Acute Kidney Injury and Heart Failure Risk with Hypoxia‐Inducible Factor–Prolyl Hydroxylase Inhibitors: A Network Meta‐Analysis

Abstract

The aim of this systematic review and network meta‐analysis was to clarify the risk of acute kidney injury (AKI) and acute heart failure associated with hypoxia‐inducible factor–prolyl hydroxylase inhibitors (HIF‐PHIs) in patients with anemia of chronic kidney disease. The certainty of evidence for each outcome was assessed using the Confidence in Network Meta‐Analysis approach. The primary outcomes were AKI and acute heart failure, while the secondary outcome was any kidney injury. Thirty‐five studies, performed between 2009 and 2025, were included in the network meta‐analysis. HIF‐PHIs were associated with a higher risk of AKI compared with erythropoiesis‐stimulating agents (ESAs) (risk ratio [RR] 1.23, 95% confidence interval [CI] 1.02–1.47, moderate certainty) and placebo (RR 1.40, 95% CI 1.11–1.75, high certainty). HIF‐PHIs were also associated with an increased risk of any kidney injury compared with placebo (RR 1.13, 95% CI 1.01–1.26, moderate certainty). ESAs ranked lower in terms of acute heart failure risk than both HIF‐PHIs and placebo, although the differences were not statistically significant. Vadadustat (RR 0.30, 95% CI 0.12–0.74, moderate certainty) and ESAs (RR 0.40, 95% CI 0.17–0.94, high certainty) were associated with a lower risk of acute heart failure compared with placebo. In conclusion, HIF‐PHIs may be associated with an increased incidence of AKI compared with placebo and ESAs. In contrast, vadadustat and ESAs might be associated with a decreased incidence of acute heart failure compared with placebo. Further robust studies should be conducted to validate these findings.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Hypoxia‐inducible factor–prolyl hydroxylase inhibitors (HIF‐PHIs) are oral agents used to treat anemia in patients with chronic kidney disease (CKD). However, their comparative safety profiles for renal complications and acute heart failure remain poorly defined.

• WHAT QUESTION DID THIS STUDY ADDRESS?

This systematic review and network meta‐analysis evaluated and compared the specific risks of acute kidney injury (AKI), any kidney injury, and acute heart failure associated with different HIF‐PHIs, erythropoiesis‐stimulating agents (ESAs), and placebos.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

Analysis of 35 studies revealed that HIF‐PHIs were associated with a significantly higher risk of AKI compared with both ESAs (moderate certainty) and placebo (high certainty). Conversely, vadadustat and ESAs might be associated with a reduced incidence of acute heart failure compared with placebo.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

This study provides insight into the potentially distinct safety profiles within the HIF‐PHI class regarding renal and cardiovascular risks. These data may support clinicians in making more tailored treatment decisions and encourage cautious renal monitoring for patients with chronic kidney disease anemia.

Anemia is a common complication in patients with chronic kidney disease (CKD), affecting an estimated 15% to 23% of this population, a significantly higher prevalence than that observed in individuals without CKD. ¹ , ² It is associated with a heightened risk of cardiovascular complications, hospital admissions, and progression to end‐stage kidney disease, which ultimately increase mortality and reduce quality of life. ² Although the pathogenesis of anemia in CKD is multifactorial, relative erythropoietin deficiency and impaired iron homeostasis are the primary underlying causes. ³ Therapeutic strategies for patients with anemia in CKD typically include iron compounds, erythropoiesis‐stimulating agents (ESAs), hypoxia‐inducible factor–prolyl hydroxylase inhibitors (HIF‐PHIs), and red blood cell transfusions. ⁴ , ⁵ Oral HIF‐PHIs stimulate erythropoiesis by inhibiting prolyl hydroxylation and enhancing erythropoietin synthesis. They are recognized as superior to placebo and noninferior to erythropoietin for hemoglobin management in both non‐dialysis and dialysis‐dependent patients with CKD. ⁶

