The efficacy and safety of roxadustat for anemia in patients with dialysis-dependent chronic kidney disease: a systematic review and meta-analysis

Abstract

Background

This meta-analysis aims to assess the efficacy and safety of roxadustat in treating anemia patients with dialysis-dependent (DD) chronic kidney disease (CKD).

Methods

We comprehensively searched 5 databases for randomized controlled trials (RCTs) investigating roxadustat for anemia in DD-CKD patients. RevMan 5.0 was used to extract and synthesize data for meta-analysis.

Results

Ten different RCTs (9 studies) and 5698 DD-CKD patients with anemia were included. Our findings revealed that when compared to the erythropoiesis-stimulating agents (ESAs) group, the roxadustat group showed increased hemoglobin levels [MD (Mean Difference) 0.25 g/dL (95%CI 0.14 g/dL to 0.36 g/dL), p < 0.00001] and improved iron-utilization by increasing serum iron [MD 1.85 µmol/L], total iron binding capacity [MD 35.73 µg/dL], transferrin saturation [MD 1.19%], and transferrin level [MD 0.40 g/L]. In addition, we found that roxadustat significantly decreased the low-density lipoprotein-cholesterol [MD −0.39 mmol/L] and total cholesterol [MD −0.6 mmol/L]. In patients with a C-reactive protein level that exceeds the upper limit of the normal range, hemoglobin levels were higher for roxadustat than for ESAs [MD 0.39 g/dL]. Treatment-emergent adverse events, treatment-emergent serious adverse events, and major adverse cardiovascular events were not significantly different between the two groups.

Conclusions

The hemoglobin levels of DD-CKD patients were significantly increased and not affected by the inflammatory state after roxadustat treatment. Roxadustat also improved iron utilization, and it was not associated with higher treatment-emergent adverse events, treatment-emergent serious adverse events, and major adverse cardiovascular events when compared to ESAs.

Keywords:

• Roxadustat
• anemia
• dialysis
• meta-analysis
• chronic kidney disease

Introduction

Renal anemia is a common complication of chronic kidney disease (CKD). The risk of anemia increases with the progression of kidney disease [1] with more than 90% of dialysis patients having the condition [2]. Anemia reduces the quality of life for dialysis patients and increases medical expenses, the incidence of cardiovascular events, hospitalization rates, and mortality [3–5]. There are many causes of renal anemia, including insufficient erythropoietin (EPO) production, reduced EPO activity, iron deficiency and other metabolic disorders, malnutrition, states of inflammation, extensive blood loss, and so on [6].

At present, the main renal anemia treatments include erythropoiesis-stimulating agents (ESAs), iron supplementation, and blood transfusions. ESAs are often the first-line treatment for renal anemia. They are analogs of EPO with characteristics including good tolerance and ease of use. Although ESAs can improve quality of life, many studies in patients with dialysis-dependent (DD) CKD have found that when the protocol goal is to reach higher target Hb levels, the mortality rate is higher in the high-dose ESA treatment groups [7]. Some meta-analyses also support the association between adverse events and high-dose ESA administration [8–10]. Therefore, more safe and more effective treatment is needed to treat renal anemia.

Hypoxia-inducible factor prolyl hydroxylase inhibitors (HIF-PHI) are newly developed, oral-based small molecule drugs for the treatment of renal anemia, of which roxadustat (FG-4592, FibroGen, Astellas, AstraZeneca) has been listed in China and Japan [6]. At present, some phase II and phase III clinical trials have proved that roxadustat can improve the hemoglobin (Hb) level of patients with non-dialysis dependent (NDD) CKD and DD-CKD.

Up to now, some authors have combined NDD-CKD and DD-CKD patients for analysis when performing a meta-analysis of roxadustat [11–13]. Because of the different populations included, the homogeneity of the data was poor. With the completion of some new clinical trials [14], we conducted this meta-analysis of randomized controlled trials (RCTs) in DD-CKD patients to evaluate the efficacy and safety of roxadustat. Our research will help physicians and nephrologists decide whether to consider using roxadustat to treat anemia in patients with DD-CKD.

Materials and methods

This review complies with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [15,16].

Search strategy

We comprehensively searched PubMed, Embase, Cochrane Library, ClinicalTrials.gov Database, and China National Knowledge Infrastructure (CNKI) from inception until July 2022 and updated the search in January 2023 for randomized controlled trials (RCTs) investigating roxadustat for anemia in DD-CKD patients without any limitations on language. We searched the reference list of all identified publications, including relevant meta-analyses and systematic reviews (Table 1).

Table 1. Search terms and results in different databases.

