Perez‐Oliva 2005; Sja'bani 1997), and one study in which there were insufficient data to determine eligibility (N0055116759).

Studies excluded from the meta‐analyses

The primary reasons for exclusion from meta‐analyses (16 studies involving 3493 participants) were that disaggregated data for different ESA types were not available separately (for example, both epoetin alfa and beta were administered within a single study arm) or that outcome data were not reported in extractable format (Akiba 2010; Arabul 2009; Chen 2008; Coyne 2000; Vanrenterghem 2002; MAXIMA Study 2007; Smith 2007; PROTOS Study 2007; RUBRA Study 2008; Shaheen 1993; Shand 1993; Sikole 1993; Alexander 2007; Teehan 1989; Watson 1990; Van Loo 1996).

Risk of bias in included studies

The risks of bias are summarised in Figure 3 and Figure 4.

Figure 3

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies. The coloured bars correspond to each risk of bias adjudication summarised across all studies. The numbers shown in the bars indicate the raw number of studies which were adjudicated the corresponding risks of bias (green = low risk; yellow = unclear risk; red = high risk). The size of each coloured box within the bars indicates the proportion of studies with the adjudicated risk. For example, there were 7 studies (13%) which were adjudicated as low risk of bias from reported sequence generation methods. The description of the risk domains considered is given in the vertical axis. *Other threats to validity include one or more of: sponsor involved in study design, analysis, or authorship; imbalance between treatment comparisons and/or premature termination of trial

Figure 4

Risk of bias summary: review authors' judgements about each risk of bias item for each included study. Each study is shown in the vertical axis and the corresponding risk of bias for each domain adjudicated by two authors is shown by coloured circles within the grid. Green (+) = low risk, yellow (?) = unclear risk, red (‐) = high risk. Other threats to validity include one or more of: sponsor involved in study design, analysis, or authorship; imbalance between treatment comparisons and/or premature termination of trial

Allocation

Sequence generation

Of 56 included studies, seven (12.5%) reported low risk methods for sequence generation (Akizawa 2011; Vanrenterghem 2002; Haag‐Weber 2009; MAXIMA Study 2007; Nissenson 2002; PATRONUS Study 2010; TREAT Study 2005).

Allocation concealment

Of 56 included studies, 10 (18%) reported adequate methods for allocation concealment (low risk of bias) (ARCTOS Study 2008; Gertz 2010; Goh 2007; Hirakata 2010; Locatelli 2001; MAXIMA Study 2007; PATRONUS Study 2010; PROTOS Study 2007; STRIATA Study 2008; Alexander 2007). The remaining 46 studies (82%) did not provide sufficient information to enable adjudication risk of bias in allocation concealment (unclear risk).

Blinding

There were 16 studies (29%) which reported that participants and investigators were blinded (Bennett 1991; Canadian EPO Study 1990; Gertz 2010; Haag‐Weber 2009; Haag‐Weber 2012; Kleinman 1989; Krivoshiev 2008; Krivoshiev 2010; Martin 2007; Nissenson 1995; Nissenson 2002; Palazzuoli 2007; Shand 1993; Spinowitz 2006; TREAT Study 2005; Watson 1990). There were 35 studies (63%) that were open‐label (high risk of bias) and the remaining three studies (5%) did not provide sufficient information to enable assessment (unclear) (Hori 2004; Kuriyama 1997; TIVOLI Study 2013). Two studies (4%) reported adequate methods of blinding outcome assessment (Canadian EPO Study 1990; TREAT Study 2005) and the remainder did not provide sufficient information to assess risk (unclear risk of bias).

Incomplete outcome data

Seven studies (13%) were judged to meet criteria for low risk of bias (fewer than 10% missing from follow‐up analyses and balanced numbers across intervention groups with similar reasons for loss to follow‐up) for low risk of incomplete outcome data bias (AMICUS Study 2007; ARCTOS Study 2008; CORDATUS Study 2011; Palazzuoli 2007; Sikole 1993; TIVOLI Study 2013; TREAT Study 2005), 31 studies (55%) were at high risk of bias, and the remaining 18 studies (32%) did not provide sufficient information to assess risk of bias (unclear risk).

Selective reporting

Outcomes of interest (mortality, cardiovascular event (fatal or nonfatal) and hypertension) were reported in 22 studies (39%) (Akizawa 2011; ARCTOS Study 2008; Bahlmann 1991; Bennett 1991; Canadian EPO Study 1990; CORDATUS Study 2011; EPOCARES Study 2010; Gertz 2010; Goh 2007; Haag‐Weber 2009; Hirakata 2010; Kleinman 1989; Klinkmann 1992; Krivoshiev 2008; Krivoshiev 2010; Locatelli 2001; Martin 2007; Milutinovic 2006; Patel 2012; STRIATA Study 2008; TREAT Study 2005; Van Loo 1996).

Other potential sources of bias

Industrial sponsor on authorship or involved in data management or analysis

There were 29 studies (51%) that reported the sponsor was involved in authorship of the study report or in data management or analysis (Allon 2002; AMICUS Study 2007; ARCTOS Study 2008; Bahlmann 1991; CORDATUS Study 2011; Coyne 2000; Coyne 2006a; Vanrenterghem 2002; Gertz 2010; Haag‐Weber 2009; Haag‐Weber 2012; Kleinman 1989; Klinkmann 1992; Krivoshiev 2010; Locatelli 2001; Martin 2007; MAXIMA Study 2007; Nissenson 2002; Patel 2012; PATRONUS Study 2010; Smith 2007; PROTOS Study 2007; RUBRA Study 2008; Spinowitz 2006; STRIATA Study 2008; Alexander 2007; TIVOLI Study 2013; TREAT Study 2005; Watson 1990).

Abstract or letter only

Seven studies (18%) were published either as an abstract or letter (Brown 1995; Coyne 2000; Coyne 2006a; Hori 2004; Smith 2007; Alexander 2007; TIVOLI Study 2013).

Imbalance in interventions

In one study, two differing ESAs were prescribed according to differing haemoglobin targets, resulting in an imbalance in treatment doses (Akizawa 2011).

Premature termination of study

Two studies were terminated early (Haag‐Weber 2012; Alexander 2007) due to development of erythropoietin antibodies (Haag‐Weber 2012) and for uncertain reasons (Alexander 2007).

Effects of interventions

See: Summary of findings for the main comparison Erythropoiesis‐stimulating agents (ESAs) for anaemia in adults with chronic kidney disease (CKD)

The summary of findings Table for the main comparison provides overall estimates of treatment effects and the quality of the available evidence for the primary efficacy (preventing blood transfusion) and safety (all‐cause mortality) outcomes and Table 1 shows the treatment estimates from pairwise and network meta‐analyses. Treatment estimates from pairwise comparisons are shown in italics in the lower left portion of each table section and treatment estimates from network analyses are shown in the upper right portion of each table section.

Open in table viewer

Table 1. Comparative effects of erythropoiesis‐stimulating agents on clinical outcomes in chronic kidney disease

Outcomes / interventions	Comparators (treatment estimate (OR (95% CI))
Outcomes / interventions	Epoetin alfa	Epoetin beta	Darbepoetin alfa	Methoxy polyethylene‐glycol epoetin beta	Biosimilar ESA	Placebo
Blood transfusion
Epoetin alfa	‐‐	2.04 (0.38‐11.0)	1.06 (0.35‐3.29)	1.14 (0.27‐4.97)	0.66 (0.19‐2.28)	0.18 (0.05‐0.59)
Epoetin beta	Not estimable	‐‐	0.52 (0.10‐2.67)	0.56 (0.11‐3.00)	0.33 (0.04‐2.60)	0.09 (0.02‐0.38)
Darbepoetin alfa	2.31 (1.34‐3.97)	Not estimable	‐‐	1.08 (0.38‐3.04)	0.62 (0.12‐3.30)	0.17 (0.05‐0.57)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	0.83 (0.17‐4.15)	0.94 (0.45‐1.95)	‐‐	0.58 (0.09‐3.92)	0.15 (0.03‐0.70)
Biosimilar ESA	0.72 (0.42‐1.22)	Not estimable	Not estimable	Not estimable	‐‐	0.27 (0.05‐1.47)
Placebo	0.07 (0.01‐0.84)	0.07 (0.03‐0.21)	0.53 (0.46‐0.63)	Not estimable	Not estimable	‐‐
All‐cause mortality
Epoetin alfa	‐‐	1.53 (0.77‐3.03)	1.17 (0.68‐2.05)	1.08 (0.54‐2.15)	0.95 (0.62‐1.44)	1.25 (0.71‐2.21)
Epoetin beta	Not estimable	̶	0.77 (0.43‐1.38)	0.71 (0.35‐1.42)	0.62 (0.29‐1.37)	0.82 (0.45‐1.48)
Darbepoetin alfa	1.12 (0.59‐2.14)	0.89 (0.38‐2.09)	‐‐	0.91 (0.60‐1.40)	0.81 (0.41‐1.61)	1.06 (0.91‐1.24)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	0.81 (0.12‐5.35)	0.90 (0.59‐1.40)	‐‐	0.88 (0.40‐1.97)	1.16 (0.74‐1.82)
Biosimilar ESA	1.04 (0.53‐2.01)	0.34 (0.04‐2.82)	Not estimable	Not estimable	‐‐	1.31 (0.65‐2.62)
Placebo	0.99 (0.14‐6.86)	0.61 (0.17‐2.15)	1.06 (0.91‐1.24)	Not estimable	Not estimable	‐‐
Fatigue
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.94 (0.57‐1.55)	Not estimable	‐‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	Not estimable	‐‐̶	Not estimable	Not estimable
Biosimilar ESA	0.18 (0.01‐3.91)	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable	‐‐
Breathlessness
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.71 (0.46‐1.10)	Not estimable	‐‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	Not estimable	‐‐	Not estimable	Not estimable
Biosimilar ESA	0.68 (0.37‐1.25)	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable	‐‐
Cardiovascular mortality
Epoetin alfa	‐‐	2.12 (0.34‐13.1)	1.48 (0.28‐7.96)	1.02 (0.16‐6.48)	0.55 (0.22‐1.38)	1.56 (0.29‐8.37)
Epoetin beta	Not estimable	‐‐	0.70 (0.12‐4.10)	0.48 (0.07‐3.31)	0.26 (0.04‐1.51)	0.74 (0.13‐4.28)
Darbepoetin alfa	2.15 (0.31‐14.9)	Not estimable	‐‐	0.69 (0.32‐1.48)	0.37 (0.06‐2.20)	1.05 (0.87‐1.26)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	0.69 (0.32‐1.48)	‐‐	0.54 (0.08‐3.74)	1.52 (0.69‐3.34)
Biosimilar ESA	0.53 (0.20‐1.35)	0.34 (0.04‐2.82)	Not estimable	Not estimable	‐‐	2.81 (0.47‐16.7)
Placebo	Not estimable	0.45 (0.06‐3.75)	1.05 (0.87‐1.26)	Not estimable	Not estimable	‐‐
Major adverse cardiovascular events
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.20 (0.01‐4.17)	Not estimable	‐‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	Not estimable	‐‐	Not estimable	Not estimable
Biosimilar ESA	0.49 (0.17‐1.47)	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	1.08 (0.95‐1.24)	Not estimable	Not estimable	‐‐
Myocardial infarction
Epoetin alfa	‐‐	Not estimable	1.04 (0.35‐3.11)	0.55 (0.05‐5.69)	1.18 (0.47‐3.02)	1.00 (0.32‐3.09)
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.87 (0.20‐3.81)	Not estimable	‐‐	0.53 (0.07‐4.18)	1.14 (0.27‐4.83)	0.97 (0.75‐1.25)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	0.47 (0.06‐3.65)	‐‐	2.17 (0.17‐27.1)	1.83 (0.18‐19.1)
Biosimilar ESA	1.23 (0.49‐3.12)	Not estimable	Not estimable	Not estimable	‐‐	0.84 (0.20‐3.65)
Placebo	3.46 (0.12‐100.51)	Not estimable	0.97 (0.75‐1.25)	Not estimable	Not estimable	‐‐
Stroke
Epoetin alfa	‐‐	4.56 (0.29‐71.8)	1.39 (0.38‐5.16)	2.36 (0.24‐23.6)	0.92 (0.39‐2.16)	2.74 (0.71‐10.5)
Epoetin beta	Not estimable	‐‐	0.31 (0.02‐4.55)	0.52 (0.02‐14.0)	0.20 (0.01‐3.61)	0.60 (0.04‐8.88)
Darbepoetin alfa	1.44 (0.37‐5.54)	Not estimable	‐‐	1.70 (0.26‐11.2)	0.66 (0.14‐3.14)	1.96 (1.40‐2.75)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	1.33 (0.17‐10.49)	‐‐	0.38 (0.03‐4.50)	1.16 (0.17‐7.90)
Biosimilar ESA	0.92 (0.39‐2.15)	Not estimable	Not estimable	Not estimable	‐‐	2.99 (0.61‐14.8)
Placebo	Not estimable	0.33 (0.01‐8.21)	1.97 (1.40‐2.76)	Not estimable	Not estimable	‐‐
Hypertension
Epoetin alfa	‐‐	0.90 (0.41‐1.95)	1.26 (0.81‐1.96)	1.18 (0.64‐2.18)	1.95 (0.97‐3.94)	2.31 (1.27‐4.23)
Epoetin beta	Not estimable	‐‐	1.41 (0.70‐2.82)	1.31 (0.63‐2.72)	2.18 (0.76‐6.22)	2.57 (1.23‐5.39)
Darbepoetin alfa	0.94 (0.62‐1.43)	1.18 (0.38‐3.69)	‐‐	0.93 (0.60‐1.45)	1.55 (0.68‐3.55)	1.83 (1.05‐3.21)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	1.38 (0.62‐3.09)	Not estimable	‐‐	1.66 (0.65‐4.21)	1.96 (0.98‐3.92)
Biosimilar ESA	1.77 (1.02‐3.09)	Not estimable	Not estimable	Not estimable	‐‐	1.18 (0.47‐2.99)
Placebo	4.10 (2.16‐7.76)	2.95 (1.19‐7.26)	1.14 (0.99‐1.32)	Not estimable	Not estimable	‐‐
End‐stage kidney disease
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	2.17 (0.37‐12.74)	Not estimable	‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	1.83 (0.66‐5.09)	‐‐	Not estimable	Not estimable
Biosimilar ESA	Not estimable	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	1.04 (0.88‐1.23)	Not estimable	Not estimable	‐‐
Vascular access thrombosis
Epoetin alfa	‐‐	0.93 (0.28‐3.10)	1.22 (0.78‐1.91)	1.04 (0.48‐2.25)	1.26 (0.45‐3.36)	1.72 (0.58‐5.16)
Epoetin beta	Not estimable	‐‐	1.30 (0.42‐4.04)	1.11 (0.38‐3.24)	1.35 (0.29‐6.34)	1.85 (0.61‐5.63)
Darbepoetin alfa	1.15 (0.73‐1.82)	Not estimable	‐‐	0.86 (0.45‐1.61)	1.04 (0.35‐3.05)	1.42 (0.50‐4.03)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	1.74 (0.49‐6.24)	0.76 (0.39‐1.47)	‐‐	1.21 (0.35‐4.22)	1.66 (0.54‐5.08)
Biosimilar ESA	1.71 (0.30‐10.00)	Not estimable	Not estimable	Not estimable	‐‐	1.37 (0.32‐5.93)
Placebo	6.40 (0.80‐51.50)	1.09 (0.28‐4.34)	1.34 (0.30‐6.01)	Not estimable	Not estimable	‐‐

