Erythropoiesis-Stimulating Agents and Mortality

• anemia
• chronic renal disease
• erythropoietin
• mortality risk
• outcomes

The introduction of recombinant human erythropoietin (rHuEpo) into clinical practice 30 years ago has dramatically improved the quality of life of countless patients with ESRD who experienced severe transfusion-dependent anemia. This treatment has largely made it possible to substantially decrease transfusion-associated complications, such as iron overload, viral infections, and alloimmunization.

Early studies showing rHuEpo-induced improvements of patient outcomes were generally underpowered. Nevertheless, the convincing initial clinical experience with rHuEpo in patients on dialysis with extremely low hemoglobin levels led to progressively more extensive use of rHuEpo in patients with ESRD and nondialysis-dependent patients with CKD with less severe anemia—despite no formal proof of benefit. Clearly, there was a need for randomized, controlled trials of sufficient sample size to demonstrate whether the treatment leads to improvement in hard patient outcomes.

A first such trial was conducted in patients on chronic hemodialysis, and its results were reported in the late 1990s¹ followed by reports from two trials of anemia correction with epoetin-α or epoetin-β in patients with CKD stage 3 or 4 in 2006^2,3 and from a trial of darbepoetin-α (darbepoetin) treatment in patients with type 2 diabetes and CKD stage 3 or 4 reported in 2009.⁴ Unexpectedly, the findings showed that full (as opposed to partial) anemia correction was associated with no reduction—or even an increase—in the risk of thromboembolic events, cardiovascular events, and death.^2,3 Even more disturbingly, the placebo-controlled study of darbepoetin versus placebo to treat patients with CKD, diabetes, and anemia also failed to show benefit in terms of hard patient outcomes.⁴ The improvement in quality of life that was clinically evident after the correction of profound anemia (raising hemoglobin from as low as 6 g/dl) was either less evident^2,4 or entirely absent³ when modestly increasing hemoglobin in patients with much higher baseline levels.

The second generation erythropoietin analogs darbepoetin (which received marketing authorization by the Food and Drug Administration and the European Medicines Agency in 2001) and continuous erythropoiesis receptor activator (CERA; approved in 2007) have a longer elimination t_1/2 compared with the parent rHuEpo products. Because these longer-acting agents are structurally different from epoetin-α and epoetin-β, the term “erythropoiesis-stimulating agents” (ESAs) was created, which also includes the parent epoetins. The main purpose of developing the new ESAs was to facilitate patient handling and acceptance of self-administered hematopoietic agents, with dosing intervals of up to once every 2 weeks for darbepoetin and up to once every month for CERA.^5,6

Today, it is well established that actively raising hemoglobin levels into normal range using high ESA doses is harmful before or after initiating dialysis therapy in the majority of patients with CKD.⁷ It is also widely believed on the basis of limited evidence that the different ESA types do not differ in benefits and harms. A Cochrane meta-analysis of 56 randomized, controlled trials in 2014 found that safety and efficacy data provided no evidence suggesting the superiority of any of the different ESA formulations.⁸ Subsequent studies comparing darbepoetin^5,6 or CERA⁹ with other ESAs led the authors to conclude that there were no discernable differences in the effects of these agents on patient-centered outcomes for patients with CKD.

In this issue of JASN, Sakaguchi et al.¹⁰ tackle the issue of whether some ESA types are superior to others in terms of survival in an impressively large registry-based cohort study of 194,698 patients receiving chronic hemodialysis or hemodiafiltration therapy in Japan. Their aim was to compare mortality risk of users of short-acting ESAs with that of users of long-acting ESAs. About 36% of patients used short-acting ESAs (26.1% used epoetin-α or epoetin-β, and 10.2% used epoetin-κ), and the other 64% of patients took long-acting ESAs (50% used darbepoetin, and 13.7% used CERA). Study outcomes were 2-year all-cause and cause-specific mortality.

Using Cox proportional hazards models, the authors found that the relative risk of death was 13% higher (a significant increase) among long-acting ESA users compared with short-acting ESA users, and they observed similar results in a propensity score–matched cohort and an inverse probability of treatment weighting analysis. In an additional subgroup analysis, they compared users of four types of ESAs: epoetin-α/epoetin-β, epoetin-κ, darbepoetin-α, and CERA. They also found in unadjusted and adjusted Cox models that the relative risk of all-cause mortality was highest among darbepoetin users followed by CERA and epoetin-κ users, and it was lowest among epoetin-α/β users. These observations, confirmed by propensity score matching and inverse probability of treatment weighting methods, imply that the decision to use a given ESA type depended essentially on each facility’s preference, not on patient characteristics. Of note, in an observational, registry-based, retrospective cohort study of 508 United States hemodialysis facilities in which patients initiating dialysis therapy used either darbepoetin or epoetin-α almost exclusively, Winkelmayer et al.⁶ found a 12% higher mortality rate among patients in the darbepoetin facilities compared with those in the epoetin-α facilities, but this difference was no longer significant after adjustment for demographic characteristics or other factors.

The study by Sakaguchi et al.¹⁰ has several strengths, including a much larger sample size than those of the analyses mentioned above, robust statistical power, and consistency of findings in various analytical models. Limitations, acknowledged by the authors, are the observational nature of the study; missing data, including lack of information on iron supplementation; absence of quality control of the registry data; potential crossover of drug use; and the single ethnic origin of the patients.

In conclusion, the observation by Sakaguchi et al.¹⁰ of a higher relative risk of mortality associated with the use of long-acting versus short-acting ESAs in patients with CKD is intriguing. It contradicts most of existing literature, but this finding is on the basis of a sample size three- to tenfold larger than that of previous studies of this issue. If long-acting ESAs do indeed cause greater harm, the underlying mechanism for this effect remains obscure. Although the statistically significant difference between long- and short-acting ESAs in these patients in relative mortality risk seems to be robust, the absence of information about increase in absolute risk renders the clinical relevance of this observation uncertain.

For now, the observation by Sakaguchi et al.¹⁰ is more of a challenge than final proof. It needs to be confirmed—or invalidated—in CKD populations in other geographic regions, ideally in prospective studies with large sample sizes.

• Copyright © 2019 by the American Society of Nephrology