Hypoxia-Inducible Factor–Prolyl Hydroxyl Domain Inhibitors: From Theoretical Superiority to Clinical Noninferiority Compared with Current ESAs?

Abstract

Anemia is a common complication of chronic kidney disease; it is mainly treated with erythropoiesis-stimulating agents (ESAs) and iron. Experimental studies extensively investigated the mechanisms involved in the body’s response to hypoxia and led to the discovery of the hypoxia-inducible factor (HIF) pathway and the enzymes regulating its function. HIF-prolyl-hydroxyl domain (PHD) inhibitors are a new class of oral drugs developed to treat anemia in chronic kidney disease. By inhibiting the function of PHD enzymes, they mimic the exposure to moderate hypoxia and stimulate the production of endogenous erythropoietin and very likely increase iron availability. Some data also suggest that their efficacy and, consequently, dose needs are less influenced by inflammation than ESAs. Overall, data from phases 2 and 3 clinical development showed efficacy in anemia correction and maintenance for all of the class molecules compared with placebo (superiority) or erythropoiesis-stimulating agents (noninferiority). Three molecules, roxadustat, vadadustat, and daprodustat, underwent extensive clinical investigation to assess their safety on hard cardiovascular end points, mortality, and special interest events (including cancer and thrombosis). Aside from vadadustat in the nondialysis population, at the prespecified primary analyses, all three molecules met the noninferiority margin for the risk of major cardiovascular events compared with erythropoiesis-stimulating agents or placebo. The reason for this discrepancy is difficult to explain. Other safety signals came from secondary analyses of some of the other randomized clinical trials, including a higher incidence of thrombosis. A more extensive clinical experience with post-marketing data on hard safety issues is needed to define better when and how to use HIF-PHD inhibitors compared with already available ESAs.

Anemia is common in CKD, with nearly 4.8 million people having this condition in the United States.¹ Its leading cause is inadequate production of endogenous erythropoietin (EPO) relative to actual hemoglobin (Hb) levels, mainly because of progressive renal damage and altered oxygen sensing in the diseased kidney due to decreased tubular oxygen consumption. During the injury process, interstitial EPO-producing cells (REP) transform into myofibroblast-like cells, producing fibrogenic molecules and inflammatory cytokines rather than EPO.² The inadequate EPO production is often associated with poor iron availability due to absolute or functional iron deficiency. Chronic inflammation is the third most crucial player contributing to anemia in CKD, leading to functional iron deficiency.

Erythropoiesis-stimulating agents (ESAs) are considered a significant achievement in medical care. However, despite efficacy in a substantial proportion of patients and generally good tolerability, ESA use is surrounded by safety concerns, including increased cardiovascular and thrombotic risk and possibly cancer recurrence.^3,4 These drawbacks are more likely to occur when ESAs are given at high doses in hyporesponsive patients or when aiming at near-to-normal Hb levels.^5–7 Furthermore, despite decades of clinical experience, a clear demonstration of improved patient outcomes, including decreased cardiovascular events and mortality, is missing when treating intermediate degrees of anemia. In addition, ESAs need a strict cold chain for transportation and storage to guarantee molecular stability and avoid immunogenicity. Furthermore, in the United States, some patients may face logistical challenges with ESA use; this could be an incentive for having an oral agent available. Finally, ESAs are expensive; for this reason, they are not still fully available and affordable in emerging countries.

The second tool of anemia management is iron. Even if it is less expensive in comparison to ESAs, both administration routes have critical issues, including poor absorption and tolerability for the oral one and the possible exposure to iron overload and toxicity when intravenous (iv) iron is given at high doses in patients with functional iron deficiency or to save on ESA doses. Infrequently, severe anaphylactic reactions can also occur, even if possibly more rarely with newer formulations.

All of these issues have been the push for the search for new agents.

Mechanism of Action and Potential Advantages of Hypoxia-Inducible Factor–Prolyl-Hydroxylase Inhibitors

Hypoxia-inducible factors (HIF) are transcription factors that play a major role in the body’s response to hypoxia.^8,9 They are heterodimers composed of an O₂-sensitive α subunit and a constitutively expressed β subunit; three different HIFα isoforms exist.¹⁰ When activated, they regulate hundreds of genes implicated in the process of erythropoiesis, angiogenesis, lipid and glucose metabolism, glycolysis, mitochondrial function, cell growth and survival, vasodilation, and cell migration, among the others. HIFs regulate gene expression not only by direct gene transcription but also through epigenetic mechanisms, especially in the case of suppression of gene expression.¹¹

HIFs are tightly regulated by the prolyl-hydroxyl domain (PHD), which require oxygen, iron, and 2-oxyglutarate (2-OG) for their catalytic activity. If the O₂ content is normal, PHD hydroxylates the HIFα subunit, which in turn is rapidly degraded.¹² Under hypoxic conditions, PHD activity decreases; HIFα becomes more available and dimerizes with the HIFβ subunit, translocates into the nucleus, and masters the response to hypoxia.

