To answer the question, “What needs to be monitored?” in anemia management, one should examine the literature in terms of “what do we know, what do we think we know, and what don't we know." In my opinion, the real issue at hand is that there is a lot we don’t know and it really affects what we should be treating and how we should be treating it. If we really knew from whence the risk of treatment with ESAs came, we would know what to monitor when we use them for treatment. We do know that anemia is an adaptive response to a chronic disease. We don't know if that adaptive response is directly related to a potential benefit from changing hemoglobin or if the benefit is indirect from a pathway that affects hemoglobin.
To tease this apart, we have many types of studies to review, including observational and randomized trials. Observational studies have uniformly provided the same conclusions regardless of population examined: higher hemoglobin is associated with better outcomes. Randomized trials, however, both contradict and agree with these conclusions. The intention-to-treat (ITT) analyses demonstrate that targeting higher hemoglobin targets cause worse cardiovascular outcomes. But the as-treated analyses still demonstrate better outcomes among those who achieve the higher hemoglobin levels.
Looking for clues
Our first qualitative clue that ESA dose might somehow confer a direct risk to outcomes was presented in the original manuscript of the Normalization of Hematocrit trial. 1 While it wasn’t obvious at the time, hindsight is of course 20/20. In this trial of ESRD patients treated to hematocrit goals of 42% vs. 30%, those in the higher goal arm didn’t do better as was originally hypothesized. While the statistics have been the subject of considerable discussion, arguably, one can conclude that this group actually did worse. However, in the “as treated” analysis, the subjects that did the best (i.e., had the lowest mortality risk) were those randomized to the higher hematocrit target who actually achieved their goal (see Figure 1). Curiously, these subjects also had a lower mortality than those achieving their target in the lower arm. Further mortality risk was neither linearly or indirectly associated with achieved hematocrit within each arm as previously demonstrated in observational studies. Rather, qualitatively it would appear that a step function existed in that if the subject who achieved their goal (regardless of what that goal was) had a risk lower than the rest of the subjects in the same target arm that couldn’t achieve their goals. These observations in a way foreshadowed the secondary analyses of CHOIR and TREAT that were subsequently published. 2, 3
After CHOIR demonstrated that targeting a hemoglobin of 13.5 gm/dl increased the risk of the composite cardiovascular outcome of death, MI, stroke, and CHF hospitalization, a secondary analysis demonstrated that controlled for target and hemoglobin achieved, the risk was entirely associated with the dose of ESA utilized. 2 On the same theme, secondary analyses were performed in TREAT comparing patients with a better hemoglobin response (who required lower doses of darbepoietin) to those with a worse hemoglobin response (who required higher doses of darbepoietin) to those who had received placebo. While no prior study had a placebo arm, it is important to note that both mathematically and clinically, there was no difference in the rates of events between patients who responded well and those who received placebo. However, the group that had the greatest risk of events (greater than those who were treated and responded and greater than those who received placebo) were those treated who didn’t respond well.
While these secondary analyses support a relationship between exogenously administered erythropoietin, one additional study further supports this relationship by demonstrating an association between endogenous erythropoietin and risk. In a population of hospitalized patients with heart failure, erythropoietin levels were measured and demonstrated to be directly associated with risk of mortality. When investigators modeled and estimated the level of erythropoietin patients were expected to have for their degree of anemia, those with a higher observed-to-predicted ratio had the highest mortality risk. 4
As much as this puzzle fits together, we must continue to be cautious as we treat anemia. The question of how anemia could be part of a protective mediator in the person with a disease resulting in inflammation needs to be understood. One hypothesis to be considered is that the decreased production of erythropoietin is the body's way of protecting itself from increasingly high levels of erythropoietin. To guard public safety, everyone should agree that the labeling for any product needs to conservatively and rigorously stick to what is known (rather than what we think we know). Translated to trial design, we “know” the results of a randomized trial. Randomized trials, by definition are the only level of evidence that prove cause and effect. Observational studies, no matter the rigor and whether based on cohorts or trial data, generate knowledge that “we think we know.” Since we “know” targeting a high hemoglobin with ESAs increases the risk of cardiovascular events and mortality, we shouldn’t target a higher hemoglobin with ESAs. Since we think we know that achieving that higher hemoglobin with smaller doses of ESA produces lower risk, it is tempting to want to discount the ITT analyses.
We are, however, stuck in a place where if we don’t try to “know” that the risk is associated with ESA dose, we will never be able to explore or offer our patients the potential benefit that could come from a higher hemoglobin. Therefore, what we monitor for therapy today still needs to be related to the conclusions of the ITT analyses; however, if we continue to explore and understand the treatment of anemia in ESRD, what we monitor tomorrow may be ESA dose or perhaps eventually erythropoietin level. Until these huge gaps in our knowledge are filled with the right studies, we need to design monitoring strategies based on what we “know.” We know targeting too high a hemoglobin is bad. It however had not been shown that a transient overshoot of the goal range is bad. So hemoglobin should be used to continuously work each patient toward their target. A transient deviation should be examined and used as a clue toward underlying changes within the patient, but may not necessarily be the most appropriate marker of poor quality of care. We “think we know” that using too high a dose is the actual mediator of this badness. When we really “know” that, we will hopefully be able to take the emphasis off of target and place it on dose, allowing patients to achieve higher hemoglobin levels without risk. But until then, lowering the target is effectively protecting the subgroup of patients who would otherwise receive these high doses from our desire to push them there. As our understanding of the treatment of anemia evolves, it is comforting to know that we are putting safety first. We must now seek to evolve our knowledge to expand the treatment of efficacy carefully without jeopardizing that safety.
Figure 1: Comparison of mortality among trials based on hemoglobin goal and ESA dose
* Mortality figures extrapolated from Figure 3.
^ Composite endpoint of death, MI, stroke, and CHF.
** Placebo arm mortality rate provided in low dose, low goal column.
^^ Mortality figures extrapolated from Figure 2. Values provided are higher erythropoietin then expected followed by erythropoietin as expected, follow by erythropoietin lower than expected.
1. Besarab et al. The Effects of Normal as Compared with Low Hematocrit Values in Patients with Cardiac Disease Who Are Receiving Hemodialysis and Epoetin. N Engl J Med 1998; 339:584-590
2. Szczech et al. Secondary analysis of the CHOIR trial epoetin- dose and achieved hemoglobin outcomes. Kidney International (2008) 74, 791–798.
3. Solomon et al. Erythropoietic Response and Outcomes in Kidney Disease and Type 2 Diabetes. NEJM 2010; 363:1146-1155
4. Belonje et al. COACH: Endogenous erythropoietin and outcome in heart failure. Circulation 2010; 121: 245-51.