Pooled Analysis of Roxadustat for Anemia in Patients With Kidney Failure Incident to Dialysis.

INTRODUCTION: Erythropoiesis-stimulating agents are associated with increased cardiovascular risk when higher doses are used toward higher hematocrit targets. Patients new to dialysis are at higher risk for morbidity and mortality. Systematic evaluation of this population was predefined in the roxadustat clinical development program. Roxadustat is a hypoxia-inducible prolyl hydroxylase inhibitor.
METHODS: Data were pooled from 3 phase 3, randomized, open-label, active-controlled trials. Eligible adults had kidney failure and initiated dialysis for 2 weeks to ≤ 4 months prior to randomization to roxadustat or epoetin alfa. Efficacy was assessed as mean change in hemoglobin from baseline averaged over weeks 28 to 52, regardless of rescue therapy. Key cardiovascular safety endpoints were major adverse cardiovascular events (MACE; all-cause mortality [ACM], myocardial infarction, and stroke), and MACE+ (MACE plus unstable angina or congestive heart failure requiring hospitalization), and ACM.
RESULTS: This study included 1530 patients with kidney failure incident to dialysis. Mean (SD) changes in hemoglobin from baseline averaged over weeks 28 to 52, regardless of rescue therapy, were 2.12 (1.45) versus 1.91 (1.42) g/dl in the roxadustat and epoetin alfa groups (least-squares mean difference: 0.22; 95% CI, 0.05 to 0.40; P = 0.0130). Risks of MACE and MACE+ were lower in the roxadustat group (hazard ratio [HR], 0.70; 95% CI, 0.51 to 0.96) than the epoetin alfa group (HR, 0.66; 95% CI, 0.50 to 0.89); the HR for ACM was 0.76 (95% CI, 0.52 to 1.11).
CONCLUSION: Roxadustat was at least as efficacious as epoetin alfa. Roxadustat had a lower risk of MACE/MACE+ in patients new to dialysis.

See Commentary on Page 559

Globally, it is estimated that 2 million patients with chronic kidney disease (CKD) receive renal replacement therapy every year; however, this may represent only a fraction of the patients who need this type of treatment. In 2016 in the United States, 726,331 people had kidney failure treated by dialysis or kidney transplantation, and 124,675 people received a pre-emptive kidney transplant. Before publication of the first US Renal Data Systems Annual Data Report (ADR), Eggers documented that survival rates after the first year of dialysis ranged from 75.3% for white patients to 82.9% for black patients. The 2007 Annual Data Report noted that “first-year death rates among incident hemodialysis patients had not changed in 11 years, while steady improvements had been noted in subsequent years on dialysis.”, The 2018 Annual Data Report showed that the first-year survival rate for patients initiating dialysis was 78%. Arguably, patients soon after starting dialysis represent a vulnerable population for whom treatment outcomes have not improved significantly in recent years.

Multiple studies have demonstrated increased cardiovascular (CV) risks associated with the use of erythropoiesis-stimulating agents (ESAs) for treating anemia in patients with CKD, including those on dialysis.6, 7, 8 Research has shown that this increased risk is associated with higher doses of ESAs needed in attempts to reach higher hematocrit targets., This research supports an unmet need for safe and effective treatment in patients with anemia of CKD, particularly in populations in need of the highest ESA doses. The incident-dialysis population has the highest mortality rate and receives the highest doses of ESAs, which further underscores the importance of considering their outcomes. To date, no clinical trials have assessed the effect of interventions for the treatment of anemia in patients starting dialysis.

In the past decade, research has established the role of hypoxia-inducible factor (HIF), a transcription factor that is the body’s main oxygen tension sensor, in orchestrating RBC production and hemoglobin response. Roxadustat (FG-4592) is a potent and reversible HIF prolyl hydroxylase inhibitor that transiently induces HIF stabilization, mimicking the natural erythropoietic response associated with transient hypoxia exposure, but occurring under normoxic conditions. The intermittent dosing strategy of roxadustat for the treatment of CKD-related anemia results in the durable maintenance of a therapeutic effect over time, without a constant effect on the HIF system.

Individually, the pivotal, phase 3 studies of roxadustat in patients on dialysis were designed to evaluate its efficacy and general safety compared with an ESA. The single studies were not powered to assess the CV safety of roxadustat. Therefore, data from the 3 pivotal phase 3 studies of roxadustat in patients on dialysis were pooled to assess the relative efficacy and CV safety of roxadustat versus epoetin alfa. Systematic evaluation of patients new to dialysis was predefined as part of the clinical development program for roxadustat.

