Hypoxia-inducible factor stabilizers: 27 228 patients studied, yet a role still undefined

ABSTRACT

With the emergence of hypoxia inducible factor–prolyl hydroxylase inhibitors (HIF-PHIs) came the hope that using these oral drugs could improve the treatment of the anemia of kidney disease. In this editorial we discuss the accumulated knowledge on these agents and the clinical context for use.

In the past, articles on the anemia of chronic kidney disease (CKD) often began with a deserved proclamation on how recombinant human erythropoietin (current terminology: erythropoiesis-stimulating agents, ESAs) had revolutionized treatment of renal anemia. Lives of patients were clearly improved, with reduced symptoms, improved quality of life and among hemodialysis patients, a greatly diminished need for blood transfusions. From 1989 to 1998 there was little question other than how high the hemoglobin (Hb) concentration should be raised (targeted) during treatment. But starting in 1998, a series of well-known studies found increased cardiovascular and thrombotic risk with treatment with ESAs to higher Hb targets (13–15 g/dL) [1–4]. Three of these studies were published between 2006–09, all in nondialysis CKD [1, 3, 4], and all with adverse risk signals. The Food and Drug Administration (FDA) responded with considerably more stringent warnings as to cardiovascular and other risks in the prescribing information (the label) for these drugs.

In isolation, this series of findings would appear to simply indicate a need to avoid high Hb targets during ESA treatment. Unfortunately, the problem was, and is, somewhat more complicated. Since more moderate Hb targets during ESA treatment had/have never been properly tested for safety, it remains, to this day, unclear whether safety risks with high Hb targets do not also extend to more intermediate Hb targets (e.g. Hb levels of 11, 12, etc.). This residual uncertainty led the FDA to claim, in Black Box Warnings, “In controlled trials, patients experienced greater risks for death, serious adverse cardiovascular reactions, and stroke when administered erythropoiesis-stimulating agents (ESAs) to target a hemoglobin level of greater than 11 g/dL.” [5] The statement is not quite correct; studies have not shown greater risk at Hb targets above 11 g/dL, but nor have studies been done to indicate that treatment to these intermediate Hb targets avoids excess risk. Along these lines, to try to understand the cause of increased risk with ESAs, post hoc analyses of major studies as well as population-based studies were performed. These studies found that higher ESA doses associated to a greater extent with adverse outcomes than did higher achieved Hb concentrations [6–9]. While only associations, these findings suggested that rather than the high Hb concentration, it could be that injection of high doses of the human hormone erythropoietin's analogues might be the reason for the observed cardiovascular safety issues with ESAs [we (S.F.) speculated on this possibility in a 2006 editorial] [10]. Perhaps treating anemia with an agent that would not cause a large surge in the serum erythropoietin concentration could provide greater cardiovascular safety. Into this environment came the hypoxia inducible factor–prolyl hydroxylase inhibitors (HIF-PHIs).

These agents effectively stimulate erythropoietin production [11], and early studies clearly demonstrated a much smaller rise in serum erythropoietin concentration than by injecting ESAs [12]. As the pivotal studies of these drugs were conceived, and as the drugs were discussed at scientific conferences, there was great excitement that the lower implied serum erythropoietin concentrations might translate to enhanced cardiovascular safety. Studies were designed in both nondialysis as well as dialysis populations and compared the drugs with ESAs as well as placebo. These studies have now been completed. In this journal, Takkavatakarn et al. report on a systematic review and meta-analysis that involved 27 338 patients treated with HIF-PHIs from 46 studies [13]. With respect to efficacy, the authors generally found a positive erythropoietic effect with HIF-PHIs, but not materially different from that with ESAs. Effects on iron parameters were somewhat different from those found with ESAs, although these findings were of unclear clinical significance. Of greatest interest, as discussed above, is whether HIF-PHIs would improve cardiovascular outcomes compared with ESAs. Takkavatakarn et al., based on their meta-analysis, state that “according to available evidence, HIF stabilizers are potentially beneficial therapy for renal anemia with a cardiovascular and mortality safety profile comparable to but not superior to current ESA.” [13] Unfortunately, the initial aspiration that ESAs and cardiovascular risk in CKD anemia treatment could be ameliorated with HIF-PHIs has not been found to be the case. This really should not be a surprising outcome of these studies, for the reason that in the studies discussed above, the classic studies where ESA safety issues were identified, the Hb targets were high: 13–15 g/dL. In contrast, in these more recent HIF-PHI compared with ESA studies, the Hb targets were much lower, only 10–12 g/dL. This is not an Hb range where ESAs have ever been found to have safety problems. With that being the case, there was no reason to expect the HIF-PHIs would have better safety than ESAs. In contrast, if the Hb targets of these studies had been higher, for example 14 g/dL, then ESAs may have been exposed for cardiovascular safety risk and it is at least possible that HIF-PHIs might have demonstrated less risk. But that is not the way the studies were designed. So, taken together, as for cardiovascular safety, while there are probably differences between the individual HIF-PHIs, it is difficult to discern any general pattern of differential safety relative to ESAs.