Three PHD enzymes exist: PHD1, PHD2 and PHD3.¹³ All of the isoforms are capable of hydroxylating the HIFα subunit¹⁴ but differ for their expression inside the cell,¹⁵ among tissues,¹⁶ and in the sensitivity to hypoxia.¹⁷ These differences translate into distinct physiologic functions.^18–21

HIF-prolyl-hydroxylase inhibitors (HIF-PHIs) mimic the body’s exposure to moderate hypoxia because they cause reversible inhibition of the enzymatic activity of PHD, prolonging the half-life of HIFα (Figure 1). All of the molecules of the class act similarly but have distinct molecular structures and vary in their potency and kinetics of inhibition of the different PHD isoforms.²² The agents with a long half-life possibly have a greater cell or in vivo potency²³ but also more unwanted off-target effects. At this time, there is no clear evidence that these molecular differences translate into different efficacy or safety.

Figure 1.

Mechanisms of action of HIF-PHI. PHD enzymes reduce their activity because of hypoxia, iron deficiency, or during treatment with HIF-PHI. Consequently, HIFα half-life increases and becomes available for entering the nucleus and heterodimerizing with the HIF1β subunit. The obtained transcription factor activates or suppresses the activity of several genes that are summarized in the bottom part of the figure. In particular, HIF activation increases iron availability and stimulates erythropoietin production, leading to increased erythropoiesis. It is also possible that the action of HIF activation on inflammation could reduce hepcidin levels and increase the erythropoietic response.

HIF-PHIs stimulate the synthesis of endogenous EPO in peritubular interstitial fibroblasts in CKD kidneys.²⁴ To what degree damaged peritubular interstitial fibroblasts can be stimulated by HIF-PHIs to produce EPO has been debated. EPO-producing cells (REP) originate from the neural crest origin that are localized in the interstitium of the renal cortex.²⁵ After injury, many of them transdifferentiate into myofibroblasts, contributing to fibrosis and losing the capability to produce EPO. EPO production can be restored after several stimuli, such as neuroprotective agents and tamoxifen.²⁵ In particular, increased HIF signaling was shown to reactivate EPO synthesis without worsening fibrosis or inflammation.²⁵ HIF-PHIs also seem to increase the pool of mesenchymal stem-like cells.²⁶

Conversely, other experimental studies in models of tissue remodeling showed that REP cells do not transdifferentiate but remain dormant; treatment with a HIF-PHI re-induced EPO expression, even in the case of kidney damage.²⁷ Recently, Kobayashi et al.²⁸ showed that in a murine model of adenosine nephropathy, treatment with molidustat induced EPO synthesis only outside fibrotic areas and only in the presence of residual renal function. They also confirmed that the process does not involve myofibroblast transdifferentiation.

HIF-PHIs also correct anemia in anephric patients²⁹ because of the stimulation of EPO production in the liver. Liver EPO production may become more relevant when there is severe scarring in the kidney, making renal EPO production less “responsive” to HIF-PHI therapy. This could translate into the need for higher doses. Of note, a new HIF-PH pan-inhibitor (TP0463518) specifically induces EPO production in the liver³⁰; the agent has not entered yet clinical development. Conversely, EPO production in the liver is of limited importance in a murine model of CKD treated with molidustat.

Two HIF inhibitors (HIF-2 specifically) are currently in advanced clinical development (PT2385) or have been recently approved for clinical use (belzutifan) for the treatment of renal cancer. In contrast to HIF-PHI, the patients treated with these agents develop anemia in a significant percentage,^31,32 indirectly confirming the crucial role of HIF-2 on erythropoiesis.

Theoretically, the stimulation of endogenous EPO at physiologic or slightly higher levels could be a potential advantage of the class with respect to traditional ESAs because exposure to high ESA doses has been associated with poor outcomes, including cardiovascular events, in patients with CKD.^33–35

In addition to erythropoiesis, the HIF system regulates iron metabolism. These steps include the control of intestinal uptake,³⁶ release from deposits through the expression of ferroportin,³⁷ and hepcidin production.³⁸ Of note, HIF-2 has a specific effect on ferroportin stabilization in the duodenum,³⁹ and on the expression of the divalent metal transporter 1 (DMT1) and ferric reductase DcytB. The overall effect is increased iron availability for erythropoiesis and reduced hepcidin levels. This is important because patients with CKD often have high hepcidin levels because of reduced renal clearance and coexistent inflammation. After HIF-PHIs, hepcidin decrease is likely an indirect consequence of increased erythropoiesis.^11,40–42 However, a direct effect has also been described.^11,43

According to these physiologic considerations, HIF-PHIs also have the potential advantage of improving iron availability in patients with functional iron deficiency and, more in general, reducing the need for iron therapy. The HIF system is also involved in regulating inflammation and the immune system in a context-dependent manner.^44,45 This could also theoretically support the efficacy of HIF-PHIs in inflamed patients and contribute to direct hepcidin decrease in addition to the indirect decrease due to EPO increase.

Efficacy on Anemia Correction

Several HIF-PHIs have undergone clinical development; some are already available for clinical use in some countries.

Overall, available phases 2 and 3 randomized controlled trials (RCTs) consistently showed the efficacy of the class in correcting and maintaining Hb levels compared with placebo or ESAs. Efficacy was confirmed across different stages of CKD (nondialysis or dialysis-dependent patients; Figure 2, A and B).