Treatment estimates for pairwise meta‐analyses are shown in italics

Figure 5 shows the networks of evidence for the safety and efficacy of ESA drugs included in the review. Each line links the treatments which have been directly compared in studies. The thickness of the line is proportional to the number of studies included in the comparison and the width of each circle is proportional to the number of participants included in the comparison. Figure 6 shows the summary treatment effects for ESAs when compared against placebo within networks.

Figure 5

Networks of the treatment efficacy and safety of ESA drugs in the treatment of anaemia in chronic kidney disease. Values lower than 1 favour the active treatment in the comparison

Figure 6

Forest plots for results from network meta‐analyses comparing ESAs versus placebo

1. Response to treatment (efficacy)

1.1 Pairwise meta‐analysis (direct comparisons)

Treatment estimates for pairwise meta‐analyses are shown in italics in Table 1.

Preventing blood transfusions

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment was provided in eight studies with 4661 participants (Bahlmann 1991; Bennett 1991; Canadian EPO Study 1990; Kleinman 1989; Patel 2012; Roth 1994; TREAT Study 2005; Van Biesen 2005). Three agents (epoetin alfa, epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The following results for blood transfusions were found:

Epoetin alfa reduced the odds of blood transfusion compared to placebo (Analysis 1.1.1 (3 studies, 196 participants): OR 0.07, 95% CI 0.01 to 0.84; I² = 81%) (Canadian EPO Study 1990; Kleinman 1989; Roth 1994) with evidence of heterogeneity that might be substantial
Epoetin beta reduced the odds of blood transfusion compared to placebo (Analysis 1.1. (2 studies, 230 participants): OR 0.07, 95% CI 0.03 to 0.21; I² = 0%) (Bahlmann 1991; Bennett 1991)
Darbepoetin alfa reduced the odds of blood transfusion compared to placebo (Analysis 1.1.3 (1 study, 4038 participants): OR 0.53, 95% CI 0.46 to 0.63) (TREAT Study 2005)
Epoetin alfa had uncertain effects on the odds of blood transfusion compared with no treatment (Analysis 1.1.4 (1 study, 157 participants): OR 3.10, 95% CI 0.16 to 58.97) (Patel 2012)
Epoetin beta had uncertain effects on the odds of blood transfusion compared with no treatment (Analysis 1.1.5 (1 study, 40 participants): OR 0.35, 95% CI 0.06 to 2.18) (Van Biesen 2005).

ESAs compared to each other

Three studies (1191 participants) compared epoetin alfa with darbepoetin alfa (Akizawa 2011; Locatelli 2001; Nissenson 2002), three studies (1823 participants) compared epoetin alfa with a biosimilar ESA (Krivoshiev 2008; Krivoshiev 2010; Martin 2007), one study (181 participants) compared epoetin beta versus methoxy polyethylene glycol‐epoetin beta (AMICUS Study 2007), and four studies (1191 participants) compared darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta (ARCTOS Study 2008; CORDATUS Study 2011; PATRONUS Study 2010; TIVOLI Study 2013).

Epoetin alfa increased the odds of blood transfusion compared to darbepoetin alfa (Analysis 1.1.6 (3 studies, 1191 participants): OR 2.31, 95% CI 1.34 to 3.97; I² = 0%) (Akizawa 2011; Locatelli 2001; Nissenson 2002)
Epoetin alfa had uncertain effects on the odds of blood transfusion compared to a biosimilar ESA (Analysis 1.1.7 (3 studies, 1823 participants): OR 0.72, 95% CI 0.42 to 1.22; I² = 0%) (Krivoshiev 2008; Krivoshiev 2010; Martin 2007)
Epoetin beta had uncertain effects on the odds of blood transfusion compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.1.8 (1 study, 181 participants): OR 0.83, 95% CI 0.17 to 4.15) (AMICUS Study 2007)
Darbepoetin alfa had uncertain effects on the odds of blood transfusion compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.1.9 (4 studies, 1191 participants): OR 0.94, 95% CI 0.45 to 1.95; I² = 44%) (ARCTOS Study 2008; CORDATUS Study 2011; PATRONUS Study 2010; TIVOLI Study 2013) with evidence of moderate heterogeneity.

Fatigue

ESAs compared to placebo or no treatment

There were no placebo or standard care‐controlled studies providing extractable data for the effects of treatment on fatigue.

ESAs compared to each other

Data for effects of ESA treatment versus another ESA on fatigue were available in three studies with 730 participants (Allon 2002; Goh 2007; Nissenson 2002).

Epoetin alfa had uncertain effects on fatigue compared to darbepoetin alfa (Analysis 1.2.1 (2 studies, 551 participants): OR 0.94, 95% CI 0.57 to 1.55; I² = 0%) (Allon 2002; Nissenson 2002)
Epoetin alfa had uncertain effects on fatigue compared to a biosimilar ESA (Analysis 1.2.2 (1 study, 179 participants): OR 0.18, 95% CI 0.01 to 3.91) (Goh 2007).

Breathlessness

ESAs compared to placebo or no treatment

There were no placebo or standard care‐controlled studies providing extractable data for the effects of treatment on breathlessness.

ESAs compared to each other

Data for effects of ESA treatment versus another ESA on breathlessness were available in three studies involving 1298 participants (Goh 2007; Haag‐Weber 2009; Nissenson 2002).

Epoetin alfa had uncertain effects on breathlessness when compared to darbepoetin alfa (Analysis 1.3.1 (1 study, 504 participants): OR 0.71, 95% CI 0.46 to 1.10) (Nissenson 2002)
Epoetin alfa had uncertain effects on breathlessness when compared to a biosimilar ESA (Analysis 1.3.2 (2 studies, 794 participants): OR 0.68, 95% CI 0.37 to 1.25; I² = 0%) (Goh 2007; Haag‐Weber 2009).

1.2 Network meta‐analysis (combination of direct and indirect comparisons)

Treatment estimates for network meta‐analyses are shown in Table 1 and network meta‐analyses for all ESAs against placebo are summarised in Figure 6. Studies grouped by comparison were deemed comparable for the effect modifiers of stage of CKD, haemoglobin target with ESA treatment, age of participants and duration of follow‐up, such that the assumption of transitivity might hold and that a network meta‐analytical approach was reasonable. However, we could not assess the comparability of treatment comparisons across different studies using statistical methods due to insufficient data. Overall, SUCRA rankings of the differing ESAs were imprecise due to sparse data rendering the analyses clinically irrelevant. Therefore, treatment rankings are not provided in the results.

Preventing blood transfusions

Blood transfusion data were provided in 19 studies (Akizawa 2011; AMICUS Study 2007; ARCTOS Study 2008; Bahlmann 1991; Bennett 1991; Canadian EPO Study 1990; CORDATUS Study 2011; Kleinman 1989; Krivoshiev 2008; Krivoshiev 2010; Locatelli 2001; Martin 2007; Nissenson 2002; Patel 2012; PATRONUS Study 2010; Roth 1994; TIVOLI Study 2013; TREAT Study 2005; Van Biesen 2005) involving 9047 participants with CKD (58.0% of the participants in this review). Most participants within the network were randomised to darbepoetin alfa or placebo due to the contribution of the large TREAT study (TREAT Study 2005).

In moderate to low quality evidence, epoetin alfa, epoetin beta, darbepoetin alfa and methoxy polyethylene glycol‐epoetin beta were all superior to placebo for preventing blood transfusion (epoetin alfa OR 0.18, 95% CI 0.05‐0.59, epoetin beta OR 0.09, 95% CI 0.02 to 0.38; darbepoetin alfa OR 0.17, 95% CI 0.05 to 0.57; methoxy polyethylene glycol‐epoetin beta OR 0.15, 95% CI 0.03 to 0.70). In very low quality evidence, biosimilar ESAs were possibly no better than placebo (OR 0.27, 95% CI 0.05 to 1.47). There were no statistical differences between all ESAs for their effects on blood transfusion in treatment estimates showing considerable uncertainty. The heterogeneity tau for this network overall was 0.89, which is consistent with moderate heterogeneity.

Fatigue

Network meta‐analysis was not possible for this outcome due to insufficient data.

Breathlessness

Network meta‐analysis was not possible for this outcome due to insufficient data.

2. Safety

2.1 Pairwise meta‐analysis (direct comparisons)

All‐cause mortality

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment on all‐cause mortality was provided in 10 studies involving 5209 participants (Bahlmann 1991; Bennett 1991; EPOCARES Study 2010; Klinkmann 1992; Kuriyama 1997; Nissenson 1995; Palazzuoli 2007; Patel 2012; Roth 1994; TREAT Study 2005). Three agents (epoetin alfa, epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The odds of all‐cause mortality were uncertain for epoetin alfa (Analysis 1.4.1 (2 studies, 235 participants): OR 0.99, 95% CI 0.14 to 6.86; I² = 0%), epoetin beta Analysis 1.4.2 (3 studies, 311 participants): OR 0.61, 95% CI 0.17 to 2.15; I² = 0%) and darbepoetin alfa (Analysis 1.4.3 (Analysis 1.4.3 (1 study, 4038 participants): OR 1.06, 95% 0.91 to 1.24) when compared with placebo (Bahlmann 1991; Bennett 1991; Nissenson 1995; Palazzuoli 2007; Roth 1994; TREAT Study 2005)
The odds of all‐cause mortality were uncertain for epoetin alfa (Analysis 1.4.4 (1 study, 157 participants): OR 1.06, 95% CI 0.39 to 2.87) and epoetin beta (Analysis 1.4.5 (3 studies, 468 participants): OR 0.69, 95% CI 0.36 to 1.33; I² = 0%) when compared with standard care (EPOCARES Study 2010; Klinkmann 1992; Kuriyama 1997).

