Roxadustat induced fetal hemoglobin in anemic CKD patients: a dual-center real-world observational cohort study

Abstract

Background

Induction of fetal hemoglobin (HbF) expression is a promising therapeutic approach for hemoglobinopathies. HbF is primarily regulated by hypoxia-inducible factor (HIF). Roxadustat, an oral HIF prolyl hydroxylase inhibitor (HIF-PHI), is utilized in the treatment of chronic kidney disease (CKD) patients with anemia. Its capacity to induce HbF has been demonstrated in cell and animal models, but not in humans.

Methods

In this dual-center, prospective, real-world, non-randomized observational cohort study, we initially enrolled 85 anemic CKD patients receiving distinct anti-anemia treatments. Following post hoc exclusion of participants with insufficient follow-up data, we analyzed the data of 52 patients, comprising those in the oral Roxadustat group (n = 27) and the subcutaneous erythropoiesis-stimulating agent (ESA) group (n = 25). HbF levels were monitored every four weeks by measuring the percentage of HbF-positive red blood cells (F-cells) and mean corpuscular fetal hemoglobin (MCHbF).

Results

Roxadustat administration showed progressive elevation of HbF levels, however, the ESA group exhibited only minor nonprogressive HbF fluctuations throughout the observation period. Significant differences in the increase of F-cells percentage and MCHbF between the two groups were observed at week 4 and amplified at week 8 and 12. Only one patient in the Roxadustat group exhibited clinically significant hypothyroidism requiring levothyroxine replacement. Safety monitoring was limited to clinically evident events reported during routine visits; consequently, subclinical abnormalities or mild adverse events were not systematically documented.

Conclusion

Our study suggested that Roxadustat has the potential to induce HbF expression in anemic CKD patients.

Study registration

The study was registered retrospectively in the Chinese Clinical Trial Registry (ChiCTR, https://www.chictr.org.cn) under the identifier ChiCTR2500096786 on February 6, 2025.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12882-026-04860-x.

Introduction

Roxadustat (FG-4592), a hypoxia-inducible factor prolyl hydroxylase inhibitor (HIF-PHI) that stabilizes HIF-α isoforms (HIF-1α, HIF-2α, HIF-3α) by inhibiting the prolyl hydroxylase domain (PHD) [1–3], is mainly used to enhance endogenous erythropoietin (EPO) production in patients with chronic kidney disease (CKD) and concomitant anemia. Beyond approved indications, its clinical applications are expanding to HIF-2α-mediated anemia spectrum disorders (e.g., pure red cell aplasia and anemia due to perioperative blood loss) [4, 5], as well as to HIF-1α-targeted interventions, including myocardial protection in acute myocardial infarction (NCT04803864), mitigation of delayed graft function and acute rejection in renal transplant recipients (ChiCTR2000032858), and cognitive improvement in maintenance hemodialysis patients (ChiCTR2300070478). However, none of these clinical studies investigated its potential for the induction of fetal hemoglobin (HbF), which is critical for hemoglobinopathies treatment.

HbF (α₂γ₂) constitutes approximately 60% to 80% of total hemoglobin in the full-term newborn and is maintained in 0.5% to 2% of red blood cells (RBCs) in adults [6–8]. Compared with HbA, HbF has a higher oxygen affinity owing to its decreased insensitivity to 2,3-bisphosphoglycerate (2,3-BPG, also called 2,3-DPG), the major modulator of hemoglobin-O2 affinity. In patients with hemoglobinopathies like β-thalassemia and sickle cell disease, the HbF levels exhibit an inverse correlation with clinical severity [9–13] Mechanistically, the γ-globin genes (HBG1 and HBG2) are known transcriptional targets of HIF. Therefore, it is hypothesized that the stabilization of HIF-mediated by HIF-PHIs like Roxadustat can activate the transcription of these genes, thereby reactivating the HbF synthesis.

The ability of HIF-PHIs to increase HbF levels was demonstrated in both cellular and animal models [14–17]. However, clinical evidence supporting this effect in human is lacking. Thus, this study aimed to evaluate the ability of Roxadustat to stimulate HbF by measuring HbF levels every four weeks in patients with anemic CKD.

Materials and methods

Study design and participants

This dual-center, prospective, real-world, non-randomized, observational cohort study initially enrolled 85 CKD patients with anemia between December 2024 and June 2025 at the Second Affiliated Hospital of Army Medical University (Xinqiao Hospital, Chongqing, China) and Chongqing Hospital of People’s Armed Police (PAP, Chongqing, China). Treatment assignment was determined entirely by the physician’s discretion based on clinical presentation and established guidelines, rather than by a randomization protocol. Consequently, participants were stratified into two cohorts: the Roxadustat group (n = 50) and the ESA group (n = 35). Patients who switched medications because of disease progression, were lost to follow-up, or provided fewer than two evaluable measurements during the study were excluded from the cohort (Fig. 1).

Fig. 1.

