Roxadustat for Treating Anemia in Patients with CKD Not on Dialysis: Results from a Randomized Phase 3 Study

Abstract

Background 

Current anemia therapies for patients with non–dialysis-dependent CKD may require injection and medical visits. Roxadustat, an oral hypoxia-inducible factor prolyl hydroxylase inhibitor, stimulates erythropoiesis and improves iron homeostasis.

Methods 

In this double-blind phase 3 study, we randomized patients with non–dialysis-dependent CKD stages 3–5 and hemoglobin <10.0 g/dl (1:1) to thrice-weekly 70-mg oral roxadustat or placebo. Doses were titrated throughout the study based on hemoglobin levels. The primary efficacy end point was mean change from baseline in hemoglobin averaged over weeks 28–52 versus placebo, irrespective of rescue therapy use. We assessed patients for adverse events.

Results 

The study included 2781 patients, 1393 who received roxadustat and 1388 who received placebo. Mean baseline hemoglobin was 9.1 g/dl for both groups. The mean change in hemoglobin from baseline was 1.75 g/dl (95% confidence interval [95% CI], 1.68 to 1.81) with roxadustat versus 0.40 g/dl (95% CI, 0.33 to 0.47) with placebo, (P<0.001). Among 411 patients with baseline elevated high-sensitivity C-reactive protein, mean change in hemoglobin from baseline was 1.75 g/dl (95% CI, 1.58 to 1.92) with roxadustat versus 0.62 g/dl (95% CI, 0.44 to 0.80) with placebo, (P<0.001). Roxadustat reduced the risk of red blood cell transfusion by 63% (hazard ratio, 0.37; 95% CI, 0.30 to 0.44). The most common adverse events with roxadustat and placebo, respectively, were ESKD (21.0% versus 20.5%), urinary tract infection (12.8% versus 8.0%), pneumonia (11.9% versus 9.4%), and hypertension (11.5% versus 9.1%).

Conclusions 

Roxadustat effectively increased hemoglobin in patients with non–dialysis-dependent CKD and reduced the need for red blood cell transfusion, with an adverse event profile comparable to that of placebo.

Clinical Trial registry name and registration number: 

Safety and Efficacy Study of Roxadustat to Treat Anemia in Patients With CKD, Not on Dialysis, NCT02174627

Visual Abstract

Anemia is frequently experienced by patients with non–dialysis-dependent CKD (NDD-CKD) and its prevalence increases with advancing CKD stage.^1–5 Anemia of CKD is associated with reduced quality of life and an increased risk of mortality and hospitalization.^6–11

The recommended therapies for anemia in patients with NDD-CKD include erythropoiesis-stimulating agents (ESAs), oral or intravenous (iv) iron, and red blood cell (RBC) transfusion.¹² However, these therapies may require injection and medical office visits, and may be associated with adverse outcomes.^12–15 RBC transfusion is associated with risks including infection and allosensitization,^15,16 and costs; thus, a treatment objective for anemia is to avoid RBC transfusion.^12,17,18 ESAs are an established therapy for anemia in patients with NDD-CKD, and although effective at increasing hemoglobin (Hb) and reducing the need for RBC transfusion, safety concerns noted with ESAs have stimulated the development of alternative treatments for anemia.^19–21 According to US product labels and international clinical practice guidelines, ESAs can be considered in patients with Hb <10 g/dl at risk of RBC transfusions.^12,17,18 ESA US product labels state that dosing should be reduced or interrupted when Hb exceeds 10 g/dl, and that supplemental iron should be administered among patients with ferritin <100 μg/L or transferrin saturation (TSAT) <20%.^22,23 Oral iron, however, is associated with gastrointestinal side effects and barriers exist to administration of iv iron in NDD-CKD.^13,24 These limitations of established therapies may contribute to estimates from international analyses that approximately 50%–90% of patients with anemia of NDD-CKD are untreated.^{4,5,11,25–30} In particular, analyses estimate that ESA use among patients with anemia of NDD-CKD in the real world ranges from 11% to 35%.^11,28–30 Accordingly, research has focused on developing alternative therapies for patients with anemia of NDD-CKD on the basis of the physiologic pathways of erythropoiesis.

Hypoxia-inducible factor prolyl hydroxylase (HIF-PH) inhibitors are a new class of drug for anemia of CKD following recent Nobel Prize–winning research delineating response mechanisms to hypoxia.³¹ Roxadustat was the first-in-class HIF-PH inhibitor to be approved by a regulatory agency to treat anemia in patients with dialysis-dependent CKD in Japan and China and in patients with NDD-CKD in China.³² Hypoxia-inducible factor (HIF) is a transcription factor that regulates expression of genes stimulating erythropoiesis. In normoxia, HIF-PH is active and promotes degradation of HIF-α.^33–35 Roxadustat reversibly stabilizes HIF-α, which promotes a coordinated erythropoietic response by activating a number of genes, including those encoding endogenous erythropoietin, enzymes of heme biosynthesis, and proteins that promote bone marrow iron availability via improved absorption and transport.^35,36 HIF-PH inhibitors are taken orally and routinely self-administered. In phase 2 and smaller, single-nation phase 3 studies among patients with NDD-CKD and anemia, roxadustat increased and maintained Hb levels,^37–39 with similar adverse events (AEs) compared with placebo.^37,38 The roxadustat clinical development program comprises both placebo- and active-controlled phase 3 clinical studies designed to assess the efficacy and safety of roxadustat in NDD-CKD. Here, we report the results of OLYMPUS, a global phase 3 study of the efficacy and safety of roxadustat in 2781 patients with anemia of NDD-CKD.

Methods

Trial Design and Oversight

This phase 3, multicenter, randomized, double-blind, placebo-controlled study (NCT02174627) was performed in 385 centers across 25 countries worldwide. The study comprised a screening period of up to 6 weeks, a treatment period of variable duration up to 4 years, and an end of study (EOS) follow-up visit (Figure 1). Treatment end date was defined on the basis of accrual of a predefined number of adjudicated cardiovascular events for a separate pooled analysis (to provide adequate power) of three phase 3, placebo-controlled roxadustat studies in NDD-CKD conducted by different sponsors (NCT02174627, NCT01887600, NCT01750190).

Figure 1.: Design of the OLYMPUS study. A minimum of two screening visits were performed at least 7 days apart. The Hb inclusion criteria were on the basis of central laboratory assessment and could be repeated during the screening period. The treatment period began on the first day of dosing with the study drug (day 1, week 0). Patients were contacted by telephone at week 1, and attended study visits every 2 weeks from week 0 to 20. After week 20, study visits occurred every 4 weeks until week 52, and then every 8 weeks until EOT. The next scheduled visit, the EOT visit, occurred as soon as possible after the target number of cardiovascular events was accrued. Treatment duration was variable for individual patients (estimated duration up to 4 years). For patients without premature study drug discontinuation, an EOS visit was performed 4 weeks after the treatment period. CV, cardiovascular; R, randomization; TIW, three times weekly; wk, week(s).

The final study protocol and informed consent form, and any amendments to them, were approved by an Independent Ethics Committee or Institutional Review Board in each participating center. All patients provided written, informed consent. This study was performed in accordance with the ethical principles of the Declaration of Helsinki and the International Council for Harmonisation Good Clinical Practice (GCP).