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in six studies involving 1205 participants (Akizawa 2011; Allon 2002; Hori 2004; Li 2008d; Locatelli 2001; Nissenson 2002), epoetin alfa was compared to a biosimilar ESA in seven studies involving 2220 participants (Goh 2007; Haag‐Weber 2009; Haag‐Weber 2012; Krivoshiev 2008; Krivoshiev 2010; Milutinovic 2006; Spinowitz 2006), epoetin beta was compared versus darbepoetin alfa in one study and 217 participants (Tolman 2005), epoetin beta was compared to methoxy polyethylene glycol‐epoetin beta in two studies involving 462 participants (AMICUS Study 2007; Chen 2012e), epoetin beta was compared to a biosimilar ESA in one study involving 290 participants (Gertz 2010) and darbepoetin alfa was compared to methoxy polyethylene glycol‐epoetin beta in four studies involving 1429 participants (ARCTOS Study 2008; CORDATUS Study 2011; PATRONUS Study 2010; STRIATA Study 2008).

The odds of all‐cause mortality with epoetin alfa were uncertain when compared to darbepoetin alfa (Analysis 1.4.6 (6 studies, 1205 participants): OR 1.12, 95% CI 0.59 to 2.14; I² = 0%) or biosimilar ESAs (Analysis 1.4.7 (7 studies, 2220 participants): OR 1.04, 95% CI 0.53 to 2.01; I² = 46%)
The odds of all‐cause mortality with epoetin beta were uncertain when compared to darbepoetin alfa (Analysis 1.4.8 (1 study, 217 participants): OR 0.89, 95% CI 0.38 to 2.09), methoxy polyethylene glycol‐epoetin beta (Analysis 1.4.9 (2 studies, 462 participants): OR 0.57, 95% CI 0.03 to 12.18; I² = 0%) or a biosimilar ESA (Analysis 1.4.10 (1 study, 290 participants): OR 0.34, 95% CI 0.04 to 2.82)
The odds of all‐cause mortality with darbepoetin alfa were uncertain when compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.4.11 (4 studies, 1429 participants): OR 0.90, 95% CI 0.59 to 1.40; I² = 0%).

Cardiovascular mortality

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment on cardiovascular mortality were provided in six studies with 4766 participants (Bahlmann 1991; Bennett 1991; EPOCARES Study 2010; Klinkmann 1992; Kuriyama 1997; TREAT Study 2005). Two agents (epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either epoetin alfa, methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The odds of cardiovascular mortality were uncertain for epoetin beta (Analysis 1.5.1 (2 studies, 260 participants): OR 0.45, 95% CI 0.06 to 3.75, I² = 0%) and darbepoetin alfa (Analysis 1.5.2 (1 study, 4038 participants): OR 1.05, 95% 0.87 to 1.26) when compared to placebo (Bahlmann 1991; Bennett 1991; TREAT Study 2005)
The odds of cardiovascular mortality were uncertain for epoetin beta (Analysis 1.5.3 (3 studies, 430 participants): OR 0.28, 95% CI 0.08 to 1.03; I² = 0%) when compared with no treatment (EPOCARES Study 2010; Klinkmann 1992; Kuriyama 1997).

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in two studies and 487 participants (Akizawa 2011; Locatelli 2001) and a biosimilar ESA in one study and 478 participants (Haag‐Weber 2009). Epoetin beta was compared to a biosimilar ESA in 1 study and 290 participants (Gertz 2010). Darbepoetin alfa was compared to methoxy polyethylene glycol‐epoetin beta in two studies and 629 participants (ARCTOS Study 2008; CORDATUS Study 2011).

The odds of cardiovascular mortality were uncertain for epoetin alfa when compared to darbepoetin alfa (Analysis 1.5.4 (2 studies, 487 participants): OR 2.15, 95% CI 0.31 to 14.91; I² = 0%) or a biosimilar ESA (Analysis 1.5.5 (2 studies, 657 participants): OR 0.53, 95% CI 0.20 to 1.35; I² = 0%)
The odds of cardiovascular mortality were uncertain for epoetin beta when compared to a biosimilar ESA (Analysis 1.5.6 (1 study, 290 participants): OR 0.34, 95% CI 0.04 to 2.82)
The odds of cardiovascular mortality were uncertain for darbepoetin alfa when compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.5.7 (3 studies, 938 participants): OR 0.69, 95% CI 0.32 to 1.48; I² = 0%).

Myocardial infarction (MI)

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment on MI were provided in three studies involving 4209 participants (Kleinman 1989; Patel 2012; TREAT Study 2005). Two agents (epoetin alfa and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either epoetin beta, methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The odds of MI were uncertain for epoetin alfa (Analysis 1.6.1 (1 study, 14 participants): OR 3.46, 95% CI 0.12 to 100.51) and darbepoetin alfa (Analysis 1.6.2 (1 study, 4038 participants): OR 0.97, 95% CI 0.75 to 1.25) when compared to placebo (Kleinman 1989; TREAT Study 2005)
The odds of MI were uncertain for epoetin alfa (Analysis 1.6.3 (1 study, 157 participants): OR 1.01, 95% CI 0.04 to 25.26) when compared to no treatment (Patel 2012).

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in two studies involving 825 participants (Akizawa 2011; Nissenson 2002), and a biosimilar ESA in two studies involving 641 participants (Goh 2007; Krivoshiev 2010). Darbepoetin alfa was compared to methoxy polyethylene glycol‐epoetin beta in two studies involving 629 participants (ARCTOS Study 2008; CORDATUS Study 2011).

The odds of MI were uncertain for epoetin alfa when compared to darbepoetin alfa (Analysis 1.6.4 (2 studies, 825 participants): OR 0.87, 95% CI 0.20 to 3.81; I² = 21%) and a biosimilar ESA (Analysis 1.6.5 (2 studies, 641 participants): OR 1.23, 95% CI 0.49 to 3.12; I² = 0%)
The odds of MI were uncertain for darbepoetin alfa when compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.6.6 (2 studies, 628 participants): OR 0.47, 95% CI 0.06 to 3.65; I² = 0%).

Stroke

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment on stroke were provided in four studies and 4334 participants (Bahlmann 1991; EPOCARES Study 2010; Patel 2012; TREAT Study 2005). Three agents (epoetin alfa, epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The odds of stroke were uncertain for epoetin beta when compared to placebo (Analysis 1.7.1 (1 study, 106 participants): OR 0.33, 95% CI 0.01 to 8.21) but were increased with darbepoetin alfa compared to placebo (Analysis 1.7.2 (1 study, 4038 participants): OR 1.97, 95% CI 1.40 to 2.76)
The odds of stroke were uncertain for epoetin alfa (Analysis 1.7.3 (1 study, 157 participants): OR 0.99, 95% CI 0.10 to 9.82) and epoetin beta (Analysis 1.7.4 (1 study, 33 participants): OR 0.20, 95% CI 0.01 to 5.39) compared to control.

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in three studies involving 996 participants (Akizawa 2011; Hirakata 2010; Nissenson 2002) and a biosimilar ESA in three studies involving 539 participants (Goh 2007; Krivoshiev 2010; Milutinovic 2006). Darbepoetin alfa was compared to methoxy polyethylene glycol‐epoetin beta in two studies and 629 participants (ARCTOS Study 2008; CORDATUS Study 2011).

The odds of stroke were uncertain for epoetin alfa versus darbepoetin alfa (Analysis 1.7.5 (3 studies, 996 participants): OR 1.44, 95% CI 0.37 to 5.54; I² = 0%) and a biosimilar ESA (Analysis 1.7.6 (3 studies, 718 participants): OR 0.92, 95% CI 0.39 to 2.15; I² = 0%)
The odds of stroke were uncertain for darbepoetin alfa when compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.7.7 (2 studies, 628 participants): OR 1.33, 95% CI 0.17 to 10.49; I² = 16%).

Hypertension

ESAs compared to placebo

Data for the effects of ESA treatment compared to placebo or no treatment on hypertension were provided in eight studies and 5058 participants (Bahlmann 1991; Bennett 1991; Canadian EPO Study 1990; Clyne 1992; Klinkmann 1992; Nissenson 1995; Patel 2012; TREAT Study 2005). Three agents (epoetin alfa, epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The odds of hypertension were increased with epoetin alfa (Analysis 1.8.1 (2 studies; 251 participants): OR 4.10, 95% CI 2.16 to 7.76; I² = 0%); epoetin beta (Analysis 1.8.2 (2 studies, 230 participants): OR 2.95, 95% CI 1.19 to 7.26; I² = 0%) and darbepoetin alfa (Analysis 1.8.3 (1 study, 4038 participants): OR 1.14, 95% CI 0.99 to 1.32) when compared to placebo (Bahlmann 1991; Bennett 1991; Canadian EPO Study 1990; Nissenson 1995; TREAT Study 2005)
The odds of hypertension were uncertain for epoetin alfa compared to no treatment (Analysis 1.8.4 (1 study, 157 participants): OR 5.31, 95% CI 0.30 to 95.20) but were increased with epoetin beta compared to no treatment (Analysis 1.8.5 (2 studies, 382 participants): OR 2.99, 95% CI 1.34 to 6.69; I² = 0%) (Clyne 1992; Klinkmann 1992; Patel 2012).

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in five studies and 1568 participants (Akizawa 2011; Coyne 2006a; Hirakata 2010; Locatelli 2001; Nissenson 2002) and a biosimilar ESA in four studies involving 1464 participants (Goh 2007; Krivoshiev 2010; Martin 2007; Milutinovic 2006). Epoetin beta was compared to darbepoetin alfa in one study and 162 participants (Tolman 2005) and methoxy polyethylene glycol‐epoetin beta in one study and 181 participants (AMICUS Study 2007). Darbepoetin alfa was compared to methoxy polyethylene glycol‐epoetin beta in five studies and 1497 participants (ARCTOS Study 2008; CORDATUS Study 2011; PATRONUS Study 2010; STRIATA Study 2008; TIVOLI Study 2013).

The odds of hypertension were uncertain for epoetin alfa compared to darbepoetin alfa (Analysis 1.8.6 (5 studies, 1568 participants): OR 0.94, 95% CI 0.62 to 1.43; I² = 45%) with evidence of moderate heterogeneity or a biosimilar ESA (Analysis 1.8.7 (4 studies, 1464 participants): OR 1.77, 95% CI 1.02 to 3.09; I² = 0%)
The odds of hypertension were uncertain for epoetin beta compared to darbepoetin alfa (Analysis 1.8.8 (1 study, 162 participants): OR 1.18, 95% CI 0.38 to 3.69)
The odds of hypertension were uncertain for epoetin beta compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.8.9 (1 study, 181 participants): OR 1.38, 95% CI 0.62 to 3.09)
The odds of hypertension were uncertain for darbepoetin alfa compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.8.10 (5 studies, 1497 participants): OR 0.94, 95% CI 0.62 to 1.42; I² = 36%) with evidence of moderate heterogeneity.

Vascular access thrombosis

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment on thrombosis of vascular access were provided in four studies and 4617 participants (Bahlmann 1991; Canadian EPO Study 1990; Klinkmann 1992; TREAT Study 2005). Three agents (epoetin alfa, epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The odds of vascular access thrombosis were uncertain for epoetin alfa (Analysis 1.9.1 (1 study, 118 participants): OR 6.40, 95% CI 0.80 to 51.50), epoetin beta (Analysis 1.9.2 (1 study, 99 participants): OR 1.09, 95% CI 0.28 to 4.34) and darbepoetin alfa (Analysis 1.9.3 (1 study, 4038 participants): OR 1.34, 95% CI 0.30 to 6.01) compared to placebo (Bahlmann 1991; Canadian EPO Study 1990; TREAT Study 2005)
The odds of vascular access thrombosis were uncertain for epoetin beta (Analysis 1.9.4 (1 study, 363 participants): OR 1.40, 95% CI 0.72 to 2.73) when compared to no treatment (Klinkmann 1992).