Methods

Trial Designs

Data were pooled from patients with CKD-related anemia who enrolled in 1 of 3 pivotal, similarly designed studies (Himalayas FGCL-4592-063 [NCT02052310], Sierras FGCL-4592-064 [NCT02273726], and Rockies D5740C00002 [NCT02174731]) who were new to dialysis (incident dialysis, defined as patients on dialysis for 2 weeks to ≤ 4 months prior to randomization [ID-DD]). All 3 trials were randomized, multicenter, open-label, epoetin alfa‒controlled phase 3 studies evaluating the efficacy of roxadustat to correct and/or maintain hemoglobin levels (Supplementary Table S1).

All study protocols were approved by relevant institutional review boards and/or ethics committees and were conducted in accordance with the tenets of the Declaration of Helsinki, the International Council for Harmonisation Guidelines for Good Clinical Practice, and all other applicable local health and regulatory requirements. All patients provided written informed consent before enrollment.

Participants

Eligible patients (aged ≥18 years) had ID-DD CKD (on dialysis for 2 weeks to ≤ 4 months prior to randomization) and CKD-related anemia and were on hemodialysis or peritoneal dialysis. This included all of the patients from the 063 study (n = 1039) and a subset of patients from the 064 (n = 71) and 002 (n = 416) studies. Study-specific inclusion criteria are detailed in Supplementary Table S1. A key exclusion criterion was recent RBC transfusion.

Interventions

After screening, eligible patients were randomized (1:1) to oral roxadustat or parenteral epoetin alfa. Each study’s drug dosing and titration procedures are provided in the Supplementary Methods.

Rescue therapy included RBC transfusion, ESAs, or transfusion and ESAs. RBC transfusion was allowed for all patients who needed rapid correction of anemia, or for whom it was considered medically necessary. ESA use was allowed for patients who met all of the following criteria: (i) hemoglobin level had not responded sufficiently after ≥2 dose increases or if the maximum dose limit had been reached; (ii) hemoglobin was <8.5 g/dl (064 and 002 only) and clinical judgment did not suggest iron deficiency or bleeding as the reason for the lack of response or rapid decrease in hemoglobin; and (iii) there was a need to reduce the risk of alloimmunization in transplant-eligible patients and/or reduce other transfusion-related risk. Each study’s protocol for IV iron supplementation differed. Study-specific details are included in the Supplementary Material.

Efficacy Outcomes

The key US Food and Drug Administration efficacy endpoint for the individual studies was mean change in hemoglobin from baseline averaged over weeks 28 to 52, regardless of rescue therapy. A key EU European Medicines Agency efficacy endpoint was the mean change in hemoglobin from baseline averaged over weeks 28 to 36, without rescue therapy within 6 weeks of and during the 8-week evaluation period.

Key secondary efficacy endpoints included: mean hemoglobin change from baseline averaged over weeks 18 to 24 in patients with high-sensitivity C-reactive protein that was higher than the upper limit of normal, mean change from baseline in low-density lipoprotein cholesterol averaged over weeks 12 to 28, mean monthly IV iron use during weeks 28 to 52, time to first RBC transfusion during treatment, mean change in mean arterial pressure averaged over weeks 8 to 12, and time to first exacerbation of hypertension up to week 52 (increase from baseline ≥20 mm Hg for systolic blood pressure [SBP] and SBP ≥170 mm Hg, or increase from baseline ≥15 mm Hg for diastolic blood pressure [DBP] and DBP ≥110.0 mm Hg).

CV Safety Endpoints

All CV safety endpoints analyzed were positively adjudicated by a central independent event review committee whose members were blinded to patients’ study-group assignment and hemoglobin level. The adjudication process is described in detail in the Supplementary Material.

The primary CV safety endpoint was time to first major adverse cardiovascular event (MACE), a composite measure of myocardial infarction, stroke, and all-cause mortality (ACM). Secondary CV safety endpoints included time to first MACE+ (composite measure of MACE plus unstable angina or congestive heart failure [CHF] requiring hospitalization) and time to ACM. Supportive endpoints included time to MACE CV mortality and MACE+ CV mortality, time to CV mortality, and time to each MACE+ component. CV safety endpoints are defined in the Supplementary Material.

The analysis period was the “on-treatment plus 7 days” (OT+7) time frame, which included events that occurred during the treatment period and within 7 days of the last dose of study drug. Using this follow-up period allowed for the ascertainment of residual adverse events, while minimizing confounding factors likely to exist after study drug discontinuation due to institution of other therapies for treatment of anemia.

Adverse Events

Safety was monitored by assessment of treatment-emergent adverse events and treatment-emergent serious adverse events during treatment and for 28 days after study drug discontinuation (OT+28) in the safety population (all randomized patients who received ≥1 dose of study drug).

