Background

Erythropoietin (EPO) is a circulating hormone produced by the kidney, that stimulates erythropoiesis by binding and activating the EPO receptors (EPOR) on erythroid progenitor cells [1]. Subjects with chronic kidney disease (CKD) often develop anemia because of decreased production of EPO resulting in insufficient erythropoiesis. The cloning of the EPO gene allowed treatment of anemia in CKD patients by stimulating erythropoiesis with rHuEpo or other erythropoiesis stimulating agents (ESAs) [2].

Chronic anemia can result in organ damage affecting the cardiovascular system, kidneys, and the central nervous system [3–6] thus anemia correction might improve outcomes. In addition, EPOR was reported in nonhematopoietic tissues including renal cells [1], with some preclinical data suggesting that ESAs may be reno-protective due activation of EPOR resulting in anti-apoptotic effects [7, 8]. Some data suggest ESAs are reno-protective through an EpoR:CD131 complex and that EPO derivatives lacking erythropoietic activity are still reno-protective [9]. Other data conflicts with both hypotheses [1, 10]. However, the possibility ESAs might mitigate the serious consequences of renal ischemia through direct (anti-apoptosis of renal cells) or indirect effects (increased oxygen delivery with increased Hb) resulted in clinical trials to assess the potential benefit of ESA treatment in humans with renal diseases, and analysis of the results of those trials is warranted.

Clinical interventions to see if there is a relationship between ESAs and renal outcomes included short-term prophylactic ESA treatment where there was a high likelihood of acute kidney injury (AKI), e.g., patients undergoing coronary artery bypass grafting (CABG) surgery. In another modality, ESA treatment at the time of surgery might mitigate the ischemic damage and delayed graft function (DGF) that occurs during the perioperative period following kidney transplant. DGF increases the risk of acute rejection, impaired graft function, and reduces long term patient and graft survival. In a third modality, treatment of CKD patients to correct anemia associated with renal failure presumes that ESA treatment might delay or prevent renal disease progression through direct anti-apoptotic effects on renal cells or indirect effects of anemia correction, eg improved oxygen delivery.

Most of the trials examining the effect of ESAs on renal patients were small, outcomes were not robust or they varied across studies. Therefore, results from individual trials were inconclusive, but meta-analyses of results from those clinical trials may allow more definitive conclusions. We reasoned further that meta-analysis of multiple modalities would add additional value. The three modalities above were selected for meta-analysis because they examined direct and/or indirect effects of ESAs on renal disease progression or renal function. We report here that meta-analyses show no significant beneficial effects in any of the modalities, suggesting that ESAs have little reno-protective benefits, at least with the patient populations examined and clinical designs employed.

Methods

We wished to assess the effect of ESAs on kidneys by analyzing data from human clinical trials where ESAs might mitigate effects of ischemia or disease progression. This necessitated comprehensive searches and identification and analysis of controlled trials with renal patients where ESAs were used to protect kidneys from ischemia or to slow renal disease progression. All trials that had relevant renal endpoints were selected and analyzed, and data was extracted from those that might test the hypothesis.

Search strategy

Literature searches were performed using OVIDSP (Wolters Kluwer companies) to access MEDLINE and other databases including Current contents, Embase and BIOSYS previews, using search terms for ESAs (EPO, erythropoietin, rHuEpo, rEpo, epoetin, darbepoetin) in combination with anemia terms (anemia, Hb, hemoglobin, hct, hematocrit), kidney or kidney injury (renal, kidney, transplant, CKD, chronic kidney disease, delayed graft function, DGF, acute kidney injury, and AKI), and terms describing possible beneficial outcome (protect, protection, reno-protection). Searches of the Clinicaltrials.gov and the Cochran database websites were performed using ESA terms combined with anemia, renal, kidney and transplant, to further identify potential papers of interest. A manual search of the reference lists in papers, review articles and other meta-analyses identified additional papers.

Trial selection/inclusion criteria

Papers considered for inclusion described human clinical data with ESA treatment and renal endpoints. Papers were rejected if they were not controlled trials, were case reports, described only preclinical data, or lacked the relevant renal endpoints. Papers with ESA treatment of renal patients on dialysis were omitted because renal disease progression was not applicable. The final list included controlled clinical trials that utilized ESAs in transplantation, AKI, and for anemia correction in predialysis CKD patients.