Pubmed
#1 roxadustat[Title/Abstract]	252
#2 FG-4592[Title/Abstract]	86
#3 dialysis[MeSH Terms]	14,7652
#4 end stage kidney disease[MeSH Terms]	10,0181
#5 renal replacement therapy[MeSH Terms]	23,1254
#6 kidney failure[MeSH Terms]	199,255
#7 #1 OR #2	291
#8 #3 OR #4 OR #5 OR #6	371,522
#9 #7 AND #8	111
#10 #9 Filters: Randomized Controlled Trial	27
Embase
#1 ‘roxadustat’/exp OR ‘roxadustat’	682
#2 #1 AND (‘human’/de OR ‘randomized controlled trial’/de)	555
#3 #1 AND ‘randomized controlled trial’/de	84
#4 #3 AND ‘chronic kidney failure’/dm	47
Cochrance
#1 (roxadustat):ti,ab,kw	160
#2 (ASP1517):ti,ab,kw	20
#3 (FG-4592):ti,ab,kw	41
#4 MeSH descriptor: [Dialysis] explode all trees	262
#5 MeSH descriptor: [Renal Replacement Therapy]  explode all trees	10,860
#6 #1 OR #2 OR #3	161
#7 #4 OR #5	11,110
#8 #6 AND #7	21
ClinicalTrials.gov
Search ‘roxadustat’: 48 studies (with results: 15 studies)
CNKI
#1 (subject=’dialysis’ OR ‘renal failure’ OR  ‘renal replacement therapy’ OR ‘end stage  renal disease’) AND (title = roxadustat)	198
#2 #1 AND (primary topic=‘renal anemia’)	50

Inclusion criteria

This meta-analysis included RCT studies that met the following criteria: (1) enrolled adult DD-CKD patients; (2) with renal anemia; (3) using roxadustat as intervention; (4) had a control group (using ESAs, such as epoetin alfa or darbepoetin alfa); 5) one or more outcome indicators were focused on Hb, ferritin, serum iron, total iron binding capacity (TIBC), transferrin saturation (TSAT), transferrin, hepcidin, low-density lipoprotein-cholesterol (LDL-c), total cholesterol (TC), C-reactive protein (CRP), Hb response rate, treatment-emergent adverse events (TEAEs), treatment-emergent serious adverse events (SAEs), and monthly intravenous (IV) iron.

Exclusion criteria

This meta-analysis excluded studies that met the following criteria: (1) non-RCT study; (2) the patient’s anemia was not caused by renal anemia; (2) the full text or the data of outcome indicators were unavailable.

Data extraction

Two researchers (Qiaoqiao Zhou and Mian Mao) independently screened and searched the relevant literature. After reading the title and abstract of each article, they excluded studies that did not meet the inclusion criteria. They carefully read the full text of the selected articles to further determine whether the study was suitable for inclusion in the analysis. In the case of differing opinions between the two researchers, a third researcher (Jing Li) participated in the decision-making process.

The extracted data included: the name of the first author of the selected article, publication year, and basic information about the study (clinical trial number, study design, location, study period, patient group, sample size, intervention measures, drug dose, study duration, study phase), patient characteristics (mean age, hemodialysis or peritoneal dialysis), and efficacy and safety outcomes (Hb, ferritin, serum iron, TIBC, TSAT, transferrin, hepcidin, LDL-c, TC, CRP, Hb response rate, TEAEs, SAEs, monthly IV iron).

Quality assessment

Two reviewers (Qiaoqiao Zhou and Mian Mao) completed the quality assessment independently. In case of disagreement, a third reviewer (Jing Li) reached a consensus through negotiation. The risk of bias in RCTs was assessed using the Cochrane Collaboration tool [17]. Meanwhile, the weighted Cohen’s kappa coefficient (κ) was used to measure agreement.

In addition, the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework was used to evaluate the quality of evidence for each study. The framework characterizes the quality of the body of evidence according to the study’s limitations, imprecision, inconsistency, indirectness, and publication bias of the main results of the studies [18].

Outcome of interest

The primary outcomes are changes from baseline in Hb levels, TEAEs, and SAEs. Secondary outcomes are mean monthly IV iron use, the incidence of major adverse cardiovascular events (MACEs), and changes from baseline in iron biomarker levels, LDL-c, TC, Hb response rates, and Hb levels in patients with a CRP level exceeding the upper limit of the normal range (ULN).

Statistical analyses

Data entry and analysis were done using Excel 2019 (Microsoft) and ReviewManager (RevMan) software version 5.0, respectively. Mean difference (MD) was used for continuous outcomes and risk ratios (RRs) for dichotomous outcomes, both are presented along with their corresponding 95% confidence intervals (CI). Statistical heterogeneity was evaluated using Mantel–Haenzel χ² tests. The fixed-effects model was utilized when there was low heterogeneity (p > 0.1 and I² <50%), whereas a random-effects model was utilized if there was high heterogeneity (p ≤ 0.10 and I² ≥ 50%). Sensitivity analysis was conducted to explore the potential reasons for the heterogeneity in treatment results. We used funnel plots to assess publication bias.