Statistical Analysis

The overall clinical trial program of patients with dialysis-dependent CKD was powered for noninferiority. Approximately 600 patients with OT+7 MACE events, with a ∼90% power to demonstrate the upper limit of the two-sided 95% confidence interval (CI), would be required to exclude 1.3, if the true hazard ratio (HR) was 1.0. This noninferiority margin was established based on the results of previous ESA clinical trials6, 7, 8 and US Food and Drug Administration guidance on antidiabetic and oncologic drugs.,

The primary analysis of time to first MACE event used a Cox regression model to obtain the HR of roxadustat versus epoetin alfa and the 95% CI. The Cox model was stratified by history of cardiovascular, cerebrovascular, or thromboembolic diseases (yes vs. no), geographic region (Europe vs. others), sex, body mass index (<30 vs. ≥30 kg/m2), and race (black vs. other). In addition, analyses by study were pooled using meta-analysis techniques. The proportional hazards assumption was checked graphically using a log-cumulative hazard plot against log-survival time. Supportive evidence was based on sensitivity and subgroup analyses, including MACE/MACE+ with CV mortality, and individual components. For the primary efficacy analysis, a multiple imputation analysis of covariance model was used, including terms for treatment group, baseline hemoglobin, and stratification factors (except screening hemoglobin ≤8.0 vs. >8.0 g/dl). A noninferiority margin for the estimated difference between treatment groups (roxadustat − epoetin alfa) of −0.75 g/dl was predefined. Details regarding the statistical analyses are provided in the Supplementary Material.

Results

Data from the intent-to-treat (ITT) population of 1530 patients with ID-DD CKD were pooled (roxadustat, n = 760; epoetin alfa, n = 770) (Figure 1). In the roxadustat group, mean treatment exposure was 1.5 patient-exposure years (PEY) per patient (up to 4.4 PEY), and total exposure was 1098.2 PEY. In the epoetin alfa group, mean exposure was 1.6 PEY (up to 4.4 PEY), and total exposure was 1189.5 PEY (Supplementary Table S2).

Figure 1

CONSORT flow diagram.

In general, baseline demographic and clinical characteristics were comparable between the treatment groups (Table 1). Overall, the mean (SD) age was 53.8 (14.7) years, and 39.5% were women. Sixty-six percent of the patients were white, and nearly 50% were from Europe. Mean (SD) baseline hemoglobin levels were comparable in the roxadustat and epoetin alfa groups (8.8 [1.2] and 8.9 [1.2] g/dl, respectively). The majority (79.2%) of patients were iron-replete, 41.6% had diabetes mellitus, and 43.2% had a history of CV disease. The majority of patients were on hemodialysis (88.6%) compared with peritoneal dialysis (11.4%). Thirty-eight percent of patients had a high-sensitivity C-reactive protein level greater than the upper limit of normal.

Table 1

Baseline demographic and clinical characteristics (ITT)

Characteristic	Roxadustat (n = 760)	Epoetin alfa (n = 770)
Age, mean (SD), year∗	53.6 (14.8)	54.0 (14.6)
Male sex, n (%)	461 (60.7)	464 (60.3)
Race, n (%)
White	508 (66.8)	505 (65.6)
Asian	116 (15.3)	127 (16.5)
Black	67 (8.8)	67 (8.7)
Other	69 (9.1)	71 (9.2)
Region, n (%)
United States	195 (25.7)	196 (25.5)
Europe	362 (47.6)	374 (48.6)
Other	203 (26.7)	200 (26.0)
Weight, mean (SD), kg	74.3 (19.3)	75.0 (19.1)
Hemoglobin, mean (SD), g/dl	8.8 (1.2)	8.9 (1.2)
Hemoglobin distribution, n (%)
<8.0 g/dl	180 (23.7)	179 (23.2)
≥8.0 g/dl	580 (76.3)	591 (76.8)
Dialysis modality, n (%)
Hemodialysis	680 (89.5)	675 (87.7)
Peritoneal dialysis	80 (10.5)	94 (12.2)
Missing	0	1 (0.1)
Duration of dialysis, mean (SD), months	2.3 (0.9)	2.3 (0.9)
Patients taking ESAs, n (%)	123 (16.2)	121 (15.7)
Hs-CRP distribution, n (%)
≤ULN	406 (53.4)	401 (52.1)
>ULN	285 (37.5)	301 (39.1)
Missing	69 (9.1)	68 (8.8)
LDL cholesterol, mean (SD), mg/dl	104.6 (39.1)	104.7 (38.1)
Iron repletion status, n (%)
Ferritin ≥100 ng/ml and TSAT ≥20%	603 (79.3)	608 (79.0)
Ferritin <100 ng/ml or TSAT <20%	155 (20.4)	162 (21.0)
Missing	2 (0.3)	0
Diabetes mellitus, n (%)	322 (42.4)	314 (40.8)
History of cardiac, cerebrovascular, or thromboembolic disease, n (%)	328 (43.2)	333 (43.2)

ESA, erythropoietin-stimulating agent; Hs-CRP, high-sensitivity C-reactive protein ITT, intent-to-treat population; LDL, low-density lipoprotein; TSAT, transferrin saturation; ULN, upper limit of normal.