Patients

Eligible patients were aged ≥18 years; had eGFR <60 ml/min per 1.73 m², corresponding to stages 3–5 CKD; were not on dialysis; and had a mean of two values of Hb <10 g/dl measured at least 7 days apart during the screening period; and ferritin ≥50 μg/L and TSAT ≥15%. Full inclusion and exclusion criteria are provided in the Supplemental Material.

Treatment

Eligible patients were randomly assigned in a 1:1 ratio to receive roxadustat or placebo in a double-blind manner, which was performed centrally using an Interactive Web Response System/Interactive Voice Response System. Randomization codes were computer generated by AstraZeneca Research & Development using GRand and a block randomization schedule comprising a block size of 4. Randomization was stratified by country to ensure a balanced global distribution. The investigator, study site staff, and patient were blinded to study treatment, but not to dose or dosing frequency. The Sponsor and designees except personnel analyzing pharmacokinetic samples were blinded to study treatment, dose and dosing frequency, and Hb values. To ensure patient safety, study personnel at local sites could not be blinded to Hb values.

The starting dose of roxadustat and placebo was 70 mg three times weekly administered orally (tablet) at least 2 days apart and no more than 4 days apart (e.g., Monday, Wednesday, Friday). Dosing was titrated as needed every 4 weeks until week 52 and every 8 weeks thereafter to achieve and maintain Hb at 11±1 g/dl. The maximum dose was capped at the lower of 3.0 mg/kg or 300 mg per dose administration. The dosing algorithm used to maintain Hb is described in the Supplemental Material. Prohibited medications used during the study are listed in the Supplemental Material. iv iron, ESA, and RBC transfusions were characterized as rescue therapy for patients with low Hb values and inadequate response to study drug, and criteria for these rescue therapies were delineated in the protocol (Supplemental Material). iv iron was permitted in patients who were intolerant or unresponsive to oral iron and who had Hb <8.5 g/dl and ferritin <100 μg/L or TSAT <20%. RBC transfusion was allowed if considered to be a medical necessity by the investigator. An ESA was permitted in patients if all of the following criteria were met: the patient had not sufficiently responded to two or more dose increases or the maximum dose limit of the study drug had been reached; the patient had Hb <8.0 g/dl; the patient did not have iron deficiency or bleeding resulting in the lack of response or decline in Hb; and there was a need to reduce the risk of alloimmunization and/or other risks associated with RBC transfusion (Supplemental Material). Treatment with an ESA was stopped when Hb reached >9 g/dl or after 4 weeks, whichever came first. Patients could receive two cycles of rescue with ESA; if they received two cycles and required ESA rescue for a third time, study drug was permanently discontinued. In patients who initiated dialysis after randomization and required an ESA, study drug was permanently discontinued.

Patients were to continue study drug until the EOS, unless there was a protocol-defined reason for study drug discontinuation, including: patient decision, an AE that the investigator thought put the patient at undue risk, severe noncompliance with the study protocol as determined by the investigator that could affect the validity of the data, pregnancy, requirement for a third cycle of rescue therapy with an ESA, and initiation of dialysis during the study for those with a need for rescue therapy with an ESA. Patients who discontinued study drug continued conventional anemia treatments deemed necessary, and study assessments or follow-up were continued whenever possible, until the EOS or study withdrawal. Patients who discontinued study drug and remained in the study could continue to undergo prespecified in-person study visits or modified follow-up, such as in-person study visits occurring at different frequencies and/or with fewer study procedures (e.g., without study laboratory assessments), or periodic telephone contact.

Visit and Measurement Schedule

Most data including serum creatinine were collected at baseline, every 4 weeks until week 52, and every 8 weeks thereafter (Figure 1). Short Form–36 (SF-36) measurements were collected at randomization and weeks 12, 28, and 52.

End Points

The primary efficacy end point was the mean change from baseline in Hb averaged over weeks 28–52, regardless of rescue therapy use. Secondary efficacy end points were assessed using a fixed sequence approach, in the following order: (1) proportion of patients achieving Hb response in the first 24 weeks without having received rescue therapy, defined as Hb ≥11.0 g/dl and increase from baseline of ≥1.0 g/dl (if >8.0 g/dl at baseline) or Hb increase from baseline of ≥2.0 g/dl (if ≤8.0 g/dl at baseline); (2) mean change from baseline in Hb averaged over weeks 28–52 among patients with baseline high-sensitivity C-reactive protein (hsCRP; assessed among patients who consented to biobank samples) greater than the upper limit of normal (ULN; 5 mg/L); (3) proportion of total time of interpolated Hb values ≥10 g/dl from weeks 28–52; (4) proportion of total time of interpolated Hb values within 10–12 g/dl from weeks 28–52; (5) mean change from baseline in serum LDL cholesterol (LDL-C) to week 24; (6) need for first rescue therapy (composite) of any of iv iron, RBC transfusion, or ESA; (7) need for first RBC transfusion as rescue therapy; (8) mean change from baseline in SF-36 Vitality subscore averaged over weeks 12–28; (9) annual rate of change in eGFR starting from week 4, using all postbaseline eGFR values before initiation of dialysis/kidney transplant; and (10) mean change from baseline in SF-36 Physical Functioning subscore averaged over weeks 12–28.

Key exploratory efficacy end points included: change in hepcidin (assessed from available biobank samples) from baseline to week 24; change from baseline in serum iron profiles (iron, ferritin, total iron binding capacity [TIBC], and TSAT) averaged over week 24 to end of treatment (EOT); and change from baseline in sitting systolic BP (SBP) and diastolic BP (DBP). TSAT was calculated as: TSAT (%) = (serum iron level × 100)/(TIBC).

The primary safety objective was to contribute adjudicated cardiovascular safety data to a separate pooled safety analysis across the NDD-CKD studies in the roxadustat phase 3 program. Safety analyses included assessment of the incidence of AEs. AEs were coded using the Medical Dictionary for Regulatory Activities version 20.0. For patients who withdrew consent, public record searches were used to confirm vital status (alive or dead), as appropriate, in accordance with local regulations.

Statistical Analyses

To contribute an adequate number of adjudicated cardiovascular safety events to the separate pooled analysis of phase 3 NDD-CKD studies (including approximately 4000 patients overall), approximately 2600 patients were planned to be randomized in OLYMPUS. The intent-to-treat (ITT) population comprised all patients who were randomized to study treatment throughout the duration of the study, irrespective of their protocol adherence and continued participation in the study. The full analysis set (FAS) included patients from the ITT analysis set who received study drug and had baseline Hb and ≥1 postdose Hb assessments. The on-treatment (OT)+28 analysis set comprised all patients who received ≥1 dose of randomized study treatment and was censored 28 days after last intake of study treatment.

The primary efficacy end point was analyzed in the ITT analysis set using missing at random–based multiple imputation analysis of covariance (ANCOVA), containing terms for treatment, baseline Hb measurement, baseline eGFR, geographic region, and cardiovascular history (Supplemental Material). Superiority of roxadustat compared with placebo was declared if the lower bound of the two-sided 95% confidence interval (95% CI) of the difference between roxadustat and placebo exceeded 0 g/dl. Overall, 14.3% and 21.2% of roxadustat and placebo datapoints, respectively, were imputed for the primary efficacy analysis of mean change from baseline in Hb (ITT analysis set) and 200 datasets with simulated data were generated accordingly (Supplemental Material). Subgroup analyses were also performed using the ITT analysis set (Supplemental Material).