According to a meta‐analysis of randomized controlled trials (RCTs), the safety profile of HIF‐PHIs regarding cardiac or kidney‐related adverse events is comparable to that of placebos and ESAs. ⁷ Nevertheless, potential risks related to hypertension and worsening kidney outcomes persist, necessitating ongoing vigilance and more rigorous clinical investigation. ⁶ , ⁸ These adverse effects may increase the risk of acute kidney injury (AKI) and acute heart failure. However, a significant gap in knowledge remains regarding these specific events across the HIF‐PHI class. In particular, the safety profiles of individual agents have not been thoroughly characterized. Accordingly, the aim of this systematic review and network meta‐analysis was to clarify the risks of AKI and acute heart failure associated with HIF‐PHIs in order to improve drug selection and the management of adverse events in patients with anemia in CKD.

MATERIALS AND METHODS

Study design

This systematic review and network meta‐analysis of RCTs was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines. ⁹ The protocol was prospectively registered in the PROSPERO database (CRD420251111586) on July 28, 2025. The inclusion criteria were adult patients with CKD and anemia. Eligible intervention included HIF‐PHIs, ESAs, or placebo. HIF‐PHIs comprised daprodustat, enarodustat, molidustat, roxadustat, and vadadustat. We excluded studies targeting pediatric populations, non‐RCTs, and those lacking a direct comparison among the eligible interventions. There were no restrictions on medication dose, follow‐up period, publication status, or article language.

Outcomes

The primary outcomes, AKI and acute heart failure, were identified using MedDRA terms; AKI was defined by “AKI” and “acute renal failure,” while acute heart failure was defined by “acute cardiac failure” and “acute left ventricular failure.” The secondary outcome was any kidney injury. These events were defined as AKI, as well as kidney‐associated events, using the terms “renal impairment,” “end‐stage renal disease,” and “renal tubular necrosis.” To minimize outcome misclassification and under‐reporting, these diverse renal events were consolidated into a composite secondary outcome, acknowledging that acute kidney injury is often captured under various diagnostic terms in clinical practice and reporting databases.

If multiple terms were identified, their combined counts were used for the analysis. The definitions of these terms were established through a preliminary literature screening, and they were included if considered plausible, following discussion among the investigators. Outcomes were collected from the results sections of the manuscripts or from the ClinicalTrials.gov website. When data were available from both sources, the results sections of the manuscripts were prioritized.

Search strategy and data collection

All studies were identified through a comprehensive search of MEDLINE via PubMed, EMBASE via ProQuest Dialog, the Cochrane Central Register of Controlled Trials (CENTRAL), and the ClinicalTrials.gov website (Figure 1 ). The search strategy is detailed in Table S1 .

Figure 1.

PRISMA flow diagram. CENTRAL, Cochrane Central Register of Controlled Trials; MEDLINE, Medical Literature Analysis and Retrieval System Online.

First, titles and abstracts were screened based on the inclusion and exclusion criteria. Second, the full texts of potentially eligible reports were reviewed, and data were extracted from eligible studies. Third, a quality assessment, a data synthesis, and statistical analyses were performed. The study procedure was independently conducted by two researchers (S.M. and M.M.). Disagreements were resolved through internal discussion or by consulting a third author (K.K.).

The following data were extracted from eligible studies: author name(s), date of publication, country, study design, population, number of outcome events and dropouts, interventions and comparisons, and participants' sex and mean age. The authors of reports that did not document certain outcomes were contacted to obtain additional unpublished data.