Age was calculated in years from birthdate to date of informed consent or date of first dose.

Efficacy Endpoints

Patients with ID-DD CKD who received roxadustat showed larger increases in hemoglobin levels over 52 weeks of treatment compared with patients who received epoetin alfa (Figure 2). For the US efficacy endpoint, the mean (SD) change in hemoglobin from baseline averaged over weeks 28 to 52, regardless of rescue therapy, was significantly higher in the roxadustat versus epoetin alfa group in the individual studies and also for the pooled analysis (mean [SD]: 2.12 [1.45] g/dl vs. 1.91 [1.42] g/dl). The least-squares mean (LSM) treatment difference was 0.22 (95% CI, 0.05 to 0.40; P = 0.0130) (Table 2). Thus, roxadustat was noninferior and superior to epoetin alfa, as there were statistically significant larger increases in hemoglobin levels from baseline averaged over weeks 28 to 52, regardless of rescue therapy. Subgroup analyses of the US efficacy analysis were similar to the full cohort (Supplementary Figure S1).

Figure 2

Hemoglobin levels by treatment arm (Full Analysis Set).

Table 2

Hemoglobin changes from baseline during weeks 28–52 regardless of rescue therapy (ITT)

	Study 063		Study 064		Study 002		Pooled data
	Roxadustat (n = 522)	Epoetin alfa (n = 521)	Roxadustat (n = 36)	Epoetin alfa (n = 35)	Roxadustat (n = 202)	Epoetin alfa (n = 214)	Roxadustat (n = 760)	Epoetin alfa (n = 770)
Mean (SD) baseline hemoglobin∗, g/dl	8.43 (1.04)	8.46 (0.96)	10.23 (0.80)	10.09 (0.91)	9.56 (1.16)	9.62 (1.24)	8.82 (1.22)	8.86 (1.19)
Mean (SD) weeks 28–52 hemoglobin†, g/dl	11.00 (0.82)	10.83 (0.88)	10.60 (0.79)	10.20 (0.78)	10.82 (0.90)	10.74 (1.01)	10.94 (0.84)	10.77 (0.92)
Mean (SD) hemoglobin change from baseline†, g/dl	2.57 (1.27)	2.36 (1.21)	0.37 (1.06)	0.10 (0.93)	1.25 (1.33)	1.12 (1.39)	2.12 (1.45)	1.91 (1.42)
MI-ANCOVA‡
LSM (SEM)	2.38 (0.04)	2.20 (0.04)	0.51 (0.16)	0.07 (0.16)	1.23 (0.08)	1.15 (0.073)	1.90 (0.06)	1.67 (0.07)
95% CI	(2.30‒2.46)	(2.12‒2.28)	(0.19‒0.83)	(−0.24 to 0.37)	(1.08‒1.39)	(1.01‒1.29))	(1.77‒2.02)	(1.55‒1.80)
LSM (SEM) difference	0.18 (0.05)		0.44 (0.23)		0.08 (0.10)		0.22 (0.09)
95% CI	(0.08‒0.29)		(0.00‒0.88)		(−0.12 to 0.29)		(0.05‒0.40)
P value	0.0005		0.0493§		0.4162§		0.0130

CI, confidence interval; ITT, intent-to-treat population; LSM, least-squares mean; MI-ANCOVA, multiple imputation analysis of covariance.

Baseline hemoglobin defined as the mean of up to 4 most recent laboratory values before the first dose of study drug.

Observed + imputed.

Treatment comparison using the multiple imputation strategy by combining the results of an ANCOVA model with baseline hemoglobin as covariate, and study; treatment; study × treatment interaction; and history of cardiovascular, cerebrovascular, or thromboembolic disease (yes vs. no) as fixed effects.

Nominal.

For the key EU efficacy endpoint, the mean (SD) change in hemoglobin from baseline averaged over weeks 28 to 36 censored for rescue therapy within 6 weeks of and during this evaluation period was significantly higher in the roxadustat versus epoetin alfa group in the individual studies and for the pooled analysis (mean [SD]: 2.37 [1.57] vs. 2.12 [1.46]). The LSM was 0.28 (95% CI, 0.11 to 0.45; P = 0.0013 [nominal]) (Table 3). Thus, roxadustat was noninferior to epoetin alfa and had a larger increase in hemoglobin levels.