For the secondary efficacy end points, a fixed sequence approach adjusted for multiple testing; formal statistical hypothesis testing was stopped once a test was accompanied by a P value ≥0.05. Secondary and key exploratory efficacy end points were analyzed using Cochran–Mantel–Haenszel tests, Cox proportional hazard regression models, ANCOVA, and mixed-effect model repeated measures; full details of the statistical methods and analysis sets are listed in Supplemental Tables 1 and 2. Need for first rescue therapy end points are presented as the number (%) of patients with the event and event rates per 100 patient-years at risk (total number of years at risk).

Details of the assessment of treatment compliance are provided in the Supplemental Material.

AEs were analyzed using the ITT analysis set, with patients censored at their individual EOS visit regardless of study drug discontinuation, at the date of withdrawal of consent or last study contact if the patient withdrew consent, or at the date of death, whichever was earliest. Patients with more than one event for the same AE category/preferred term were counted once in that category/preferred term. The AE exposure-adjusted event rate (per 100 patient-years) was calculated as follows:

[Number of patients with AEs/(the total number of days at risk for that AE across all patients in given group/365.25)] × 100.

All data were analyzed using SAS software version 9.4 (SAS Institute, Cary, NC).

Results

Patients

Overall, 5222 patients were screened. Of these, 2441 (46.7%) were not randomized, most commonly due to patients not meeting all inclusion/exclusion criteria (n=2336 [95.7%]; Figure 2). From June 26, 2014 to May 11, 2017, 2781 patients were randomized to roxadustat (n=1393) or placebo (n=1388). Twenty patients were excluded due to incorrect randomization (n=4) or significant GCP violations (n=16) in obtaining or recording the data that might affect the validity of the data; therefore, the ITT population comprised 1384 and 1377 patients receiving roxadustat and placebo, respectively (Figure 2). The FAS included 2728 patients (roxadustat n=1371; placebo n=1357) and the OT+28 population included 2760 patients (roxadustat n=1384; placebo n=1376). The predominant reason for exclusion from the FAS was the absence of postdose Hb data (n=32).

Figure 2.: Patient disposition. ^aBecause of GCP violations. Study-specific discontinuation criteria included patients who required dialysis initiation and ESA rescue therapy. When possible, patients who discontinued treatment were followed for concomitant medications, AEs, vital status, and hospitalization.

Baseline demographic and clinical characteristics were similar between the treatment groups (Table 1). Mean baseline Hb overall was 9.1 g/dl and mean eGFR was 19.7 and 20.0 ml/min per 1.73 m² in the roxadustat and placebo groups, respectively. At baseline, approximately 58% of patients had ferritin >100 μg/L and TSAT >20% (Table 1). Iron supplements were taken by 1777 of 2761 (64.4%) patients during the study, including oral bivalent (n=1283 of 2761 [46.5%]) and oral trivalent (n=188 of 2761 [6.8%]) iron (Supplemental Table 3).

Table 1. - Baseline demographic and clinical characteristics (ITT analysis set)

Characteristic	Roxadustat (n=1384)	Placebo (n=1377)
Age, yr
Mean (SD)	60.9 (14.7)	62.4 (14.1)
Sex, n (%)
Female	820 (59.2)	774 (56.2)
Race, n (%)
White	623 (45.0)	611 (44.4)
Black	112 (8.1)	115 (8.4)
Asian	544 (39.3)	538 (39.1)
Native Hawaiian or Pacific Islander	0	2 (0.1)
American Indian or Alaska Native	24 (1.7)	29 (2.1)
Other	81 (5.9)	82 (6.0)
Geographic region, n (%)
USA	343 (24.8)	340 (24.7)
Canada	26 (1.9)	24 (1.7)
Latin America	206 (14.9)	205 (14.9)
Asia	522 (37.7)	521 (37.8)
Europe	287 (20.7)	287 (20.8)
Weight, kg
Mean (SD)	69.9 (18.5)	70.6 (18.8)
BMI, kg/m2
Mean (SD)	26.7 (6.0)	26.9 (6.1)
Comorbidities, n (%)
Hypertension	1274 (92.1)	1280 (93.0)
Type 2 diabetes mellitus	737 (53.3)	771 (56.0)
Dyslipidemia	683 (49.3)	691 (50.2)
Coronary artery disease	160 (11.6)	178 (12.9)
Cardiac failure congestive	151 (10.9)	155 (11.3)
Hb, g/dl
Mean (SD)	9.1 (0.7)	9.1 (0.7)
Hb, g/dl, n (%)
≤8	129 (9.3)	131 (9.5)
>8 to ≤9	386 (27.9)	402 (29.2)
>9	869 (62.8)	844 (61.3)
eGFR a , ml/min per 1.73 m2
Mean (SD)	19.7 (11.7)	20.0 (11.7)
eGFR, ml/min per 1.73 m2, n (%)
<10	291 (21.0)	283 (20.6)
10 to <15	300 (21.7)	315 (22.9)
15 to <30	534 (38.6)	520 (37.8)
30 to <45	201 (14.5)	196 (14.2)
45 to <60	55 (4.0)	59 (4.3)
≥60	3 (0.2)	4 (0.3)
LDL-C, mg/dl
Mean (SD)	94.4 (43.4)	92.4 (42.0)
Ferritin >100 μg/L and TSAT >20%, n (%)	809 (58.5)	799 (58.0)
hsCRP, mg/dl
Mean b (SD)	0.7 (1.5)	0.7 (1.8)
hsCRP >ULN b , n (%)	227 (16.4)	209 (15.2)
SBP, mmHg
Mean (SD)	134.4 (13.3)	135.5 (12.7)
DBP, mmHg
Mean (SD)	74.5 (9.1)	74.1 (9.3)
Most likely cause of CKD, n (%) c
Diabetic nephropathy	614 (44.9)	602 (44.2)
Ischemic/hypertensive nephropathy	207 (15.2)	192 (14.1)
Chronic GN	169 (12.4)	155 (11.4)
Other primary or secondary GN	75 (5.5)	82 (6.0)
Cystic kidney disease	74 (5.4)	62 (4.6)
Chronic interstitial nephritis	36 (2.6)	27 (2.0)
Chronic pyelonephritis (infectious)	30 (2.2)	44 (3.2)
FSGS	28 (2.0)	17 (1.2)
IgA nephropathy	19 (1.4)	23 (1.7)
Obstructive nephropathy	17 (1.2)	24 (1.8)
Membranous nephropathy	13 (1.0)	10 (0.7)
Lupus nephritis	7 (0.5)	2 (0.1)
Renal artery stenosis	2 (0.1)	1 (<0.1)
Minimal change	2 (0.1)	6 (0.4)
Obstructive uropathy	1 (<0.1)	0
Not specified	25 (1.8)	15 (1.1)
Unknown	124 (9.1)	137 (10.1)
Other	92 (6.7)	117 (8.6)
Missing	18	16

BMI, body mass index.

^aCalculated using the four-variable Modification of Diet in Renal Disease equation.

^bhsCRP quantified from stored biomarker samples obtained at randomization (n=753 for roxadustat, n=717 for placebo); ULN is 5 mg/l (0.5 mg/dl).

^cPercentages were on the basis of the number of patients with nonmissing data (n=1366 for roxadustat, n=1361 for placebo).