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in three studies and 1084 participants (Coyne 2006a; Hirakata 2010; Nissenson 2002) and a biosimilar ESA in two studies and 823 participants (Martin 2007; Milutinovic 2006). Epoetin beta was compared to methoxy polyethylene glycol‐epoetin beta in one study and 181 participants (AMICUS Study 2007). Darbepoetin alfa was compared to methoxy polyethylene glycol‐epoetin beta in one study and 489 participants (PATRONUS Study 2010).

The odds of vascular access thrombosis were uncertain for epoetin alfa when compared to darbepoetin alfa (Analysis 1.9.5 (3 studies, 1084 participants): OR 1.15, 95% CI 0.73 to 1.82; I² = 0%) or a biosimilar ESA (Analysis 1.9.6 (2 studies, 823 participants): OR 1.74, 95% CI 0.30 to 10.00; I² = 38%) with evidence of moderate heterogeneity
The odds of vascular access thrombosis were uncertain for epoetin beta when compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.9.7 (1 study, 181 participants): OR 1.74, 95% CI 0.49 to 6.24)
The odds of vascular access thrombosis were uncertain for darbepoetin alfa when compared to methoxy polyethylene glycol‐epoetin beta ((Analysis 1.9.8 (1 study, 489 participants): OR 0.76, 95% CI 0.39 to 1.47).

End‐stage kidney disease

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment on ESKD were provided in four studies and 4161 participants (Brown 1995; EPOCARES Study 2010; Kuriyama 1997; TREAT Study 2005). Three agents (epoetin alfa, epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

The odds of ESKD were uncertain for darbepoetin alfa compared to placebo (Analysis 1.10.1 (1 study, 4038 participants): OR 1.04, 95% CI 0.88 to 1.23) (TREAT Study 2005)
The odds of ESKD were uncertain for epoetin alfa (Analysis 1.10.2 (1 study, 17 participants): OR 0.27, 95% CI 0.03 to 2.12) and epoetin beta (Analysis 1.10.3 (2 studies, 106 participants): OR 0.40, 95% CI 0.08 to 1.93; I² = 26%) (Brown 1995; EPOCARES Study 2010; Kuriyama 1997) when compared to no treatment with evidence of moderate heterogeneity.

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in two studies and 492 participants (Akizawa 2011; Hirakata 2010). Darbepoetin alfa was compared to methoxy polyethylene glycol‐epoetin beta in one study and 305 participants (CORDATUS Study 2011).

The odds of ESKD were uncertain for epoetin alfa when compared to darbepoetin alfa (Analysis 1.10.4 (2 studies, 492 participants): OR 2.17, 95% CI 0.37 to 12.74; I² = 48%) with evidence of moderate heterogeneity
The odds of ESKD were uncertain for darbepoetin alfa when compared to methoxy polyethylene glycol‐epoetin beta (Analysis 1.10.5 (1 study, 305 participants): OR 1.83, 95% CI 0.66 to 5.09).

Major cardiovascular events

ESAs compared to placebo

Data for effects of ESA treatment compared to placebo or no treatment on major cardiovascular events were provided in three studies and 4228 participants (EPOCARES Study 2010; Patel 2012; TREAT Study 2005). Three agents (epoetin alfa, epoetin beta and darbepoetin alfa) were assessed against placebo or no treatment. No study evaluated either methoxy polyethylene glycol‐epoetin beta or a biosimilar ESA with placebo or standard care.

Darbepoetin alfa may increase odds of major cardiovascular events compared to placebo (Analysis 1.11.1 (1 study, 4038 participants): OR 1.08, 95% CI 0.95 to 1.24) (TREAT Study 2005)
Epoetin alfa (Analysis 1.11.2 (1 study, 157 participants): OR 2.40, 95% CI 0.29 to 20.11) and epoetin beta (Analysis 1.11.3 (1 study, 33 participants): OR 0.61, 95% CI 0.07 to 4.98) have uncertain effects on major cardiovascular events when compared to no treatment (EPOCARES Study 2010; Patel 2012).

ESAs compared to each other

Epoetin alfa was compared to darbepoetin alfa in one study and 321 participants (Akizawa 2011) and a biosimilar ESA in one study and 462 participants (Krivoshiev 2010).

The odds of a major cardiovascular event was uncertain for epoetin alfa when compared to darbepoetin alfa (Analysis 1.11.4 (1 study, 321 participants): OR 0.20, 95% CI 0.01 to 4.17) and a biosimilar ESA (Analysis 1.11.5 (1 study, 462 participants): OR 0.49, 95% CI 0.17 to 1.47).

2.2 Network meta‐analysis (combination of direct and indirect comparisons)

Treatment estimates for network meta‐analyses are shown in Table 1 and network meta‐analyses for all ESAs against placebo are summarised in Figure 6. Studies grouped by comparison were deemed comparable for the effect modifiers of stage of CKD, haemoglobin target with ESA treatment, age of participants and duration and follow‐up, such that the assumption of transitivity might hold and that a network meta‐analytical approach was reasonable. Overall, SUCRA rankings of the differing ESAs were imprecise due to sparse data rendering the analyses clinically irrelevant. Therefore, treatment rankings are not provided in the results.

All‐cause mortality

All‐cause mortality data were provided in 31 studies (Akizawa 2011; Allon 2002; AMICUS Study 2007; ARCTOS Study 2008; Bahlmann 1991; Bennett 1991; Chen 2012e; CORDATUS Study 2011; EPOCARES Study 2010; Gertz 2010; Goh 2007; Haag‐Weber 2009; Haag‐Weber 2012; Hori 2004; Klinkmann 1992; Krivoshiev 2008; Krivoshiev 2010; Kuriyama 1997; Li 2008d; Locatelli 2001; Milutinovic 2006; Nissenson 1995; Nissenson 2002; Palazzuoli 2007; Patel 2012; PATRONUS Study 2010; Roth 1994; Spinowitz 2006; STRIATA Study 2008; Tolman 2005; TREAT Study 2005) involving 11,024 participants with CKD (70.7% of the participants in this review). Most participants within the network were randomised to darbepoetin alfa or placebo due to the contribution of the large TREAT study (TREAT Study 2005). Effects of all ESA formulations on the odds of death from any cause were imprecise when compared with placebo or other ESA drug and were not statistically significant but there was considerable uncertainty in the comparative treatment effects. The heterogeneity tau for the network was 0.0, indicating no statistical evidence of heterogeneity.

Cardiovascular mortality

Cardiovascular mortality were provided in 14 studies (Akizawa 2011; ARCTOS Study 2008; Bahlmann 1991; Bennett 1991; CORDATUS Study 2011; EPOCARES Study 2010; Gertz 2010; Goh 2007; Haag‐Weber 2009; Klinkmann 1992; Kuriyama 1997; Locatelli 2001; STRIATA Study 2008; TREAT Study 2005) in 7138 participants (45.8% of the participants in this review). Effects of all ESA formulations on the odds of death caused by a cardiovascular event were imprecise when compared with placebo or other ESA drug and were not statistically significant but there was considerable uncertainty in the comparative treatment effects. The heterogeneity tau for the network was 0%, indicating no statistical evidence of heterogeneity.

Myocardial infarction

Nine studies provided data for one or more MI outcomes (Akizawa 2011; ARCTOS Study 2008; CORDATUS Study 2011; Kleinman 1989; Krivoshiev 2010; Nissenson 2002; Patel 2012; TREAT Study 2005) in 6303 participants (40.4% of the participants in this review). Effects of all ESA formulations on the odds of MI were imprecise when compared with placebo or other ESA drug and were not statistically significant but there was considerable uncertainty in the comparative treatment effects. The heterogeneity tau for the network was 0.0, indicating no statistical evidence of heterogeneity.

Stroke

There were 12 studies that provided data for one or more stroke events (Akizawa 2011; ARCTOS Study 2008; Bahlmann 1991; CORDATUS Study 2011; EPOCARES Study 2010; Goh 2007; Hirakata 2010; Krivoshiev 2010; Milutinovic 2006; Nissenson 2002; Patel 2012; TREAT Study 2005) in 6676 participants (42.8% of the participants in this review). Effects of all ESA formulations on the odds of stroke were imprecise when compared with placebo or other ESA drug and were not statistically significant except for the comparison between darbepoetin alfa and placebo (OR 1.96, 95% CI 1.40 to 2.75), but there was considerable uncertainty in the comparative treatment effects. The heterogeneity tau for the network was 0.0, indicating no statistical evidence of heterogeneity.

Hypertension

Hypertension data were provided in 24 studies (Akizawa 2011; AMICUS Study 2007; ARCTOS Study 2008; Bahlmann 1991; Bennett 1991; Canadian EPO Study 1990; Clyne 1992; CORDATUS Study 2011; Coyne 2006a; Goh 2007; Hirakata 2010; Klinkmann 1992; Krivoshiev 2010; Martin 2007; Milutinovic 2006; Nissenson 1995; Nissenson 2002; Patel 2012; PATRONUS Study 2010; STRIATA Study 2008; TIVOLI Study 2013; Tolman 2005; TREAT Study 2005) in 9930 participants (63.7% of participants in this review). All proprietary ESA drugs were significantly worse than placebo for the odds of inducing hypertension (epoetin alfa OR 2.31, 95% CI 1.27 to 4.23; epoetin beta OR 2.57, 95% CI 1.23 to 5.39; darbepoetin alfa OR 1.83, 95% CI 1.05 to 3.21) except for methoxy polyethylene glycol‐epoetin beta for which the treatment estimate was less precise and marginally included the possibility of no effect (OR 1.96, 95% CI 0.98 to 43.92). The effects on biosimilar ESAs on the odds of hypertension were less certain (OR 1.18, 95% CI 0.47 to 2.99). The heterogeneity tau for the network was 0.37, consistent with low heterogeneity.

End‐stage kidney disease

The network for this outcome provided no closed loops of evidence (Figure 5) and conventional pairwise meta‐analysis was the primary source of evidence for this outcome, showing generally imprecise estimates of comparative treatment effects.

Vascular access thrombosis

Eleven studies provided data for one or more episodes of vascular access thrombus (AMICUS Study 2007; Bahlmann 1991; Canadian EPO Study 1990; Coyne 2006a; Hirakata 2010; Klinkmann 1992; Martin 2007; Milutinovic 2006; Nissenson 2002; PATRONUS Study 2010; TREAT Study 2005) in 7194 participants (46.1% of the participants in this review). Effects of all ESA formulations on the odds of vascular access thrombosis were imprecise when compared with placebo or other ESA drug and were not statistically significant but there was considerable uncertainty in the comparative treatment effects. The heterogeneity tau for the network was 0.0, indicating no statistical evidence of heterogeneity.

Major cardiovascular events

3. Assessment of heterogeneity and inconsistency within network analyses

There was important clinical diversity in studies based on the age of the participants, stage of CKD and duration of treatment. Treatment estimates from direct and indirect evidence in networks with closed loops did not show evidence of statistical inconsistency except for three of the five loops for hypertension (Table 2). However, the results for inconsistency were very imprecise as individual direct and indirect estimates were themselves imprecise and so the possibility of inconsistency in network analyses for other outcomes could not be excluded. When comparing a common heterogeneity variance in networks and with empirical distributions of heterogeneity variances specific to the outcome and types of treatment being compared, networks for blood transfusion (τ = 0.89) and hypertension (τ = 0.37) possessed heterogeneity variances that indicated the presence of low to moderate heterogeneity. Similarly, when evaluating the inconsistency in the networks as a whole, there was an indication that global inconsistency was present within the networks for blood transfusion (Chi² = 6.38; P = 0.01) and hypertension (Chi² = 6.40; P = 0.04). We therefore downgraded the credibility of the evidence provided by these two networks as the risk of important inconsistency was high. Meta‐regression to explore potential sources of inconsistency was not possible due to sparse data for direct treatment comparisons.