Results

Study characteristics

From 455 potentially relevant articles, we selected 36 for full-text review according to our inclusion and exclusion criteria. In the end, 9 studies [14,19–26] that consisted of a total of 10 RCTs were included. One of the studies had 2 different durations (6 weeks and 19 weeks) of roxadustat treatment and involved 2 different groups of patients [25]. The PRISMA flowchart shows the inclusion and exclusion process and the reasons for exclusion (Figure 1). The main characteristics of the studies included are presented in Table 2.

Figure 1. PRISMA 2020 flow diagram for updated systematic reviews, which included searches of databases, registers, and other sources.

Table 2. Subject demographics and baseline characteristics of the included studies in the meta-analysis.

Study ID	Clinical trial number	Location	Single/ multicenter	Study period	Groups	Number of patients	Age(years) mean ± SD	Roxadustat dose	Study duration	Phases	Patients type
Provenzano 201601	NCT01147666	USA	M	2010–2012	Roxadustat	41	55.8 ± 13.4	1.0, 1.5, 1.8, or 2.0 mg/kg TIW	6 weeks	2	HD
					Epoetin alfa	13	59.5 ± 10.1
Provenzano 201602	NCT01147666	USA	M	2010–2012	Roxadustat	67	56.9 ± 12.1	1.0, 1.5, 1.8, or 2.0 mg/kg TIW	19 weeks	2	HD
					Epoetin alfa	23	57.0 ± 11.6
Chen 2017	NCT01596855	China	M	2011–2012	Roxadustat	74	50.0 ± 12.8	1.1–1.8,1.5–2.3 or 1.7–2.3 mg/kg, TIW	6 weeks	2	HD
					Epoetin alfa	22	53.8 ± 10.0
Chen 2019	NCT02652806	China	M	2015–2016	Roxadustat	204	54.7 ± 13.3	100mg (weight 45 to 60Kg), 120 mg (weight >60 Kg); TIW	27 weeks	3	HD + PD
					Epoetin alfa	100	53.2 ± 13,1
Akizawa 2020	NCT02952092	Japan	M	2016–2018	Roxadustat	150	64.6 ± 11.7	70 mg or 100 mg	24 weeks	3	HD
					Darbepoetin alfa	151	64.9 ± 10.1
Charytan 2021	NCT02273726	USA	M	2015–2018	Roxadustat	370	57.6 ± 13.6	70, 100, 150, or 200 mg TIW based on the patient’s prescribed prestudy ESA dose	52 weeks	3	HD + PD
					Epoetin alfa	371	58.4 ± 13.3
Csiky 2021	EudraCT number 2013-001497-16	Europe and USA	M	2014–2018	Roxadustat	414	61.0 ± 13.8	20, 50, 100 mg TIW based on the patient’s prescribed prestudy ESA dose	104 weeks	3	HD + PD
					Epoetin alfa or Darbepoetin alfa	420	61.8 ± 13.4
Provenzano 2021	NCT02052310	USA, Europe, South America, and Asia	M	2014–2018	Roxadustat	522	53.8 ± 14.7	70 mg (weight < =70Kg), 100 mg (weight >70–+–160 Kg); TIW	52 weeks	3	HD + PD
					Epoetin alfa	521	54.3 ± 14.6
Hou 2022	ChiCTR2000035054	China	S	2019–2020	Roxadustat	86	48 ± 12	100 mg (weight 45 to 60 Kg), 120 mg (weight >60 Kg); TIW	24 weeks	not mention	PD
					ESAs	43	48.3 ± 13
Fishbane 2022	NCT02174731	USA, Europe, Asia, Australia, Latin America	M	2014–2018	Roxadustat	1051	53.5 ± 15.3	70–200 mg TIW, based on prior ESA dose	52 weeks	3	HD + PD
					Epoetin alfa	1055	54.5 ± 15.0

Risk of bias

The risk of bias for the included studies was assessed using the Cochrane Risk-of-Bias tool (Figure 2). All the studies were clinical RCTs. One study had an overall low risk of bias [19]. In terms of interventions, the experimental group was given oral roxadustat and the control group was given ESAs subcutaneously or intravenously, so most trials could not be double-blind but were rather open-labelled research [14,20–26]. Five studies described in detail its random sequence generation and allocation concealment methods [14,20,23,26]. Since the observation indicators were objective data, the blinding of outcome assessment for all the studies was judged to be at low risk. Only 1 trial did not report all prespecified outcome indicators [22], so it had a high risk of reporting bias. Table S1 showed κ = 0.609 to 1.00, which indicated that there was a high inter-rater agreement for risk of bias assessment.

Figure 2. Cochrane Risk-of-Bias tool.