Quality assessment

The Cochrane Risk of Bias 2 tool was used to assess the risk of bias in each study. ¹⁰ Based on potential biases, each study was classified as having “high risk,” “low risk,” or “some concerns.” In addition, the certainty of evidence for each outcome was assessed using the Confidence in Network Meta‐Analysis (CINeMA) approach, which is based on the GRADE approach for network meta‐analysis. ¹¹ This approach incorporates within‐study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence. Certainty was categorized as “very low,” “low,” “moderate,” or “high.” The CINeMA web application was used to evaluate the confidence in the evidence. ¹²

Network meta‐analysis

A frequentist network meta‐analysis was performed using a random‐effects model to account for any significant heterogeneity among the studies. Risk ratios (RRs) and 95% confidence intervals (CIs) for each outcome compared were calculated based on the population size. Analyses were performed for the entire class of HIF‐PHIs as well as for each individual HIF‐PHI. The potential for publication bias was assessed by visually inspecting the funnel plot and applying Egger's test. ¹³ All statistical analyses were performed using R ver. 4.5.0 (R Foundation for Statistical Computing, Vienna, Austria) with the netmeta package (ver. 3.2–0).

Subgroup analysis

First, the subgroup analysis focusing on non‐dialysis populations was restricted to studies involving only non‐dialysis patients and evaluated the risks of AKI, any kidney injury, and acute heart failure. Second, a similar analysis was performed specifically for dialysis populations (including hemodialysis and peritoneal dialysis) to assess the risk of acute heart failure. In the dialysis population, AKI and any kidney injury outcomes were not assessed because they are not clinically applicable to this group.

Sensitivity analysis

Several sensitivity analyses were conducted for the primary outcomes. First, trials were stratified into placebo‐controlled and active‐comparator studies. Second, we evaluated the impact of study length by restricting the analysis to trials with a duration of 50 weeks or longer. Third, analyses were conducted comparing outcomes derived exclusively from published manuscripts against those retrieved from ClinicalTrials.gov registry records. Finally, incidence rates based on person‐years were calculated, and network meta‐analyses were performed using incidence rate ratios (IRRs).

RESULTS

Study selection and characteristics

After a comprehensive search, 11,637 potentially eligible records were identified (Figure 1 ). Ultimately, 35 studies conducted between 2009 and 2025 were included in the analysis. The characteristics of the included studies are summarized in Table 1 , while specific terms for each outcome are detailed in Table S2 . Twenty‐one studies (60%) had an open‐label design, 3 (9%) were single‐blinded, and 11 (31%) were double‐blinded. Geographically, 16 studies (46%) were conducted in multiple countries, 10 (29%) in Asia, 6 (17%) in North America, and 3 (9%) in Europe. AKI was reported in 19 studies (54%), any kidney injury in 32 (91%), and acute heart failure in 20 (57%).

Table 1.