CONSORT flow diagram. Enrollment and outcomes of the study. Of 130 CKD patients with anemia assessed, 45 patients met exclusion criteria (18 not meeting inclusion criteria, 27 declined participation). Among the 85 eligible patients, 50 were assigned to the Roxadustat group and 35 to the ESA group. Primary outcome data were obtained for 52 patients at 12 weeks, 27 in the Roxadustat group and 25 in the ESA group

Inclusion and exclusion criteria

Study participants were selected based on a set of predefined inclusion and exclusion criteria to ensure the reliability and relevance of the findings.

Inclusion criteria

Diagnosis of CKD with concomitant anemia required the following criteria:

1. Confirmed CKD according to KDIGO criteria (GFR < 90 mL/min/1.73 m² for > 3 months).

2. Anemia defined as HGB < 120 g/L (female) or <130 g/L (male) (WHO 2023);

3. Adult age (≥18 years old);

4. Willingness to provide written informed consent.

Exclusion criteria

1. Receipt of HbF inducer or hematopoietic modulators (e.g., Yisui Shengxue granules, hydroxyurea, thalidomide, hypomethylating agents, and progestogens) within 3 months before enrollment.

2. Concurrent unaddressed anemia from other causes.

3. Comorbid thrombocytopenia (PLT < 100 × 10⁹/L) or thrombocythemia (PLT > 450 × 10⁹/L).

4. Severe cardiopulmonary or cerebrovascular conditions.

5. Recent exposure to stressors like extreme temperatures, surgery, poisoning, trauma, or high-altitude environments.

6. Pregnancy, lactation, or unwillingness to use contraception among women of childbearing age.

7. Concurrent participation in other clinical trials or studies.

8. Organic heart disease, coronary insufficiency, significant arrhythmias, resting heart rate > 100/min, or severe hypertension.

9. Respiratory conditions, including bronchodilator use, asthma, interstitial lung disease, chronic obstructive pulmonary disease, respiratory insufficiency, or active tuberculosis.

10. History of specialized physical training.

Study size

The study sample sizes for the Roxadustat and ESA groups were calculated using G*Power 3.1 software (Düsseldorf, Germany). Our preliminary data showed that the percentage of F-cells in patients with long-term use of Roxadustat and ESA was 4.01 ± 1.52% and 1.27 ± 0.99%, respectively, with a Cohen’s d of 2.13. Experimental data in vitro [14] showed that the percentage of F-cells was 16.1 ± 7.9% before and 37.1 ± 13.2% after Roxadustat treatment, with a Cohen’s d of 1.93. A high Cohen’s d value (the effect size > 1.00) can lead to errors in sample size estimation. Following the ICH E9 regulatory standards for clinical study designs, the implementation of sensitivity corrections is strongly advised to mitigate potential bias. The sample size calculation was based on Cohen’s d = 0.80 with a one-sided α = 0.05 and 80% test power, using the Mann–Whitney U test (two groups), which resulted in a requirement of at least 21 participants per group. Cohen’s d of 0.80 denotes a large effect that captures the minimum between-group difference in clinical studies and trials. A total of 85 patients were included, with an estimated dropout rate of 30%.

Drug administration

Patients in the Roxadustat group received oral Roxadustat capsules (FibroGen China; 150 mg/capsule; NMPA Approval No. H20180023) thrice a week. The control group received subcutaneous ESA (recombinant human erythropoietin (rhEPO); 3SBIO Inc., Shenyang, China; NMPA Approval No. S19980009) at 10,000 IU per dose, twice weekly. Other adjunctive anti-anemia regimens, including oral folic acid tablets (5 mg/d; Tianjin Lisheng Pharmaceutical Co., Ltd.; NMPA Approval No. H20200215) and iron polysaccharide complex capsules (150 mg/d, Schwarz Pharma Manufacturing Inc., Zhuhai, NMPA Approval No. H20150627) were routinely administered to both groups.

Observational outcomes

The primary outcomes were the changes in HbF levels from baseline to week 12, including the percentage of F-cells and MCHbF. Secondary outcomes focused on adverse events (AEs) necessitating urgent therapeutic intervention. Safety data were primarily collected from outpatient medical records and patient-initiated reports. Routine hematological assessments for hepatic or thyroid functions were not performed as planned, nor were blood pressure measurements or cardiovascular imaging examinations. During each outpatient visit or patient-initiated contact, attending physicians specifically inquired about symptoms potentially related to thyroid events, with all reported symptoms documented verbatim. Hospital admissions were identified by cross-referencing the electronic discharge database.

Ethical approval declarations

The study was registered in the Chinese Clinical Trial Registry (ChiCTR2500096786, PID 258,883) and was conducted according to the principles of the Declaration of Helsinki. The study protocol was approved by the Medical Ethics Committees of the Second Affiliated Hospital of the Army Medical University (Ref: 2024-Yan-308–01) and the Chongqing Hospital of PAP (Ref: WJYY2024-1210). All the participants have signed the written consent form.