Meta-analysis

Primary outcomes

1. Efficacy outcome: change from baseline in Hb levels (g/dL)

Ten RCTs compared the change from baseline in Hb levels with roxadustat versus ESAs for DD-CKD patients, and included 5580 participants. As shown in Figure 3(A), the Hb level was higher for roxadustat than for ESAs [MD 0.25 g/dL (95%CI 0.14 g/dL to 0.36 g/dL), p < 0.00001]. However, the results of the heterogeneity test showed that these studies were heterogeneous (p < 0.0001). Therefore, we conducted a further sensitivity analysis. Although the participants in these studies were dialysis patients, 3 studies [14,20,26] included incident dialysis patients (on dialysis for ≥ 2 weeks but ≤ 4 weeks) while the other 7 studies were all stable dialysis patients (on dialysis for > 4 weeks). Therefore, we divided these 10 studies into two groups for analysis. As shown in Figure 3(B), the Hb levels of the 2 groups were higher for roxadustat than for ESAs. For the 7 studies involving stable dialysis patients, we performed sensitivity (leave-one-out) analysis to evaluate the impact of each study on the overall effect of Hb level. As shown in Table S2, when the study of Akizawa 2020 was left out, the results of the heterogeneity test showed p = 0.12 > 0.1, and the Hb level was still higher for roxadustat than for ESAs [MD 0.28 g/dL (95%CI 0.22 g/dL to 0.34 g/dL), p < 0.00001].

Figure 3. (1) Forest plot for the change from baseline in Hb level (g/dL). (A) all RCTs; (B) sensitivity analysis for the stable dialysis patients and incident dialysis patients. (2) Funnel chart for the change from baseline in Hb level

2. Safety outcome

2.1 TEAEs

Eight studies reported overall TEAEs. There were 1953 (79.6%) overall TEAEs in roxadustat group and 1794 (81.7%) in the ESAs group. There was no significant difference between the two groups in terms of TEAEs [RR 1.02 (95%CI 1.00 to 1.05), p = 0.09] (Figure 4(1)).

Figure 4. (1) Forest plots of treatment-emergent adverse events (TEAEs). (2) Forest plots of common treatment-emergent adverse events (TEAEs). (A). hypertension; (B.) hyperkalemmia; (C) diarrhea; (D) muscle spasms; (E) upper respiratory tract infections; (F) arteriovenous fistula thrombosis; (G) arteriovenous fistula site complication. (3) Funnel chart of treatment-emergent adverse events (TEAEs).

The most common TEAEs were hypertension, hyperkalemia, diarrhea, muscle spasms, upper respiratory tract infections, arteriovenous fistula thrombosis, and arteriovenous fistula site complications. These data are presented as forest plots in Figure 4(2). There were no significant heterogeneities in all the TEAEs. Meanwhile, our results did not show a significant difference in the risk ratio of these TEAEs.

2.2 SAEs

Nine studies reported SAEs. There was no significant difference between the two groups in terms of SAEs [RR 1.05 (95%CI 0.99 to 1.12), p = 0.11] (Figure 5).

Figure 5. (1) Forest plots of treatment-emergent serious adverse events (SAEs). (2) Funnel chart of treatment-emergent serious adverse events (SAEs).

Secondary outcomes

1. Efficacy outcome

1.1 Changes from baseline in iron biomarker levels

We have unified the unit of each iron biomarkers. As shown in Figure 6, when compared to the ESAs group, the roxadustat group showed increased serum iron [MD 1.85 µmol/L (95%CI 1.21 µmol/L to 2.50 µmol/L), p < 0.00001], TIBC [MD 35.73 µg/dL (95%CI 17.40 µg/dL to 54.06 µg/dL), p = 0.0001], TSAT [MD 1.19% (95%CI 0.14% to 2.24%), p = 0.03], and transferrin [MD 0.40 g/L (95%CI 0.30 g/L to 0.50 g/L), p < 0.00001]. There was no difference between the roxadustat and the ESAs group regarding ferritin [MD −17.16 µg/L (95%CI −46.67 µg/L to 12.35 µg/L), p = 0.25] and hepcidin [MD −15.76 ng/ml (95%CI −32.55 ng/ml to 1.04 ng/ml), p = 0.07] levels.

Figure 6. Forest plot for the change from baseline in iron biomarkers. (A) ferritin (μg/L) (B) iron (μmol/L) C. TIBC (μg/dL) (D) TSAT (%) (E) transferrin (g/L) (F) hepcidin (ng/mL).

In these 6 groups of data, except TSAT, the other 5 groups of data had heterogeneity. All RCTs allowed participants to take iron supplements orally, and some RCTs also allowed IV iron supplements. Therefore, after excluding the RCTs with IV iron supplementation [14,20,23,6], we conducted a sensitivity analysis on these 6 groups of data. The results of the meta-analysis showed that except TIBC, the other 5 groups of data showed homogeneity, but the statistical significance of these 6 groups of data before and after sensitivity analysis was consistent and had no impact on the results.

1.2 Monthly IV iron use (mg)

The data on monthly IV iron supplementation was reported in 4 RCTs with 4413 participants. As shown in Figure 7, there was no significant heterogeneity (p = 0.27, I²=19%). The roxadustat group had a much lower dose than the ESAs group [MD −24.39 mg (95%CI −33.64 mg to −15.15 mg), p < 0.00001].

Figure 7. Forest plot for the monthly IV iron use (mg).