As of January 2023, there are several countries in which HIF-PHI drugs have received regulatory approval for marketing, yet the role of the drugs in the treatment of CKD anemia remains uncertain. As oral agents, the greatest need for HIF-PHIs is in the nondialysis CKD population. In the USA, primarily as a result of logistical issues, anemia and iron deficiency are grossly undertreated. More nondialysis patients receive blood transfusions than ESA treatment [14]. Despite high rates of iron deficiency, a basic health need, few patients are treated with intravenous iron. Remarkably, in the 2 years prior to initiating dialysis treatment, 40% of patients receive blood transfusions [14] and this may reduce the opportunity for subsequent transplantation.

As oral agents, the HIF-PHIs could be ideal for the treatment of anemia among patients with nondialysis CKD. However, this hinges to a great extent on the cardiovascular safety of these drugs. Usually, we rely on meta-analyses to synthesize data and help bring understanding to important questions such as HIF-PHI safety. While the research approach in the meta-analysis of Takkavatakarn et al. [13] was excellent in most regards, we have somewhat less confidence in the inferences that may be drawn from their cardiovascular analyses [13]. In nondialysis CKD, by far the largest contribution of patients for the meta-analysis came from three key studies, of roxadustat, vadadustat and daprodustat (together over 10 000 patients) [15–17]. Unfortunately, the cardiovascular results of these studies were highly complex and as a result may not fit well in the neatly blended approach of meta-analysis, as will be explained in the next three paragraphs.

While Takkavatakarn et al. [13] reported no difference in cardiovascular risk with HIF-PHIs, in the roxadustat study, there was actually a finding of increased risk for death among roxadustat patients compared with placebo [15]. Findings like this led to the FDA not approving roxadustat for marketing in the USA. But the study procedures and patient disposition with roxadustat vs placebo were remarkably difficult to analyze. There was an extraordinary disparity between the number of patients who dropped out of the study and discontinued treatment between the roxadustat and placebo arms (Table 1) [18]. A large number of these nondialysis patients in the study reached end-stage kidney disease. Upon starting dialysis, patients on roxadustat were much more likely to continue on study drug and accrue adverse events including death at the high rates typical of dialysis. In contrast, anemia treatment with placebo was much less likely to be successful after initiating dialysis. Therefore, patients on placebo withdrew from study and stopped accruing adverse events. This led to a major discrepancy between the rate of exposure to adverse events between the groups. The result was a remarkable degree of analytic confusion. In contrast to the findings of Takkavatakarn et al., the FDA found major adverse cardiovascular events to be significantly increased with roxadustat compared with placebo (pooled from three studies) with an on-treatment analysis {hazard ratio (HR) 1.38 [95% confidence interval (CI) 1.11, 1.70]} [18]. Presumably, this was due to the bias of roxadustat patients remaining on dialysis while placebo patients withdrew from the study. To adjust for this, the sponsor utilized an intention-to-treat analysis censored at the start of dialysis, and found no significant difference between the groups [HR 1.02 (95% CI 0.86, 1.021)]. This degree of analytic confusion led to difficult questions. Most relevant to the Takkavatakarn et al. study is whether the differential dropout and the profound effect on measured cardiovascular outcomes could be captured in a meta-analysis.

Table 1:

Disparity in patient dropout and drug exposure in the three major roxadustat nondialysis studies.

	Roxadustat (N = 2386)	Placebo (N = 1884)
Patient disposition
Intention-to-treat population, n	2391	1886
Completed treatment, n (%)	1485 (62.2)	769 (40.8)
Discontinued treatment early, n (%)	901 (37.8)	1115 (59.2)
Drug exposure
Duration of exposure (weeks)
Mean	84.6	64.3
Median	87.1	57.1
By study (patient-years)	3870.7	2323.2
001 Study	2263.1	1747.6
060 Study	1134.9	377.3
608 Study	472.7	198.3

As discussed in the text, the roxadustat key phase 3 nondialysis studies provide a unique and difficult analytic challenge. It is not uncommon in research to have higher patient dropout rates in the investigational product group than in control groups. Yet here the dropout rate was much higher in the placebo group. Part of this was due to more difficulty treating anemia with placebo. But a very important group of patients were those who started dialysis, which occurred at a high rate. Patients on roxadustat tended to stay in study after starting dialysis and accrued cardiovascular and other adverse events at the high rates typical of patients on dialysis. In contrast, placebo patients dropped out of the study at high rates after starting dialysis, and patients’ adverse events were not counted in the analysis. This created a major analytic problem. The sponsor provided a number of alternative ways to look at the data to mitigate this difficult problem. Ultimately, they were unable to convince the FDA and the drug did not receive approval.