Overall, 1300 patients discontinued study drug prematurely. Fewer patients in the roxadustat group (n=499 of 1384 [36.1%]) discontinued study drug versus the placebo group (n=801 of 1376 [58.2%]); the difference was observed early and maintained throughout the study (Figure 3, Supplemental Figure 1). Post hoc analyses showed a 51% lower risk of study drug discontinuation with roxadustat versus placebo (hazard ratio [HR] 0.49; 95% CI, 0.44 to 0.55; nominal P<0.001). In post hoc analyses, factors associated with study drug discontinuation were baseline eGFR <15 ml/min per 1.73 m² and initiation of dialysis after randomization. Fewer patients in the roxadustat arm discontinued study drug versus placebo among those with baseline eGFR <15 ml/min per 1.73 m² (n=250 of 591 [42.3%] versus n=421 of 597 [70.5%]; HR 0.42; 95% CI, 0.36 to 0.49; nominal P<0.001) and those initiating dialysis while on study drug (n=174 of 457 [38.1%] versus n=185 of 272 [68.0%]; HR 0.36; 95% CI, 0.29 to 0.45; nominal P<0.001). Withdrawal of consent was more common in placebo- than roxadustat-treated patients (n=128 [9.3%] versus n=84 [6.1%], respectively). Vital status at the end of the study was confirmed for 2756 of 2781 patients (99.1%), with ten (0.7%) roxadustat- and 15 (1.0%) placebo-treated patients having unknown vital status at the end of the study.

Figure 3.: Time to premature study drug discontinuation by treatment arm and baseline eGFR <10 ml/min per 1.73 m² (OT+28 analysis set). KM percentages were calculated at 24 months. Permanent discontinuation criteria: patient decision or investigator decision (AE, severe noncompliance, need for three or more cycles of ESA rescue therapy, or initiation of dialysis and need for ESA rescue therapy). Assessment of premature study drug discontinuation by baseline eGFR <10 ml/min per 1.73 m² was performed post hoc. KM, Kaplan–Meier; N, number of patients in that treatment arm.

Dosing

Duration of exposure was longer with roxadustat than placebo (mean [SD] 19.62 [10.39] versus 15.24 [10.48] months, respectively; median [interquartile range] 20.80 [11.38–27.55] versus 14.57 [5.52–23.51] months, respectively). The most common roxadustat doses were in the range of 40–100 mg orally three times weekly (Supplemental Table 4). Mean (SD) weekly dose, calculated as the total cumulative dose/actual treatment duration (in weeks), of roxadustat was 207.6 (116.7) mg, with a median (interquartile range) of 182.3 (124.1–267.3) mg. During treatment, 1204 (87.0%) patients in the roxadustat group compared with 577 (41.9%) patients in the placebo group had at least one dose reduction, whereas approximately 90% of all patients had at least one dose increase (Supplemental Table 5). Compliance (≥75%) to treatment was observed in 1330 (96.1%) patients with roxadustat and 1330 (96.7%) patients with placebo.

Primary Efficacy End Point

The adjusted least-squares mean (LSM) change from baseline in Hb averaged over weeks 28–52 was significantly greater with roxadustat versus placebo (+1.75 g/dl [95% CI, 1.68 to 1.81] versus +0.40 g/dl [95% CI, 0.33 to 0.47], respectively; difference +1.35 g/dl; 95% CI, 1.27 to 1.43; P<0.001) (Figure 4A). An initial separation of Hb levels for roxadustat versus placebo was observed after randomization and sustained until EOT (Figure 4C).

Figure 4.: Hb end points. (A) change from baseline in Hb averaged over weeks 28–52; (B) change from baseline in Hb averaged over weeks 28–52 by baseline iron replete status; (C) mean Hb levels by visit; (D) proportion of patients achieving Hb response; (E) change from baseline in Hb averaged over weeks 28–52 in patients with elevated hsCRP at baseline. Error bars are 95% CIs. (A) ITT analysis set. Change in Hb from baseline to mean value during weeks 28–52 was analyzed using ANCOVA with the following fixed-effect covariates at baseline: Hb, baseline eGFR, cardiovascular/cerebrovascular/thromboembolic history, geographic region (USA versus ex-USA), and treatment arm. Adjusted LSMs, their difference, and corresponding 95% CI were generated from datasets where missing data were imputed using missing at random–based multiple imputation by treatment arm with baseline Hb, baseline eGFR, cardiovascular/cerebrovascular/thromboembolic history, and geographic region (USA versus ex-USA) as predictor variables. Observed values up to the EOT visit if treatment was completed; the EOS visit if patient discontinued treatment; or date of withdrawal of consent, last contact, or death if patient withdrew consent, was lost to follow-up, or died; and imputed values up to death of patient were used to derive the mean from weeks 28 to 52. (B) ITT analysis set. Iron replete defined as ferritin >100 μg/l and TSAT >20%. (C) ITT analysis set. Week 0 on the x axis refers to the baseline value. (D) FAS. Hb response was defined as Hb ≥11.0 g/dl and Hb increase from baseline by ≥1.0 g/dl for patients with baseline Hb >8.0 g/dl, or Hb increase from baseline by ≥2.0 g/dl for those with baseline Hb ≤8.0 g/dl, at two consecutive visits (with available data) separated by at least 5 days during the first 24 weeks of treatment without having received rescue therapy before Hb response. Statistical analysis was on the basis of the Cochran–Mantel–Haenszel test, adjusting for baseline Hb (≤8, >8 g/dl), baseline eGFR (≤30, >30 ml/min per 1.73 m²), cardiovascular/cerebrovascular/thromboembolic history, and geographic region (USA versus ex-USA). Patients who had discontinued study drug or taken rescue medication before response were considered nonresponders. (E) ITT analysis set. hsCRP was quantified from available stored biomarker samples obtained at randomization. ULN is 5 mg/l (0.5 mg/dl). Data were analyzed analogously to the primary analysis.

Secondary Efficacy End Points

A significantly greater proportion of patients receiving roxadustat (77.0%) achieved an Hb response versus placebo (8.5%) during the first 24 weeks of treatment (relative risk 9.12; 95% CI, 7.63 to 10.89; P<0.001) (Figure 4D). Among 411 patients with baseline hsCRP >ULN, the adjusted LSM change from baseline in Hb averaged over weeks 28–52 was significantly greater with roxadustat versus placebo (+1.75 g/dl [95% CI, 1.58 to 1.92] versus +0.62 g/dl [95% CI, 0.44 to 0.80], respectively; difference +1.13 g/dl; 95% CI, 0.91 to 1.35; P<0.001) (Figure 4E). The proportions of total time of interpolated Hb values ≥10 g/dl and 10–12 g/dl from weeks 28 to 52 were significantly greater with roxadustat versus placebo (P<0.001) (Table 2).