Laboratory analyses

Blood samples were analyzed by the Department of Laboratory Medicine using standardized protocols (ISO 15189:2022). Complete blood count testing was conducted using a Sysmex XN-9000 automated hematology analyzer (Sysmex Corporation, Kobe, Japan), Mindray BC-7500 CRP-integrated hematology analyzers (Mindray Bio-Medical Electronics Co., Ltd., Shenzhen, China) at Xinqiao Hospital, and Mindray BC-6800 Plus analyzers at PAP Hospital. Biochemical parameters were quantified using Mindray BS-2800 M automated biochemistry analyzers and Mindray BS-2200 M systems, with all instruments undergoing daily quality control procedures per the manufacturer’s specifications.

Sample preparation and flow cytometric analysis

Peripheral venous blood samples (20 µL) from the residual specimens for complete blood count were washed with PBS, fixed in 0.1% glutaraldehyde, and permeabilized with Triton X-100. Immunostaining was performed in a 30 µL reaction volume containing 1.5 µL of an APC-conjugated anti-human fetal hemoglobin monoclonal antibody [18–20] (Clone HbF-1; Catalog No. MHFH05, Thermo Fisher Scientific) and an APC-conjugated Mouse IgG1 Isotype Control (Cat. No. MA518093, Invitrogen) as an isotype control. Incubation for 40 minutes in the dark at room temperature, with brief, gentle vortexing every 20 minutes. Following a wash with 0.1% BSA-PBS and resuspended in 250 µL of the same buffer, a minimum of 10,000 single-cell events were recorded per sample. Data acquisition was conducted on a CytoFLEX SRT flow cytometer (Beckman Coulter) calibrated with CytoFLEX Ready to Use Daily QC Fluorospheres (Catalog No. C65719). Daily QC procedures confirmed optimal optical alignment and fluidics, specifically verifying that the 660 nm detection channel and DD1 parameters satisfied rigorous quality control specifications before sample analysis. To minimize technical variability, instrument settings, including PMT voltages, were standardized across all time points. Data were analyzed using FlowJo software (v10.8.1). The gating strategy involved debris exclusion (FSC-H vs. SSC-H) and doublet removal (FSC-H vs. FSC-A) to isolate single cells, within which F-cell populations were identified by plotting HbF-APC fluorescence (FL8-H) versus SSC-H, with gates set based on matched isotype controls (Fig. S1).

Outcome measures

Pretreatment baseline parameters, including demographics (e.g., age and sex), medical history, hemoglobin levels, CKD stage, and mean corpuscular hemoglobin (MCH), were formally documented. Outcomes regarding HbF levels (indicated by the percentage of F-cells and MCHbF) were assessed at baseline and at weeks 4, 8, and 12 post-treatment initiation. HbF levels were evaluated based on the percentage of F-cells and MCHbF. Quantitative assessment of MCHbF was performed using the following formula: MCHbF (pg/RBC) = (percentage of F cells × MCH)/100. Adverse events necessitating urgent clinical intervention were documented, with data primarily collected from outpatient medical records and patient-initiated reports.

Statistical analysis

Normally distributed values are presented as mean ± standard deviation (SD), and non-normally distributed data are presented as medians and interquartile range (IQR). Data distribution was assessed using the Shapiro–Wilk normality test. We used the Wilcoxon signed-rank test to compare the percentage of F-cells and MCHbF in the study and reported the results as median with interquartile range (IQR). For categorical proportions, the chi-square test (χ2) or Fisher’s exact test was used. The influencing factors were analyzed using the linear mixed model (LMM) [21–23], incorporating baseline clinical and laboratory characteristics as covariates. These models included treatment group, time, and their interaction as regression terms, while adjusting for baseline clinical and laboratory characteristics as covariates to account for potential confounders. To account for the within-subject correlation inherent in repeated measures data, a random intercept was introduced at the subject level (with slopes modeled as fixed effects). The general form of the model is specified as follows: Yit=β0+β1Timet+β2Groupi+β3(Timet × Groupi) +γXi+b0i+ϵit, where Yit represents the observed value for the i-th subject at time t; Xi denotes the vector of baseline covariates; b0i∼N (0, τ2) represents the individual random intercept; and ϵit∼N (0, σ2) represents the residual error. Additionally, the intraclass correlation coefficient (ICC) was calculated to quantify between-patient heterogeneity and validate the necessity of the mixed-effects model structure. Statistical significance was set at p value < 0.05(two-sided α). Data visualization was performed using R version 4.3.2 (Vienna, Austria).

Results

Baseline characteristics of participants

To study the induction effect of Roxadustat on HbF, we initially enrolled 85 participants with anemic CKD from two hospitals (Xinqiao Hospital, n = 68; PAP Hospital, n = 17) between December 2024 and June 2025 and divided them into two groups: patients treated with Roxadustat (n = 50) and those treated with an erythropoiesis-stimulating agent (ESA, n = 35). After a 12-week follow-up, a total of 52 patients were included in the final analytical cohort (Roxadustat, n = 27; ESA n = 25). (Fig. 1). The baseline characteristics between two groups showed no significant differences between the Roxadustat and ESA groups (p > 0.05, Table 1). The characteristics, including the percentage of F-cells, MCHbF, hemoglobin concentration (HGB), white blood cell count (WBC), platelet count (PLT), age, sex, anemia grade, CKD stage, and primary disease. The clinical characteristics of completed participants are shown in Table S1.