Figure 2.

Mean Hb changes from baseline in nondialysis and dialysis patients. (A) Nondialysis patients. Mean Hb changes from baseline in the main phase 3 randomized clinical trials testing HIF-PHI compared with placebo or darbepoetin alfa. (B) Dialysis patients. Mean Hb changes from baseline in the main phase 3 randomized clinical trials testing HIF-PHI compared with ESA.

The single molecules underwent clinical development through various size, duration, and geographical distribution RCTs. Three agents (roxadustat, vadadustat, daprodustat) were tested in large RCTs performed globally on thousands of people. Conversely, other drugs, such as molidustat, enarodustat, and desidustat, were tested in smaller phase 3 RCTs with a shorter follow-up (no more than 1 year) and mainly located in Asia.⁴⁶ Considering that the molecules have different molecular structures and have not been compared in head-to-head RCTs, no information is available on the relative efficacy of one agent over the other.

Nondialysis Patients with CKD

Roxadustat was mainly compared with placebo. According to a pooled analysis of three RCTs on 4277 patients,⁴⁷ roxadustat given at a starting dose of ≥70 mg three time a week on average increased Hb of 1.9 g/dl from baseline over 52 weeks, demonstrating superiority compared with placebo. Roxadustat also significantly reduced the need for blood transfusions compared with placebo (6.1 versus 20.4 per 100 patient-exposure years, respectively) and the need for rescue therapy (9% in the roxadustat versus 31% in the placebo group).⁴⁷ Of note, in the first 12 weeks of treatment, the roxadustat-treated group experienced a relatively sharp increase in Hb levels (nearly 2.5 g/dl from baseline), possibly due to the choice of a relatively high starting dose.

The efficacy of vadadustat and daprodustat was studied exclusively compared with ESAs (darbepoetin alfa). In the Efficacy and Safety Study to Evaluate Vadadustat for the Correction of Anemia in Subjects with Non-dialysis-dependent Chronic Kidney Disease (PRO₂TECT), vadadustat was shown noninferior to darbepoetin alfa on Hb levels in a mixed population of ESA-naïve and ESA-treated patients (mean Hb changes from baseline to weeks 24–36 of 1.43±0.05 g/dl for vadadustat and 1.38±0.05 g/dl for darbepoetin alfa).⁴⁸ The patients in the vadadustat group had a slightly higher increase in Hb levels than darbepoetin alfa in the first 12 weeks; the trend was more evident in the conversion group. The percentage of patients who received red-cell transfusions was similar in the two treatment groups.⁴⁸

Daprodustat was tested globally in a mixed population of 3872 ESA-naïve and ESA-treated patients in the Anemia Studies in Chronic Kidney Disease: Erythropoiesis via a Novel Prolyl Hydroxylase Inhibitor Daprodustat-Non-Dialysis (ASCEND-ND) study.⁴⁹ The comparator was darbepoetin alfa. The mean±SEM change in Hb level from baseline to weeks 28–52 was similar in the two groups (0.74±0.02 g/dl for daprodustat and 0.66±0.02 g/dl for darbepoetin alfa), meeting noninferiority. A sharp increase of Hb levels >2 g/dl was observed in ≤2% of the patients in both groups.

Dialysis Population

Roxadustat was mainly compared with other ESAs (epoetin alfa in the majority of the RCTs), showing noninferiority⁵⁰; efficacy was well maintained across incident and prevalent dialysis patients.⁵⁰ Most studies used a starting dose of 70 mg three times a week for a body weight ≤70 kg or 100 mg for weights >70 kg. Of note, data obtained in Japan and China suggest that lower starting doses of roxadustat could be needed in hemodialysis.⁵¹ However, these populations are known for having lower dose needs in general.

Vadadustat was compared with darbepoetin alfa in two different trials enrolling incident and prevalent dialysis patients for a total of nearly 4000 patients⁵²; noninferiority was shown in maintaining Hb levels in the incident and prevalent patients. Smaller Japanese phase 3 RCTs showed similar efficacy data.⁵³

The Anemia Studies in Chronic Kidney Disease: Erythropoiesis via a Novel Prolyl Hydroxylase Inhibitor Daprodustat–Dialysis (ASCEND-D) compared daprodustat to epoetin alfa (hemodialysis) or darbepoetin alfa (peritoneal dialysis) in 2964 prevalent patients.⁵⁴ Hb levels were stable in both groups over a median follow-up period of 2.5 years. During the first 4 weeks, a slightly higher percentage of patients receiving daprodustat had Hb increases >2 g/dl compared with ESAs (4.1% and 1.6%, respectively).

Patients receiving peritoneal dialysis represent a small percentage of the enrolled patients in the large RCTs. Akizawa et al.⁵⁵ tested the efficacy and safety of roxadustat in a small, noncomparative RCT of 56 patients randomized to roxadustat 50 or 70 mg three times a week (ESA naive) or 70 or 100 mg three times weekly (previous ESA users). Overall, the trial showed an adequate Hb maintenance rate inside a Hb target of 10–12 g/dl of 92.3% and 74% in ESA-naive and ESA-converted patients. The small sample size of the ESA-naive subgroup (n=13) does not allow a reliable analysis of the efficacy of the single doses. Two other small Japanese RCTs suggested the possibility that lower roxadustat doses could be sufficient in peritoneal dialysis patients.^56,57 However, this specific population could have had per se lower dose needs (type of dialysis, small body size).