1.3 Change from baseline in LDL-c level and TC level (mmol/L)

Four studies reported LDL-c and 3 studies reported TC. As shown in Figure 8 and Figure 9, the heterogeneity test of the data in these two figures was not statistically significant, which indicates that these studies were homogeneous. Compared with the ESAs group, the roxadustat group had a much lower level of LDL-c [MD −0.39 mmol/L (95%CI −0.47 mmol/L to −0.31 mmol/L), p < 0.00001] and TC [MD −0.6 mmol/L (95%CI −0.7 mmol/L to −0.49 mmol/L), p < 0.00001].

Figure 8. Forest plot for the change from baseline in LDL-c (mmol/L).

Figure 9. Forest plot for the change from baseline in TC (mmol/L).

1.4 Hb response rate

Hb response was defined as (1) subjects whose Hb levels were maintained at no less than 0.5 g/dL below their mean baseline value during the last 2 weeks of the 6-week dosing period [22]; (2) a mean Hb level, averaged over weeks 23 through 27, that was no lower than 1.0 g per deciliter below baseline [21]; (3) the number (percentage) of patients whose Hb levels did not decrease by > 0.5 g/dL from their baseline[25]; (4) Hb level ≥ 10.0 g/dL and change in Hb from baseline ≥1.0 g/dL during the whole study period [24]; (5) patients (%) with mean Hb 10–12 g/dL from weeks 28–36[20]; (6) proportion of patients achieving mean Hb levels 10.0–12.0 g/dL during weeks 28–36 without having received rescue therapy within 6 weeks prior to and during weeks 28–36[23]. Seven RCTs reported the rate with 3151 participants. As shown in Figure 10, there was no significant heterogeneity (p = 0.24, I²=25%). However, there was no difference between the two groups [RR 1.03 (95%CI 0.99 to 1.07), p = 0.09] (Figure 10).

Figure 10. Forest plot for the Hb response rate.

1.5 Change from baseline in Hb levels of patients with CRP > ULN (g/dL)

The change from baseline in Hb levels of patients with CRP > ULN were available in 4 RCTs with 606 participants. These 4 groups of data were both heterogeneous (p = 0.04, I²=63%) and not statistically significant (p = 0.33) (Figure 11(A)). We performed a sensitivity analysis as shown in Figure 11(B). After the data from one study [26] was excluded, the results of the heterogeneity test showed that p = 0.34, and the Hb level was higher for roxadustat than for ESAs [MD 0.39 g/dL (95%CI 0.03 g/dL to 0.75 g/dL), p = 0.03].

Figure 11. Forest plot for the change from baseline in Hb level of patients with CRP > ULN (g/dL). (A) all RCTs; (B) sensitivity analysis for the remaining RCTs.

2. Safety outcome

2.1 The incidence of major adverse cardiovascular events (MACEs)

We compared the incidence of MACEs from 5 RCTs with 4858 participants. As shown in Figure 12, there was no significant heterogeneity (p = 0.88, I²=0%) and no difference between the two groups [RR 1.04 (95%CI 0.85 to 1.28), p = 0.70].

Figure 12. Forest plots of major adverse cardiovascular events (MACE).

Publication bias

Publication bias was examined by funnel plots in Figures 3–5.

GRADEpro for the outcomes

The classification of evidence quality is used to ensure the comparability of each system evaluation. Cochrane systematic evaluation recommends the use of "Grades of Recommendation, Assessment, Development and Evaluation", namely GRADE standard. This standard is a comprehensive evidence-grading system formulated by GRADE Working Group from the perspective of users and has been adopted by many international organizations and associations, including WHO and Cochrane Collaboration Network. Cochrane collaboration group recommends that GRADE standard to be used to evaluate the quality of evidence for each outcome in its systematic evaluation, which is divided into four levels: high, moderate, low and extremely low. To enable the easy use of the GRADE system to quantitatively rate evidence and provide a reasonable recommendation, the GRADE working group has designed GRADEprofile Guideline Development Tool (GRADEpro GDT).

We evaluated all outcome indicators using GRADEpro (http://gradepro.org). The 5 degradation factors of evidence quality evaluation were risk of bias, inconsistency of results, indirectness of evidence, imprecision, and publication bias. After a comprehensive analysis, we found that the outcome indicators with high certainty were Hb, serum iron, transferrin, LDL-c, TC, and MACEs. The outcome indicators with moderate certainty were TIBC, Hb levels of patients with CRP > ULN, IV iron, and Hb response rate. The certainty of the remaining outcome indicators was low (Table 3).

Table 3. GRADEpro.