Open in table viewer

Table 2. Evaluation of consistency using loop specific approach

Treatments included in the loop of evidence	Inconsistency factor*	95% CI
All‐cause mortality
Epoetin alfa – epoetin beta – darbepoetin alfa – biosimilar ESA	0.87	0.00‐3.32
Epoetin beta – darbepoetin alfa – placebo	0.40	0.00‐1.82
Epoetin alfa – epoetin beta – biosimilar ESA – no treatment	0.66	0.00‐3.36
Epoetin beta – darbepoetin alfa – methoxy polyethylene glycol‐epoetin beta	0.02	0.00‐2.08
Epoetin alfa – epoetin beta – biosimilar ESA – placebo	0.64	0.00‐3.99
Epoetin alfa – darbepoetin alfa – placebo	0.17	0.00‐2.17
Epoetin alfa – epoetin beta – darbepoetin alfa – no treatment	0.16	0.00‐1.58
Epoetin alfa – epoetin beta – placebo – no treatment	0.07	0.00‐2.54
Transfusion
Epoetin alfa – epoetin beta – placebo – no treatment	2.09	0.00‐6.91
Epoetin alfa – darbepoetin alfa – placebo	1.97	0.00‐4.20
Epoetin beta – darbepoetin alfa ‐ methoxy polyethylene glycol‐epoetin beta ‐ placebo	1.26	0.00‐3.39
Myocardial infarction
Epoetin alfa – darbepoetin alfa – placebo	1.13	0.00‐4.37
Hypertension
Epoetin alfa – darbepoetin alfa – placebo	1.55	0.26‐2.84
Epoetin alfa – epoetin beta – darbepoetin alfa – no treatment	2.03	0.00‐4.66
Epoetin beta – darbepoetin alfa – placebo	1.56	0.73‐2.38
Epoetin alfa – epoetin beta – placebo – no treatment	2.15	0.00‐4.91
Epoetin beta – darbepoetin alfa – methoxy polyethylene glycol‐epoetin beta	2.49	0.76‐4.22
Vascular access thrombosis
Epoetin beta – darbepoetin alfa – placebo	1.32	0.00‐3.86
Epoetin beta – darbepoetin alfa – methoxy polyethylene glycol‐epoetin beta – placebo	0.98	0.00‐3.35

*The inconsistency factor is the absolute difference in the log odds ratio estimated from indirect and direct treatment comparisons and is reported together with the 95% confidence interval. A 95% confidence interval that includes zero indicates that the result is compatible with zero inconsistency between effect estimates using indirect (network meta‐analysis) and direct (conventional pairwise meta‐analysis) treatment comparisons. We used the'ifplot' command in STATA to estimate inconsistency (Chaimani 2013) allowing for all comparisons within a loop to share a common heterogeneity variance

4. Subgroup and sensitivity analyses

Subgroup and meta‐regression analyses to explore potential sources of heterogeneity and inconsistency in networks were precluded by sparse data for direct treatment comparisons (four studies or fewer for all comparisons). Differences in treatment estimates between studies and between direct and indirect evidence in network analyses may have been due to differing prescribing approaches, changing use of ESA across time, differing policies for blood transfusions and outcome adjudication and differing stages of CKD in contributing studies.

5. Grading of the evidence

When grading our confidence in the evidence using the methods of Del Giovane 2012, we first generated contribution matrices for the networks providing evidence for the primary outcomes (preventing blood transfusions (Figure 7) and all‐cause mortality (Figure 8)). In these matrices, the size of each square is proportional to the weight attached to each direct summary effect (horizontal axis) for the estimation of each network summary effect (vertical axis) with the numbers re‐expressing weights as percentages. We then evaluated risk of bias assessments for treatment comparisons obtained in network meta‐analyses proportional to study contributions (Figure 9 for the outcome of preventing blood transfusions and Figure 10 for the outcome of all‐cause mortality). We then considered the overall study limitations obtained from risk of bias assessments, imprecision in estimated treatment effects and inconsistency within networks.

Figure 7

Contributions matrix: percentage contribution of each direct estimate to the network meta‐analysis estimates for all‐cause mortality. Rows correspond to network meta‐analysis odds ratios (separated for mixed and indirect evidence) and the columns correspond to direct meta‐analysis odds ratios. The size of the shaded boxes are proportional to the percentage contribution of each direct estimate to the network meta‐analysis and to the entire network (lowest row). The last row shows the number of included direct comparisons. The names of the treatment comparisons are shown in the first column. For example, information for the network estimate of epoetin alfa versus darbepoetin alfa is derived from both direct and indirect evidence (generating a mixed estimate. Of this mixed network estimate, trials directly comparing epoetin alfa versus darbepoetin alfa contribute 44.8 % of the information to the network estimate of effect and trials directly comparing epoetin alfa versus placebo contribute 4.6% of the network estimated effect, etc. We used the 'netweight' command in STATA to generate the plot. The contribution matrix shows how much each direct comparison in the network contributes to each network (mixed or indirect) estimate

Figure 8

Contributions matrix: percentage contribution of each direct estimate to the network meta‐analysis estimates for preventing blood transfusions. Rows correspond to network meta‐analysis odds ratios (separated for mixed and indirect evidence) and the columns correspond to direct meta‐analysis odds ratios. The size of the shaded boxes are proportional to the percentage contribution of each direct estimate to the network meta‐analysis and to the entire network (lowest row). The last row shows the number of included direct comparisons. The names of the treatment comparisons are shown in the first column. For example, information for the network estimate of epoetin alfa versus darbepoetin alfa is derived from both direct and indirect evidence (generating a mixed estimate. Of this mixed network estimate, trials directly comparing epoetin alfa versus darbepoetin alfa contribute 78.7% of the information to the network estimate of effect and trials directly comparing epoetin alfa versus placebo contribute 7.2% of the network estimated effect, etc. We used the 'netweight' command in STATA to generate the plot. The contribution matrix shows how much each direct comparison in the network contributes to each network (mixed or indirect) estimate

Figure 9

Study limitations distribution for each network estimate for pairwise comparisons of erythropoiesis‐stimulating agents on the primary safety outcome (all‐cause mortality). Calculations are based on the contributions of direct evidence to the network estimates and the overall risks of bias from all bias domains (sequence generation, allocation concealment, blinding of participants and investigators, blinding of outcome assessment, attrition from follow‐up, selective outcome reporting and other sources of bias) within studies contributing to the direct evidence. The colours represent risk (green, low; yellow unclear; red, high). The direct comparisons are described in the vertical axis

Figure 10

Study limitations distribution for each network estimate for pairwise comparisons of erythropoiesis‐stimulating agents on the primary efficacy outcome (preventing blood transfusions). Calculations are based on the contributions of direct evidence to the network estimates and the overall risks of bias from all bias domains (sequence generation, allocation concealment, blinding of participants and investigators, blinding of outcome assessment, attrition from follow‐up, selective outcome reporting and other sources of bias) within studies contributing to the direct evidence. The colours represent risk (green, low; yellow unclear; red, high). The direct comparisons are described in the vertical axis

For each comparison, the quality of the evidence for preventing blood transfusions and all‐cause mortality was frequently downgraded from high‐quality due to important study limitations (summary of findings Table for the main comparison). Our confidence in the treatment estimates was generally moderate or low for comparisons of ESAs against placebo and was low or very low quality particularly for direct comparisons between two ESAs.

Discussion

available in

Summary of main results

Clinical guidelines recommend ESA treatment to avoid blood transfusions and anaemia‐related symptoms for patients with CKD (NICE 2011; KDIGO 2013). However, whether all the available ESAs are equally effective and safe has not been adequately evaluated by individual RCTs and is central to informed patient choices and rational pharmaceutical policy. To date, no studies have compared methoxy polyethylene glycol‐epoetin beta or biosimilar ESAs directly against placebo or have provided head‐to‐head comparisons for darbepoetin alfa and methoxy‐polyethylene glycol‐epoetin beta. This review of the effects of ESA treatment for anaemia in CKD included 56 studies involving 15,596 randomised adult participants and provides the first evidence for the comparative efficacy and safety of all ESAs in the setting of CKD.

Network meta‐analysis showed that all proprietary ESAs (epoetin alfa, epoetin beta, darbepoetin alfa, and methoxy polyethylene glycol‐epoetin beta) prevented blood transfusions compared to placebo with ORs ranging between 0.09 and 0.18. Our confidence in these beneficial effects was considered generally moderate or low using the GRADE approach. The efficacy of biosimilar ESAs for preventing blood transfusions was less certain and included the possibility of no effect; our confidence in this evidence was very low. The estimated treatment effects of the differing ESAs in head‐to‐head comparisons on preventing blood transfusions were imprecise and we could not be sure whether any of the formulations were similar or different for their effects on this outcome. The comparative effectiveness of different ESA formulations against each other or placebo on other potentially beneficial effects of therapy (such as reductions in fatigue and breathlessness) was inconclusive due to sparse data and the inconsistent methods used to report these outcomes in existing studies.

Although ESAs as a drug class are known to increase the odds of vascular and mortality outcomes when used to target higher haemoglobin levels (Palmer 2010; Phrommintikul 2007), the comparative safety of the available agents against each other and placebo is uncertain. All proprietary ESAs (epoetin alfa, epoetin beta, darbepoetin alfa, and methoxy polyethylene glycol‐epoetin beta) increased the odds of hypertension to a similar extent when compared with placebo, but estimated effects for biosimilar ESAs were much less precise. Greater precision in estimates for darbepoetin alfa (due to the availability of data from the large TREAT Study 2005 which provided 26% of all participants included in this review) suggested that darbepoetin alfa increased the odds of stroke compared to placebo but there was no clear evidence that this or other potential harms of darbepoetin alfa differed from those of other ESA derivatives. We could not discern any differences in the differing ESAs compared with placebo or each other for their effects on all‐cause mortality, cardiovascular mortality, myocardial infarction or vascular access thrombosis either in direct evidence or direct and indirect treatment comparisons in network analyses. Networks of evidence for the outcomes of end‐stage kidney disease and major cardiovascular events did not provide any closed loops of evidence; information from direct comparisons in available studies were uninformative and we could not identify whether ESAs differed from each other with regards to these outcomes.

The risks of bias in the contributing studies and imprecise treatment estimates limited our overall confidence in the results and inconsistency between direct and indirect evidence in analyses for preventing blood transfusion and hypertension reduced the credibility of treatment estimates derived by network analysis. Evidence from network meta‐analyses comparing ESA formulations against placebo were generally moderate or low quality while treatment estimates for head‐to‐head comparisons between ESAs were down‐graded to generally very low quality. Thus, presently there is no evidence for a preferred ESA to treat anaemia in CKD based on considerations of efficacy or safety including for biosimilar ESA preparations

Overall completeness and applicability of evidence

We included all eligible studies to February 2014 of ESA formulations that are currently available for the treatment of anaemia in CKD, but we did not include other potential interventions for anaemia in this setting such as iron supplementation. The information presented in this review is derived primarily from studies of epoetin alfa and darbepoetin alfa which contributed to outcome data for 5290 participants (33.9%) in the network of evidence for all‐cause mortality and 4749 participants (30.5%) in the network of evidence for preventing blood transfusions, due to a large contribution from the placebo‐controlled RCT of darbepoetin alfa which contributed 4038 participants (TREAT Study 2005). Data for biosimilar ESAs were limited to nine studies providing 2709 (17.4%) of participants in the meta‐analyses. Due to the few studies and data for biosimilar ESAs, we elected to combine information for all the different non‐proprietary ESAs studied (alfa, delta, omega, theta, zeta), but recognise that these agents may have important clinical and biological differences in their effects. In addition, selective outcome reporting reduced our confidence in the estimated treatment effects.

Participants involved in the meta‐analyses were nearly equally distributed between those treated with dialysis or those with milder forms of CKD, however participants who were recipients of a kidney transplant for treatment of end‐stage kidney disease were rarely involved in contributing studies and the findings of this review may not be directly applicable to this clinical setting. In addition, it was unclear how many participants in studies in dialysis were conducted in the setting of peritoneal dialysis.

While we aimed to incorporate many patient‐important outcomes including symptoms of anaemia such as breathlessness and fatigue, other relevant outcomes for participants were not included (such as health‐related quality of life) as these are infrequently reported in anaemia studies in CKD and are frequently at risk of selective reporting (Clement 2009). In addition, the core outcomes of greatest importance to clinicians and patients in the management of CKD remain poorly explored, in comparison to other clinical specialities (such as rheumatology (the OMERACT (Outcome Measures in Rheumatology) initiative; www.omeract.org/)), and as such we could not align our review with the outcomes considered most important to consumers, health professionals and other stakeholders. We deliberately did not include analysis of treatment effects on haemoglobin levels as this is a surrogate outcome that adds little if anything to our understanding of clinical outcomes.