Vadadustat and daprodustat were tested at standard doses in two small Japanese RCTs dedicated to patients on peritoneal dialysis; the two molecules effectively maintained Hb levels inside the Hb target range of 11–13 g/dl.^58,59

Kidney Transplant Recipients

Patients with a kidney transplant have been excluded from phases 2 and 3 RCTs. This is not surprising for several reasons. First, this is a delicate patient population for potential adverse events. Second, HIF-PHIs could influence the immune system,⁶⁰ and little experimental research has been conducted to investigate whether these agents could either enhance or reduce the tolerability of the graft. Third, pharmaceutical interactions with immunosuppressive agents need careful evaluation. According to two case series, roxadustat effectively increased Hb levels in patients with a kidney transplant.^61,62 No graft rejections were reported.

Patients with Inflammation and Hyporesponse to Treatment

A hyporesponse to ESAs is a clinical challenge for several reasons. High ESA doses have been associated with poor outcomes.^5–7 Even if the relationship is biased because hyporesponse is mediated by increased inflammation, comorbidity burden, and more severe anemia, a high ESA dose per se can cause harm by activating the EPO receptors outside erythropoiesis. Among unwanted effects, thrombogenic activation of endothelial cells and platelets could occur, possibly explaining the increased risk of thrombotic and cardiovascular events in this patient population after ESA therapy. The fact that HIF-PHIs do not expose to high EPO peaks even in the high dose range could then represent a potential advantage.^63,64 Inadequate anemia correction also contributes to poor quality of life and heart performance. Finally, the need for higher ESA doses increases treatment costs.

Some phase 2 RCTs suggested that HIF-PHI dose needs were unrelated to C-reactive protein (CRP) levels.⁶⁴ Similar findings were confirmed by some phase 3 RCTs. Chen et al.⁶⁵ reported that the mean Hb levels and doses were similar across baseline CRP values in dialysis patients treated with roxadustat; in those with high CRP levels, roxadustat obtained higher Hb changes from baseline compared with epoetin alfa (0.9±1.0 g/dl versus 0.3±1.1 g/dl) despite a dose decrease for roxadustat and a dose increase for epoetin alfa. Similarly, in the HIMALAYAS trial,⁶⁶ dose needs for roxadustat were unrelated to CRP levels. Conversely, epoetin alfa-treated patients with high baseline hs-CRP required higher doses to maintain comparable Hb levels.⁶⁶ In the Roxadustat in the Treatment of Anemia in Chronic Kidney Disease Patients Not Requiring Dialysis (ALPS) trial,⁶⁷ roxadustat was superior to placebo independently from CRP levels, with a slightly lower percentage of patients achieving Hb response in those with CRP above the upper limit of normality. Similar observations were made in other RCTs of roxadustat and vadadustat using ESAs as a comparator in the nondialysis^48,68,69 and dialysis populations.^52,66,70 However, the percentage of patients truly hyporesponsive to ESAs or with a high degree of inflammation was low. Zhou et al.⁷¹ reported 32 hyporesponsive dialysis patients who were shifted from ESAs to roxadustat; nearly half of the patients achieved Hb values in the target range. As one might expect, they differentiated from the nonresponders for having a lower degree of systemic inflammation.

A pilot study specifically tested the efficacy of daprodustat in 15 hyporesponsive patients.⁷² Unfortunately, efficacy was below expectations partly because of insufficient compliance. More recently, in the ASCEND-ND trial,⁴⁹ daprodustat was shown more effective than darbepoetin alfa only in high-intermediate quintiles of baseline hs-CRP but not in the lower or the highest ones. In the Anemia Studies in Chronic Kidney Disease: Erythropoiesis via a Novel Prolyl Hydroxylase Inhibitor Daprodustat-Dialysis (ASCEND-D) trial,⁵⁴ 12.2% of the enrolled patients fulfilled the definition of ESA hyporesponsiveness (previous therapy with epoetin alfa doses ≥450 IU/kg per week). According to prespecified subgroup analyses, daprodustat was shown to be more effective than ESAs, either in the hyporesponsive patients or those receiving a previous ESA dose ≥7000 IU/week.

Overall, the possibility that HIF-PHIs could be more effective than ESAs in hyporesponsive patients is intriguing. However, well-designed studies testing this hypothesis are needed. It is still unknown whether single molecules could act differently; available information is rather influenced by the chosen starting doses.

HIF-PH Inhibition and Iron Metabolism

HIF stabilization increases iron availability; HIF-PHIs could also be effective in patients with iron deficiency or improve iron utilization from stores.