Studies included in the meta‐analysis

References	Comorbidities	Region	Study type	Intervention and comparison	Number of ITT patients	Mean age, years (male, %)	Follow‐up period (weeks)
Akizawa et al. 23	CKD stage 1–5 without HD	Asia	Individual (double‐blinded)	Roxadustat 50–100 mg TIW Placebo	107	63 (45)	24
Akizawa et al. 24	CKD stage 1–5 without HD	Asia	Individual (open‐label)	Roxadustat 20–300 mg TIW Darbepoetin	332	64 (71)	26
Barratt et al. 25	CKD stage 3–5 without HD	Europe	Individual (open‐label)	Roxadustat 70–100 mg TIW Darbepoetin alfa	616	67 (45)	104
Besarab et al. 26	CKD stage 3–4	North America	Individual (single‐blinded)	Roxadustat 0.7–2.0 mg/kg BIW or TIW Placebo	116	68 (48)	4
Brigandi et al. 27	CKD stage 3–5 with HD	Multinational	Individual (single‐blinded)	Daprodustat 10–100 mg Placebo	70	NA	4
Charytan et al. 28	HD	Asia	Individual (open‐label)	Roxadustat 70–400 mg/TIW Epoetin alfa	740	58 (55)	56
Chen et al. 29	HD	Asia	Individual (open‐label)	Roxadustat 20–250 mg TIW Epoetin alfa	304	50 (60)	27
Chen et al. 30	CKD stage 3–5 without HD	Asia	Individual (double‐blinded)	Roxadustat 20–250 mg TIW Placebo	152	54 (38)	8
Chertow et al. 31	CKD stage 3–5 without HD	Multinational	Individual (open‐label)	Vadadustat 300–600 mg/day Darbepoetin	3,471	66 (44)	94
Coyne et al. 32	CKD stage 3–5 without HD	Multinational	Individual (double‐blinded)	Roxadustat 70–100 mg TIW Placebo	916	65 (41)	89
Csiky et al. 33	HD and PD	Europe	Individual (open‐label)	Roxadustat 20–100 mg TIW Epoetin alfa or darbepoetin alfa	834	62 (58)	52
Eckardt et al. 34	HD and PD	Multinational	Individual (open‐label)	Vadadustat 150–600 mg/day Darbepoetin alfa	3,902	57 (60)	75
Fishbane et al. 15	CKD stage 3–5 without HD	Multinational	Individual (double‐blinded)	Roxadustat 70–300 mg TIW Placebo	2,761	62 (43)	52
Fishbane et al. 35	HD and PD	Multinational	Individual (open‐label)	Roxadustat 70–200 mg TIW Epoetin alfa	2,101	55 (60)	52
Fliser et al. 36	eGFR 30–59 mL/min/1.73 m2	Europe	Individual (single‐blinded)	Epoetin beta pegol Placebo	235	63 (63)	96
Holdstock et al. 37	CKD stage 3–5 without HD	Multinational	Individual (double‐blinded)	Daprodustat 0.5–5 mg/day Placebo	72	69 (54)	4
Holdstock et al. 38	CKD stage 3–5 without HD	Multinational	Individual (open‐label)	Daprodustat 1–4 mg/day rhEPO	250	66 (42)	28
Johansen et al. 39	CKD stage 3–5 without HD	Multinational	Individual (double‐blinded)	Daprodustat Placebo	614	67 (43)	28
Macdougall et al. 40	CKD stage 3–5 without HD	Multinational	Individual (open‐label)	Molidustat 25–150 mg/day Darbepoetin alfa	124	69 (53)	15
Martin et al. 41	CKD stage 3–4	North America	Individual (double‐blinded)	Vadadustat 240–630 mg/day Placebo	91	65 (43)	6
Meadowcroft et al. 42	HD	Multinational	Individual (double‐blinded)	Daprodustat 4–12 mg/day Placebo	216	60 (65)	28
NCT03799627 43	HD	North America	Individual (open‐label)	Vadadustat Epoetin alfa	175	NA (64)	12
Nangaku et al. 44	CKD stage 3–5 with HD	Asia	Individual (double‐blinded)	Vadadustat 150–600 mg/day Placebo	51	71 (61)	4
Nangaku et al. 45	CKD stage 3–5 without HD	Asia	Individual (open‐label)	Daprodustat 1–24 mg/day Epoetin beta pegol	299	70 (62)	52
Nangaku et al. 46	CKD stage 3–5 without HD	Asia	Individual (open‐label)	Vadadustat 150–600 mg/day Darbepoetin alfa	304	72 (49)	52
Pergola et al. 47	CKD stage 3–5 without HD	North America	Individual (double‐blinded)	Vadadustat 150–600 mg/day Placebo	210	67 (47)	20
Pfeffer et al. 48	eGFR 20–60 mL/min/1.73 m2	Multinational	Individual (double‐blinded)	Darbepoetin alfa Placebo	4,023	68 (43)	116
Provenzano et al. 49	HD and PD	Multinational	Individual (open‐label)	Roxadustat 70–100 mg TIW Epoetin alfa	1,039	54 (59)	52
Shutov et al. 50	CKD stage 3–5 without HD	Multinational	Individual (open‐label)	Roxadustat 70–100 mg TIW Placebo	594	63 (46)	104
Singh et al. 51	HD and PD	North America	Individual (open‐label)	Daprodustat 4–12 mg/day Epoetin alfa or darbepoetin alfa	2,956	59 (57)	52
Singh et al. 52	CKD stage 3–5 without HD	Multinational	Individual (open‐label)	Daprodustat 1–24 mg/day Darbepoetin alfa	3,870	68 (44)	54
Singh et al. 53	HD and PD	Multinational	Individual (open‐label)	Daprodustat 1–24 mg/day Darbepoetin alfa	312	54 (62)	52
Toka et al. 54	HD	North America	Individual (open‐label)	Vadadustat 300–1,200 mg TIW Epoetin beta	451	62 (58)	50
Yamamoto et al. 14	CKD stage 3–5 without HD	Asia	Individual (open‐label)	Molidustat 25–150 mg/day Darbepoetin	164	71 (61)	52
Yamamoto et al. 55	CKD stage 3–5 without HD	Asia	Individual (open‐label)	Molidustat 4–150 mg/day Darbepoetin	161	72 (62)	52

