Hepcidin-guided screen-and-treat interventions against iron-deficiency anaemia in pregnancy: a randomised controlled trial in The Gambia

Background: WHO recommends daily iron supplementation for pregnant women, but adherence is poor because of side-effects, effectiveness is low, and there are concerns about possible harm. The iron-regulatory hormone hepcidin can signal when an individual is ready-and-safe to receive iron. We tested whether a hepcidin-guided screen-and-treat approach to combat iron-deficiency anaemia could achieve equivalent efficacy to universal administration, but with lower exposure to iron. Methods: We did a three-arm, randomised, double-blind, non-inferiority trial in 19 rural communities in the Jarra West and Kiang East districts of The Gambia. Eligible participants were pregnant women aged 18–45 years at between 14 weeks and 22 weeks of gestation. We randomly allocated women to either WHO’s recommended regimen (ie, a daily UN University, UNICEF, and WHO international multiple-micronutrient preparation [UNIMMAP] containing 60 mg iron), a 60 mg screen-and-treat approach (ie, daily UNIMMAP containing 60 mg iron for 7 days if weekly hepcidin was <2·5 μg/L or UNIMMAP without iron if hepcidin was ≥2·5 μg/L), or a 30 mg screen-and-treat approach (ie, daily UNIMMAP containing 30 mg iron for 7 days if weekly hepcidin was <2·5 μg/L or UNIMMAP without iron if hepcidin was ≥2·5 μg/L). We used a block design stratified by amount of haemoglobin at enrolment (above and below the median amount of haemoglobin on every enrolment day) and stage of gestation (14–18 weeks vs 19–22 weeks). Participants and investigators were unaware of the random allocation. The primary outcome was the amount of haemoglobin at day 84 and was measured as the difference in haemoglobin in each screen-and-treat group compared with WHO's recommended regimen; the non-inferiority margin was set at −5·0 g/L. The primary outcome was assessed in the per-protocol population, which comprised all women who completed the study. This trial is registered with the ISRCTN registry, number ISRCTN21955180. Findings: Between June 16, 2014, and March 3, 2016, 498 participants were randomised, of whom 167 were allocated to WHO's recommended regimen, 166 were allocated to the 60 mg per day screen-and-treat approach, and 165 were allocated to the 30 mg per day screen-and-treat approach. 78 participants were withdrawn or lost to follow-up during the study; thus, the per-protocol population comprised 140 women assigned to WHO's recommended regimen, 133 allocated to the 60 mg screen-and-treat approach, and 147 allocated to the 30 mg screen-and-treat approach. The screen-and-treat approaches did not exceed the non-inferiority margin. Compared with WHO's recommended regimen, the difference in the amount of haemoglobin at day 84 was −2·2 g/L (95% CI −4·6 to 0·1) with the 60 mg screen-and-treat approach and −2·7 g/L (–5·0 to −0·5) with the 30 mg screen-and-treat approach. Adherence, reported side-effects, and adverse events were similar between the three groups. The most frequent side-effect was stomachache, which was similar in the 60 mg screen-and-treat group (82 cases per 1906 person-weeks) and with WHO's recommended regimen (81 cases per 1974 person-weeks; effect 1·0, 95% CI 0·7 to 1·6); in the 30 mg screen-and-treat group the frequency of stomachache was slightly lower than with WHO's recommended regimen (58 cases per 2009 person-weeks; effect 0·7, 95% CI 0·5 to 1·1). No participants died during the study. Interpretation: The hepcidin-guided screen-and-treat approaches had no advantages over WHO's recommended regimen in terms of adherence, side-effects, or safety outcomes. Our results suggest that the current WHO policy for iron administration to pregnant women should remain unchanged while more effective approaches continue to be sought. Funding: Bill & Melinda Gates Foundation and the UK Medical Research Council. Full details of the study design are in the appendix (pp 1, 2) and the published trial protocol.22 The HAPn study is a randomised, double-blind, proof-of-concept, non-inferiority trial to assess WHO's recommended daily iron regimen with two screen-and-treat approaches. We did the study in 19 rural communities in the Jarra West and Kiang East districts of The Gambia. In these locations, anaemia is common and malaria endemicity is low, heterogeneous, and seasonal. Nurse midwives and fieldworkers identified and screened pregnant women at their first antenatal care visits (day 0) at two health facilities (Soma Health Centre, Soma Town, Jarra West; and Kaiaf Health Centre, Kaiaf Town, Kiang East), obtained informed consent, and gathered demographic information. Women aged 18–45 years were eligible for randomisation if gestational age was 14–22 weeks. Gestational age was assessed by either self-reported first date of last menstrual period or, if the woman could not recall this information, by fundal height. We excluded women if they were unlikely to remain in the area for the duration of the study, had severe anaemia (haemoglobin concentration <70 g/L), had a serious illness, had chronic disease, or self-reported a history of previous pregnancy complications (eg, repeated miscarriage or abortions, pre-eclampsia or eclampsia). At enrolment (day 0), women were provided with long-lasting insecticide-treated bed nets. Any woman found to have a concentration of haemoglobin lower than 70 g/L during the trial was treated as per the Gambian national protocol. The trial was approved by the Medical Research Council (MRC) Unit The Gambia Scientific Coordinating Committee (SCC), Joint Gambia Government MRC ethics committee (SCC 1357, amendments L2014.56v2), and the London School of Hygiene & Tropical