As summarized by a meta-analysis,⁷³ in general, HIF-PHIs decrease hepcidin, ferritin, increase total iron-binding capacity and decrease transferrin saturation in some instances, with no effect on serum iron. However, increased erythropoiesis could have partially influenced these findings. Indeed, the effect seems less evident for molecules that used a relatively less potent starting dose.⁵²

Clinical data on treatment efficacy according to iron stores or on the needs for iron therapy are less clear. Early in 2016, Besarab et al.⁷⁴ published a small, phase 2 study of 60 incident dialysis patients treated with roxadustat and randomized to no iron, oral iron, or iv iron and showed the efficacy of roxadustat regardless of iron status and iron treatment. However, Hb increases were smaller in patients receiving no iron, and the follow-up was relatively short (12 weeks).

Iron-deficient patients (serum ferritin ≤100 ng/ml and/or transferrin saturation ≤20%) were studied only in RCTs with roxadustat; in nondialysis, roxadustat also effectively corrected and maintained Hb levels over 52 weeks in those who were iron deficient at study entry.⁶⁸ However, no information was given on whether higher roxadustat doses were needed compared with those who were noniron deficient. No data have been published for dialysis patients yet.

All of the other molecules were tested in RCTs that enrolled patients with sufficient iron stores (even if serum ferritin of <200 ng/ml could be considered iron deficiency for hemodialysis patients).

Some RCTs showed lower iv iron use with roxadustat than placebo^67,68,75 or darbepoetin alfa⁶⁹ in nondialysis patients. This is of interest, considering that patients in the placebo group were more anemic and thus more likely to be prescribed iron therapy. In the dialysis population, the majority of RCTs with roxadustat reported less iron use compared with ESAs^50,66,70,76; this is less evident for the other molecules of the class.^52,54 The higher Hb increase obtained with roxadustat than the other molecules should also be considered.

Altogether, available information seems to support lower iron requirements during treatment with HIF-PHIs (at least for roxadustat). However, the fact that iron therapy was not standardized by study protocols complicates data interpretation.

We should ask ourselves to what extent less iron need is a true benefit. This is likely true for the nondialysis population or patients on peritoneal dialysis or home hemodialysis. Oral iron is often little tolerated in this subset, and iv iron involves several practical issues, including possible damage of peripheral veins in the perspective of the creation of an arteriovenous fistula. It follows that compliance to oral iron is small, and iv iron is underprescribed. In patients on dialysis, this is less true because iv iron can be easily administered as needed at each dialysis session. The real need there is to improve iron availability from existing deposits for those having high hepcidin and ferritin levels because this subset of patients is possibly exposed to tissue damage from iron overload⁷⁷ and oxidative stress from the administration of iv iron.⁷⁸

Major Adverse Cardiovascular Events and Thrombotic Risk

Great expectation was put on HIF-PHIs for the possibility of improving outcomes in terms of survival and reduced risk for major adverse cardiovascular events (MACE) or thrombosis. This anticipation was driven by the different mechanism of action of the class compared with ESAs, encompassing several aspects. As detailed above, HIF-PHIs expose patients to lower EPO concentrations and could also correct anemia in hyporesponsive patients while using relatively low doses. HIF-PHIs improve or have a neutral effect on some traditional cardiovascular risk factors. In some experimental models, HIF-PHIs reduce BP.^79,80 Data from RCTs are not consistent in this respect. HIF-PHIs also reduce serum cholesterol because they inhibit cholesterol synthesis⁸¹ and enhance intestinal excretion.⁸² Interestingly, fatty acids are physiologic modulators of HIF in parallel to oxygen.⁸³ In RCTs, roxadustat consistently decreased total and LDL cholesterol.^50,66,70,75 However, HDL cholesterol was decreased as well, even if to a lower extent. The effect of other molecules is lower, possibly because of different levels of inhibition on the three PHD isoforms.

No data exist on whether HIF-PHIs could influence lipoprotein(a), which is highly atherogenic. This is important because assays for LDL cholesterol measure cholesterol content in both LDL and lipoprotein(a) particles.⁸⁴ In addition, lipoprotein(a) becomes elevated in patients with CKD because of reduced renal clearance. HIF-PHIs could also reduce glucose levels and improve insulin sensitivity.⁸⁵

In addition to metabolic aspects, the HIF system is involved in the adaptation to acute and chronic ischemia, atherogenesis, and vascular calcification. However, data from experimental studies are conflicting because they are strongly influenced by the experimental conditions and by the degree of hypoxia of the affected tissues, possibly activating the three HIF and PHD isoforms differently.^86–89

Finally, the HIF system could also affect platelet count and their state of activation with opposing effects. Recently, an experimental study did not show any effect on platelet production, activation, and thrombus formation in five out of six nephrectomized rats.⁹⁰ Similar data were obtained in patients with CKD versus healthy volunteers.⁹⁰ As it is known for ESAs,⁹¹ changes in platelet count are possibly influenced by the degree of hematopoiesis and iron status.⁹²

After regulatory requirements for drug approval, MACE risk has been studied globally by large, phase 3 RCTs of roxadustat, vadadustat, and daprodustat, enrolling thousands of patients. These RCTs tested MACE inside the Hb targets recommended by Kidney Disease Improving Global Outcomes guidelines⁹³ and the European Renal Best Practice position paper.⁹⁴ However, there were subtle differences among the trial designs (Figure 3). In addition to MACE, other cardiovascular end points and events of special interest were considered for prespecified secondary analyses. These include MACE+ (MACE plus unstable angina and heart failure requiring hospitalization), thrombosis, and cancer. Data were obtained across different CKD categories (nondialysis, incident to dialysis, and maintenance dialysis) and using different comparators (placebo or ESA molecules). All of the RCTs had a median follow-up of >18 months (possibly not long enough for assessing hard end points, especially considering that many RCTs had a high rate of treatment discontinuation). The inferiority margins set by regulatory agencies were not the same for the three molecules and across world regions and changed over time for daprodustat (Figure 3).