Table 3

Hemoglobin changes from baseline during weeks 28–36 censored for rescue therapy (PPS)

	Study 063		Study 064		Study 002		Pooled data
	Roxadustat (n = 490)	Epoetin alfa (n = 468)	Roxadustat (n = 31)	Epoetin alfa (n = 29)	Roxadustat (n = 152)	Epoetin alfa (n = 172)	Roxadustat (n = 673)	Epoetin alfa (n = 669)
Mean (SD) baseline hemoglobin∗, g/dl	8.43 (1.04)	8.43 (0.96)	10.29 (0.80)	10.08 (0.92)	9.54 (1.17)	9.65 (1.25)	8.77 (1.20)	8.82 (1.20)
Mean (SD) weeks 28–36 hemoglobin, g/dl	11.13 (1.06)	10.94 (1.02)	10.71 (0.76)	10.17 (0.96)	10.86 (1.10)	10.85 (1.16)	11.07 (1.05)	10.89 (1.06)
Mean (SD) hemoglobin change from baseline, g/dl	2.70 (1.42)	2.50 (1.27)	0.48 (1.26)	0.08 (0.99)	1.31 (1.57)	1.15 (1.42)	2.37 (1.57)	2.12 (1.46)
MMRM
LSM (SEM)	2.59 (0.05)	2.39 (0.06)	0.42 (0.20)	−0.02 (0.20)	1.23 (0.09)	1.18 (0.08)	2.17 (0.06)	1.89 (0.06)
95% CI	(2.48–2.70)	(2.28–2.50)	(0.03–0.81)	(−0.41 to 0.36)	(1.06–1.40)	(1.02–1.34)	(2.05–2.30)	(1.77–2.02)
LSM (SEM) difference	0.20 (0.08)†		0.44 (0.28)‡		0.05 (0.12)‡		0.28 (0.09)‡
95% CI	(0.05–0.35)		(−0.10 to 0.99)		(−0.19 to 0.28)		(0.11–0.45)
P value	0.0090		0.1123		0.6898§		0.0013

CI, confidence interval; MMRM, mixed model of repeated measures; LSM, least-squares mean; PPS, per protocol set.

Baseline hemoglobin defined as the mean of up to 4 most recent laboratory values before the first dose of study drug.

Treatment comparison made using an MMRM with baseline hemoglobin as a covariate, and treatment, visit, visit × treatment interaction, and randomization stratification factors, except mean qualifying screening hemoglobin (≤8.0 vs. >8.0 g/dl), as fixed effects.

Treatment comparison using the MMRM with baseline hemoglobin as covariate, and study; treatment; visit, visit × treatment interaction; study × treatment interaction; and history of cardiovascular, cerebrovascular, or thromboembolic disease (yes vs. no) as fixed effects.

Nominal.

Key Secondary Efficacy Endpoints

In patients with baseline high-sensitivity C-reactive protein greater than the upper limit of normal, mean (SD) changes in hemoglobin from baseline averaged over weeks 18 to 24 were 2.14 (1.34) in the roxadustat group and 2.11 (1.47) g/dl in the epoetin alfa group, corresponding to an LSM difference of 0.27 (95% CI, −0.04 to 0.58; P = 0.09).

Mean (SD) changes from baseline in low-density lipoprotein (LDL) cholesterol averaged over weeks 12 to 28 in the roxadustat group versus epoetin alfa group were −22.57 (29.94) versus −4.79 (27.89) mg/dl. The LSM treatment difference was −17.50 mg/dl (95% CI, −22.22 to −12.78 mg/dl; P < 0.0001).

Mean (SD) monthly IV iron use over weeks 28 to 52 in the roxadustat group and epoetin alfa group was 53.57 (143.10) mg versus 70.22 (173.33) mg per patient-exposure months (P < 0.0001).

The percentages of patients who received RBC transfusions during the study were 6.1% and 6.7% of patients in the roxadustat and epoetin alfa groups. The exposure-adjusted incidence rates were similar in the roxadustat and epoetin alfa groups (4.2 and 4.3 per 100 PEY, respectively). The HR for time to first RBC transfusion was 0.99 (95% CI, 0.66 to 1.47).

The mean (SD) change from baseline in mean arterial pressure averaged over weeks 8 to 12 in the roxadustat group and epoetin alfa group was −0.05 (9.00) and 1.03 (9.24) mm Hg, corresponding to a LSM treatment difference of −0.35 (95% CI, −1.65 to 0.95).