BIW, twice weekly; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; HD, hemodialysis; ITT, intention‐to‐treat; NA, not applicable; PD, peritoneal dialysis; rhEPO, recombinant human erythropoietin; TIW, three times weekly.

Quality assessment

The risk of bias was assessed for the primary and secondary outcomes in each study, and the same results were obtained for both outcome types from the same study. The assessments are detailed in Table S3 . Twenty‐four studies (69%) were judged to have a “high” risk of bias because they employed an open‐label or single‐blinded design. The remaining 11 studies (31%) were judged to have a “low” risk of bias. Visual inspection of the funnel plots and Egger's regression test revealed no significant publication bias across all assessed outcomes (Figures S1–S6 ). The CINeMA results are provided in Table S4 .

Network meta‐analysis

Nineteen studies were included in the network meta‐analysis of AKI (Figure 2 a ). HIF‐PHIs were associated with an increased risk of AKI compared with ESAs (RR 1.23, 95% CI: 1.02–1.47, moderate certainty) and placebo (RR 1.40, 95% CI 1.11–1.75, high certainty) (Figure 3 a ). In contrast, ESAs were not associated with an increased risk of AKI compared with placebo (RR 1.14, 95% CI 0.90–1.44, moderate certainty). In the evaluation of each individual HIF‐PHI, roxadustat, daprodustat, and vadadustat had a higher AKI risk than both ESAs and placebo. In particular, roxadustat was associated with an increased risk of AKI compared with placebo (RR 1.53, 95% CI 1.14–2.06, high certainty).

Figure 2.

Network plot of treatment comparisons. The node size indicates the total number of patients in a treatment group, and the line width reflects the number of trials directly comparing the connected groups, which are also displayed numerically next to each line. (a) Acute kidney injury (AKI), (b) any kidney injury, and (c) acute heart failure. ESA, erythropoiesis‐stimulating agents; HIF‐PHI, hypoxia‐inducible factor–prolyl hydroxyl domain inhibitor.

Figure 3.

League tables of acute kidney injury, any kidney injury, and acute heart failure. The top row presents the pooled analysis results for all HIF‐PHIs, while the bottom row displays the results for individual HIF‐PHIs. The green columns display the treatments and their respective P‐scores. The estimates in the colored cells are derived from the network meta‐analysis, while those in the white cells are derived from pairwise meta‐analyses. The numbers represent risk ratios (95% confidence intervals), and bold indicates statistical significance. (a) Acute kidney injury (AKI), (b) any kidney injury, and (c) acute heart failure. NA, not applicable. ESA, erythropoiesis‐stimulating agents; HIF‐PHI, hypoxia‐inducible factor–prolyl hydroxyl domain inhibitor.