Table 1.

Baseline clinical and laboratory characteristics of patients

Characteristic	Overall(n = 52)	Roxadustat (n = 27)	ESA (n = 25)	P-value1
F-cells % (median IQR)	1.16 (0.55, 1.90)	0.77 (0.44, 1.88)	1.39 (0.82, 1.91)	0.2
MCHbF pg (median IQR)	0.34 (0.14, 0.57)	0.21 (0.12, 0.45)	0.44 (0.25, 0.68)	0.067
Age y mean (SD)	53.62 (12.44)	51.59 (12.09)	55.80 (12.67)	0.28
HGB g/L (median IQR)	92 (80, 103)	87 (76, 100)	95 (82, 113)	0.071
WBC 109/L (median IQR)	6.24 (4.78, 7.76)	6.24 (4.89, 8.24)	6.42 (4.67, 7.75)	0.85
PLT 1012/L (median IQR)	179 (143, 226)	179 (143, 234)	174 (142, 207)	0.42
Sex %(n)				0.58
Male	59.62% (31.00)	55.56% (15.00)	64.00% (16.00)
Female	40.38% (21.00)	44.44% (12.00)	36.00% (9.00)
Anemia Grade %(n)2				0.19
Mild	48.08% (25.00)	40.74% (11.00)	56.00% (14.00)
Moderate	44.23% (23.00)	55.56% (15.00)	32.00% (8.00)
Severe	7.69% (4.00)	3.70% (1.00)	12.00% (3.00)
CKD stage %(n)3				0.86
Stage 2–3	21.15% (11.00)	18.52% (5.00)	24.00% (6.00)
Stage 4	19.23% (10.00)	22.22% (6.00)	16.00% (4.00)
Stage 5	59.62% (31.00)	59.26% (16.00)	60.00% (15.00)
Treatment %(n)				0.095
Dialysis	50.00% (26.00)	62.96% (17.00)	36.00% (9.00)
non-Dialysis	50.00% (26.00)	37.04% (10.00)	64.00% (16.00)
Primary Disease %(n)				0.56
Type2 Diabetes	65.38% (34.00)	70.37% (19.00)	60.00% (15.00)
Non-Diabetes	34.62% (18.00)	29.63% (8.00)	40.00% (10.00)

¹χ2, Fisher’s exact test, or Wilcoxon rank sum test when appropriate

²Anemia severity was defined according to the hemoglobin level: mild, 90–120 g/L; moderate, 60–89 g/L; and severe, 30–59 g/L

³CKD phase was defined according to the eGFR value: stage 2, 60–89 mL/min/1.73 m²; stage 3a, 45–59 mL/min/1.73 m²; stage 3b, 30–44 mL/min/1.73 m²; stage 4, 15–29 mL/min/1.73 m²; and stage 5, <15 mL/min/1.73 m² or dialysis

However, 33 participants (Roxadustat: 23, 46%; ESA: 10, 29%) were excluded due to the lack of at least two HbF measurements, resulting in an attrition rate that exceeded our expectations. To systematically assess potential selection bias, we analyzed the characteristics of the enrolled and excluded populations (Tables S2–S5). First, regarding the initially enrolled cohort (n = 85), the Roxadustat group exhibited lower baseline hemoglobin (p = 0.012) and a higher proportion of moderate anemia (p = 0.017) compared to the ESA group (Table S2). Second, regarding attrition in the overall population (n = 85), patients lost to follow-up had higher baseline MCHbF levels compared to those who completed the study (p = 0.048; Table S3). Third, subgroup analysis revealed that this attrition difference was primarily driven by the Roxadustat group (Table S4). Specifically, patients lost to follow-up in the Roxadustat group exhibited significantly higher baseline F-cells% (p = 0.018) and MCHbF (p = 0.007) than those who remained. This indicates that the retained cohort did not possess a baseline advantage. Fourth, in the ESA group (Table S5), a difference was observed in dialysis status (p = 0.027), with non-dialysis patients being more frequently lost to follow-up, likely attributable to their reduced frequency of clinical contact. Overall, these analyses indicated that despite the high attrition rate, the exclusion of participants did not introduce a selection bias favoring Roxadustat.