Table 2. - Secondary efficacy end points

End Point	Roxadustat (n=1384)		Placebo (n=1377)		Relative Risk/HR/Difference in LSM Changes (95% CI) a	P Value
	N	Number (%)/Adjusted LSM (95% CI)/Event Rate per 100 PY (Total Number of Yr at Risk) a	n	Number (%)/Adjusted LSM (95% CI)/Event Rate per 100 PY (Total Number of Yr at Risk) a
1. Proportion with Hb response during the first 24 wk of treatment, %	1371	1055 (77.0)	1357	112 (8.5)	9.12 (7.63 to 10.89)	<0.001
2. Change from baseline in Hb averaged over wk 28–52 in patients with baseline hsCRP >ULN, g/l	213	1.75 (1.58 to 1.92)	198	0.62 (0.44 to 0.80)	1.13 (0.91 to 1.35)	<0.001
3. Proportion of total time of interpolated Hb values ≥10 g/dl over wk 28–52, %	1220	0.82 (0.80 to 0.85)	1145	0.33 (0.31 to 0.35)	0.50 (0.47 to 0.52)	<0.001
4. Proportion of total time of interpolated Hb values 10–12 g/dl over wk 28–52, %	1220	0.70 (0.68 to 0.72)	1145	0.28 (0.26 to 0.30)	0.42 (0.40 to 0.45)	<0.001
5. Mean change in LDL-C from baseline to wk 24, mg/dl	1147	−14.58 (–16.67 to –12.49)	1133	−0.70 (–2.78 to 1.37)	−13.88 (–16.37 to –11.39)	<0.001
6. Event rate for first rescue therapy (composite)	254	11.90 (2134.88)	574	39.76 (1443.49)	0.26 (0.23 to 0.31)	<0.001
7. Event rate for first rescue RBC transfusion	176	7.98 (2206.74)	320	19.61 (1631.69)	0.37 (0.30 to 0.44)	<0.001

Hb response was defined as Hb ≥11.0 g/dl and Hb increase from baseline by ≥1.0 g/dl for patients with baseline Hb >8.0 g/dl, or Hb increase from baseline by ≥2.0 g/dl for patients with baseline Hb ≤8.0 g/dl, at two consecutive visits (with available data) separated by at least 5 d during the first 24 wk of treatment without having received rescue therapy (RBC transfusion, ESA, or iv iron) before Hb response. PY, patient yr at risk.

^aNo. of patients (%) and relative risk (95% CI) are presented for secondary end point 1; adjusted LSM change (95% CI) and difference in LSM changes (95% CI) are presented for secondary end points 2, 3, 4, and 5; and event rate per 100 patient-yr at risk (total number of yr at risk) and HR (95% CI) are presented for secondary end points 6 and 7.

Mean LDL-C levels at baseline were similar for roxadustat (94.50 mg/dl) and placebo (92.52 mg/dl). The adjusted LSM change in LDL-C from baseline to week 24 was significantly greater with roxadustat versus placebo (−14.58 mg/dl versus −0.70 mg/dl, respectively; difference −13.88 mg/dl; 95% CI, −16.37 to −11.39; P<0.001) (Table 2).

Fewer roxadustat- versus placebo-treated patients received ≥1 rescue therapy for the overall composite, and for each individual component (RBC transfusion, iv iron, or ESA; Figure 5). Event rates for use of first rescue therapy and first rescue RBC transfusion were lower with roxadustat than placebo (Figure 5, Table 2). Roxadustat achieved risk reductions versus placebo of 74% for rescue therapy overall, 63% for RBC transfusion, 59% for iv iron, and 87% for ESA (all P<0.001) (Figure 5).

Figure 5.: Rescue therapy end points. OT+28 analysis set. Data shown are for first rescue therapy (composite of individual rescue therapies). Event rates are presented per 100 patient-years. For the overall (composite) end point, rescue therapy refers to iv iron, RBC transfusion, or ESA. RBC transfusion was allowed if rapid correction of anemia was required or if deemed a medical necessity by the investigator. The HR, 95% CI, and P value comparing the treatment arms were estimated using a Cox proportional hazards model with baseline Hb, baseline eGFR, cardiovascular/cerebrovascular/thromboembolic history, and geographic region (USA versus ex-USA) included in the model. 95% CIs were from Wald and ties and were calculated using the Efron method. Time to first rescue therapy was calculated as (date of first rescue therapy, or date of censoring if no rescue therapy was taken)−(date of first dose of study drug)+1. Patients who did not take any rescue therapy were censored at the earliest occurrence of 28 days after their last intake of study drug, or the date of withdrawal of consent or last study contact if the patient withdrew consent or was lost to follow-up, or the date of death.

Adjusted LSM change from baseline in SF-36 Vitality subscore was 1.59 with roxadustat and 1.15 with placebo (difference +0.44; 95% CI, −0.11 to 0.99; P=0.120) and in SF-36 Physical Functioning subscore was 0.14 with roxadustat and −0.39 with placebo (difference +0.52; 95% CI, 0.0 to 1.05; nominal P=0.051; Supplemental Table 6).

Annual rate of change in eGFR starting from week 4 was −3.70 ml/min per 1.73 m² with roxadustat and −3.19 ml/min per 1.73 m² with placebo (difference −0.51 ml/min per 1.73 m²; 95% CI, −1.00 to −0.01; nominal P=0.046; Supplemental Table 6).

Subgroup Analysis of the Primary Efficacy End Point

Results from the subgroup analyses were consistent with the main analysis, including across baseline eGFR and Hb categories (Figure 6). Hb improvement with roxadustat versus placebo was similar among the 42% of patients with ferritin ≤100 μg/L and/or TSAT ≤20% (+1.76 g/dl versus +0.43 g/dl, respectively) compared with patients with values above those thresholds (+1.71 g/dl versus +0.39 g/dl, respectively) (both P<0.001; Figure 4B).

Figure 6.: Change from baseline in Hb (g/dl) averaged over weeks 28–52, by patient subgroups (ITT analysis set). Treatment difference is for roxadustat versus placebo. Error bars are 95% CIs. Change in Hb from baseline to average over weeks 28–52 was analyzed using an ANCOVA model with the following fixed-effect covariates at baseline: Hb and eGFR values in continuous scales; cardiovascular/cerebrovascular/thromboembolic history (yes/no); geographic region (USA versus ex-USA); and treatment arm, subgroup, and treatment by subgroup interaction as fixed effects. Adjusted LS means, their difference, and corresponding 95% CI were generated from datasets where missing data were imputed using missing at random-based multiple imputation with baseline Hb, baseline eGFR, cardiovascular/cerebrovascular/thromboembolic history, and geographic region (USA versus ex-USA) as predictor variables. hsCRP was quantified from available stored biomarker samples obtained at randomization; ULN is 5 mg/l (0.5 mg/dl); iron replete defined as ferritin >100 μg/l and TSAT >20%. LS, least squares.

Exploratory End Points

Mean (SD) baseline hepcidin was 163.16 (116.94) ng/ml for roxadustat and 155.45 (111.83) ng/ml for placebo (Table 3). Compared with baseline, hepcidin at week 24 was reduced with roxadustat and increased with placebo (adjusted LSM [SD] change from baseline −35.94 [116.69] ng/ml versus +9.42 [115.78] ng/ml, respectively) (Table 3). Adjusted LSM changes in serum iron, ferritin, TIBC, and TSAT from baseline averaged over week 24 to EOT were as follows: serum iron was increased with roxadustat and reduced with placebo (+6.63 versus −1.07 µg/dl, respectively; difference 7.70 µg/dl; 95% CI, 5.82 to 9.58) (Figure 7A, Table 3); serum ferritin was reduced with roxadustat and increased with placebo (−37.10 versus +17.45 µg/l, respectively; difference −54.55 µg/l; 95% CI, −71.68 to −37.42) (Figure 7B, Table 3); serum TIBC was increased with roxadustat and reduced with placebo (+30.79 versus −3.82 µg/dl, respectively; difference 34.61 µg/dl; 95% CI, 31.29 to 37.93) (Figure 7C, Table 3); and there was no difference in change from baseline in TSAT (−0.83% versus −0.26%, respectively; difference −0.57%; 95% CI, −1.31 to 0.18) (Figure 7D, Table 3). Iron profile parameters analyzed according to baseline quartile are shown in Supplemental Figure 2. The largest reductions in serum ferritin with roxadustat treatment were seen in those patients with the highest baseline values.