Data extraction

The data was recovered by SE and reviewed by ZE. Recovered data included the study characteristics, study location, length of study, ESA treatment, nature of the comparator arm, number of subjects in each arm, time intervals and definitions of renal endpoints. Results were grouped according to study type (patients presenting with or at risk of AKI, studies with kidney transplant patients, and CKD patients undergoing anemia correction). For trials involving AKI, data collected for meta-analysis was the number of patients with AKI and number of patients with renal recovery following AKI. Other endpoints recovered from those trials were any creatinine-based or enzymatic markers that were measures of renal function or renal injury. With kidney transplant studies the measures recovered for meta-analysis were incidence of DGF within the first week post-surgery and graft loss/survival over a 1 year period. Other data collected were any creatinine-based data, incidence of proteinuria, and enzymatic-based markers of renal injury. The meta-analysis endpoint in anemia correction trials was incidence of progression to renal replacement therapy (RRT; progression to dialysis or kidney transplant) at any time during the study. Other data recovered were, estimated glomerular filtration rate (eGFR), serum creatinine (sCr), and their rate of change over time, and incidence of proteinuria. All the trial information and secondary measures are summarized in Tables 2, 3 and 4. The data used in meta-analysis are shown in Figs. 3, 4, 5 and 6.

Data extracted to assess trial quality (bias) included randomization, concealment of allocation, masking of patients and clinicians, documentation of dropouts and withdrawals, and whether analysis was by intention-to-treat.

Statistical analysis

Data were summarized using Comprehensive Meta-Analysis Software (V2) (Biostat, Inc., Englewood, NJ, USA). A random-effects model was used because it assumes treatment effects are not identical in all studies. However, results of analyses using a fixed-effects model, which assumes that the treatment effect is the same in each study and that differences in results are due only to chance, are also provided when the I2 statistic was not equal to zero. Risk ratios (RR) and 95% confidence intervals were calculated to compare results for patients treated with ESA with the control group. Heterogeneity or inconsistency across studies was assessed using Cochrane’s Q (p-value) and the I2 statistic. The p-value for the z-test comparing treatment groups was also determined.

Results

Description of searches and study selection criteria

The titles of papers from the searches were reviewed, and abstracts examined. Papers with potential relevance to ESAs, human clinical trials and tissue protection were recovered. This process resulted in 4056 papers. The selection and rejection process for these papers is shown in Fig. 1. Papers describing non-human studies, were reviews, were not clinical trials, lacked renal endpoints, were not in English, did not include a term for anemia, Hb or an ESA in the paper, or they did not otherwise fulfill the inclusion criteria were excluded. The resulting 309 papers described clinical trials with ESA-treated subjects that fell into 3 categories, at risk or presenting with AKI, ESA-treated kidney transplant patients and patients undergoing anemia correction with ESAs. Papers describing trials on dialysis patients, trials lacking a control group, trials that did not use ESAs, or were case studies, were omitted. Choukroun 2012 [11] was an anemia correction trial on renal transplant patients and not CKD patients so it was omitted. In 3 trials, ESAs were given prior to renal transplant [12–15] and omitted because there could be no direct effect of ESA on the ischemic transplanted kidney. Duplications were identified; Oh 2012 [16] was a reanalysis of Song 2009 [17] and Revicki 1995 [18] was a follow-up of Roth 1994 [19]. The Park (2005) [20] and Olweny (2012) [21] trials were excluded from meta-analysis because they were retrospective trials without AKI endpoints. 33 papers published between 1989 and 2015 remained, and their characteristics and extracted data are summarized in Tables 2, 3 and 4. Measures of renal function (sCr, eGFR, and enzymatic) varied, (methods and times), or were not reported in many papers. Therefore, we chose not to perform meta-analyses using those markers but instead summarize available data in the tables. Meta-analyses (Forrest plots) using the selected hard endpoints, are shown in Figs. 3, 4, 5 and 6.

Fig. 1
figure 1

Flow chart of study selection

Full size image

Risk of bias assessment

Trial quality (potential bias) was evaluated utilizing Jadad [22] and Cochrane recommendations. With the exception of Kamar 2010 [23] (which was a observational trial) all the trials used in meta-analysis were RCTs. Risk of bias assessment is shown in Table 1 and Fig. 2. Most trials provided an ITT analysis with reporting of lost patients. The trials also had adequate methods to randomly distribute subjects into intervention vs control groups. Blinding of subject distribution and blinding of outcome to assessors was inadequate in most trials, particularly the anemia correction trials. However, the hard renal endpoints used in these meta-analyses are strengths. Most AKI and transplant trials were double-blinded with few dropouts, while the anemia correction trials were mostly open-label with variable numbers of dropouts. Overall, the trials had a risk of bias that was considered acceptable and thus results from meta-analysis would be informative.