Percentage of F-cells

To evaluate HbF levels, we analyzed the percentages of F cells and MCHbF. The percentage of F-cells denotes the fraction of HbF-expressing RBCs within the total RBC population, indicating the distribution of F-cells in the bloodstream. Before Roxadustat treatment, the median percentage of F-cells in anemic CKD patients was 0.77%, and it was significantly increased to 2.29%, 2.66%, and 4.34% after medication for 4, 8, and 12 weeks, respectively; however, it remained stable in patients treated with ESA (Fig. 2A). Notably, in patient #26, who was the only one diagnosed with non-transfusion-dependent β-thalassemia, the percentage of F-cells increased from 1% to 6.77% after 12 weeks of medication with Roxadustat (Fig. S2).

Fig. 2.

HbF levels dynamics against Roxadustat and ESA treatment. (A) dynamics in F-cells percentage post-treatment in patients who received Roxadustat vs ESA. (B) fitted curve for the F-cells percentage between two groups. (C) intergroup comparison of F-cells percentage increment. (D) dynamics in MCHbF post-treatment in patients who received Roxadustat vs ESA. (E) fitted curve for MCHbF between two groups. (F) intergroup comparison of MCHbF increment. (box: median with interquartile range; whiskers: min to max; timepoints: baseline, week 4, week 8, and week 12)

Curve-fitting analysis showed that the percentage of F cells increased over time in patients treated with Roxadustat (Fig. 2B). There was also a significant difference (p < 0.001) in the increase in F-cells between the Roxadustat and ESA groups during the observation period (Fig. 2B). We next analyzed the increase in F-cell percentage at weeks 4, 8, and 12 compared to baseline. Figure 2C shows that, compared with the ESA group, the increase in the percentage of F-cells in patients in the Roxadustat group was significantly higher during the observational period (Table S6). These results indicate that Roxadustat treatment was associated with a modest increase percentage of F-cells in patients with CKD and anemia.

MCHbF

MCHbF quantifies HbF in individual RBCs and provides a more accurate assessment of the ability of newborn RBCs to synthesize HbF than the percentage of F-cells. Additionally, MCHbF is less affected by factors such as blood dilution from transfusions, fluid infusions, and dialysis, thereby enhancing its utility in monitoring drug efficacy. Therefore, we quantified HbF levels in these patients using the MCHbF method. It also showed a significant increase of median MCHbF from baseline to week 12 (baseline: 0.21 pg/RBC; week 4: 0.67 pg/RBC; week 8: 0.79 pg/RBC; week 12: 1.13 pg/RBC) in the Roxadustat group patients, however, the MCHbF levels in the ESA group patients remained unchanged during the observation period (Fig. 2D). We then performed a curve fitting analysis and found a similar trajectory to the percentage of F-cells in the Roxadustat and ESA groups (Fig. 2E). The increase in MCHbF levels in Roxadustat-treated patients was significantly higher than that in ESA-treated patients at weeks 4, 8, and 12 (Table S7). By week 4, a substantial difference in the increase in MCHbF levels emerged between the groups, with progressively strengthened statistical significance over time (Fig. 2F). These findings suggested that Roxadustat may enhance HbF synthesis moderately in the erythrocytes of patients with CKD and anemia.

Response rate and influencing factors

There were 10, 14, and 17 patients in the Roxadustat group who were examined for F-cell percentage at weeks 4, 8, and 12 (Table S1), respectively. Notably, the response to Roxadustat varied among patients. Initially, after 4-, 8-, and 12-week medication with Roxadustat, 60% (6/10), 71.43% (10/14), and 94.12% (16/17) of patients had F-cell percentages > 2% in all RBCs, respectively (Table S1). Notably, at week 12, there were nine patients (52.94%) who achieved F-cell percentages > 4% (Table S1). In addition, the proportions of patients with >1% increase in F-cell percentage were 50% (5/10), 64.29% (9/14), and 82.35% (14/17), at weeks 4, 8, and 12, respectively (Fig. 3A). Also, 52.94% (9/17) of the Roxadustat-treated patients achieved an increase in MCHbF > 1 pg/RBC (Fig. 3B).

Fig. 3.

Increase in HbF levels in every subject post Roxadustat treatment. Waterfall-style bars display the increase in F-cells percentage (A) and MCHbF (B) at 4-, 8-, and 12-weeks post Roxadustat treatment. Subject units are arranged along the x-axis; the dark gray, light gray, and yellow denote weeks 4, 8, and 12, respectively