Table 3. - Hepcidin and iron profile exploratory end points (ITT analysis set)

End Point a	Roxadustat (n=1384)			Placebo (n=1377)			LSM Treatment Difference (95% CI)	P Value
	n	Mean Baseline	Adjusted LSM (95% CI)	n	Mean Baseline	Adjusted LSM (95% CI)
Hepcidin, ng/ml	658	163.16	−35.94 (–44.87 to –27.02)	604	155.45	9.42 (0.18 to 18.66)	−45.36 (–56.24 to –34.49)	<0.001
Serum iron, µg/dl	1201	67.82	6.63 (5.09 to 8.17)	1050	67.20	−1.07 (–2.67 to 0.53)	7.70 (5.82 to 9.58)	<0.001
Serum ferritin, µg/l	1200	248.32	−37.10 (–51.11 to –23.09)	1050	241.35	17.45 (2.86 to 32.05)	−54.55 (–71.68 to –37.42)	<0.001
TSAT, %	1199	29.93	−0.83 (–1.44 to –0.22)	1047	29.54	−0.26 (–0.89 to 0.38)	−0.57 (–1.31 to 0.18)	0.134
TIBC, µg/dl	1200	230.44	30.79 (28.08 to 33.50)	1050	232.46	−3.82 (–6.65 to –0.99)	34.61 (31.29 to 37.93)	<0.001

^aLSM change from baseline to week 24 for hepcidin; LSM change from baseline to mean during wk 24 to EOT for iron, ferritin, TSAT, and TIBC. Hepcidin was quantified from stored biomarker specimens obtained at baseline and wk 24.

Figure 7.: Serum iron parameters by visit. (A) Iron; (B) ferritin; (C) TIBC; (D) TSAT (ITT analysis set). Error bars are 95% CIs. Baseline is defined as the last measurement before randomization. 95% CI of the mean is on the basis of the normal distribution.

Adjusted LSM changes in SBP were similar with roxadustat and placebo (1.24 mmHg versus 2.06 mmHg, respectively; difference −0.82 mmHg; 95% CI, −1.73 to 0.09). The adjusted LSM changes in DBP were +0.51 mmHg with roxadustat and −0.32 mmHg with placebo (difference +0.82 mmHg; 95% CI, 0.31 to 1.34).

Safety

Interpretation of AEs was on the basis of the ITT analysis set. The proportions of patients with any AE were similar between the roxadustat and placebo groups (89.8% versus 88.3%, respectively) (Table 4). The most commonly reported AEs by preferred term were ESKD, urinary tract infection (UTI), pneumonia, and hypertension (Table 5). Most AEs by preferred term had similar time at risk–adjusted event rates for both treatment groups; the only AE with an event rate difference of >2.0 events per 100 patient-years between roxadustat and placebo was UTI (6.8 versus 4.2 events per 100 patient-years, respectively) (Table 5).

Table 4. - AEs summary (ITT analysis set)^a

AE Category	Roxadustat (n=1384)			Placebo (n=1377)
	Patients with Event	%	Event Rate (per 100 PY) b	Patients with Event	%	Event Rate (per 100 PY) b
Any AE	1243	89.8	182.9	1216	88.3	171.9
Any serious AE (including events with an outcome of death)	795	57.4	42.1	749	54.4	40.0
Any AE with an outcome of death	262	18.9	9.3	213	15.5	7.8
All-cause mortality c	284	20.5	9.6	245	17.8	8.4
Any AE in the cardiac disorders SOC d	316	22.8	12.7	293	21.3	12.0
Any serious AE in the cardiac disorders SOC d	174	12.6	6.5	157	11.4	6.0

Patients with multiple events in the same category were counted only once in that category. Patients with events in more than one category were counted once in each of those categories. PY, patient yr at risk; SOC, system organ class.

^aIncluded AEs with an onset date on or after the date of randomization and up to and including the EOS visit; or date of last contact or withdrawal of consent, if before the EOS visit.

^bCalculated as [No. of patients with AEs divided by (the total number of days at risk for that AE across all patients in given group divided by 365.25)] × 100.

^cIncludes deaths from public record searches.

^dAnalyses performed post hoc.

Table 5. - Most common AEs (≥5%) by preferred term (ITT analysis set)^a

Preferred Term	Roxadustat (n=1384)			Placebo (n=1377)
	Patients with Event	%	Event Rate (per 100 PY) b	Patients with Event	%	Event Rate (per 100 PY) b
ESKD	290	21.0	11.7	282	20.5	11.8
UTI	177	12.8	6.8	110	8.0	4.2
Pneumonia	165	11.9	6.2	130	9.4	4.9
Hypertension	159	11.5	6.1	125	9.1	4.8
Edema peripheral	149	10.8	5.7	111	8.1	4.3
Diarrhea	144	10.4	5.5	119	8.6	4.6
Nausea	125	9.0	4.8	104	7.6	4.1
Hyperkalemia	118	8.5	4.4	95	6.9	3.6
Cough	105	7.6	3.9	69	5.0	2.6
Viral upper respiratory tract infection	101	7.3	3.8	106	7.7	4.1
Upper respiratory tract infection	96	6.9	3.6	76	5.5	2.9
Headache	94	6.8	3.5	82	6.0	3.1
Constipation	92	6.6	3.4	88	6.4	3.4
Hypoglycemia	91	6.6	3.4	73	5.3	2.8
Gastritis	81	5.9	3.0	72	5.2	2.7
Azotemia	80	5.8	2.9	73	5.3	2.7
Dyspnea	78	5.6	2.9	74	5.4	2.8
Bronchitis	78	5.6	2.9	73	5.3	2.8
Vomiting	78	5.6	2.9	69	5.0	2.6
Dizziness	77	5.6	2.9	87	6.3	3.4
AKI	75	5.4	2.7	47	3.4	1.7
Asthenia	74	5.3	2.7	75	5.4	2.8
Arthralgia	73	5.3	2.7	54	3.9	2.0
Back pain	72	5.2	2.6	57	4.1	2.1
Pruritus	68	4.9	2.5	80	5.8	3.1
Edema	66	4.8	2.4	69	5.0	2.6

Percentages were on the basis of the No. of patients in the ITT analysis set in that treatment arm. PY, patient yr at risk.

^aIncluded AEs with an onset date on or after the date of randomization and up to and including the EOS visit; or date of last contact or withdrawal of consent, if before the EOS visit.

^bCalculated as [No. of patients with AEs divided by (the total No. of days at risk for that AE across all patients in given group divided by 365.25)] × 100.

The frequency of serious AEs was similar for roxadustat and placebo (57.4% versus 54.4%, respectively) (Table 4). The most commonly reported serious AEs by preferred term were ESKD, pneumonia, and azotemia, although cardiovascular AEs were also observed in this category (Supplemental Table 7).

In an overview of cardiovascular AEs, the proportion of roxadustat- and placebo-treated patients with AEs was 22.8% versus 21.3%, respectively, and with serious AEs was 12.6% versus 11.4%, respectively, in the cardiac disorders system organ class (Table 4). The complete listing of serious AEs in the cardiac disorders system organ class is provided in Supplemental Table 8.

