Effect of Ferric Citrate versus Ferrous Sulfate on Iron and Phosphate Parameters in Patients with Iron Deficiency and CKD

Study Setting and Population

From September 2016 to October 2018, we conducted an investigator-initiated, randomized, open-label study of 60 participants with CKD. Financial support was provided as an independent research grant by Keryx Biopharmaceuticals, Inc., a wholly-owned subsidiary of Akebia Therapeutics, Inc. The study was approved (F160318006) by the University of Alabama at Birmingham (UAB) Institutional Review Board for Human Use, and all participants provided written informed consent. This study complied with Clinicaltrials.gov registration requirements (registration number NCT02888171).

Participants were recruited from the nephrology outpatient clinics at UAB. Inclusion criteria included being ≥18 years of age, having an eGFR 15–45 ml/min per 1.73 m², and having a transferrin saturation (TSAT) ≤30% and a ferritin ≤300 ng/ml (2). Exclusion criteria included hemoglobin concentrations >13 g/dl (which was changed from >11 g/dl in the original protocol to enhance recruitment of men); severe anemia, defined as hemoglobin <8.0 g/dl in men and <7.0 g/dl in women; a known disorder of iron homeostasis; known cause of anemia other than iron deficiency or kidney disease; irritable bowel disease or inflammatory bowel disease; cirrhosis or the presence of alanine aminotransferase, aspartate aminotransferase, or bilirubin concentrations three times above the normal range; serum phosphate <3.0 mg/dl; symptomatic gastrointestinal bleeding within 12 weeks before the screening visit; receipt of any form of kidney replacement therapy; pregnancy or lactation in women; currently receiving nutritional vitamin D in dosages >2000 IU/d; receipt of any erythropoesis-stimulating agent within 4 weeks of the screening visit; receipt of any intravenous iron agent within 8 weeks of screening; receipt of a blood transfusion within 4 weeks of screening; known allergies or severe adverse reactions to prior oral iron therapy; current use of oral phosphate binders; and current use of calcitriol or any other activated vitamin D analog. Participants who were taking oral iron supplements at the time of the screening visit were allowed to participate but had to undergo a wash-out period of at least 4 weeks before randomization.

Study Protocol and Intervention

All study visits occurred on the Clinical Research Unit of the Center for Clinical and Translational Science at UAB. Participants were randomized 1:1 to receive either oral ferric citrate (2 g by mouth three times a day with meals, equivalent to 1260 mg of elemental iron per day) or ferrous sulfate (325 mg by mouth three times a day, equivalent to 195 mg of elemental iron per day) for a period of 12 weeks (Supplemental Figure 1). Randomization was accomplished by study staff using a computer-generated randomization code with permuted blocks (block size of 4–6), stratified by presence or absence of severe iron deficiency (defined as a TSAT ≤20% and a ferritin ≤100 ng/ml). Blood was obtained at baseline and at weeks 2, 6, and 12 to assess the effect of ferric citrate versus ferrous sulfate on primary and secondary outcome variables. Adherence to randomized medication assignment was assessed as the proportion of participants who took at least two thirds of the study medication, as determined by manually counting pills left in study medication bottles brought to each follow-up visit.

Outcomes of Interest

The primary estimand was the change in TSAT and serum ferritin from baseline to the end of treatment. Measurements of TSAT and ferritin were done using standard assays in the main UAB hospital laboratory. Secondary estimands included the change in hemoglobin, FGF23, and hepcidin. Hemoglobin measurements were done in the main UAB hospital laboratory using standard equipment. Both intact fibroblast growth factor 23 (iFGF23) and cFGF23 were measured in duplicate in plasma samples in the UAB Metabolism Core Laboratory/Human Physiology Core Laboratory using established ELISAs (Quidel, San Clemente, CA), with coefficients of variation (CVs) <5%. Quantitative measurements of bioactive 25-hepcidin were obtained in sera using a competitive ELISA (Intrinsic LifeSciences, La Jolla, CA) (12). The intra-assay CVs of hepcidin were 11%–19% in the lower range of hepcidin values (<30 ng/ml) and 4%–11% in the upper range of hepcidin values (>30 ng/ml), and the corresponding interassay CVs were 12%–40% in the lower range of hepcidin values and 2%–10% in the higher range. In addition, erythroferrone concentrations were measured in serum samples as a post hoc exploratory analysis using an established ELISA (Intrinsic LifeSciences), with CVs <15%. Because of the expense of measurement, hepcidin and erythroferrone were measured in a subset of participants (the first 40 individuals who completed the study, 20 randomized to both ferric citrate and ferrous sulfate) at three time points (baseline, 2 weeks, and 12 weeks). In exploratory post hoc analyses, intact parathyroid hormone concentrations were measured in serum samples from each visit using an established ELISA (Quidel) and inflammatory cytokines (IFN-γ, IL-10, IL-17A, IL-1β, IL-6, TNF-α, IL-17E/IL-25, IL-17F, IL-21, and IL22) were measured in serum samples at each visit using electrochemiluminescence assays (U-PLEX TH17 Combo 2; Meso Scale Discovery, Rockville, MD).

Statistical Analyses

Using a modified intention-to-treat analysis, all participants who were randomized, received at least one dose of study drug, and attended at least one postbaseline visit were included in the analysis. We used linear mixed-effects models to compare primary and secondary efficacy end points by randomization group, using compound symmetry to account for repeated measures within each participant. In these models, randomized group, time, and group×time terms were analyzed as fixed effects, and participants were analyzed as random effects. The primary outcome was the difference in the mean change in each outcome variable from baseline to 12 weeks. We also examined differences in the rate of change in each analyte across all time points by testing the statistical significance of group×time interaction terms in the model. In these latter models, when there was a significant effect of time, we localized individually significant changes in postbaseline time points by comparing with baseline values in the mixed linear models. We natural log-transformed FGF23 concentrations to achieve a normal distribution. The proportion of participants who experienced an adverse event at any time after the baseline visit were compared by randomization group using the Fisher exact test. Two-tailed P values <0.05 were prespecified as statistically significant. All analyses were completed using SAS version 9.4 (SAS Institute, Cary, NC).

Sample Size Calculation

The sample size estimation was powered to detect a significant difference in the mean change in TSAT and ferritin in response to oral ferric citrate versus ferrous sulfate challenge. In a prior study, treatment with ferric citrate for 12 weeks raised serum ferritin by 67% as compared with placebo (2). To detect a more modest 40% difference in the rise of ferritin in individuals taking ferrous sulfate versus ferric citrate over a similar period of intervention, we would need to recruit 48 individuals overall (24 per arm), assuming a type 1 error of 0.05 and a power of 80%. To account for 20% dropout, we aimed to recruit a total of 60 participants. We estimated that this would provide 80% power to detect at least a 20% difference in the change in TSAT between treatment groups. For these estimates, we assumed that ferritin and TSAT were normally distributed and that the difference in their mean change over time would be normally distributed.