Certainty assessment							№ of patients		Effect		Certainty	Importance
№ of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Roxadustat	ESAs	Relative (95% CI)	Absolute (95% CI)
Hb (g/dL)
10	Randomized trials	Serious	Not serious	Not serious	Not serious	Strong association all plausible residual confounding would reduce the demonstrated effect dose response gradient	2904	2676	–	MD 0.25 higher (0.14 higher to 0.36 higher)	⨁⨁⨁⨁ High	CRITICAL
Teaes (%)
8	Randomized trials	Serious	Very serious	Not serious	Not serious	All plausible residual confounding would reduce the demonstrated effect	1953/2454 (79.6%)	1794/2196 (81.7%)	RR 1.02 (1.00 to 1.05)	16 more per 1,000 (from 0 fewer to 41 more)	⨁⨁◯◯ Low	CRITICAL
Saes (%)
9	Randomized trials	Serious	Very serious	Not serious	Not serious	All plausible residual confounding would reduce the demonstrated effect	1027/2976 (34.5%)	951/2713 (35.1%)	RR 1.05 (0.99 to 1.12)	18 more per 1,000 (from 4 fewer to 42 more)	⨁⨁◯◯ Low	CRITICAL
ferritin (μg/L)
8	Randomized trials	serious	Serious	Not serious	Serious	All plausible residual confounding would reduce the demonstrated effect	1442	1225	–	MD 17.16 lower (46.67 lower to 12.35 higher)	⨁⨁◯◯ Low	IMPORTANT
Iron (μmol/L)
9	Randomized trials	serious	Not serious	Not serious	Not serious	All plausible residual confounding would reduce the demonstrated effect	1856	1645	–	MD 1.85 higher (1.21 higher to 2.5 higher)	⨁⨁⨁⨁ High	IMPORTANT
TIBC (μg/dl)
8	Randomized trials	serious	Not serious	Not serious	Serious	All plausible residual confounding would reduce the demonstrated effect	1486	1274	–	MD 35.73 higher (17.4 higher to 54.06 higher)	⨁⨁⨁◯ Moderate	IMPORTANT
Tsat (%)
8	Randomized trials	serious	Very serious	Not serious	Not serious	All plausible residual confounding would reduce the demonstrated effect	1441	1224	–	MD 1.19 higher (0.14 higher to 2.24 higher)	⨁⨁◯◯ Low	IMPORTANT
transferrin (g/L)
3	Randomized trials	serious	not serious	not serious	not serious	all plausible residual confounding would reduce the demonstrated effect	370	267	–	MD 0.4 higher (0.3 higher to 0.5 higher)	⨁⨁⨁⨁ High	IMPORTANT
hepcidin (ng/mL)
7	Randomized trials	serious	serious	not serious	serious	all plausible residual confounding would reduce the demonstrated effect	1351	1178	–	MD 15.76 lower (32.55 lower to 1.04 higher)	⨁⨁◯◯ Low	IMPORTANT
IV iron (mg)
4	Randomized trials	serious	not serious	not serious	serious	all plausible residual confounding would reduce the demonstrated effect	2190	2223	–	MD 24.39 lower (33.64 lower to 15.15 lower)	⨁⨁⨁◯ Moderate	IMPORTANT
LDL-c (mmol/L)
4	Randomized trials	serious	not serious	not serious	not serious	all plausible residual confounding would reduce the demonstrated effect	921	851	–	MD 0.39 lower (0.47 lower to 0.31 lower)	⨁⨁⨁⨁ High	IMPORTANT
TC (mmol/L)
3	Randomized trials	serious	not serious	not serious	not serious	all plausible residual confounding would reduce the demonstrated effect	507	431	–	MD 0.6 lower (0.7 lower to 0.49 lower)	⨁⨁⨁⨁ High	IMPORTANT
Hb response rate (%)
7	Randomized trials	serious	serious	not serious	not serious	all plausible residual confounding would reduce the demonstrated effect	1336/1683 (79.4%)	1115/1467 (76.0%)	RR 1.03 (0.99 to 1.07)	23 more per 1000 (from 8 fewer to 53 more)	⨁⨁⨁◯ Moderate	IMPORTANT
Hb of high CRP (g/dL)
4	Randomized trials	serious	serious	not serious	not serious	all plausible residual confounding would reduce the demonstrated effect	321	285	–	MD 0.2 higher (0.21 lower to 0.61 higher)	⨁⨁⨁◯ Moderate	IMPORTANT
MACE (%)
5	randomized trials	serious	not serious	not serious	not serious	all plausible residual confounding would reduce the demonstrated effect	176/2462 (7.1%)	166/2396 (6.9%)	RR 1.04 (0.85 to 1.28)	3 more per 1,000 (from 10 fewer to 19 more)	⨁⨁⨁⨁ High	IMPORTANT

CI: confidence interval; MD: mean difference; RR: risk ratio.

Discussion

In our meta-analysis, we included 9 RCTs with 10 datasets and 5698 DD-CKD participants to assess the efficacy and safety of roxadustat. To our best knowledge, this is the first meta-analysis to evaluate the efficacy and safety of roxadustat for anaemia in dialysis-dependent chronic kidney disease patients. Furthermore, we included the latest RCT results with the largest sample size by Fishbane reported in 2022. The results of our meta-analysis showed that compared with ESAs, roxadustat may increase the level of Hb, serum iron, TIBC, TSAT, and transferrin, while reducing LDL-c, TC, and the necessary dosage of IV iron. Even in states of inflammation, roxadustat may correct renal anemia without having to increase the oral dosage. In addition, roxadustat possessed a similar safety background as ESAs in terms of TEAEs, SAEs, and MACEs. All outcome indicators were evaluated using GRADEpro as providing ratings above low level.