Adapted from the US FDA Briefing Documents for the 15 July 2021 Advisory Board on Roxadustat [18].

The confusion did not end with roxadustat. A key study of vadadustat compared with darbepoetin alfa was also in nondialysis CKD [16]. In contrast to the results of Takkavatakarn et al., in this large study the rate of major adverse cardiovascular events (MACE) was increased with vadadustat compared with darbepoetin [HR 1.17 (95% CI 1.01, 1.36)]. Yet again, the results were almost unfathomably complex. The increase in MACE events with vadadustat occurred almost exclusively among patients outside of the USA, with no statistical difference among US patients. Despite extensive analysis, nothing emerged to clearly explain this discrepancy. Furthermore, in the key phase 3 study in dialysis, there was no difference in MACE between vadadustat and darbepoetin [19]. It is possible that the increase in overall MACE in the nondialysis population represented a false-positive safety finding, but of course it is also possible that the increase in MACE with vadadustat was real and just lacked a good explanation. Again, a very confusing result that does not neatly blend well into a meta-analysis. And again, the FDA did not approve vadadustat.

Finally, there is the case of daprodustat, compared in nondialysis CKD with darbepoetin alfa [17]. The results of the intention-to-treat analysis show no increase in MACE or deaths with daprodustat. But in an on-study analysis, there was a higher incidence of a first MACE event with daprodustat (14.1%) vs darbepoetin (10.5%) [HR 1.40 (95% CI 1.17, 1.68)]. This and other concerns led an FDA advisory panel on 26 October 2022 to vote 14 to 5 against approval in nondialysis CKD. Yet, the authors explained that the on-treatment analysis was greatly complicated by “different dosing intervals in the daprodustat and darbepoetin alfa groups, which resulted in different observation periods.” [17] Again, complex results that simply do not fit well into the neat blender of a meta-analysis. The studies of these three drugs dominate the meta-analysis of Takkavatakarn et al. [13] because of their great size compared with other studies. Yet, the incredible complexity of each study made it impossible for the FDA to approve any of the drugs for treatment in nondialysis CKD (daprodustat was approved for use in dialysis on 1 February 2023). These are just not the kind of studies that can be simply “dropped” into the clean, neat and misleadingly clean world of the meta-analysis.

Takkavatakarn et al. [13] noted no observed differences in death and MACE between HIF-PHIs and control groups. Indeed, if this meta-analysis was read in isolation then it would seem very surprising that the FDA failed to approve any of these agents for nondialysis CKD in the USA. But what is the truth as to the cardiovascular safety of HIF-PHIs in nondialysis CKD? Is it possible, with over 10 000 patients, that there is still no discernable answer? Are Takkavatakarn et al. correct that further studies are needed? Could further studies provide a clearer answer to the questions raised on HIF-PHI cardiovascular safety in nondialysis CKD? Would pharmaceutical companies invest the huge sums of money required to do further and perhaps larger studies? Would agencies like the National Institutes of Health conduct such studies? The answer to all of these questions is probably no. So, as of early 2023 we are somewhat stuck. The anemia of nondialysis CKD is grossly undertreated and patients are hurt by the resulting need for blood transfusions and the detrimental subsequent effect on kidney transplantation. The lack of an oral agent for nondialysis CKD anemia is a most important unmet need. HIF-PHIs fit this need, but the questions as to safety remain and will continue to be controversies with unresolved answers. As good as the Takkavatakarn et al. study is, it cannot adequately answer primary questions as to HIF-PHI cardiovascular safety, particularly in the nondialysis CKD population.

We shall don our editorialists’ hats and give an opinion on this difficult subject. We believe having an oral drug to treat nondialysis CKD anemia has tremendous potential value to improve patients’ lives, reduce transfusions, and perhaps improve transplant availability and outcomes. So, we place our fulcrum for balancing risk and benefit in a position that would allow for taking on some lower grade risk. We are of the opinion that the HIF-PHI phase 3 studies do not show a clear safety deficit compared with control groups. However, we acknowledge that there is substantial residual uncertainty. We agree with the UK, European Union and Asian countries that have initially approved these drugs. The drugs are not approved for nondialysis CKD in the USA. While we do not agree with the FDA rejection of the drugs (daprodustat is approved for dialysis CKD) , we understand that they are unlikely to change their decision. Yet, to improve the lives of patients with nondialysis CKD there is so much to be gained with an oral agent. With any marketing of HIF-PHIs, we believe that there needs to be appropriate risk management and mitigation strategies. This could take many forms, but at the core there could be a serious, large post-marketing patient registry. While lacking in precision, it would allow for a general sense that HIF-PHIs are used in practice with no major signals of potential harm.

CONFLICT OF INTEREST STATEMENT

J.H.N. and D.A.M.: none. S.F.: research and consulting for AstraZeneca, Fibrogen, Akebia and GSK, all involved in development of HIF-PHI drugs.