Figure 3.

Graphical representation of the Hb target and noninferiority margins used during the main RCT testing HIF-PHI in nondialysis and dialysis patients. RCTs testing roxadustat used a Hb target of 10–12 g/dl worldwide. RCTs with vadadustat had a lower Hb target for the United States than other countries. Daprodustat trials had the same Hb target worldwide (10–11 g/dl). Noninferiority margins were 1.2–1.25 for daprodustat, 1.25 for vadadustat, and 1.3 for roxadustat. MACE, major adverse cardiovascular events.

In the nondialysis population, a pooled analysis of RCTs showed noninferiority of roxadustat compared with placebo.⁵⁰ One trial also showed noninferiority for MACE of roxadustat compared with darbepoetin alfa but had insufficient statistical power.⁶⁹ The Food and Drug Administration (FDA) performed an independent analysis on MACE risk in the roxadustat trials and underlined several methodologic issues.⁹⁵ Patients randomized to placebo were more likely to be more severely anemic and dropped out earlier from RCTs, reducing their observation time, despite a possible higher cardiovascular risk. According to on-treatment analysis, an overall higher hazard ratio (HR) for MACE was shown in patients randomized to roxadustat (1.38; 95% confidence interval [CI], 1.11 to 1.7), including a higher risk for all-cause mortality.

In the ASCEND-ND trial,⁴⁹ daprodustat was noninferior to darbepoetin alfa on MACE risk (HR=1.03; 95% CI, 0.89 to 1.19) at intention-to-treat analysis. However, on-treatment MACE analysis showed a higher MACE incidence for daprodustat than darbepoetin alfa (14.1% and 10.5%, respectively; HR=1.40; 95% CI, 1.17 to 1.68). Nevertheless, post hoc analyses considering different dosing intervals were consistent with intention-to-treat analysis. Of note, nearly 20% of the patients discontinued the drug before completing 1 year of follow-up in both groups.

Quite unexpectedly, in the PRO₂TECT study,⁴⁸ the patients randomized to vadadustat had a higher HR for MACE compared with darbepoetin alfa, exceeding the noninferiority margin (1.17; 95% CI, 1.01 to 1.36); this excess risk was largely due to nonfatal myocardial infarction and death from noncardiovascular causes. According to a prespecified secondary analysis, the increased risk for MACE was present for those living outside the United States but not for the other locations (worldwide study). The result is difficult to interpret. Two different Hb targets were used in the two subgroups (10–11 g/dl in the United States; 10–12 g/dl in other countries). However, the same Hb target was also applied to patients randomized to darbepoetin alfa. Moreover, compared with other HIF-PHI molecules, Hb changes were smoother after vadadustat therapy. It is possible that other factors were more important, such as different patient characteristics and practice patterns, as reported for regional differences in ESA response in the Trial to Reduce Cardiovascular Events with Aranesp Therapy (TREAT) study.⁹⁶ In this respect, a negative influence of inflammation and higher dose needs do not seem a likely explanation. Indeed, patients from the United States have usually a higher comorbidity burden (and consequently inflammation) than those living outside the United States. Moreover, the data from PRO₂TECT⁴⁸ are not consistent with those obtained in dialysis,⁵¹ despite a similar trial design and the higher dose needs in the dialysis population.

RCTs consistently showed noninferiority for MACE risk in dialysis patients compared with ESAs.^50,52,54 For roxadustat trials, the pooled on-treatment analysis of four RCTs found a nonsignificant HR of 1.09 (95% CI, 0.95 to 1.26), not crossing the inferiority margin of 1.3.⁵⁰ The period of observation of the analysis was “on-treatment plus 7 days”.⁵⁰ The FDA made a sensitivity analysis, also including the events occurring during the study regardless of treatment exposure, and found a nearly statistically significant trend toward increased MACE risk for roxadustat (HR=1.14; 95% CI, 1 to 1.3).⁹⁵ The difference was driven mainly by all-cause mortality and myocardial infarction; stroke was neutral.⁹⁵ Of note, mean drug exposure was shorter for roxadustat than epoetin alfa.⁹⁵ However, data from this analysis should be interpreted with caution, given the heterogeneity of the included trials.