The percentage of patients who experienced an exacerbation of hypertension up to week 52 in the roxadustat group and epoetin alfa group was 25.9% (33.6 per 100 PEY) and 25.4% (32.0 per 100 PEY) (HR, 1.02; 95% CI, 0.84 to 1.25).

Primary Cardiovascular Safety Endpoint

During a mean of 1.4 years of treatment exposure in the roxadustat group and a mean of 1.6 years of treatment exposure in the epoetin alfa group, the risk for MACE was lower in the roxadustat versus epoetin alfa group (HR, 0.70; 95% CI, 0.51 to 0.96; P = 0.029) (Figures 3a and 4b).

Figure 3

Forest plot of cardiovascular safety analyses in incident-dialysis patients on dialysis for 2 week to ≤4 months prior to randomization. (a) MACE using all-cause mortality. (b) MACE using cardiovascular mortality. (c) Individual events. ∗OT-7: events that occurred during the treatment period and within 7 days of the last dose of study drug. †PEY for each patient = (last dose date − first dose date + 1) / 365.25. Incidence rate (per 100 PEY) = 100 × number of subjects with events / PEY. ‡HR derived using a meta-analysis method combining individual study log-HRs with weights inversely proportional to the variants of the study-specific log-HRs. §P value reported when the point estimate of HR was <1.0. ACM, all-cause mortality; CHF, congestive heart failure; CI, confidence interval; CVM, cardiovascular mortality; HR, hazard ratio; MACE, major adverse cardivascular event; MACE+, MACE plus unstable angina and CHF requiring hospitalization; MI, myocardial infarction; OT+7, on treatment plus 7 days after last study drug; PEY, patient-exposure years.

Figure 4

Kaplan-Meier curves for MACE, MACE+, and ACM. ACM, all-cause mortality; CHF, congestive heart failure; CI, confidence interval; CVM, cardiovascular mortality; HR, hazard ratio; MACE, major adverse cardivascular event; MACE+, MACE plus unstable angina and CHF requiring hospitalization; MI, myocardial infarction; OT+7, on treatment plus 7 days after last study drug.

Secondary Cardiovascular Safety Endpoints

The risk for MACE+ was lower in the roxadustat group versus epoetin alfa group (HR, 0.66; 95% CI, 0.50 to 0.89; P = 0.005) (Figures 3a and 4b). The HR for ACM was 0.76 (95% CI, 0.52, 1.11; P = 0.15) (Figures 3a and 4c).

Supportive Cardiovascular Safety Endpoints

The risk for MACE CV mortality was significantly lower in the roxadustat group versus epoetin alfa group (HR, 0.63; 95% CI, 0.43 to 0.92; P = 0.018). Likewise, the risk for MACE+ CV mortality was significantly lower in the roxadustat group versus epoetin alfa group (HR, 0.60; 95% CI, 0.43 to 0.83; P = 0.002). The HR for CV mortality was 0.63 (95% CI, 0.37 to 1.05; P = 0.075) (Figure 3b).

Although analyses of the MACE and MACE+ components were not powered for noninferiority, the results for the individual components of MACE and MACE+ are generally consistent with those for the composite measures. The risk of stroke was lower in the roxadustat versus epoetin alfa group (HR, 0.41; 95% CI, 0.18 to 0.94; P = 0.035) (Figure 3c). A decomposition of MACE and MACE+ endpoints into the first occurrence of its components shows broadly similar rates between the treatment groups, with ACM being the most common MACE and MACE+ event (data not shown).

Subgroup Analysis

Subgroup analyses of MACE showed clinically consistent results (Supplementary Figure S2). Among US-based patients, the HR for the roxadustat group versus epoetin alfa group was 0.38 (95% CI, 0.20 to 0.73).

Eighty percent of patients in the roxadustat (611 of 760) and epoetin alfa (619 of 766) groups experienced at least 1 treatment-emergent adverse event. These adverse events, occurring in ≥5% of patients in either treatment group, are summarized in Table 4. Treatment-emergent serious adverse events, occurring in ≥1% of patients in either treatment group, are indicated in Table 5.