Thirty‐two studies were included in the network meta‐analysis of any kidney injury (Figure 2 b ). HIF‐PHIs were associated with an increased risk of any kidney injury compared with placebo (RR 1.13, 95% CI 1.01–1.26, moderate certainty, Figure 3 b ). In contrast, ESAs were not associated with an increased risk of any kidney injury compared with placebo (RR 1.09, 95% CI 0.96–1.23, moderate certainty). An evaluation of each individual HIF‐PHI revealed that all HIF‐PHIs ranked higher in terms of any kidney injury than ESAs and placebo. Molidustat was associated with an increased risk of any kidney injury compared with roxadustat (RR 1.75, 95% CI 1.01–3.04, low certainty), vadadustat (RR 1.77, 95% CI 1.04–3.04, moderate certainty), ESAs (RR 1.78, 95% CI 1.05–3.04, moderate certainty), and placebo (RR 1.96, 95% CI 1.13–3.41, moderate certainty). Roxadustat was associated with an increased risk of any kidney injury compared with placebo (RR 1.12, 95% CI 1.01–1.24, high certainty).

Nineteen studies were included in the network meta‐analysis of acute heart failure (Figure 2 c ). ESAs ranked lower in terms of acute heart failure risk than both HIF‐PHIs and placebo. Specifically, all HIF‐PHIs and ESAs ranked lower for acute heart failure risk than for placebo. Vadadustat was associated with a decreased risk of acute heart failure compared with daprodustat (RR 0.32, 95% CI 0.16–0.63, moderate certainty, Figure 3 c ) and placebo (RR 0.30, 95% CI 0.12–0.74, moderate certainty). ESAs were also associated with a decreased risk of acute heart failure compared with daprodustat (RR 0.42, 95% CI 0.22–0.78, moderate certainty) and placebo (RR 0.40, 95% CI 0.17–0.94, high certainty).

Subgroup analysis

Regarding the primary analysis of AKI, four of the 19 studies included a mixed population of dialysis and non‐dialysis patients; thus, a subgroup analysis was restricted to non‐dialysis patients across the remaining 15 RCTs. The findings were consistent with the main analysis (Figures S7A and S8A ). Similarly, for the secondary analysis of any kidney injury, 10 of the 32 studies with mixed populations were excluded, and a subgroup analysis of the remaining 22 RCTs confirmed consistent results for any kidney injury (Figures S7B and S8B ). Regarding the primary analysis of acute heart failure, 9 of 19 studies included a population of dialysis patients; thus, a subgroup analysis was restricted to non‐dialysis patients across the remaining 10 RCTs. The observed trend for acute heart failure was similar to that in the primary analysis, but the association was no longer statistically significant in this subgroup (Figures S7C and S8C ).

In addition, a subgroup analysis was performed specifically for dialysis populations (including both hemodialysis and peritoneal dialysis) across nine studies to assess the risk of acute heart failure. While a similar trend was observed, the association was no longer statistically significant in this subgroup (Figures S9 and S10 ).

Sensitivity analysis

First, trials were stratified into placebo‐controlled and active‐comparator studies. HIF‐PHIs were associated with an increased risk of AKI compared with placebo (RR 1.53, 95% CI 1.15–2.04, Figure S11A ). However, no statistically significant differences in the risk of AKI were observed for the comparisons of ESAs vs. placebo or HIF‐PHIs vs. ESAs (Figure S11B,C ). In addition, no statistically significant differences in the risk of acute heart failure were found in any pairwise comparisons among HIF‐PHIs, ESAs, and placebo (Figure S12 ). Second, when restricting the analysis to trials with a duration of 50 weeks or longer, the findings were consistent with the primary analysis for both AKI and acute heart failure (Figure S13 ). Sensitivity analysis for trials with a duration of less than 50 weeks could not be performed due to the limited number of studies. Third, analyses were conducted comparing outcomes derived exclusively from published manuscripts against those retrieved from ClinicalTrials.gov registry records. In the analysis of published manuscript data, findings for both AKI and acute heart failure were consistent with the primary analysis (Figure S14 ). For ClinicalTrials.gov data, while acute heart failure results remained consistent with the primary analysis (Figure S15B ), the treatment ranking for AKI differed substantially from the main findings (Figure S15A ). However, the 95% confidence intervals were wide, suggesting that this discrepancy was likely attributable to a lack of statistical power. Finally, in terms of person‐time analysis, IRRs for AKI were significantly higher for HIF‐PHIs compared with both placebo (IRR 1.39, 95% CI 1.10–1.76, Figure S16B ) and ESAs (IRR 1.23, 95% CI 1.02–1.48, Figure S16C ). Furthermore, IRRs for acute heart failure tended to be lower for both ESAs (IRR 0.46, 95% CI 0.18–1.15, Figure S17B ) and HIF‐PHIs (IRR 0.51, 95% CI 0.22–1.21, Figure S17C ) compared with placebo. However, the incidence rate for AKI was higher in the placebo group than in the HIF‐PHI and ESA groups, whereas the incidence rate for acute heart failure was lower with placebo than with HIF‐PHIs and ESAs.