Quality of the evidence

Risks of bias

Risks of bias in the included studies was generally high or unclear for more than half of studies in all of the risk of bias domains we assessed, limiting our confidence in the estimated treatment effects from these data. No study was low risk for allocation concealment, blinding of outcome assessment and attrition from follow‐up. Allocation concealment, in which investigators are unaware of the treatment allocation for individual participants, was reported using low risk methods in only 10 (18%) studies. Blinding of outcome assessment was clearly documented as low risk in two studies (4%), and differences in haemoglobin levels between groups made it likely that investigators were aware of the treatment allocation in the remaining placebo‐ or no treatment‐controlled studies. Follow‐up data was incomplete (for more than 10% of randomised participants and/or markedly discrepant between treatment arms) in 31 (55%) studies and unclearly documented in a further 18 (32%) studies.

Heterogeneity

Evidence for moderate to substantial heterogeneity was present for many pairwise meta‐analysis beyond that expected from random variation; however, analyses lacked power for subgroup or meta‐regression analyses due to the small number of studies (≤ four studies) in these meta‐analyses. There were substantial differences in the treatment effects in studies comparing epoetin alfa with placebo on blood transfusion that may have related to the characteristics of the patient populations or transfusion policies within the studies. Evidence of moderate heterogeneity was common in treatment effects estimated from studies comparing darbepoetin with methoxy polyethylene glycol‐epoetin beta, which limited our confidence in estimated treatment effects for myocardial infarction and hypertension. In addition, moderate to substantial heterogeneity in treatment estimates for epoetin alfa compared with darbepoetin alfa were present in analyses for all‐cause mortality, hypertension and end‐stage kidney disease.

Inconsistency

Notably, there was important clinical diversity in the included studies and evidence of inconsistency between treatment effects estimated from direct evidence (within head‐to‐head studies) and mixed evidence (from both direct and indirect evidence) generated using network analyses for the outcomes of blood transfusion and hypertension. Our confidence in the results obtained from network meta‐analysis was reduced by these differences. Importantly, as data were generally sparse, our ability to ascertain evidence of inconsistency was relatively low, and important inconsistency within analyses could not be excluded.

Potential biases in the review process

While this review was prepared using a sensitive electronic search strategy to identify eligible studies, was conducted according to a prespecified protocol and is reported using Cochrane Collaboration methods, the review has limitations which should be considered when interpreting the results. First, relatively few data were available for most comparisons resulting in inconclusive evidence for many outcomes including cardiovascular events, anaemia‐related symptoms, and end‐stage kidney disease. Second, while the included studies appeared similar in their treatment approaches to anaemia, there was evidence of heterogeneity in treatment effects estimated by individual studies and inconsistency between direct and indirect evidence that reduced the credibility of estimated treatment safety and efficacy which could not be explored meaningfully using subgroup or meta‐regression analyses. Third, data for a single study (the TREAT Study 2005 study) dominated many of the analyses and most studies had high or unclear risks of bias for key domains. Finally, outcome data for patient‐important outcomes were not available in most studies or were reported ad hoc and therefore reduced our confidence in the reliability of these treatment effects. This was particularly the case for reporting of quality of life domains such as fatigue; the reporting of this outcome did not allow many data to be included in analyses.

In addition, we originally included network analyses for the effects of higher and lower haemoglobin targets with ESAs in the peer‐reviewed protocol of this review. As these comparisons are adequately addressed by existing meta‐analyses, and network analyses are unlikely to provide more information for these comparisons than conventional pairwise meta‐analyses and hinder the readability and usefulness of this review, we have not included the results of these networks in the final version of this review.

Agreements and disagreements with other studies or reviews

To our knowledge, this is the first systematic review directly comparing different ESA preparations against each other or placebo in the setting of kidney disease and is the first to evaluate the evidence for biosimilar ESA formulations against proprietary ESAs and placebo. Previous systematic reviews of ESA drugs to treat anaemia in CKD have largely focused on treatment strategies that compare higher with lower haemoglobin targets using the same ESA or ESAs against placebo or no treatment (Palmer 2010; Phrommintikul 2007; Strippoli 2006) which have generally shown increased cardiovascular events and mortality in patients who have been prescribed ESA treatment to achieve a higher haemoglobin target. The findings in this review are similar to an earlier Cochrane meta‐analysis comparing darbepoetin alfa against placebo or other ESAs which found that darbepoetin alfa reduced transfusion without showing beneficial effects on mortality or quality of life, while the treatment effects of darbepoetin alfa compared to other ESAs were uncertain (Palmer 2014).

Our findings agree with a review of RCTs comparing the efficacy and safety of epoetin (alfa or beta) with darbepoetin alfa in patients undergoing cancer treatment (AHRQ 2006). In analyses of head‐to‐head studies comparing epoetin with darbepoetin (7 studies, 1415 participants), epoetin versus placebo or no treatment (48 studies, 4518 participants) and darbepoetin versus placebo or no treatment (4 studies, 598 participants), those authors found that there was no statistical difference between epoetin alfa or beta and darbepoetin alfa on need for blood transfusion or risk of thromboembolic events and that both epoetin alfa or beta and darbepoetin alfa were better than placebo or no treatment at preventing blood transfusions, although many analyses showed evidence of important heterogeneity. Similar to our analyses, the authors reported that many studies were not designed to evaluate survival and that reporting of quality of life outcomes was frequently unusable for analyses. In an updated report dated 2013, treatment effects for darbepoetin alfa compared to epoetin alfa or beta on preventing blood transfusions, on‐study mortality, and thromboembolic events remained inconclusive and information for quality of life was assessed as low quality (AHRQ 2013).

While a study of the comparative effectiveness of IV iron preparations on patient outcomes for adults who have ESKD is ongoing, as this is a non‐randomised analysis, it is unlikely that treatment effects will be sufficiently free of confounding by treatment indication and other patient and health services related characteristics to inform clinical practice and policy (Boulware 2012) and does not address the use of ESAs.

Figure 1

Overview of anaemia in chronic disease

Navigate to figure in ReviewOpen in new tab

Figure 2

Study flow diagram.

Navigate to figure in ReviewOpen in new tab

Figure 3

Navigate to figure in ReviewOpen in new tab

Figure 4

Navigate to figure in ReviewOpen in new tab

Figure 5

Networks of the treatment efficacy and safety of ESA drugs in the treatment of anaemia in chronic kidney disease. Values lower than 1 favour the active treatment in the comparison

Navigate to figure in ReviewOpen in new tab

Figure 6

Forest plots for results from network meta‐analyses comparing ESAs versus placebo

Navigate to figure in ReviewOpen in new tab

Figure 7

Navigate to figure in ReviewOpen in new tab

Figure 8

Navigate to figure in ReviewOpen in new tab

Figure 9

Navigate to figure in ReviewOpen in new tab

Figure 10

Navigate to figure in ReviewOpen in new tab

Analysis 1.1

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 1 Blood transfusion.

Navigate to figure in ReviewOpen in new tab

Analysis 1.2

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 2 Fatigue.

Navigate to figure in ReviewOpen in new tab

Analysis 1.3

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 3 Breathlessness.

Navigate to figure in ReviewOpen in new tab

Analysis 1.4

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 4 All‐cause mortality.

Navigate to figure in ReviewOpen in new tab

Analysis 1.5

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 5 Cardiovascular mortality.

Navigate to figure in ReviewOpen in new tab

Analysis 1.6

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 6 Myocardial infarction.

Navigate to figure in ReviewOpen in new tab

Analysis 1.7

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 7 Stroke.

Navigate to figure in ReviewOpen in new tab

Analysis 1.8

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 8 Hypertension.

Navigate to figure in ReviewOpen in new tab

Analysis 1.9

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 9 Vascular access thrombosis.

Navigate to figure in ReviewOpen in new tab

Analysis 1.10

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 10 End‐stage kidney disease.

Navigate to figure in ReviewOpen in new tab

Analysis 1.11

Comparison 1 ESA versus ESA or placebo/no treatment, Outcome 11 Major cardiovascular events.

Navigate to figure in ReviewOpen in new tab

Summary of findings for the main comparison. Erythropoiesis‐stimulating agents (ESAs) for anaemia in adults with chronic kidney disease (CKD)

ESAs for anaemia in adults with CKD
Intervention	Comparison/intervention	Nature of the evidence	Confidence in the evidence	Reasons for downgrading our confidence in the evidence*	Network treatment estimate OR (95% CI)
Preventing blood transfusion
Epoetin alfa	Placebo	Mixed	Low	Study limitations (‐1) Inconsistency (‐1)	0.18 (0.05 to 0.59)
Epoetin beta	Placebo	Mixed	Low	Study limitations (‐1) Inconsistency (‐1)	0.09 (0.02 to 0.38)
Darbepoetin alfa	Placebo	Mixed	Moderate	Inconsistency (‐1)	0.17 (0.05 to 0.57)
Methoxy polyethylene glycol‐epoetin beta	Placebo	Indirect	Low	Study limitations (‐1) Inconsistency (‐1)	0.15 (0.03 to 0.70)
Biosimilar ESA	Placebo	Indirect	Very low	Study limitations (‐1) Imprecision (‐1) Inconsistency (‐1)	0.27 (0.05 to 1.47)
Epoetin alfa	Epoetin beta	Indirect	Very low	Study limitations (‐1) Imprecision (‐1) Imprecision (‐1)	2.04 (0.38 to 11.0)
Epoetin alfa	Darbepoetin alfa	Mixed	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	1.06 (0.35 to 3.29)
Epoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Indirect	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	1.14 (0.27 to 4.97)
Epoetin alfa	Biosimilar ESA	Mixed	Very low	Study limitations (‐1) Imprecision (‐1) Imprecision (‐1)	0.66 (0.19 to 2.28)
Epoetin beta	Darbepoetin alfa	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.52 (0.10 to 2.67)
Epoetin beta	Methoxy polyethylene glycol‐epoetin beta	Mixed	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.56 (0.11 to 3.00)
Epoetin beta	Biosimilar ESA	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.33 (0.04 to 2.60)
Darbepoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Mixed	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	1.08 (0.38 to 3.04)
Darbepoetin alfa	Biosimilar ESA	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.62 (0.12 to 3.30)
Methoxy polyethylene glycol‐epoetin beta	Biosimilar ESA	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	0.58 (0.09 to 3.92)
All‐cause mortality
Epoetin alfa	Placebo	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	1.25 (0.71 to 2.21)
Epoetin beta	Placebo	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.82 (0.45 to 1.48)
Darbepoetin alfa	Placebo	Mixed	Moderate	Imprecision (‐1)	1.06 (0.91 to 1.24)
Methoxy polyethylene glycol‐epoetin beta	Placebo	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	1.16 (0.74 to 1.82)
Biosimilar ESA	Placebo	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	1.31 (0.65 to 2.62)
Epoetin alfa	Epoetin beta	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	1.53 (077 to 3.03)
Epoetin alfa	Darbepoetin alfa	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	1.17 (0.68 to 2.05)
Epoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Indirect	Very low	Study limitations (‐1) Inconsistency (‐1) Imprecision (‐1)	1.08 (0.54 to 2.15)
Epoetin alfa	Biosimilar ESA	Mixed	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	0.95 (0.62 to 1.44)
Epoetin beta	Darbepoetin alfa	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.77 (0.43 to 1.38)
Epoetin beta	Methoxy polyethylene glycol‐epoetin beta	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.71 (0.35 to 1.42)
Epoetin beta	Biosimilar ESA	Mixed	Low	Study limitations (‐1) Imprecision (‐1)	0.62 (0.29 to 1.37)
Darbepoetin alfa	Methoxy polyethylene glycol‐epoetin beta	Mixed	Very low	Study limitations (‐2) Imprecision (‐1)	0.91 (0.60 to 1.40)
Darbepoetin alfa	Biosimilar ESA	Indirect	Low	Study limitations (‐1) Imprecision (‐1)	0.81 (0.41 to 1.61)
Methoxy polyethylene glycol‐epoetin beta	Biosimilar ESA	Indirect	Very low	Study limitations (‐2) Inconsistency (‐1) Imprecision (‐1)	0.88 (0.40 to 1.97)
CI: Confidence interval; OR: Odds Ratio *There was moderate heterogeneity in the network for preventing blood transfusion (τ = 0.89 which was between the 50th and 75th quartile of empirical distributions of heterogeneity variances specific to the type of outcome and types of treatments being compared) (Turner 2012) We did not downgrade for reasons of indirectness or publication bias as insufficient studies contributed to network treatment estimates to draw meaningful conclusions. We downgraded for inconsistency when the network did not include a closed loop of evidence for the comparison and accordingly the presence of inconsistency could not be excluded.
GRADE Working Group grades of evidence (GRADE: Rating the quality of evidence 2011) High quality: We are very confidence that the true effect lies close to that of the estimate of effect Moderate quality: We are moderately confident in the estimate of effect: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of effect Very low quality: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect

Summary of findings for the main comparison. Erythropoiesis‐stimulating agents (ESAs) for anaemia in adults with chronic kidney disease (CKD)

Navigate to table in Review

Table 1. Comparative effects of erythropoiesis‐stimulating agents on clinical outcomes in chronic kidney disease

Outcomes / interventions	Comparators (treatment estimate (OR (95% CI))
Outcomes / interventions	Epoetin alfa	Epoetin beta	Darbepoetin alfa	Methoxy polyethylene‐glycol epoetin beta	Biosimilar ESA	Placebo
Blood transfusion
Epoetin alfa	‐‐	2.04 (0.38‐11.0)	1.06 (0.35‐3.29)	1.14 (0.27‐4.97)	0.66 (0.19‐2.28)	0.18 (0.05‐0.59)
Epoetin beta	Not estimable	‐‐	0.52 (0.10‐2.67)	0.56 (0.11‐3.00)	0.33 (0.04‐2.60)	0.09 (0.02‐0.38)
Darbepoetin alfa	2.31 (1.34‐3.97)	Not estimable	‐‐	1.08 (0.38‐3.04)	0.62 (0.12‐3.30)	0.17 (0.05‐0.57)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	0.83 (0.17‐4.15)	0.94 (0.45‐1.95)	‐‐	0.58 (0.09‐3.92)	0.15 (0.03‐0.70)
Biosimilar ESA	0.72 (0.42‐1.22)	Not estimable	Not estimable	Not estimable	‐‐	0.27 (0.05‐1.47)
Placebo	0.07 (0.01‐0.84)	0.07 (0.03‐0.21)	0.53 (0.46‐0.63)	Not estimable	Not estimable	‐‐
All‐cause mortality
Epoetin alfa	‐‐	1.53 (0.77‐3.03)	1.17 (0.68‐2.05)	1.08 (0.54‐2.15)	0.95 (0.62‐1.44)	1.25 (0.71‐2.21)
Epoetin beta	Not estimable	̶	0.77 (0.43‐1.38)	0.71 (0.35‐1.42)	0.62 (0.29‐1.37)	0.82 (0.45‐1.48)
Darbepoetin alfa	1.12 (0.59‐2.14)	0.89 (0.38‐2.09)	‐‐	0.91 (0.60‐1.40)	0.81 (0.41‐1.61)	1.06 (0.91‐1.24)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	0.81 (0.12‐5.35)	0.90 (0.59‐1.40)	‐‐	0.88 (0.40‐1.97)	1.16 (0.74‐1.82)
Biosimilar ESA	1.04 (0.53‐2.01)	0.34 (0.04‐2.82)	Not estimable	Not estimable	‐‐	1.31 (0.65‐2.62)
Placebo	0.99 (0.14‐6.86)	0.61 (0.17‐2.15)	1.06 (0.91‐1.24)	Not estimable	Not estimable	‐‐
Fatigue
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.94 (0.57‐1.55)	Not estimable	‐‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	Not estimable	‐‐̶	Not estimable	Not estimable
Biosimilar ESA	0.18 (0.01‐3.91)	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable	‐‐
Breathlessness
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.71 (0.46‐1.10)	Not estimable	‐‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	Not estimable	‐‐	Not estimable	Not estimable
Biosimilar ESA	0.68 (0.37‐1.25)	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable	‐‐
Cardiovascular mortality
Epoetin alfa	‐‐	2.12 (0.34‐13.1)	1.48 (0.28‐7.96)	1.02 (0.16‐6.48)	0.55 (0.22‐1.38)	1.56 (0.29‐8.37)
Epoetin beta	Not estimable	‐‐	0.70 (0.12‐4.10)	0.48 (0.07‐3.31)	0.26 (0.04‐1.51)	0.74 (0.13‐4.28)
Darbepoetin alfa	2.15 (0.31‐14.9)	Not estimable	‐‐	0.69 (0.32‐1.48)	0.37 (0.06‐2.20)	1.05 (0.87‐1.26)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	0.69 (0.32‐1.48)	‐‐	0.54 (0.08‐3.74)	1.52 (0.69‐3.34)
Biosimilar ESA	0.53 (0.20‐1.35)	0.34 (0.04‐2.82)	Not estimable	Not estimable	‐‐	2.81 (0.47‐16.7)
Placebo	Not estimable	0.45 (0.06‐3.75)	1.05 (0.87‐1.26)	Not estimable	Not estimable	‐‐
Major adverse cardiovascular events
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.20 (0.01‐4.17)	Not estimable	‐‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	Not estimable	‐‐	Not estimable	Not estimable
Biosimilar ESA	0.49 (0.17‐1.47)	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	1.08 (0.95‐1.24)	Not estimable	Not estimable	‐‐
Myocardial infarction
Epoetin alfa	‐‐	Not estimable	1.04 (0.35‐3.11)	0.55 (0.05‐5.69)	1.18 (0.47‐3.02)	1.00 (0.32‐3.09)
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	0.87 (0.20‐3.81)	Not estimable	‐‐	0.53 (0.07‐4.18)	1.14 (0.27‐4.83)	0.97 (0.75‐1.25)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	0.47 (0.06‐3.65)	‐‐	2.17 (0.17‐27.1)	1.83 (0.18‐19.1)
Biosimilar ESA	1.23 (0.49‐3.12)	Not estimable	Not estimable	Not estimable	‐‐	0.84 (0.20‐3.65)
Placebo	3.46 (0.12‐100.51)	Not estimable	0.97 (0.75‐1.25)	Not estimable	Not estimable	‐‐
Stroke
Epoetin alfa	‐‐	4.56 (0.29‐71.8)	1.39 (0.38‐5.16)	2.36 (0.24‐23.6)	0.92 (0.39‐2.16)	2.74 (0.71‐10.5)
Epoetin beta	Not estimable	‐‐	0.31 (0.02‐4.55)	0.52 (0.02‐14.0)	0.20 (0.01‐3.61)	0.60 (0.04‐8.88)
Darbepoetin alfa	1.44 (0.37‐5.54)	Not estimable	‐‐	1.70 (0.26‐11.2)	0.66 (0.14‐3.14)	1.96 (1.40‐2.75)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	1.33 (0.17‐10.49)	‐‐	0.38 (0.03‐4.50)	1.16 (0.17‐7.90)
Biosimilar ESA	0.92 (0.39‐2.15)	Not estimable	Not estimable	Not estimable	‐‐	2.99 (0.61‐14.8)
Placebo	Not estimable	0.33 (0.01‐8.21)	1.97 (1.40‐2.76)	Not estimable	Not estimable	‐‐
Hypertension
Epoetin alfa	‐‐	0.90 (0.41‐1.95)	1.26 (0.81‐1.96)	1.18 (0.64‐2.18)	1.95 (0.97‐3.94)	2.31 (1.27‐4.23)
Epoetin beta	Not estimable	‐‐	1.41 (0.70‐2.82)	1.31 (0.63‐2.72)	2.18 (0.76‐6.22)	2.57 (1.23‐5.39)
Darbepoetin alfa	0.94 (0.62‐1.43)	1.18 (0.38‐3.69)	‐‐	0.93 (0.60‐1.45)	1.55 (0.68‐3.55)	1.83 (1.05‐3.21)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	1.38 (0.62‐3.09)	Not estimable	‐‐	1.66 (0.65‐4.21)	1.96 (0.98‐3.92)
Biosimilar ESA	1.77 (1.02‐3.09)	Not estimable	Not estimable	Not estimable	‐‐	1.18 (0.47‐2.99)
Placebo	4.10 (2.16‐7.76)	2.95 (1.19‐7.26)	1.14 (0.99‐1.32)	Not estimable	Not estimable	‐‐
End‐stage kidney disease
Epoetin alfa	‐‐	Not estimable	Not estimable	Not estimable	Not estimable	Not estimable
Epoetin beta	Not estimable	‐‐	Not estimable	Not estimable	Not estimable	Not estimable
Darbepoetin alfa	2.17 (0.37‐12.74)	Not estimable	‐	Not estimable	Not estimable	Not estimable
Methoxy polyethylene‐glycol epoetin beta	Not estimable	Not estimable	1.83 (0.66‐5.09)	‐‐	Not estimable	Not estimable
Biosimilar ESA	Not estimable	Not estimable	Not estimable	Not estimable	‐‐	Not estimable
Placebo	Not estimable	Not estimable	1.04 (0.88‐1.23)	Not estimable	Not estimable	‐‐
Vascular access thrombosis
Epoetin alfa	‐‐	0.93 (0.28‐3.10)	1.22 (0.78‐1.91)	1.04 (0.48‐2.25)	1.26 (0.45‐3.36)	1.72 (0.58‐5.16)
Epoetin beta	Not estimable	‐‐	1.30 (0.42‐4.04)	1.11 (0.38‐3.24)	1.35 (0.29‐6.34)	1.85 (0.61‐5.63)
Darbepoetin alfa	1.15 (0.73‐1.82)	Not estimable	‐‐	0.86 (0.45‐1.61)	1.04 (0.35‐3.05)	1.42 (0.50‐4.03)
Methoxy polyethylene‐glycol epoetin beta	Not estimable	1.74 (0.49‐6.24)	0.76 (0.39‐1.47)	‐‐	1.21 (0.35‐4.22)	1.66 (0.54‐5.08)
Biosimilar ESA	1.71 (0.30‐10.00)	Not estimable	Not estimable	Not estimable	‐‐	1.37 (0.32‐5.93)
Placebo	6.40 (0.80‐51.50)	1.09 (0.28‐4.34)	1.34 (0.30‐6.01)	Not estimable	Not estimable	‐‐
Treatment estimates for pairwise meta‐analyses are shown in italics

Table 1. Comparative effects of erythropoiesis‐stimulating agents on clinical outcomes in chronic kidney disease

Navigate to table in Review

Table 2. Evaluation of consistency using loop specific approach

Treatments included in the loop of evidence	Inconsistency factor*	95% CI
All‐cause mortality
Epoetin alfa – epoetin beta – darbepoetin alfa – biosimilar ESA	0.87	0.00‐3.32
Epoetin beta – darbepoetin alfa – placebo	0.40	0.00‐1.82
Epoetin alfa – epoetin beta – biosimilar ESA – no treatment	0.66	0.00‐3.36
Epoetin beta – darbepoetin alfa – methoxy polyethylene glycol‐epoetin beta	0.02	0.00‐2.08
Epoetin alfa – epoetin beta – biosimilar ESA – placebo	0.64	0.00‐3.99
Epoetin alfa – darbepoetin alfa – placebo	0.17	0.00‐2.17
Epoetin alfa – epoetin beta – darbepoetin alfa – no treatment	0.16	0.00‐1.58
Epoetin alfa – epoetin beta – placebo – no treatment	0.07	0.00‐2.54
Transfusion
Epoetin alfa – epoetin beta – placebo – no treatment	2.09	0.00‐6.91
Epoetin alfa – darbepoetin alfa – placebo	1.97	0.00‐4.20
Epoetin beta – darbepoetin alfa ‐ methoxy polyethylene glycol‐epoetin beta ‐ placebo	1.26	0.00‐3.39
Myocardial infarction
Epoetin alfa – darbepoetin alfa – placebo	1.13	0.00‐4.37
Hypertension
Epoetin alfa – darbepoetin alfa – placebo	1.55	0.26‐2.84
Epoetin alfa – epoetin beta – darbepoetin alfa – no treatment	2.03	0.00‐4.66
Epoetin beta – darbepoetin alfa – placebo	1.56	0.73‐2.38
Epoetin alfa – epoetin beta – placebo – no treatment	2.15	0.00‐4.91
Epoetin beta – darbepoetin alfa – methoxy polyethylene glycol‐epoetin beta	2.49	0.76‐4.22
Vascular access thrombosis
Epoetin beta – darbepoetin alfa – placebo	1.32	0.00‐3.86
Epoetin beta – darbepoetin alfa – methoxy polyethylene glycol‐epoetin beta – placebo	0.98	0.00‐3.35
*The inconsistency factor is the absolute difference in the log odds ratio estimated from indirect and direct treatment comparisons and is reported together with the 95% confidence interval. A 95% confidence interval that includes zero indicates that the result is compatible with zero inconsistency between effect estimates using indirect (network meta‐analysis) and direct (conventional pairwise meta‐analysis) treatment comparisons. We used the'ifplot' command in STATA to estimate inconsistency (Chaimani 2013) allowing for all comparisons within a loop to share a common heterogeneity variance