Table 4

TEAEs occurring in ≥5% of ID-DD patients in either treatment group (OT+28)∗

Preferred term†	Roxadustat (N = 760)		Epoetin alfa (N = 766)
	n (%)	PEY = 1098.2 incidence rate (per 100 PEY)‡	n (%)	PEY = 1189.5 incidence rate (per 100 PEY)‡
Hypertension	112 (14.7)	10.2	105 (13.7)	8.8
Diarrhea	87 (11.4)	7.9	51 (6.7)	4.3
Arteriovenous fistula thrombosis	73 (9.6)	6.6	55 (7.2)	4.6
Headache	67 (8.8)	6.1	50 (6.5)	4.2
Muscle spasms	66 (8.7)	6.0	46 (6.0)	3.9
Hypotension	65 (8.6)	5.9	46 (6.0)	3.9
Hyperphosphatemia	53 (7.0)	4.8	36 (4.7)	3.0
Nausea	52 (6.8)	4.7	35 (4.6)	2.9
Pneumonia	51 (6.7)	4.6	55 (7.2)	4.6
Arteriovenous fistula site complication	42 (5.5)	3.8	51 (6.7)	4.3
Vomiting	39 (5.1)	3.6	21 (2.7)	1.8
Hyperkalemia	32 (4.2)	2.9	40 (5.2)	3.4

ID-DD, incident-dialysis dependent; PEY, patient-exposure years; TEAE, treatment-emergent adverse event.

OT+28: a TEAE occurred (or a pre-existing condition worsened) during the treatment period and within 28 days of the last dose of study drug. Patients with >1 event in a category were counted once for that category.

Based on Medical Dictionary for Regulatory Activities, version 20.0.

PEY for each patient = (last dose − date first dose date + 1) / 365.25. Incidence rate (per 100 PEY) = 100 × number of patients with events / PEY.

Table 5

TESAEs occurring in ≥1% of patients in either treatment group (OT+28)∗

Preferred term†	Roxadustat (n = 760)		Epoetin alfa (n = 766)
	n (%)	PEY = 1098.2 incidence rate (per 100 PEY)‡	n (%)	PEY = 1189.5 incidence rate (per 100 PEY)‡
Any TESAE	318 (41.8)	29.0	318 (41.5)	26.7
Arteriovenous fistula thrombosis	44 (5.8)	4.0	26 (3.4)	2.2
Pneumonia	36 (4.7)	3.3	38 (5.0)	3.2
Peritonitis	20 (2.6)	1.8	16 (2.1)	1.3
Sepsis	18 (2.4)	1.6	11 (1.4)	0.9
Fluid overload	16 (2.1)	1.5	13 (1.7)	1.1
Hypertensive crisis	11 (1.4)	1.0	17 (2.2)	1.4
Device-related infection	10 (1.3)	0.9	6 (0.8)	0.5
Urinary tract infection	10 (1.3)	0.9	6 (0.8)	0.5
Acute myocardial infarction	9 (1.2)	0.8	18 (2.3)	1.5
Cardiac failure congestive	9 (1.2)	0.8	10 (1.3)	0.8
Arteriovenous fistula site complication	8 (1.1)	0.7	3 (0.4)	0.3
Coronary artery disease	8 (1.1)	0.7	4 (0.5)	0.3
Hypotension	8 (1.1)	0.7	7 (0.9)	0.6
Death	6 (0.8)	0.5	10 (1.3)	0.8
Gangrene	6 (0.8)	0.5	8 (1.0)	0.7
Hyperkalemia	6 (0.8)	0.5	9 (1.2)	0.8
Cellulitis	5 (0.7)	0.5	8 (1.0)	0.7
Cardiac arrest	4 (0.5)	0.4	9 (1.2)	0.8

PEY, patient-exposure years; TESAE, treatment-emergent serious adverse event.

OT+28: a treatment-emergent adverse event occurred (or a pre-existing condition worsened) during the treatment period and within 28 days of the last dose of study drug. Patients with >1 event in a category were counted once for that category.

Medical Dictionary for Regulatory Activities, version 20.0.

PEY for each patient = (last dose date − first dose date + 1) / 365.25. Incidence rate (per 100 PEY) = 100 × number of patients with events / PEY.

Discussion

For patients with CKD, during their first year on dialysis---the incident period---the risk for morbidity and mortality is high. In a controlled clinical trial of patients with prevalent dialysis, the patient sampling technique (i.e., recruitment) is one of convenience; patients were screened and invited to enroll based on factors allowing them to survive through their incident period to be eligible for enrollment. This survival bias imposed in a prevalent cohort must be recognized, given that US-based patients have a 3-year survival rate of only 57%. The current pooled analysis includes data from patients who were on dialysis for 2 weeks to ≤4 months and received only a limited amount of ESA and who then underwent comparative evaluation of study treatments for a mean of 1.5 years, which mitigates this bias. Thus, our findings provide insight into not only the prevalent period (most frequently evaluated in clinical trials) but also the incident period, when patients are at highest risk for adverse events in general, and CV events in particular. Importantly, the incident period is generally when a vast majority (∼80%) of patients on dialysis initiate anemia therapy, as <15% of US-based patients were treated with an ESA in the 12 months before dialysis initiation. Inclusion of patients during the incident period also encompassed a broader patient population than did the conversion studies with median dialysis vintage of 3 years by also including those who were less likely to survive the initial years of dialysis treatment.