The LMMs were employed to account for initial imbalances in the cohort and to control for the potential confounding effects of baseline characteristics on treatment outcomes. By including baseline characteristics as covariates, the model statistically adjusted for these factors, ensuring that the estimated treatment effects were not driven by confounding variables. Specifically, LMMs were fitted with treatment group, follow-up time, and their interaction as regression terms, along with a random intercept for each patient to account for within-patient correlation across repeated measurements. As shown in Tables 2 and 3, the anti-anemia treatment regimen was the primary factor influencing HbF levels. The percentage of F-cells increased 0.27% faster every 4 weeks in the Roxadustat group (β = 0.27, p < 0.05) (Table 2). Variance component estimates from the random-intercept model indicated substantial within-patient correlation, with an ICC of 0.653. This suggests that approximately 65.3% of the total residual variance in HbF was attributable to between-patient heterogeneity, while 34.7% reflected within-patient variability and measurement error (random intercept variance τ2 = 1.49). And MCHbF increased 0.08 pg/RBC faster every 4 weeks (β = 0.08, p < 0.001) (Table 3). The ICC for the MCHbF model was approximately 0.64, reflecting substantial baseline heterogeneity among patients (random intercept variance τ2 = 0.12). In addition, a significant positive correlation between baseline HGB and MCHbF (β = 0.01, p < 0.05) was found, and each 1 g/L increase in HGB corresponded to a 0.008 pg/RBC rise in MCHbF (Table 3). Concurrently, a positive association trend between HGB and F-cell percentage was noted (β = 0.02, p = 0.06) (Table 2), suggesting that higher baseline HGB may predict greater longitudinal increases in HbF levels, with statistical significance likely achievable through cohort expansion. Notably, a potential negative association was observed between age and percentage of F-cells (β = −0.04, p = 0.06) (Table 2), suggesting that older age may correlate with reduced HbF synthesis capacity; however, this trend was not observed in MCHbF (β = −0.01, p = 0.11) (Table 3). Other covariates showed no significant associations with longitudinal changes in HbF levels.

Table 2.

Influencing factors of F-cells percentage

Effect	Term	Estimate*	Std. error	95% CI-low	95% CI-high	Statistic	P-value
fixed	(Intercept)	1.45	1.71	−1.97	4.87	0.85	0.40
fixed	Weeks	0.00	0.04	−0.08	0.08	−0.09	0.93
fixed	Treatment drug	−0.11	0.77	−1.64	1.42	−0.14	0.89
fixed	HGB	0.02	0.01	0.00	0.05	1.95	0.06
fixed	WBC	−0.04	0.06	−0.17	0.09	−0.64	0.52
fixed	PLT	0.00	0.00	−0.01	0.01	−0.04	0.96
fixed	SEX	0.68	0.46	−0.24	1.59	1.48	0.14
fixed	Grade	0.53	0.36	−0.20	1.27	1.47	0.15
fixed	CKD stage	−0.24	0.30	−0.84	0.36	−0.81	0.42
fixed	Dialysis	−0.32	0.45	−1.23	0.58	−0.71	0.48
fixed	Age	−0.04	0.02	−0.07	0.00	−1.95	0.06
fixed	HbF baseline	0.09	0.16	−0.24	0.41	0.52	0.60
fixed	Weeks and treatment drug	0.27	0.07	0.13	0.41	3.81	0.00

^*Estimated values correspond to β coefficients, β coefficients represent the change per 4-week treatment interval

Table 3.

Influencing factors of MCHbF

Effect	Term	Estimate*	Std. error	95% CI-low	95% CI-high	Statistic	P-value
fixed	(Intercept)	0.32	0.49	−0.66	1.31	0.66	0.51
fixed	Weeks	0.00	0.01	−0.03	0.02	−0.22	0.83
fixed	Treatment drug	−0.05	0.22	−0.49	0.38	−0.25	0.80
fixed	HGB	0.01	0.00	0.00	0.02	2.31	0.03
fixed	WBC	−0.01	0.02	−0.05	0.02	−0.75	0.46
fixed	PLT	0.00	0.00	0.00	0.00	−0.55	0.58
fixed	SEX	0.22	0.13	−0.03	0.48	1.75	0.09
fixed	Grade	0.17	0.10	−0.04	0.38	1.59	0.12
fixed	CKD stage	−0.04	0.08	−0.21	0.13	−0.49	0.63
fixed	Dialysis	−0.14	0.13	−0.39	0.12	−1.06	0.29
fixed	Age	−0.01	0.01	−0.02	0.00	−1.64	0.11
fixed	MCHbF baseline	0.05	0.15	−0.24	0.35	0.36	0.72
fixed	Weeks and treatment drug	0.08	0.02	0.04	0.12	3.77	0.00

^*Estimated values correspond to β coefficients, β coefficients represent the change per 4-week treatment interval

Subgroup analysis

To further validate the positive correlation between baseline HGB and post-treatment HbF levels, especially MCHbF, we conducted subgroup analyses stratified by anemia severity (mild: 90–120 g/L; moderate-to-severe: 30–89 g/L) within the Roxadustat group. The results revealed that a significant increase in HbF levels after medication was achieved earlier in patients with mild anemia (week 4) than in those with moderate-to-severe anemia (weeks 8 or 12) (Fig. S3). Consistent with the LMM results, this temporal pattern was particularly evident for MCHbF, confirming its strong association with baseline HGB.

Adverse events

Safety assessment was event-driven, focusing specifically on adverse events leading to treatment discontinuation or modification and severe cardiovascular events requiring hospitalization. Routine physiological parameters (e.g., blood pressure, liver function, or thrombotic markers) were not systematically monitored. In the Roxadustat group, one patient discontinued therapy due to the development of symptomatic hypothyroidism. Following the cessation of Roxadustat and the initiation of short-term levothyroxine replacement, thyroid function normalized. No other patients in either group exhibited clinical symptoms suggestive of overt hypothyroidism, nor were there any hospital admissions due to cardiovascular or thrombotic events.