The likelihood of death was ascertained in two populations: one that included only patients remaining in the trial (excluding deaths after withdrawal of consent) and one that included public record searches to increase ascertainment of as many deaths as possible during the study period, including deaths that occurred after withdrawal of consent but before EOS (Table 4). Mortality rates for the roxadustat and placebo groups with more complete ascertainment were 9.6 and 8.4 events per 100 patient-years, respectively (Table 4).

The overall mean (SD) change from baseline in serum potassium was similar between the roxadustat and placebo groups (+0.05 [0.522] mmol/l versus +0.01 [0.541] mmol/l, respectively). The proportion of patients with postbaseline serum potassium concentrations ≥6.5 mmol/l before initiation of dialysis was 10.3% with roxadustat and 6.8% with placebo (Supplemental Table 9).

Discussion

The clinical development program for roxadustat for patients with anemia of NDD-CKD includes placebo-controlled trials and trials comparing roxadustat with ESA treatment. In this large, global, phase 3 study of patients with anemia of NDD-CKD, roxadustat was superior to placebo in increasing Hb. Compared with placebo, roxadustat reduced the need for RBC transfusions, iv iron, and ESAs. RBC transfusions are common in patients with NDD-CKD and anemia, have increased in frequency after US label changes, and carry potential risks including infection and allosensitization.^12,15,16,40 Roxadustat treatment led to a 63% relative risk reduction compared with placebo, and an absolute risk reduction of 11.63 transfusions per 100 patient-years. Similar Hb improvements with roxadustat were observed among all patient subpopulations assessed including patients with elevated or normal baseline hsCRP levels. Studies have demonstrated an inverse association between the inflammatory marker hsCRP and ESA responsiveness.^41,42 Similar Hb improvements were observed with roxadustat treatment regardless of baseline iron status. The clinical significance of roxadustat’s efficacy in patients with low baseline iron stores is supported by the relatively high (42%) prevalence of patients with baseline ferritin ≤100 μg/l and/or TSAT ≤20% in OLYMPUS. Collectively, these data show that roxadustat is effective at increasing Hb and decreasing risk of RBC transfusions in patients with NDD-CKD, with efficacy seen regardless of baseline characteristics such as elevated hsCRP or low iron stores.

Improvements in iron availability and mobilization may have contributed to the Hb response observed with roxadustat. Consistent with previous studies,^37–39 roxadustat lowered serum hepcidin—a key regulator of iron homeostasis⁴³—and increased serum iron levels and serum TIBC. The observed reduction in serum ferritin with roxadustat has been reported previously,^37–3839,44 and in this study was most prominent in patients with the highest baseline values. Overall, the reduction in hepcidin, increased serum iron, and, in particular, increased TIBC with roxadustat support improved absorption, mobilization, and utilization of iron for erythropoiesis in roxadustat-treated patients, particularly in the setting of the increase in Hb from baseline regardless of baseline iron repletion status and the observed decreased need for iv iron.

Observational studies have shown that patients with anemia of NDD-CKD frequently have a high incidence of adverse clinical outcomes.⁹ In OLYMPUS, the safety profile of roxadustat was comparable to placebo, as was seen in previous placebo-controlled roxadustat studies.^37–39 More patients receiving roxadustat who had low eGFR or initiated dialysis, which are associated with higher risk of death and cardiovascular events,^6–8 remained on roxadustat versus placebo. To minimize potential bias due to differential dropout, the ITT analysis set and time at risk–adjusted event rates were utilized. Event rates of UTI and pneumonia appeared higher for roxadustat than placebo, although there is no clear biologic mechanism for these observations.

Because cardiovascular risk is increased in this patient population, event rates for AEs and serious AEs in the cardiac disorders system organ class were assessed. Analysis of cardiovascular events is limited by the size of the study and will be analyzed using a pooled analysis of multiple roxadustat studies; however, rates were generally comparable for roxadustat and placebo, both overall and for individual cardiac serious AEs. Importantly, OLYMPUS allowed for a prospective and complete ascertainment of patient mortality using public record searches, even after withdrawal of consent, and vital status was confirmed in >99% of patients. These results showed that mortality rates between roxadustat and placebo were 9.6 and 8.4 events per 100 patient-years, respectively. OLYMPUS was neither individually powered to assess nor prospectively planned as a standalone study to determine cardiovascular safety and mortality risk with roxadustat. The findings will be further assessed in the fully powered analysis of the pooled population of the phase 3, placebo-controlled roxadustat NDD-CKD studies.

Some safety data were difficult to interpret in the setting of differential study drug discontinuation. For example, a numerically greater number of patients with roxadustat than placebo had ≥1 postbaseline serum potassium concentration of ≥6.5 mmol/l. Because patients who discontinued study drug prematurely and remained in the study with modified follow-up often did not continue central laboratory testing, OT data were used to assess serum potassium outlier values. However, these data are confounded by patients at higher risk for hyperkalemia being more likely to remain on roxadustat than placebo. Notably, placebo-treated patients with baseline eGFR <15 ml/min per 1.73 m², which is associated with an approximate twofold and fivefold increased risk of hyperkalemia versus eGFR 15–29 ml/min per 1.73 m² and 30–59 ml/min per 1.73 m², respectively,⁴⁵ were approximately 2.5-times more likely to prematurely discontinue study drug versus roxadustat-treated patients. Therefore, and given the overall severe CKD of the studied population, mean changes in serum potassium between treatments are meaningful to assess the potential effects of roxadustat. Overall, mean changes in serum potassium from baseline between treatments did not differ, suggesting that roxadustat does not affect serum potassium. Similarly, numeric decreases and increases of <1 mmHg for roxadustat versus placebo in SBP and DBP, respectively, suggest that roxadustat does not meaningfully affect BP values.

For clinical trials among patients with anemia of NDD-CKD, such as the Trial to Reduce Cardiovascular Events with Aranesp^® Therapy (TREAT), the use of a placebo control has been considered to provide greater evidentiary power than use of an active control and to allow valid assessment of safety and efficacy.^17,46 For studying efficacy, the placebo control means that roxadustat was assessed against a comparator not expected to increase Hb or prevent RBC transfusion. For studying safety, the use of placebo allows a rigorous assessment of noninferiority for cardiovascular safety via separate pooled analyses of the phase 3 NDD-CKD studies, and comparison with a therapy without known AEs. Although ESA is also a valid comparator, interpretation of safety against an active comparator with reported cardiovascular safety concerns when dosed to achieve normal or near normal Hb levels^19–21 can be complex. Boxed warnings in the US product labels of approved ESAs state that no trial has identified an Hb target level, ESA dose, or dosing strategy that does not increase the risks of death and other cardiovascular events.^22,23 The lack of an active comparator in OLYMPUS precludes direct comparison of roxadustat with ESAs. However, the roxadustat clinical development program comprises several placebo- and active-controlled phase 3 clinical studies, and the efficacy and safety of roxadustat compared with ESA among patients with anemia of NDD-CKD have been evaluated in the completed phase 3, active-controlled DOLOMITES study (NCT02021318). In DOLOMITES, 616 patients with NDD-CKD stages 3–5 and anemia were randomized to receive roxadustat (n=323) or darbepoetin alfa (n=293). Among these patients, roxadustat was noninferior to darbepoetin alfa in the correction of Hb levels (Hb response rates: roxadustat 89.5% versus darbepoetin alfa 78.0%; difference: 11.5%; 95% CI, 5.7% to 17.4%) during the first 24 weeks of treatment and safety profiles were comparable.