DISCUSSION

This network meta‐analysis of RCTs suggests that HIF‐PHIs may be associated with an increased risk of AKI compared with both placebo and ESAs, with high to moderate certainty, in patients with anemia in CKD. When a broader definition of kidney outcomes (any kidney injury) was included in the analysis, the results remained directionally consistent, though the certainty decreased to moderate to low. HIF‐PHIs are generally not considered to increase the risk of AKI. Given the high proportion of open‐label studies, these results should be interpreted with caution, and this association requires further investigation. Interestingly, RCTs reported that molidustat is associated with a higher incidence of CKD worsening compared with ESAs, and that roxadustat decreases the estimated glomerular filtration rate to a greater extent compared with placebo. ¹⁴ , ¹⁵ HIF‐PHIs have also been associated with a higher risk of thrombotic events, malignancy, and sepsis. ⁶ , ⁸ These findings may explain how HIF‐PHIs possibly increase the risk of AKI. In contrast, ESAs did not increase the risk of AKI compared with placebo. Similarly, a Cochrane systematic review previously concluded that ESAs do not reduce the risk of AKI compared with placebo. ¹⁶

Roxadustat was associated with an increased risk of AKI compared with placebo. Roxadustat has been reported to be associated with an increased incidence of kidney failure and a faster rate of eGFR decline compared with placebo. ¹⁷ In addition, molidustat was associated with an increased risk of any kidney injury compared with roxadustat, vadadustat, ESAs, and placebo. Some RCTs have linked molidustat to a higher incidence of worsening CKD but have not generally demonstrated an elevated risk overall. ¹⁸ Interestingly, although all HIF‐PHIs share a common core mechanism of action, individual agents differ markedly in their selectivity profiles for prolyl hydroxylase domain isoforms—such as the predominant inhibition of prolyl hydroxylase domain‐2 by molidustat compared with the pan‐prolyl hydroxylase domain inhibition exhibited by roxadustat. ¹⁹ Although the biological mechanisms underlying these divergent clinical findings remain incompletely understood, such pharmacological variations could plausibly exert distinct effects on renal tissue oxygenation and cellular stress pathways. ¹⁷ Nonetheless, robust clinical evidence directly linking these mechanistic distinctions to differential renal outcomes remains scarce, underscoring the need for targeted mechanistic and head‐to‐head clinical investigations. Furthermore, there is substantial variability in the incidence rate of renal adverse events associated with HIF‐PHIs across clinical trials (Figure S16A ). While such discrepancies may partly stem from differences in study design and patient populations, the primary drivers remain unclear. Therefore, our findings should be interpreted with caution.