Hypoxia-inducible factor (HIF) is a dimer consisting of α and β subunits, which participate in the regulation of the oxygen concentration in the body. HIF-α is present as 1α, 2α and 3α subtypes. HIF-2α is mainly distributed in the kidney and is needed for the production of red blood cells. Prolyl hydroxylase (PHD) family is involved in the regulation of HIF. Under a normal oxygen state, HIF-α is recognized by PHD and quickly degraded by proteasome [27]. Under hypoxic conditions, the decreased PHD activity leads to less degradation of HIF-α. The intact HIF-α and HIF-β combine to form a dimer to up-regulate a series of genes[6]. HIF-PHI strongly regulates the production of red blood cells, which is achieved mainly by promoting the production of EPO and the expression of its receptor and reducing the level of hepcidin to promote the expression of iron metabolism-related proteins [28,29]. Roxadustat is a HIF-PHI. On 17 December 2018, it was officially approved by the State Drug Administration of China to treat CKD anemia in DD and NDD patients.

The primary efficacy outcome of our meta-analysis showed that roxadustat significantly increased the Hb level of dialysis patients when compared with the ESAs group (Figure 3(A), p < 0.00001). Considering that some of the participants in the study were incident dialysis patients, which may lead to the heterogeneity of the data when we conducted a sensitivity analysis, the subjects of the study were patients with stable dialysis. The final analysis result was still that roxadustat could significantly improve Hb levels (Figure 3(B), Table S2). This result was also consistent with some pooled analyses of previous reports [11,13,30,31]. However, there was no significant difference in the Hb response rate between the two groups (p = 0.09).

The primary safety outcome of our meta-analysis suggests that roxadustat does not increase the incidence of TEAEs and SAEs (p = 0.09, p = 0.57, respectively). In recent years, researchers had paid great attention to cardiovascular safety in clinical trials. The FDA regards cardiovascular safety as major adverse cardiovascular events (MACEs) – death, myocardial infarction, and stroke[31,32]. Therefore, MACEs were also of concern in our meta-analysis. As shown in Figure 12, roxadustat did not increase MACEs when compared to ESAs (p = 0.70). This result was consistent with a pooled analysis of four phase 3 studies by Barratt et al. [30]. They reported that roxadustat was non-inferior to ESAs in terms of risk for MACEs [HR 1.09 (95%CI 0.95 to 1.26)] and MACE+ (MACE plus unstable angina or congestive heart failure) [HR 0.98 (95%CI 0.86 to 1.11)].

In addition to EPO deficiency, iron deficiency is also a key factor in renal anemia in CKD patients, especially dialysis patients [33,34]. These patients may have absolute or functional iron deficiency, which can be diagnosed by TSAT and ferritin. More and more evidence shows that excessive hepcidin could explain the causes of iron metabolism disorders in many CKD patients [35]. Hepcidin is mainly produced in the liver. It can reduce the level of plasma iron by inhibiting the absorption of iron by the small intestine and the release of circulating iron by macrophages, leading to anemia [36]. After hypoxia activates the HIF pathway, the expression of hepcidin in the liver will be significantly reduced [37,38], which helps to increase the synthesis of transferrin, enhance the intestinal absorption of iron [39,40], and also increase iron binding to promote the transport of iron to the bone marrow [41]. Studies have shown that roxadustat can stimulate the synthesis of endogenous EPO, thereby affecting and regulating the body’s iron metabolism and increasing the level of Hb [42,43]. In our meta-analysis, the roxadustat group showed increased iron, TIBC, TSAT, and transferrin (p < 0.05, respectively). There was no difference between the roxadustat and the ESAs group regarding ferritin and hepcidin. These results may suggest that roxadustat could increase the level of Hb by upregulating serum iron, TIBC, TSAT, and transferrin. Since some studies allowed participants to use IV iron, in order to ensure the homogeneity of the data, we conducted a further sensitivity analysis on iron metabolism indicators after excluding these studies, but the results were still consistent with those before excluding them. Although iron supplementation could effectively increase the Hb level of dialysis patients and reduce the dosage of ESAs and the need for blood transfusion [44], excessive intravenous iron infusion causes oxidative stress injury and iron overload, which can lead to damage to the heart, liver, spleen and other important organs[6]. As shown in Figure 7, the dosage of IV iron in the roxadustat group was less than that in the ESAs group. These outcomes may reveal that roxadustat could improve iron metabolism, increase the absorption and utilization of oral iron, and reduce the dosage of IV iron.