Overall, the efficacy analysis of data pooled from patients with ID-DD CKD shows that roxadustat resulted in larger increases in hemoglobin when compared with epoetin alfa. For the key US efficacy endpoint, roxadustat was noninferior and superior to epoetin alfa for increasing hemoglobin from baseline averaged over weeks 28 to 52, regardless of rescue therapy, with a statistically significant larger change from baseline. Pooled results from patient subgroups, based on major demographic and clinical characteristics, were consistent with the overall findings. For the key EU efficacy endpoint, roxadustat was noninferior and superior to epoetin alfa and achieved a larger hemoglobin increase from baseline averaged over weeks 28 to 36 censored for rescue therapy within 6 weeks of and during the 8-week treatment period. In patients with baseline high-sensitivity C-reactive protein greater than the upper limit of normal, hemoglobin changes from baseline averaged over weeks 18 to 24 were greater in the roxadustat group compared with the epoetin alfa group. Between weeks 12 and 28, LDL cholesterol levels were significantly reduced in the roxadustat group compared with the epoetin alfa group. In general, the pooled efficacy results are consistent with those from the individual studies.

In terms of CV safety, the risks for MACE and MACE+ were significantly lower in the roxadustat group compared with the epoetin alfa group. Several analyses focusing on individual component events, and their various combinations, were mostly consistent with the main results.

Consistent with the general tolerability profile for roxadustat, most patients in the roxadustat and epoetin alfa groups had at least 1 treatment-emergent adverse event.

From an analytical perspective, the ID-DD population is a subgroup of the overall DD population who participated in the roxadustat phase 3 clinical trial. As such, this was a prespecified subgroup of interest, resulting in proactive efforts to recruit the >1500 patients who had newly initiated dialysis. Nevertheless, it is appropriate to acknowledge that, although the sample size is the largest of ID-DD patients studied to date, it is smaller than the overall population and subject to variability in the point estimate related to its size.

From a clinical perspective, the ID-DD population represents the “universe” of all dialysis patients who begin dialysis. All patients on dialysis are “incident to dialysis” at some point in time. Because different subgroups experience different mortality rates (e.g., age is directly correlated with mortality, and people with diabetes also experience greater risk), the snapshot of patients on dialysis---when categorized as such by vintage---changes with time. It can be argued that the results from a study of patients with a mean vintage of several years may not be generalizable to patients who did not survive to that point at baseline. Furthermore, because anemia therapy is typically initiated during the incident period in the majority of patients starting dialysis, and the use of anemia treatment is generally long term, this population provides information on the full spectrum of patients’ dialysis experience. The evaluation of anemia therapy started in the incident period and continued into the prevalent dialysis period, as done in this pooled analysis, provides a clinically meaningful comparison of roxadustat versus epoetin alfa.

Lower rates of MACE and MACE+ were observed with roxadustat in this clinical study setting where patients treated with roxadustat or epoetin alfa generally required initiation of treatment and correction of anemia, followed by maintenance treatment. MACE+ results are consistent with those of the overall DD population, where a reduction in MACE+ risk with roxadustat was also observed. Although these studies were not designed to identify the mechanism(s) for these findings, one potential explanation may be the lower erythropoietin exposure with roxadustat versus treatment with erythropoietin analogs, and it may be notable that previous studies have demonstrated reduced ESA resistance, CV events, and mortality in patients on dialysis living at high altitude.

This pooled analysis of data from patients with ID-DD CKD enrolled in the phase 3 clinical trials of roxadustat was defined a priori to evaluate the efficacy and safety of this therapy in this vulnerable and representative population of patients new to dialysis, when most patients initiate anemia treatment. Roxadustat not only corrected and maintained hemoglobin at least as efficaciously as epoetin alfa, but also showed a decreased requirement for IV iron supplementation. Most importantly, roxadustat versus epoetin alfa decreased CV risk, as evidenced by the 30% and 34% risk reductions for MACE and MACE+, respectively.

Disclosure

RP serves as a consultant for AstraZeneca, DaVita, and FibroGen, Inc., and holds stock in DaVita. SF has received research funding from and has consulted for Akebia, AstraZeneca, and FibroGen. LS, RL, KGS, MZ, TTL, and K-HPY are employees of FibroGen and hold stock and/or stock options in the company. MTH was an employee of AstraZeneca at the time of this study; he continues to hold stock in AstraZeneca. LF and JH are employees of AstraZeneca and hold stock and/or stock options in AstraZeneca.