Both vadadustat and ESAs were associated with a reduced incidence of acute heart failure compared with placebo. Anemia is associated with worse outcomes in patients with heart failure. Reduced tissue oxygen delivery, caused by anemia, triggers a cascade of hemodynamic and neurohormonal alterations. This ultimately elevates the myocardial workload, potentially leading to adverse left ventricular remodeling and left ventricular hypertrophy. ²⁰ Indeed, anemia is an independent risk factor for the development of heart failure among patients with CKD. ²¹ However, previous research found that ESAs did not improve clinical outcomes, such as all‐cause mortality or hospitalization for worsening heart failure, in patients with systolic heart failure and mild‐to‐moderate anemia. ²² Subgroup analyses comparing dialysis and non‐dialysis patients revealed that the statistical significance observed in the overall population was not maintained in either dialysis or non‐dialysis cohorts, with no significant differences detected between these subgroups. In the present study, only vadadustat demonstrated a statistically significant reduction in acute heart failure risk among the HIF‐PHIs. Whether this represents a drug‐specific finding or a broader class effect across HIF‐PHIs remains to be determined. Although further study is needed, our findings suggest that the correction of anemia with HIF‐PHIs or ESAs might mitigate compensatory hemodynamic overload, thereby reducing the risk of acute heart failure in patients with CKD, particularly by preventing its initial onset.

This study has several strengths. First, given that this meta‐analysis was based on RCTs, the certainty of evidence for most of the main results was rated as high to moderate. Increased risks of AKI and any kidney injury, as well as a decreased risk of acute heart failure, were generally observed across both the entire class and for individual HIF‐PHI agents. Although molidustat showed no increased risk of AKI, a significant risk was observed for any kidney injury. This discrepancy may be attributed to differences in the definitions used across studies. Second, AKI and acute heart failure were clearly defined in most RCTs based on MedDRA. Standardizing these definitions helps to mitigate potential bias associated with inconsistent terminology across studies. Third, while the results of the pairwise and network meta‐analyses were generally consistent, the network approach successfully addressed areas where pairwise data were conflicting or missing.

Nevertheless, this study has several limitations. First, because this meta‐analysis prioritized studies designed for efficacy, it may be underpowered to detect rare or specific adverse events compared with trials specifically designed for safety outcomes. While crude incidence rates and network IRRs yielded seemingly contrasting patterns for AKI and acute heart failure, the IRR estimates should be considered more authoritative. Crude incidence rates do not adjust for differences in total exposure time or potential trial‐level confounding across treatment arms, whereas the IRR incorporates person‐time directly into the network meta‐analysis framework. Second, a high risk of bias was identified in several RCTs because the distinct routes of administration—intravenous for ESAs and oral for HIF‐PHIs—prevented the implementation of a blinded design. Consequently, the certainty of evidence assessed by the GRADE approach ranged from moderate to low for most outcomes. Therefore, the findings of this study should be interpreted with caution, particularly when applying them to clinical decision‐making. Third, owing to the limited number of studies, we were able to perform subgroup analyses for individual HIF‐PHIs only. The inability to perform additional subgroup or dose–response meta‐analyses limits the overall robustness of the findings. Fourth, because “any kidney injury” was defined using various renal adverse‐event terms across the aggregated RCT data, duplicate events may have been counted for a single patient, and this variability in the definition may have also contributed to study heterogeneity. Because the dataset lacks the granularity to identify duplicate events, this represents a substantial methodological constraint. Consequently, although the findings of this study provide important insights, their robustness is constrained, and thus the safety signals detected here should be rigorously validated in further clinical studies.

CONCLUSIONS

The findings of this study suggest that HIF‐PHIs may increase the incidence of AKI compared with placebo and ESAs. In contrast, vadadustat and ESAs might be associated with a reduced incidence of acute heart failure compared with placebo. These differences in safety profiles may influence the clinical selection criteria in adult patients with anemia in CKD. Given the limitations of the current data, further robust studies should be conducted to clarify and validate these findings.

FUNDING

No funding was received for this work.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

Satoru Mitsuboshi, Makoto Morizumi, Shungo Imai, Satoko Hori, and Kazumasa Kotake wrote the manuscript; Satoru Mitsuboshi and Kazumasa Kotake designed the research; Satoru Mitsuboshi and Makoto Morizumi performed the research and analyzed the data.

Supporting information

Table S1.

Data S1.

DATA AVAILABILITY STATEMENT

All data that support the findings are included within the manuscript.