In the INNO₂VATE trials, a MACE occurred in 420 (21.6%) of the patients randomized to vadadustat and in 449 (23%) of those in the darbepoetin alfa group, with a neutral HR (0.96, intention-to treat basis; 95% CI, 0.84 to 1.1).⁵² Similar neutral HRs were shown for secondary cardiovascular end points.⁵² The ASCEND-D trial also analyzed MACE risk on intention-to-treat basis and showed a neutral HR (0.93; 95% CI, 0.81 to 1.07).⁵⁴

Considering that ESA treatment could increase the risk for thrombotic events, this was carefully monitored during RCTs. Moreover, hypoxia itself is thrombogenic,^97,98 and the HIF system can be activated in prothrombotic conditions.⁹⁹ Nevertheless, studies performed at high altitudes do not seem to support increased thrombogenicity¹⁰⁰ or increased platelet count.¹⁰¹ A metanalysis of 30 studies (n=13,146) showed a higher risk for thrombotic events when using HIF-PHIs than placebo or ESAs.⁴⁶ Similar safety signals came from some but not all RCTs. Increased risk of vascular access thrombosis was also reported.⁴⁶ Some of the possible thrombogenic risks of HIF-PHIs could be due to a sharp increase in Hb levels at treatment start; however, thrombosis was also reported in some RCTs with smoother Hb increases. Moreover, thrombotic events were not observed only at treatment start but also during the maintenance phase of anemia therapy. In this respect, the analysis by the FDA of the roxadustat trials showed a weak association between roxadustat dose and the probability of thrombotic events in dialysis-dependent patients⁹⁵; no clear association was shown for epoetin alfa.⁹⁵ The association between Hb values at the time of the event is less clear, particularly in the dialysis population.⁹⁵

Other Safety Issues

Because of rapid and unorganized growth, ischemia and necrosis are common in neoplastic masses. Accordingly, the HIF system is upregulated in several cancers and is among the targets of anticancer therapy.¹⁰² Moreover, the HIF system upregulates vascular endothelial growth factor (VEGF)—a potent mitogenic and angiogenic factor. However, the role of HIF and the degree of PHD activity is context dependent because PHD deficiency is protective in some cases.¹⁰³ Data from toxicological studies aimed at assessing the oncogenic potential of HIF-PHI molecules are reassuring.^104,105 Moreover, no significant increases in VEGF have been shown in clinical studies with HIF-PHIs.^80,106,107 However, the exact meaning of these changes is of difficult interpretation because no clear cutoff for normal values exists for VEGF levels. Data from phase 3 RCTs are overall reassuring in this respect.⁴⁶ but long treatment exposure is needed to assess a possible oncogenic effect. A safety signal around cancer worsening came from the ASCEND-ND trial,⁴⁹ which showed a significantly higher rate of cancer-related death or tumor progression or recurrence in the daprodustat group than in the darbepoetin alfa group (HR=1.47; 95% CI, 1.03 to 2.1). However, absolute numbers were relatively low (82 versus 67) and post hoc analyses considering dosing frequencies attenuated the imbalance for cancer events.

VEGF has been implicated in the pathogenesis of several eye diseases, including diabetic retinopathy and macular degeneration. Eye examination was not performed routinely during RCTs with HIF-PHIs, except for a small trial that specifically performed blinded ophthalmologic visits at baseline and follow-up.¹⁰⁸ No increased risk was reported after daprodustat use over 48 weeks, with only a few cases occurring in both treatment groups.¹⁰⁸ Conversely, in the ASCEND-ND trial,⁴⁹ a slightly higher incidence of proliferative retinopathy, macular edema, or choroidal neovascularization was observed for daprodustat (n=54; 2.8%) compared with darbepoetin alfa (n=46; 2.4%; HR=1.22; 95% CI, 0.83 to 1.81).

Chronic hypoxia is a model of pulmonary hypertension; studies in mice demonstrated the role of the HIF system in hypoxic pulmonary hypertension.¹⁰⁹ No significant data are available about the possible development or worsening of pulmonary hypertension after HIF-PHIs because of the disease’s rarity.

According to experimental studies, HIF-mediated pathways may contribute to cyst progression in polycystic kidney disease.¹¹⁰ For this reason, several RCTs did not enroll patients with this disease or foresaw ultrasound monitoring. While waiting for further knowledge, HIF-PHIs are probably not the best therapeutic option for polycystic kidney disease.

Quite unexpectedly, a higher incidence of severe infection, sepsis/septic shock, and other bacterial infections have been reported for roxadustat in the nondialysis population compared with placebo.⁹⁵ The same was not reported for other molecules.^48,49,52,54

Experimental data testing HIF-PHIs in models of CKD progression gave conflicting results.^111–113 Overall, large RCTs with HIF-PHIs showed noninferiority of the class compared with placebo⁴⁸ or other ESAs^49,69 on the risk of reaching ESKD. However, many of the enrolled patients had CKD at later stages, reducing the potential for demonstrating any nephroprotective effect or safety signal of the class on this aspect.

Hyperkalemia can occur more frequently with roxadustat. However, this effect looks mild. Its pathophysiological background is unclear but is possibly related to metabolic acidosis.⁹⁵

Roxadustat could suppress thyrotropin secretion.^114,115 The possible explanation is that roxadustat is similar to triiodothyronine and binds to thyroid hormone receptor β at a higher affinity than triiodothyronine.¹¹⁶ This effect needs monitoring because subclinical hypothyroidism is often observed in CKD and has been associated with higher cardiovascular risk and all-cause mortality in CKD.¹¹⁷

Finally, according to the FDA review,⁹⁵ other adverse events were reported more commonly with roxadustat than placebo (as expected) or to the comparator ESA; some of them, such as esophageal and gastric erosions, could negatively affect compliance. Of note, they were reported more frequently in patients receiving daprodustat in the ASCEND-ND trial.⁴⁹ At present, a similar detailed safety profile has not been made public yet for the other HIF-PHI molecules (except on safety data reported in the appendixes of the published papers).