Another reason that placebo might be considered is that several reports estimate that many patients with anemia of NDD-CKD in the real world are untreated^{4,5,11,25–30} and ESA use is relatively uncommon in treated patients.^11,28–30 In particular, a US study from 2018 showed that approximately 11%–13% of patients with anemia of NDD-CKD received an ESA; of these, <5% received ESA treatment consistently.²⁸ This likely reflects problems with affordability, physical access to injectable therapy, and interpretation of current guideline recommendations including the recommendation that ESAs are used intermittently and to prevent RBC transfusions, given their risk of cardiovascular events.^19–21 The high discontinuation rate and need for rescue therapy in the placebo arm in OLYMPUS support the need for effective anemia treatment in patients with NDD-CKD.

The OLYMPUS study has several strengths that further our knowledge of roxadustat from the previous phase 3 NDD-CKD study conducted in China.³⁸ The OLYMPUS study had a longer follow-up period than that previously assessed and enrolled a larger and more diverse patient population, which is important for the generalizability of the data. This large patient population allowed for the analysis of patient subgroups, such as by baseline eGFR and Hb categories and iron repletion status, and additional secondary analyses than were possible in earlier studies. Another strength of OLYMPUS is that, compared with the Correction of Hemogloblin and Outcomes in Renal Insufficiency (CHOIR) and TREAT ESA studies,^20,21 OLYMPUS enrolled a broader, real-world patient population. For instance, baseline Hb and eGFR were lower in patients in OLYMPUS (mean: 9.1 g/dl and approximately 20 ml/min per 1.73 m²) compared with CHOIR (mean: 10.1 g/dl and 27 ml/min per 1.73 m²) and TREAT (median: 10.5 g/dl and 34 ml/min per 1.73 m²).^20,21 Evidence from the CKD Outcomes and Practice Patterns Study suggests that up to 45% of patients with NDD-CKD and Hb <10 g/dl have stage 5 CKD, confirming the clinical applicability of the OLYMPUS study.⁴

Although the inclusion of patients with severe disease was an advantage of OLYMPUS, it also affected data interpretation. The rate of study drug discontinuation was relatively high in both treatment arms but higher in the placebo group. AEs were an uncommon reason for study drug discontinuation in both treatment groups. Post hoc analyses identified low eGFR and initiation of dialysis after randomization as contributors to study drug discontinuation overall and to differences in study drug discontinuation by treatment group. Although the study allowed continuation of roxadustat after dialysis initiation, study drug discontinuations may have been driven by administrative challenges relating to standardized processes of anemia treatment at dialysis centers, avoidance of the added complexity of experimental drug administration, and patients who may require hospitalization and have other medical complexities making strict adherence to the protocol difficult. Similar to TREAT,²¹ patients in OLYMPUS who discontinued study drug (but did not withdraw from study) were followed for as long as possible and remained in the ITT analysis until study closure/withdrawal, using follow-up options designed, in part, to minimize withdrawal of consent. The rate of study withdrawal in OLYMPUS was low (7.8%) compared with the ESA studies CHOIR (38%) and PEARL (approximately 24%),^20,47 and vital status at the end of the study was confirmed in >99% of randomized patients. However, although the ITT population allowed for a more balanced evaluation of safety, this population could also contain reporting bias for AEs due to the higher withdrawal of consent in placebo patients and possible under-reporting of some events (e.g., nonserious AEs) in patients with modified post-treatment follow-up.

OLYMPUS had limitations. Study personnel at local study sites could not be blinded to Hb values; therefore, Hb increases could have in some cases been considered by patients and physicians to be suggestive of treatment allocation and potentially affected continuation in the study or reporting of AEs. No assessment of site personnel or patients was performed to examine perception of treatment allocation. Median exposures of 20.80 and 14.57 months in the roxadustat and placebo groups, respectively, do not allow for the assessment of roxadustat in this population over a time period longer than the maximum treatment duration of 4 years.

In conclusion, oral roxadustat was more effective than placebo in correcting anemia, reduced the risk of RBC transfusions, and had comparable safety to placebo in this international, double-blind, phase 3 study. Roxadustat was effective in a broad range of patients with NDD-CKD, including patients with inflammation, and patients with more advanced CKD (lower baseline eGFR), more severe anemia (lower baseline Hb levels), and lower iron stores (low baseline ferritin and TSAT) than those included in historical NDD-CKD trials with ESAs. The findings support anemia treatment using roxadustat, an oral medication that can be administered in the home setting and without requirement for routine iv iron supplementation, which may simplify anemia treatment for patients with NDD-CKD.

Disclosures

M.A. El-Shahawy reports ownership interest in Paramount Hope Dialysis Center, East LA Dialysis Center; research funding from AstraZeneca, Pfizer, Bayer, UCB, and Sanofi; honoraria from AstraZeneca; being a scientific advisor or membership from AstraZeneca and Bayer; and speakers bureau from AstraZeneca. S. Fishbane reports receiving research support and consulting fees from AstraZeneca; consultancy agreements from Akebia, Cara Therapeutics, and FibroGen; research funding from Cara, Gilead, and Merck; and honoraria from Akebia and AstraZeneca. L. Frison, N.J. Guzman, and D.J. Little are employees of AstraZeneca. N.J. Guzman and D.J. Little also report ownership interest in AstraZeneca. M.T. Houser is employed by and has ownership interest in AstraZeneca. R. Pecoits-Filho reports receiving consulting fees paid to his employer from Akebia and AstraZeneca for participation in advisory boards and educational events, and research grants from Fresenius Medical Care. R. Pecoits-Filho also reports consultancy agreements with Rethrophin; being a scientific advisor or member with the Kidney Disease: Improving Global Outcomes Executive Board, International Society of Nephrology Executive Board, and Standardised Outcomes in Nephrology Executive Committee; serving on editorial boards for American Journal of Kidney Diseases, Blood Purification, Nephrology, Peritoneal Dialysis International, Hemodialysis International, and the Brazilian Journal of Nephrology; and serving on the speakers bureau with AstraZeneca and Novo Nordisk. P.E. Pergola reports receiving research support and consulting fees from AstraZeneca. P.E. Pergola also reports consultancy agreements with Akebia Therapeutics, Ardelyx, Bayer, Corvidia Therapeutics, Gilead Sciences, Reata Pharmaceuticals, and Tricida; reports ownership interest in Unicycive Therapeutics; and reports receiving research funding as principal investigator or subinvestigator on multiple clinical trials (the contracts are with his practice, not individual); being a scientific advisor or member with Ardelyx and Unicycive; and serving on the speakers bureau with AstraZeneca. B.P. Van reports receiving research support and consulting fees from AstraZeneca. B.P. Van also reports honoraria from Astellas, AstraZeneca, Boehringer Ingelheim, DiethelmKellerSiberHegner, Kalbe International, Nanogen (Vietnam), Otsuka, Pfizer, Servier, and Tedis; being a scientific advisor or member with Nguyen Tri Phuong University Hospital (Vietnam); and serving on the speakers bureau with Astellas, AstraZeneca, Boehringer Ingelheim, DKSH, Kalbe International, Otsuka, Pfizer, Sanofi, Servier, and Tedis.

Funding

This study was supported by AstraZeneca.

Data Sharing Statement

Data underlying the findings described in this manuscript can be requested in accordance with AstraZeneca’s data sharing policy, described at https://astrazenecagroup-dt.pharmacm.com/DT/Home by accessing www.vivli.org.