Medicine (LSHTM) ethics committee (no 7168). The trial was overseen by a data safety monitoring board, trial steering committee, and trial monitor, and it was done according to Good Clinical Practice standards supervised by the MRC Unit The Gambia at LSHTM (MRCG@LSHTM) Clinical Trials Office. All participants gave written informed consent. At screening (day 0), eligible women were randomly allocated (1:1:1) using computer-generated numbers to one of three intervention arms: (1) WHO's recommended regimen of daily supplementation with UN University, UNICEF, and WHO international multiple micronutrient capsules (UNIMMAP) containing 60 mg iron as ferrous fumarate (the reference group); (2) weekly screening of plasma hepcidin for 12 weeks, every time succeeded by either daily supplementation for 7 days with UNIMMAP containing 60 mg iron if the concentration in plasma of hepcidin was less than 2·5 μg/L or daily supplementation for 7 days with UNIMMAP containing no iron if hepcidin levels were 2·5 μg/L or higher (the 60 mg screen-and-treat group); or (3) screen-and-treat supplementation as described for the 60 mg screen-and-treat group but with UNIMMAP containing 30 mg iron (the 30 mg screen-and-treat group). Calculation of the hepcidin threshold of less than 2·5 μg/L to define ready-and-safe to receive iron has been described previously.19 Randomisation was based on a permuted block design (block size of nine) with stratification by haemoglobin (above and below the median concentration of haemoglobin of the respective enrolment day) and gestational age (14–18 weeks or 19–22 weeks), to account for natural differences in haematological and iron status. Participants and the research team (except for the data manager) were unaware of group allocation and supplementation type throughout the fieldwork. Supplements were prepacked weekly by the field coordinator using computer-generated lists accounting for each participant's preceding hepcidin value. UNIMMAP was produced in three variants containing 60 mg, 30 mg, or no iron (DSM Nutritional Products, Johannesburg, South Africa) as identical gelatine capsules, packed in tubs under Good Manufacturing Practice conditions. All formulations also contained 400 μg folic acid and 13 other micronutrients (appendix p 3). Participants were instructed to take one capsule a day with water or another drink. The intervention started at day 0 (the day of screening, enrolment, and randomisation) and continued for 84 days or until delivery, whichever came first. At screening (day 0), qualified personnel recorded the participant's medical history, did a medical examination, and collected a sample of venous blood (5–7 mL) for field measurement of haemoglobin (HemoCue Hb301 analyser; HemoCue, Ängelholm, Sweden) and to do the malaria rapid test (Alere Bioline Malaria Ag Pf, Abbot, Seoul, South Korea). If a blood sample was positive for Plasmodium falciparum infection on the malaria rapid test, we followed up with microscopy to confirm the presence of P falciparum parasites. Blood samples were transferred on ice to the laboratory at MRCG@LSHTM Keneba fieldstation for a full blood count (Medonic M Series; Boule Diagnostics, Spånga, Sweden) and assessment of plasma hepcidin. Plasma hepcidin was assayed by an ELISA with a detection range of 0·049–25·0 μg/L (hepcidin-25 [human] EIA Kit; Peninsula Laboratories International, San Carlos, CA, USA). The assay was validated as part of a worldwide harmonisation exercise.23 Hepcidin was quantified as single measurements to allow results within 24 h after blood collection and because of cost (appendix p 4). We also measured amounts in serum of ferritin, iron, unbound iron binding capacity, transferrin saturation, soluble transferrin receptor, C-reactive protein, and α1-acid glycoprotein, using an automated analyser (Cobas Integra 400 plus; Roche Diagnostics, Rotkreuz, Switzerland). On day 2 and weekly thereafter, every participant was seen by a fieldworker who counted remaining supplements, measured axillary temperature, recorded self-reported side-effects, and gave the next week's supply of tablets. At day 14, day 49, and day 84, venous blood (5–7 mL) was gathered for assessments and processing, as described for day 0. At day 7 and weekly thereafter (except when venous blood was collected), field staff collected fingerprick capillary blood samples. At every timepoint, haemoglobin was measured by the HemoCue analyser, P falciparum infection was measured by the malaria rapid test, and hepcidin was assayed to ascertain subsequent allocation of iron or no iron in the two screen-and-treat groups. To maintain masking of the treatment allocation, participants in the reference group also had weekly fingerprick blood samples collected and hepcidin concentrations analysed, even though the results did not affect subsequent supplement allocation. At day 0, day 14, day 49, and day 84, we used freshly washed red blood cells for malaria growth assays. Remaining plasma was stored at −20°C for iron and bacterial growth assays. Day 14 was selected for the ex-vivo malaria susceptibility assays as a time when there would most likely be a high level of reticulocytosis. Day 49 was then selected as the midpoint between days 14 and endpoint at day 84. Reticulocyte counts were assessed by fluorescence-activated cell sorting of CD71-positive cells. Gambian national guidelines stipulate that pregnant women should receive intermittent preventive treatment against malaria with sulfadoxine and pyrimethamine, beginning with the first dose at 16 weeks of gestation and then at least two other doses with an interval of 1 month between them. To ensure no interference with the malaria susceptibility assays, we arranged for