Discussions

To our knowledge, this is the first clinical study of Roxadustat in elevating HbF levels in humans, which aligns with earlier pre-clinical evidence in murine sickle cell models, FG-4592 increased γ-globin mRNA and raised the percentage of F-cells by 1.47 ± 0.45 folds [15]. Another animal study using rhesus monkeys also demonstrated that the HIF-PHI FG-2216 elevated the percentage of F-cells from 0.7 ± 0.2% to 3.3 ± 1.7% [16]. While providing initial clinical proof-of-concept, our findings are primarily hypothesis-generating regarding the precise in vivo mechanisms of HIF stabilization on HbF synthesis in humans. Nevertheless, collectively, these data support the hypothesis that Roxadustat-mediated HIF pathway activation may serve as a potential therapeutic strategy for hemoglobinopathies, warranting dedicated future investigation.

Roxadustat inhibits HIF prolyl hydroxylase, a mechanism that theoretically circumvents the myelosuppression frequently observed with chemotherapy-based agents. In contrast to well-characterized HbF modulators such as hydroxyurea, decitabine, and thalidomide, Roxadustat exhibited no reports of severe adverse reactions common to those agents, including myelosuppression, reproductive toxicity, neurotoxicity, or metabolic disturbances [24–31]. While systematic biochemical screening was not performed, routine clinical follow-up revealed no widespread severe adverse events. Consistent with reports of reversible hypothyroidism in larger CKD trials [32, 33], one participant in our study was diagnosed with hypothyroidism after presenting with clinical symptoms, leading to treatment discontinuation. The participant received thyroid hormone replacement therapy, and thyroid function recovered. In contrast to invasive and complex modalities aiming for HbF reactivation or cellular replacement, including lentiviral vector-mediated BCL11A editing (NCT03282656) [34, 35] and allogeneic hematopoietic stem cell transplantation [36, 37], Roxadustat may serve as a more accessible HbF-inducing agent. It avoids the prohibitive costs and donor constraints of these complex therapies, offering a pragmatic solution with improved accessibility.

Although this study was conducted in patients with CKD and concomitant anemia rather than patients with hemoglobinopathies, and the observed median increments—3.22% in F-cells percentage and 1.05 pg/RBC in MCHbF—remain below the ≥12% F-cells or ≥ 4 pg/RBC values generally considered clinically meaningful for β-thalassemia or sickle-cell disease, we still regard Roxadustat as a promising HbF inducer. Although modest HbF induction is unlikely to be sufficient as monotherapy, its oral availability and non-myelosuppressive profile make it an attractive adjunctive agent for combination regimens. Strategies combining Roxadustat with established hemoglobinopathies treatment agents such as hydroxyurea or pomalidomide have attracted particular attention. Experimental evidence in vitro confirmed that the co-administration of Roxadustat and hydroxyurea synergistically elevates HbF levels more than hydroxyurea alone [14]. However, no clinical trials have investigated the efficacy and safety of this combination therapy. Two ongoing clinical trials are evaluating Roxadustat-based combination therapies: luspatercept for refractory MDS (NCT06006949) and rapamycin for pure red cell aplasia (ChiCTR2400088827). These studies demonstrated the superior efficacy of the combination over the respective monotherapies and provided additional translational support for the feasibility of Roxadustat-hydroxyurea co-administration.

Several limitations of the present study should be acknowledged. First, the study was limited by its short duration, relatively small sample size, and the absence of a randomized control group. Consequently, the findings regarding HbF induction in anemic CKD patients are preliminary and cannot be directly extrapolated to patients with hemoglobinopathies without further validation in dedicated clinical trials. Second, the prolonged 4-week follow-up intervals coupled with premature 12-week termination obscured the HbF induction dynamics of Roxadustat, precluding plateau identification, although LMM analysis confirmed sustained elevation. Third, potential selection bias regarding treatment allocation represents a notable limitation of this observational study. Specifically, the initially enrolled Roxadustat group presented with significantly more severe anemia (lower HGB and higher grade) compared to the ESA group (Table S2). Additionally, the study observed a loss to follow-up rate exceeding 30%, which reduced the final analytical sample to 52 patients and compromised the representativeness of the cohort. Based on available records, this high attrition may be attributed to economic factors, as Roxadustat is a relatively newer and costlier agent than traditional ESAs, and geographical distance, making regular visits to the study center logistically challenging. Crucially, no evidence indicated that these exclusions were driven by adverse events. While the final cohort exhibited balanced characteristics, this substantial attrition combined with the initial imbalance, suggests potential confounding by indication, where physicians preferentially prescribed Roxadustat to patients with refractory or more severe anemia. Although we employed LMM analysis to statistically adjust for baseline characteristics, residual confounding from unmeasured covariates cannot be fully ruled out, which may influence the comparative effectiveness estimates. Fourth, LMM and subgroup analyses indicated a potential HGB-HbF association based solely on current data but did not provide conclusive evidence. We did not explore the factors affecting HbF levels in detail. We did not assess the expression of γ-globin regulatory genes (e.g., BCL11A, KLF1, HBS1L-MYB), which are known modulators of hydroxyurea responsiveness [38, 39]. Hence, we could not determine whether these polymorphisms contributed to individual differences in responses to Roxadustat treatment. Finally, safety monitoring was restricted to passive surveillance based on patient self-reports and routine medical records, precluding systematic assessment. Consequently, the evaluation of key safety parameters, specifically blood pressure, thrombosis, cardiovascular events, hospitalizations, and liver and thyroid function, was not performed at standardized intervals but was driven solely by clinical indication. Notably, routine thyroid screening was not conducted for the majority of participants. As subclinical or asymptomatic abnormalities were likely uncaptured, our findings reflect only clinically evident events rather than a comprehensive safety profile.