Some RCTs also observed the effect of roxadustat on blood lipids, so we performed a meta-analysis on LDL-c and TC. As shown in Figures 8 and 9, roxadustat significantly reduced LDL-c and TC levels in patients. Therefore, when using statins at the same time, we should pay attention to reducing the dose and monitoring the adverse reactions of statins.

Some studies had shown that inflammation can inhibit the body’s response to ESAs [45]. We were interested to understand whether inflammation has the same effect on roxadustat. Four studies observed a change from baseline in Hb level of patients with a CRP level exceeding ULN. Since the participants of one of the studies were patients newly initiating dialysis [26], we excluded that study and conducted a meta-analysis of the remaining 3 groups of stable dialysis patients [19,21,24]. As shown in Figure 11(B), the Hb level was higher for roxadustat than for ESAs (p = 0.03). In these studies, compared to patients with normal CRP levels, patients with elevated CRP had a lower Hb response, despite receiving a higher dose of ESAs. However, a similar inflammatory state did not affect the Hb response and dosage of roxadustat. This result indicates that roxadustat could increase Hb level even in the inflammatory state, and the dosage may not be affected. This provides a new way for nephrologists to treat anemia in dialysis patients with concomitant inflammatory conditions.

Our meta-analysis suggests that roxadustat is an effective and safe drug for the treatment of anemia in DD-CKD patients according to the data of these RCTs. Due to the different mechanisms of roxadustat and ESAs in regulating Hb and iron metabolism, roxadustat provides a new choice for nephrologists to treat renal anemia in DD-CKD patients. Many genes unrelated to erythropoiesis are regulated by HIF, and their activity may be affected by HIF-PHI, resulting in some adverse reactions, such as the possible impact on tumors [46]. In the treatment of renal anemia, the higher the target value of hemoglobin, the closer the high-dose ESAs are to all-cause mortality, hypertension, stroke, and thrombotic events [47]. It is still not clear whether the adverse reactions of roxadustat will increase with increasing dosage, or what is the optimal target value of Hb for renal anemia. It should be noted that the current large-scale clinical trials were sponsored by the industry. Our results showed that the Hb level was higher for roxadustat than for ESAs, and the MD was 0.25 g/L. This value may not have obvious clinical significance, even though previous reports showed this value as a primary superiority outcome. Because we are not sure that with uncontrolled variable of dosing, whether this outcome is relevant or significant enough. Our meta-analysis cannot replace an influential randomized controlled trial. However, in the case of a small sample size and a limited number of studies, it may solve clinical problems that individual research results cannot answer.

To our knowledge, this is the first meta-analysis to evaluate the efficacy and safety of roxadustat for anemia in DD-CKD patients. Compared with previous similar meta-analyses [11,12,48], our analysis has several strengths. First, our analysis included the latest research [14], with a larger sample size and longer follow-up time, and focuses on DD-CKD patients only. Second, in addition to analyzing Hb, we also analyzed more indicators of iron biomarkers and Hb changes in the inflammatory state, and then sensitivity analysis was conducted to test the stability of these outcomes. Third, we focused on MACEs in the process of analyzing TEAEs and SAEs, which are of concern to some authors. Fourth, we evaluated all outcome indicators using GRADEpro. Our meta-analysis also has some limitations. First, although, to our best knowledge, this is the first meta-analysis to evaluate the efficacy and safety of roxadustat for anemia in DD-CKD patients, we did not register this study in PROSPERO or Cochrane library in advance. Second, the dialysis patients in our meta-analysis included hemodialysis patients and peritoneal dialysis patients; we did not analyze the data of these two groups of patients separately, which may add to the heterogeneity of our meta-analysis. Third, epoetin alfa and darbepoetin alfa are two different kinds of erythropoietin in the control group, and there was no unified standard for the dose of roxadustat in the intervention group, which may bias the outcome.

Conclusions

In this meta-analysis of RCTs, after oral administration of roxadustat, the Hb level of DD-CKD patients was significantly increased and was not affected by states of inflammation. Roxadustat may also improve iron utilization by increasing iron, TIBC, TSAT, and transferrin while decreasing blood lipid levels. In addition, roxadustat was not associated with higher TEAEs, SAEs, and MACEs when compared to ESAs.

Authors’ contributions

QZ and MM were involved in the research idea and study design. QZ, JL and FD were responsible for data acquisition. QZ and MM were involved in data analysis and interpretation. JL and FD were involved in supervision or mentorship. All authors read and approved the final manuscript.

Acknowledgements

We thank Jing Ma (Office of Cancer Prevention and Treatment, Sichuan Clinical Research Center for Cancer, Sichuan Cancer Hospital & Institute, Sichuan Cancer Center, Affiliated Cancer Hospital of University of Electronic Science and Technology of China, Chengdu, China) for providing methodological assistance for this manuscript.

Disclosure statement

All authors have no conflicts of interest to declare. The authors declare that this article has not been published previously in whole or in part.

Data availability

The data used to support the findings of this study are available from the corresponding author upon request.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.