participants to receive their first dose of sulfadoxine and pyrimethamine immediately after the blood sample was taken on day 49. We measured ex-vivo growth rates of P falciparum parasites in fresh red blood cells and of four sentinel bacterial species (Staphylococcus epidermidis, Staphylococcus aureus, Salmonella enterica, and Escherichia coli) in heat-inactivated serum as proxy safety indices, using methods described previously (appendix p 4).15, 24 The bacterial species were selected as frequent causes of sepsis in low-income settings and as representing a range of iron-acquisition mechanisms. Assays for S epidermidis proved unreliable, with frequent absence of any growth, so these findings have been excluded from the results. The technical reasons for this lack of growth were discovered in hindsight and insufficient samples were available to rerun the tests. We monitored participants until delivery, and the outcomes of the pregnancy were registered for both mother and child (postnatal check-up within 72 h after delivery). When possible, reasons for a participant being lost to follow-up were recorded. The primary outcome was the amount of haemoglobin at day 84, measured as the difference in haemoglobin in each screen-and-treat group compared with the reference group (WHO's recommended regimen). Secondary outcomes were the prevalence at day 84 of anaemia, iron deficiency, and iron-deficiency anaemia, the total iron dose administered over the 84-day study period, adverse events, and adherence to the assigned strategy.22 Anaemia was defined as an amount of haemoglobin less than 11 g/dL. Iron deficiency was defined as a concentration in plasma of ferritin lower than 15 μg/L if C-reactive protein was lower than 5 mg/L, or plasma ferritin lower than 30 μg/L if C-reactive protein was higher than 5 mg/L. Iron-deficiency anaemia was defined as an amount of haemoglobin lower than 11 g/dL and plasma ferritin less than 15 μg/L when C-reactive protein was less than 5 mg/L, or haemoglobin lower than 11 g/dL and plasma ferritin less than 30 μg/L when C-reactive protein was higher than 5 mg/L and the ferritin index [soluble transferrin receptor:log10-ferritin] was greater than 2·0). Adverse events were defined as any untoward or unfavourable medical occurrence, including signs and symptoms associated temporally with the research procedure or trial intervention, whether considered related to the woman's participation in the research or not. Serious adverse events were investigated by a doctor and defined as any adverse event that was life-threatening or resulted in death or required admission to hospital or prolongation of admission, was a persistent or relevant disability or incapacity, was a congenital anomaly or birth defect, or was a reported maternal death, miscarriage, or stillbirth. Adherence was calculated as described in the appendix (p 5). For our sample size calculation, we used data from a previous study in neighbouring villages25 to analyse haemoglobin concentrations, which yielded an SD of 12·8 g/L. This value was used to calculate a sample size of 154 participants for each of the three arms, using a one-sided α of 2·5% with a conservative Bonferroni-type correction. Initially, a total sample size of 462 pregnant women was calculated, assuming less than 10% loss to follow-up. With a non-inferiority margin of −5·0 g/L, this number was used to provide 80% power to establish that the 60 mg screen-and-treat approach is non-inferior to WHO's recommended regimen, the 30 mg screen-and-treat approach is non-inferior to WHO's recommended regimen, and the 30 mg screen-and-treat approach is non-inferior to the 60 mg screen-and-treat approach. To ensure that the study was done across different seasons and to ensure that detailed monitoring could be achieved, we enrolled study participants in six cohorts starting from June, 2014, then roughly every 3–4 months afterwards, from September, 2014, January, 2015, April, 2015, August, 2015, and December, 2015. After the first two cohorts were enrolled, permission was obtained from the ethics committee to increase the sample size to 498, because loss to follow-up exceeded 10%. Per-protocol analysis was used to assess non-inferiority of the primary outcome. All missing values and outliers present after data lock (on March 13, 2017) were maintained. In the intention-to-treat analysis, missing values were replaced by multiple imputation (appendix p 5). Intervention effects on continuous variables were measured as the difference in mean estimates, with logarithmic transformation (ln) as appropriate. A modified intention-to-treat analysis was also done (excluding participants withdrawn before the first dose of iron supplementation), and groups were compared using linear regression analysis, with intervention entered as a dummy-coded categorical variable. For analyses of bacterial growth, differences between timepoints were assessed by repeat measures ANOVA and Scheffé's post-hoc tests. Differences between study groups were assessed by χ2 test. The number of adverse events was too low to allow meaningful analysis by type of adverse event. For every woman, we summed the counts for various types of adverse events. We used negative binomial regression to assess group differences in observed counts. Negative binomial regression was used instead of Poisson regression to account for overdispersion (ie, where the variance exceeds the mean). Effect sizes thus obtained are reported as the relative change in observed counts. Adherence was assessed as the extent to which the participant's history of supplementation coincided with the prescribed supplementation (appendix pp 5–7). The funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.