Table 1 Assessment of Risk of Bias of Randomized Controlled Trials
Full size table
Fig. 2
figure 2

Risk of bias graph

Full size image

Outcomes and meta-analysis

AKI trials

Nine trials were identified [16, 20, 21, 24–29] that assessed whether ESAs might reduce the risk of AKI (Table 2). In 8 trials the subjects underwent cardiac surgery (coronary artery grafting, or valvular heart surgery involving cardiopulmonary bypass) and in 1 trial the subjects underwent partial nephrectomy. The combined number of subjects was 1020; 490 in the ESA groups and 530 in the control groups. The trial sizes ranged from 71 to 187 subjects. The number of ESA administrations were small (1 or 2) so there were little/no changes in Hb (Table 2).

Table 2 AKI studies
Full size table

The endpoint tested in the meta-analysis was the number of patients that developed AKI within 2–7 days (>50% increase serum creatinine, or >0.3 mg/dl increase, AKIN definition). Four of the trials were performed by overlapping members of the same study groups [16, 17, 27, 29]. Song (2009) and Oh (2012) analyzed the same 71 patients and patient data, but used different definitions of AKI. They increased the duration of observation to 72 instead of 48 h, and therefore had different numbers of patients that progressed to AKI. We used the determinations from Oh (2012) because it is more recent and the definition used is more complete (AKIN).

Overall 107 of 367 (29%) of the subjects developed AKI in the ESA groups, with 133 of 357 (37%) in the control groups (Fig. 3). The RR slightly favored the ESA arm, but it did not reach statistical significance using either the random effects (0.79 [0.55, 1.14]), or fixed effects models (0.85 [0.69, 1.05]). Heterogeneity was high (I2 = 60%), 3 trials showed benefit in the ESA arm, while the other 4 were neutral, or favored the control arm. This heterogeneity is further apparent when other renal endpoints were examined (Table 2). In 1 trial [20] there was no difference in renal recovery, in 4 trials there was no difference in creatinine-based markers. However, in a 5th mixed results were reported. In a 6th creatinine markers favored slightly (p = 0.054) the ESA group and in the 7th, creatinine-based markers favored the ESA group. In 3 trials there was no difference in eGFR between groups, while in another trial, eGFR was improved in the ESA arm. Overall the secondary outcome analyses using non-creatinine-based renal biomarkers did not demonstrated significant reno-protection by ESAs. In 3 trials urine or plasma NGAL or serum cystatin C) were the same in both groups; in the 4th, urinary NGAL was lower in the ESA arm, although the significance of this difference is uncertain.

Fig. 3
figure 3

ESAs and incidence of AKI in patients at risk for AKI

Full size image

Renal transplant trials

Reinstitution of blood flow in cadaveric or live donor kidneys activates a sequence of events that results in renal injury, which may result in the development of DGF. DGF can translate into a decrease in long-term graft survival. In most ESA trials in transplant patients [14, 23, 30–36], DGF was defined as a requirement for dialysis within 7 days of the transplant [37]. In trials where multiple definitions were presented, data according to this definition was used. However, in some papers the definition of DGF was not disclosed, or an alternate measure was used (Table 3). The trial sizes were small to moderate in size (29–181 subjects). Like AKI trials, the number of ESA administrations were limited with little/no change in Hb.

Table 3 Kidney transplant studies
Full size table

A meta-analysis with 450 subjects utilizing the DGF endpoint (7 trials), is shown in Fig. 4. DGF developed in 92 of 223 (41%) in the ESA arms and 106 of 227 (47%) in the control arms. The RR was neutral using random or fixed effects models (0.96 [0.83, 1.10]. Heterogeneity was low (I2 = 0%).