Table 2. Evaluation of consistency using loop specific approach

Navigate to table in Review

Comparison 1. ESA versus ESA or placebo/no treatment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Blood transfusion Show forest plot	19		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

1.1 Epoetin alfa versus placebo	3	196	Odds Ratio (M‐H, Random, 95% CI)	0.07 [0.01, 0.84]
1.2 Epoetin beta versus placebo	2	230	Odds Ratio (M‐H, Random, 95% CI)	0.07 [0.03, 0.21]
1.3 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	0.53 [0.46, 0.63]
1.4 Epoetin alfa versus control	1	157	Odds Ratio (M‐H, Random, 95% CI)	3.10 [0.16, 58.97]
1.5 Epoetin beta versus no control	1	40	Odds Ratio (M‐H, Random, 95% CI)	0.35 [0.06, 2.18]
1.6 Epoetin alfa versus darbepoetin alfa	3	1191	Odds Ratio (M‐H, Random, 95% CI)	2.31 [1.34, 3.97]
1.7 Epoetin alfa versus biosimilar ESA	3	1823	Odds Ratio (M‐H, Random, 95% CI)	0.72 [0.42, 1.22]
1.8 Epoetin beta versus methoxy polyethylene glycol‐epoetin beta	1	181	Odds Ratio (M‐H, Random, 95% CI)	0.83 [0.17, 4.15]
1.9 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	4	1191	Odds Ratio (M‐H, Random, 95% CI)	0.94 [0.45, 1.95]
2 Fatigue Show forest plot	3		Odds Ratio (IV, Random, 95% CI)	Subtotals only

2.1 Epoetin alfa versus darbepoetin alfa	2	551	Odds Ratio (IV, Random, 95% CI)	0.94 [0.57, 1.55]
2.2 Epoetin alfa v biosimilar ESA	1	179	Odds Ratio (IV, Random, 95% CI)	0.18 [0.01, 3.91]
3 Breathlessness Show forest plot	3		Odds Ratio (IV, Random, 95% CI)	Subtotals only

3.1 Epoetin alfa versus darbepoetin alfa	1	504	Odds Ratio (IV, Random, 95% CI)	0.71 [0.46, 1.10]
3.2 Epoetin alfa versus biosimilar ESA	2	794	Odds Ratio (IV, Random, 95% CI)	0.68 [0.37, 1.25]
4 All‐cause mortality Show forest plot	31		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

4.1 Epoetin alfa versus placebo	2	235	Odds Ratio (M‐H, Random, 95% CI)	0.99 [0.14, 6.86]
4.2 Epoetin beta versus placebo	3	311	Odds Ratio (M‐H, Random, 95% CI)	0.61 [0.17, 2.15]
4.3 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	1.06 [0.91, 1.24]
4.4 Epoetin alfa versus control	1	157	Odds Ratio (M‐H, Random, 95% CI)	1.06 [0.39, 2.87]
4.5 Epoetin beta versus control	3	468	Odds Ratio (M‐H, Random, 95% CI)	0.69 [0.36, 1.33]
4.6 Epoetin alfa versus darbepoetin alfa	6	1205	Odds Ratio (M‐H, Random, 95% CI)	1.12 [0.59, 2.14]
4.7 Epoetin alfa versus biosimilar ESA	7	2220	Odds Ratio (M‐H, Random, 95% CI)	1.04 [0.53, 2.01]
4.8 Epoetin beta versus darbepoetin alfa	1	217	Odds Ratio (M‐H, Random, 95% CI)	0.89 [0.38, 2.09]
4.9 Epoetin beta versus methoxy polyethylene glycol‐epoetin beta	2	462	Odds Ratio (M‐H, Random, 95% CI)	0.81 [0.12, 5.35]
4.10 Epoetin beta versus biosimilar ESA	1	290	Odds Ratio (M‐H, Random, 95% CI)	0.34 [0.04, 2.82]
4.11 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	4	1429	Odds Ratio (M‐H, Random, 95% CI)	0.90 [0.59, 1.40]
5 Cardiovascular mortality Show forest plot	14		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

5.1 Epoetin beta versus placebo	2	260	Odds Ratio (M‐H, Random, 95% CI)	0.45 [0.06, 3.75]
5.2 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	1.05 [0.87, 1.26]
5.3 Epoetin beta versus no treatment	3	430	Odds Ratio (M‐H, Random, 95% CI)	0.28 [0.08, 1.03]
5.4 Epoetin alfa versus darbepoetin alfa	2	487	Odds Ratio (M‐H, Random, 95% CI)	2.15 [0.31, 14.91]
5.5 Epoetin alfa versus biosimilar ESA	2	657	Odds Ratio (M‐H, Random, 95% CI)	0.53 [0.20, 1.35]
5.6 Epoetin beta versus biosimilar ESA	1	290	Odds Ratio (M‐H, Random, 95% CI)	0.34 [0.04, 2.82]
5.7 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	3	938	Odds Ratio (M‐H, Random, 95% CI)	0.69 [0.32, 1.48]
6 Myocardial infarction Show forest plot	9		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

6.1 Epoetin alfa versus placebo	1	14	Odds Ratio (M‐H, Random, 95% CI)	3.46 [0.12, 100.51]
6.2 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	0.97 [0.75, 1.25]
6.3 Epoetin alfa versus no treatment	1	157	Odds Ratio (M‐H, Random, 95% CI)	1.01 [0.04, 25.26]
6.4 Epoetin alfa versus darbepoetin alfa	2	825	Odds Ratio (M‐H, Random, 95% CI)	0.87 [0.20, 3.81]
6.5 Epoetin alfa versus biosimilar	2	641	Odds Ratio (M‐H, Random, 95% CI)	1.23 [0.49, 3.12]
6.6 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	2	628	Odds Ratio (M‐H, Random, 95% CI)	0.47 [0.06, 3.65]
7 Stroke Show forest plot	12		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

7.1 Epoetin beta versus placebo	1	106	Odds Ratio (M‐H, Random, 95% CI)	0.33 [0.01, 8.21]
7.2 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	1.97 [1.40, 2.76]
7.3 Epoetin alfa versus control	1	157	Odds Ratio (M‐H, Random, 95% CI)	0.99 [0.10, 9.82]
7.4 Epoetin beta versus control	1	33	Odds Ratio (M‐H, Random, 95% CI)	0.20 [0.01, 5.39]
7.5 Epoetin alfa versus darbepoetin alfa	3	996	Odds Ratio (M‐H, Random, 95% CI)	1.44 [0.37, 5.54]
7.6 Epoetin alfa versus biosimilar ESA	3	718	Odds Ratio (M‐H, Random, 95% CI)	0.92 [0.39, 2.15]
7.7 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	2	628	Odds Ratio (M‐H, Random, 95% CI)	1.33 [0.17, 10.49]
8 Hypertension Show forest plot	24		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

8.1 Epoetin alfa versus placebo	2	251	Odds Ratio (M‐H, Random, 95% CI)	4.10 [2.16, 7.76]
8.2 Epoetin beta versus placebo	2	230	Odds Ratio (M‐H, Random, 95% CI)	2.95 [1.19, 7.26]
8.3 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	1.14 [0.99, 1.32]
8.4 Epoetin alfa versus control	1	157	Odds Ratio (M‐H, Random, 95% CI)	5.31 [0.30, 95.20]
8.5 Epoetin beta versus no treatment	2	382	Odds Ratio (M‐H, Random, 95% CI)	2.99 [1.34, 6.69]
8.6 Epoetin alfa versus darbepoetin alfa	5	1568	Odds Ratio (M‐H, Random, 95% CI)	0.94 [0.62, 1.43]
8.7 Epoetin alfa versus biosimilar ESA	4	1464	Odds Ratio (M‐H, Random, 95% CI)	1.77 [1.02, 3.09]
8.8 Epoetin beta versus darbepoetin alfa	1	162	Odds Ratio (M‐H, Random, 95% CI)	1.18 [0.38, 3.69]
8.9 Epoetin beta versus methoxy polyethylene glycol‐epoetin beta	1	181	Odds Ratio (M‐H, Random, 95% CI)	1.38 [0.62, 3.09]
8.10 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	5	1497	Odds Ratio (M‐H, Random, 95% CI)	0.94 [0.62, 1.42]
9 Vascular access thrombosis Show forest plot	11		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

9.1 Epoetin alfa versus placebo	1	118	Odds Ratio (M‐H, Random, 95% CI)	6.40 [0.80, 51.50]
9.2 Epoetin beta versus placebo	1	99	Odds Ratio (M‐H, Random, 95% CI)	1.09 [0.28, 4.34]
9.3 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	1.34 [0.30, 6.01]
9.4 Epoetin beta versus control	1	362	Odds Ratio (M‐H, Random, 95% CI)	1.40 [0.72, 2.73]
9.5 Epoetin alfa versus darbepoetin alfa	3	1084	Odds Ratio (M‐H, Random, 95% CI)	1.15 [0.73, 1.82]
9.6 Epoetin alfa versus biosimilar	2	823	Odds Ratio (M‐H, Random, 95% CI)	1.74 [0.30, 10.00]
9.7 Epoetin beta versus methoxy polyethylene glycol‐epoetin beta	1	181	Odds Ratio (M‐H, Random, 95% CI)	1.74 [0.49, 6.24]
9.8 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	1	489	Odds Ratio (M‐H, Random, 95% CI)	0.76 [0.39, 1.47]
10 End‐stage kidney disease Show forest plot	7		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

10.1 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	1.04 [0.88, 1.23]
10.2 Epoetin alfa versus control	1	17	Odds Ratio (M‐H, Random, 95% CI)	0.27 [0.03, 2.12]
10.3 Epoetin beta versus control	2	106	Odds Ratio (M‐H, Random, 95% CI)	0.40 [0.08, 1.93]
10.4 Epoetin alfa versus darbepoetin alfa	2	492	Odds Ratio (M‐H, Random, 95% CI)	2.17 [0.37, 12.74]
10.5 Darbepoetin alfa versus methoxy polyethylene glycol‐epoetin beta	1	305	Odds Ratio (M‐H, Random, 95% CI)	1.83 [0.66, 5.09]
11 Major cardiovascular events Show forest plot	5		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only

11.1 Darbepoetin alfa versus placebo	1	4038	Odds Ratio (M‐H, Random, 95% CI)	1.08 [0.95, 1.24]
11.2 Epoetin alfa versus control	1	157	Odds Ratio (M‐H, Random, 95% CI)	2.40 [0.29, 20.11]
11.3 Epoetin beta versus control	1	33	Odds Ratio (M‐H, Random, 95% CI)	0.61 [0.07, 4.98]
11.4 Epoetin alfa versus darbepoetin alfa	1	321	Odds Ratio (M‐H, Random, 95% CI)	0.20 [0.01, 4.17]
11.5 Epoetin alfa versus biosimilar epoetin	1	462	Odds Ratio (M‐H, Random, 95% CI)	0.49 [0.17, 1.47]

Comparison 1. ESA versus ESA or placebo/no treatment

Navigate to table in Review

Cochrane Review language

Website language

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

PICOs

PICOs

Population

Intervention

Comparison

Outcome

Plain language summary

The relative safety and effectiveness of different epoetin drugs for treating anaemia in people with chronic kidney disease

Visual summary

Authors' conclusions

Implications for practice

Implications for research

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Inclusion criteria

Exclusion criteria

Types of interventions

Types of outcome measures

Primary outcomes

Response to treatment

Safety

Secondary outcomes

Response to treatment

Safety

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Relative treatment effects

Relative treatment rankings

Assessment of clinical and methodological heterogeneity within treatment comparisons

Assessment of transitivity across treatment comparisons

Data synthesis

Methods for direct treatment comparisons

Methods for indirect and mixed comparisons

Assessment of statistical heterogeneity

Assumptions when estimating heterogeneity

Measures and tests for heterogeneity

Assessment of statistical inconsistency

Local approaches for evaluating inconsistency

Global approaches for evaluating inconsistency

Investigation of heterogeneity and inconsistency

Sensitivity analysis

Summary of findings table

Results

Description of studies

Results of the search

Included studies

Other studies

Excluded studies

Studies excluded from the meta‐analyses

Risk of bias in included studies

Allocation

Sequence generation

Allocation concealment

Blinding

Incomplete outcome data