Conclusion and Future Perspective

All HIF-PHI molecules are already available for clinical use in some Asian countries or in Chile because local regulatory authorities have less stringent policies. In Japan, lower roxadustat starting doses (50 mg three times a week) than those used in clinical development are recommended for reducing the risk of overshooting.

Roxadustat has received the marketing authorization from the European Medicines Agency (EMA) in 2021.¹¹⁸ Among considerations on clinical safety, the EMA underlined that the cardiovascular and mortality risk was similar to ESA-treated patients on the basis of data from the correction of Hb study. However, methodological and study design issues complicate interpretation in other data pools (in comparison with placebo and stable dialysis patients). The product information contains detailed information on MACE, MACE+, and all-cause mortality from clinical trials factors and circumstances that can increase this risk. From this perspective, it is recommended not to switch dialysis patients who are stable on ESA therapy without clear clinical reasons. Conversely, the FDA declined its approval in 2021 in its current form and sought further clinical trials to analyze safety. The discrepancy is partly due to different policies between the two agencies on end points, pooling of the phase 3 clinical trials, and analyses of the results. In particularly, the EMA evaluated two more studies^78,96 than the FDA, both containing darbepoetin alfa as a comparator and not completed yet at the time of the US submission.

The FDA also declined the approval of vadadustat for safety reasons (noninferiority in MACE in nondialysis patients, increased risk of vascular access thrombosis in dialysis patients, and drug-induced liver injury).¹¹⁹ The approval decision is still pending from the EMA. Daprodustat has recently started its application to the FDA and EMA.

More extended observation is necessary for being completely reassured on cancer risks, although the accumulating data are promising overall.

On these premises, the availability of a new class of effective drugs with an entirely different mechanism of action besides ESAs increases the options for doctors and patients. The oral administration, associated with better iron availability, makes HIF-PHIs particularly interesting in nondialysis, peritoneal dialysis, and home hemodialysis patients. LDL cholesterol reduction coupled with a possible reduced inflammation and lower EPO levels could theoretically lead to a lower cardiovascular burden in the long term, at least for nondialysis patients.

RCTs have not demonstrated a lower cardiovascular burden with HIF-PHIs to date, and this is unlikely to be demonstrated in future trials because they will not have a stronger statistical power than the large phase 3 registration trials.

Oral administration, possibly coupled with less need for oral iron, could be perceived as an improvement in quality of life by many nondialysis patients. By contrast, oral administration could increase the pill burden and negatively affect treatment compliance. For this reason, oral administration is less likely to be considered an advantage in the dialysis setting. However, patient preferences are often hard to foresee.

More generally, anemia correction with HIF-PHI could improve quality of life. Unfortunately, this aspect has not really been considered by RCTs in the field. The preliminary findings of the Anemia Studies in Chronic Kidney Disease: Erythropoiesis via a Novel Prolyl Hydroxylase Inhibitor Daprodustat in Non-dialysis Subjects Evaluating Hemoglobin and Quality of Life (ASCEND-NHQ)¹²⁰ showed a statistically significant improvement of the vitality score (fatigue) in comparison with placebo in ESA-naïve nondialysis patients. Not only were achieved Hb levels markedly differed in the two groups, but also the Hb target (11–12 g/dl) was higher than that of the other ASCEND trials and of that indicated in the request for FDA approval (10–11 g/dl).

In dialysis patients, a persistent improvement in the overall status was reported for roxadustat compared with darbepoetin alfa but without a significant difference in health-related quality of life questionnaires.⁷⁶ More specific patient-related outcome measures are possibly needed to demonstrate improved quality of life and patient well-being after HIF-PHI.

Considering the unique mechanism of action of HIF-PHIs, we would have generally expected improved outcomes compared with present ESAs; in general, only noninferiority for safety was reached. However, the likely benefit in inflamed patients suggests that HIF-PHIs could be the treatment of choice for anemia in hyporesponsive inflamed patients. RCTs comparing HIF-PHIs with ESAs are still needed to evaluate safety, specifically targeting the ESA-hyporesponsive population.

Disclosures

L. Del Vecchio reports participating on advisory boards for Astellas, GSK, and Travere, and received speaker fees directly or indirectly supported by Amgen, Astellas, and Vifor-Fresenius. F. Locatelli reports being a member of advisory boards for Accelsior, Akebia, Amgen, Astellas, AstraZeneca, Baxter, GSK, Norgine, Otsuka, Roche, Travere, and Vifor- Pharma; speaker at meetings supported by unrestricted grants from Amgen, Astellas, Roche, and Vifor Pharma; consultancy for Amgen and Baxter; honoraria from Amgen, Astellas, and Baxter; and other interests or relationships with ERA, DOPPS, KDIGO, and SIN.

Funding

None.

Author Contributions

Both authors contributed equally to the conception, writing, and revision of the present article.