Fig. 4
figure 4

ESAs and DGF in patients undergoing kidney transplant

Full size image

Meta-analysis of long term graft loss over 1 year in four trials showed similar outcomes (Fig. 5). Fifteen of 221 subjects (6.8%) had graft loss in the ESA arms and 21 of 241 (8.7%) in the control arms. The RR (0.78 [0.41, 1.48]) slightly favored the ESA arm but did not reach statistical significance. Heterogeneity was low (I2 = 0%). Excluding the retrospective study [23] reduced the apparent benefit with 9/139 (6.5%) in the ESA arm and 10/142 (7.0%) having graft loss, and the RR was closer to neutral, but with a larger range (0.90 [0.37, 2.15]).

Fig. 5
figure 5

ESAs and graft loss in patients undergoing kidney transplant

Full size image

In the 7 trials, additional renal outcomes were reported that showed no differences between ESA and no-ESA groups (Table 3). These included creatinine-based endpoints (6 trials), eGFR (3 trials), proteinuria (1 trial), histological indices in graft biopsies at 6 weeks and 6 months post-transplant (1 trial), and low molecular weight urinary protein AKI biomarkers (NGAL and IL-18) (1 trial) [34].

Anemia correction trials

CKD patients are often anemic, and ESA treatment to increase and maintain Hb levels is long-term. Therefore, analysis of ESA anemia correction clinical trials is a potentially useful method to assess the effect of Hb increases, and oxygen delivery to renal tissues, on renal disease progression.

In the 19 anemia correction trials identified, CKD patients were typically divided into 2 groups; those remaining at their starting Hb (control) and those where ESAs were used to target a higher Hb. ESAs in the 19 trials [18, 38–55] were typically given 1-3 times per week to raise and maintain target Hb levels (Table 4). The achieved Hb levels in most trials were 11–13.5 g/dL, with increases of 1–2.5 g/dL above the starting level. Trial duration ranged from 2 to 48 months. Many subjects in the lower Hb groups received ESAs, but at lower doses. In some trials, there was no ESA treatment of patients in the control groups. We performed meta-analysis on all trials and a separate meta-analysis of trials where subjects in the control groups did not receive ESAs (Fig. 6).

Table 4 Anemia correction studies
Full size table
Fig. 6
figure 6

ESAs in anemic CKD patients. The 18 trials were divided into 2 groups. In 6 trials there was no ESAs administered in the control group. In 12 trials some patients in the control groups were given ESAs. The RR and range for each group (filled diamonds) and the overall RR (open diamond) are shown

Full size image

Patients that progressed to RRT included those that began dialysis or received a transplant. In one trial a patient withdrew because of sepsis and AKI [48]. This event was included in the RRT endpoint of that study. No patients progressed to dialysis in either arm of the Lim 1989 [42] trial making it unsuitable for inclusion in a meta-analysis with a RRT endpoint.

The remaining 18 anemia correction trials had a combined total of 8020 subjects; 3964 in the treatment arm (higher Hb) and 4056 in the comparator (low Hb control) arm. Trials were of varying size; 3 had over 600 subjects. The initial and achieved Hbs in the 2 groups are shown in Table 4.

Overall, 768 (19.4%) of subjects in the treatment arm and 786 (19.3%) in the control arm, progressed to RRT (Fig. 6). With meta-analysis, the RR (random effects) of progression to RRT was 1.04 [0.91, 1.18] with low heterogeneity (I2 = 25.0%). This lack of effect on disease progression is supported in 18 trials by other assessments of change in renal function, including proteinuria, or creatinine based markers where there were no significant differences reported between groups (Table 4). However, in one trial time to a doubling in serum creatinine was significantly slower in the ESA group (Kuriyama 1997) [41]. This anemia correction meta-analysis does not assess direct ESA effects per se because subjects in both arms may have received ESAs. However, Hb levels increased in the ESA treatment/high Hb arms. Thus the absence of benefit argues that anemia correction per se is not reno-protective.

In 6 of the 18 anemia correction trials, subjects in the comparator arm did not receive ESAs [18, 19, 38–43]. These trials included a total of 268 subjects. 42 of 129 in the ESA group (33%) and 60 of 139 in the control group (43%) progressed to dialysis. Meta-analysis showed a trend towards improvement in the progression to RRT in the ESA treatment group but this did not reach statistical significance; the RR according to the random effects model was 0.79 [0.6, 1.04] (Fig. 6). The result was similar using the mixed effects model. Heterogeneity was low. Measures of serum creatinine over time showed no statistical difference in 6 of the 7 trials. Thus this select analysis also does not support either direct or indirect (anemia correction) beneficial effect on renal disease progression by ESAs.