Future studies should quantify the HbF induction capacity of Roxadustat beyond the CKD cohort, such as hemoglobinopathy cohorts through dose-escalation trials, establish synergistic dosing thresholds with standard therapies (e.g., hydroxyurea), and prioritize defining minimal effective doses for combinatorial regimens to optimize clinical benefit-toxicity profiles. Furthermore, we aim to implement single-nucleotide polymorphism genotyping of γ-globin regulatory genes to elucidate genetic determinants underlying treatment response heterogeneity. This pharmacogenomic strategy will enable clinical trial designers to prescreen participants for Roxadustat adjunctive therapy, thereby facilitating precise dosing optimization based on individual genetic profiles, particularly through genotype-guided regimens.

Conclusion

In conclusion, this 12-week prospective observational study suggested that Roxadustat modestly induces HbF synthesis in patients with anemic CKD. Additionally, progressive MCHbF accumulation confirmed sustained modulation of erythroid precursors. These findings should be viewed as hypothesis-generating rather than definitive proof of efficacy. They provide a rationale for exploring HIF-PHIs as potential HbF-inducing agents, though their safety profile requires confirmation in rigorous prospective clinical trials. We hypothesize that Roxadustat may serve as an adjunct to HU to enhance HbF induction, potentially mitigating the toxicity associated with high-dose HU monotherapy. However, safety data cannot be directly extrapolated from this CKD cohort. Therefore, prospective clinical trials in patients with sickle cell disease and thalassemia are warranted. Future research should prioritize assessing the safety and efficacy of HIF-PHI/HU combination therapy as a primary endpoint, while secondary endpoints should focus on determining the optimal HIF-PHI dosage for HbF induction. Furthermore, integrating γ-globin promoter polymorphism profiling could help elucidate the pharmacogenetic determinants of HbF modulation, thereby facilitating the development of genotype-specific therapeutic strategies.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

HIF

Hypoxia-inducible factor

PHD

Prolyl hydroxylase

PHI

Prolyl hydroxylase inhibitor

HbF

Fetal hemoglobin

MCHbF

Mean corpuscular fetal hemoglobin

ESA

Erythropoiesis-stimulating agent

LMM

Linear mixed model

ICC

Intraclass correlation coefficient

Author contributions

Conceptualization, Q.R., Z.J.L., and L.C.; methodology, W.G., M.C., Z.Y.L., C.T., L.H., X.W., Y.J., T.Y., X.L., X.Y., Y.X. and L.X.; validation, W.G., C.T., Y.J., and X.L.; formal analysis, W.G., C.T., L.H., and X.W.; investigation, W.G., M.C., Z.Y.L. and X.Y.; resources, W.G., Z.Y.L. and W.D.; data curation, Q.R., T.Y.; writing—original draft preparation, W.G., M.C. and Z.Y.L.; writing—review and editing, M.C., Q.R., Z.J.L. and L.C.; visualization, M.C., Z.Y.L. and Y.J.; supervision, L.X. and Z.J.L.; project administration, L.X. and Q.R.; funding acquisition, L.C. and M.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by grants from the National Key Research and Development Program of China (2024YFA1107103), Medical Basic Research Innovation Center of the Ministry of Education for Myeloid Acute Radiation Syndrome (ARSBIC-B-202402), Chongqing Science and Health Joint Medical Scientific Research Project (2025MSXM098), Science and Technology Research Program of the Chongqing Education Commission (KJZD-M202312802), and Chongqing Natural Science Foundation (CSTB2022NSCQ-MSX0190).

Data availability

The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.

Materials availability

Not applicable.

Declarations

Ethics approval and consent to participate

The study was conducted according to the principles of the Declaration of Helsinki. The study protocol was approved by the Medical Ethics Committees of the Second Affiliated Hospital of the Army Medical University (Ref: 2024-Yan-308–01) and the Chongqing Hospital of PAP (Ref: WJYY2024-1210). The informed consent forms were obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.