Discussion

We assessed potential beneficial effects of ESA treatment on acute or chronic renal disease. One potential benefit is that ESAs might increase renal tissue survival and therefore renal function following ischemic events due to an interaction of ESAs with receptors resident on the surface of renal cells resulting in an anti-apoptotic effect. Alternatively, there may be mitigation of the negative effects of anemia, since anemia is associated with an increased risk of renal disease progression and allograft loss over the long term [56, 57]. However, these meta-analyses showed no clear benefit of short-term ESAs in AKI and transplant trials, where there was little change in Hb levels, arguing an absence of direct benefit. There was also no significant ESA benefit in longer-term anemia correction trials, regardless of whether the comparator group received or did not receive ESAs. Thus there appeared to be little short or long-term reno-protective benefit of ESAs, via direct (via activation of EPOR or via an interaction of ESA with an EPOR:CD131 hybrid receptor [9]) or indirect (increased Hb) mechanisms.

The lack of clear benefit of ESAs on renal disease is consistent with earlier meta-analyses. A meta-analysis with patients at risk for AKI showed no benefit of ESAs on incidence of AKI [58]. Another meta-analyses of effects of ESAs on CKD patients also showed no clear benefit on progression to RRT, comparing ESA treatment to no treatment [59] or comparing high vs low Hb targets [60, 61], nor was there was an association between ESA dose and annual GFR change or progression to ESRD [62].

Overall and to date, the potential cyto-protective effects of ESAs reported in animal models have generally not translated into benefit in humans, according to other studies examining benefit with other ischemic tissues [63]. There was no significant benefit of ESAs on infarct size in a meta-analyses of patients with acute ST-segment elevation myocardial infarction [64, 65], and no effect on nonfatal heart related events in a meta-analysis of ESA-treated patients with heart failure [66]. There was also no difference in a meta-analysis of retinopathy of prematurity in infants treated with ESAs [67]. There was no benefit of either ESA or increased Hb in an ESA trial on patients with traumatic brain injury [68, 69], and there was no benefit in a phase 3 trial with ESA treatment of stroke patients [70]. Taken together, these observations suggest that ESAs may not have the broad, robust, non-hematopoietic protective abilities described by some investigators, at least not in humans.

The gap between preclinical reports of benefit of ESAs in animals, and the absence of similar robust benefit in humans, has several explanations. Dose and dose regimens may be different, or the animal studies used homogeneous animal types under controlled conditions that cannot be mimicked in the clinic. Another possibility is that a benefit may have been unobservable because of the trial designs used. In this AKI meta-analysis the subjects were primarily cardiac patients and did not have only ischemia to the kidney as in animal studies and therefore may be immune to potential reno-protective ESA benefits.

There could also be other induced mechanisms that may confound the outcome data. For example, sepsis can affect outcomes and blood pressure can increase with ESA treatment and can negatively correlate with renal outcomes [71, 72]. However, control of blood pressure did not affect progression to ESRD in a clinical trial [73].

Alternatively, the beneficial conclusions of preclinical animal studies need to be reconsidered. There are many reports in animals showing a lack of effect of ESAs [1, 74]. The reno-protective hypothesis assumes that EPOR is present, and functional, at significant level on the surface of renal cells. However reports of EPOR presence are either assumed according to responses in tissue culture and in animals, or based on western or immunohistochemistry studies with anti-EPOR antibodies now shown to be nonspecific [75]. Recently a specific antibody to EPOR was discovered and western blots on renal tissue showed few, if any, detectable EPOR raising further questions about the validity of the hypothesis [10].

These meta-analyses have limitations. Majorities of included trials were small, single center, and had modest event rates. The anemia correction trials were larger, but conclusions around direct effects were confounded by the frequent use of ESAs in the comparator arm, though trials where the comparator arm did not receive ESAs similarly showed no benefit. Within each grouping (CKD progression, AKI, transplantation) there were differences in patient selection, treatment regimen and outcome definition. Finally, the meta-analyses were based on aggregated, not individual patient level data, which precluded adjustments for confounding factors such as age and comorbidities.

Conclusions

In contrast to some preclinical studies demonstrating reno-protection by ESAs in animals, anemia correction, prophylaxis or post-injury intervention with ESAs provided no significant clinical reno-protection in humans. This suggests that ESAs may not have robust, nor reproducible direct, or indirect, benefits on renal function.