Protocol and statistical analysis plan for a randomized controlled trial of the effect of intravenous iron on anemia in Malawian pregnant women in their third trimester (REVAMP – TT)

Ethics and dissemination

National Health Sciences Research Committee of Malawi Approval – NHSRC REF. Number: 20/11/2622. Human Research Ethics committee at WEHI – HRE REF. Number: 20/25.

Trial registration number

Australia and New Zealand Clinical Trial Registry – ANZCTR 12621001239853.

Protocol

This manuscript summarizes the trial protocol and statistical analysis plan (SAP).

In short, pregnant women attending antenatal care at eight primary health centres in Zomba district, Malawi (Figure 1), will be prescreened. Women in their third trimester, with negative malaria rapid diagnostic tests (RDT) and a capillary hemoglobin of 10 g/dL or less will be invited to be enrolled in the trial (Figure 2). Women will be randomized to receive either intravenous FCM or standard-of-care oral iron. All trial procedures and follow-up times are detailed in Table 1. This protocol describes the main trial looking at pregnant women in their third trimester and following them until one-month postpartum, after which we will report on the primary and secondary objectives. Follow-up of mothers and infants from one-month postpartum to 12 months postpartum consists of a range of exploratory economic, biological, cognitive and clinical outcomes. These analyses are beyond the scope of this protocol, as are the exploratory outcomes collected up to 12-months postpartum and will be reported separately. The trial design is summarized in Figure 2.

Figure 1. Map of Malawi with location of Zomba district, the coordinating trial site in Zomba (red) and the recruiting government primary health centers (white).

Figure prepared in QGIS2.10.

Figure 2. REVAMP-TT Trial schema.

Trial design with visit timings, visit numbers (standardized to match those of REVAMP trial – P.02/18/2357) and main trial activities are represented. FCM, Ferric Carboxymaltose; iv, intravenous; mth, months; REVAMP-TT, Randomized controlled trial of the Effect of intraVenous iron on Anemia in Malawian Pregnant women - Third Trimester; V0, visit 0; V1, visit 1; V4, visit 4; V7, visit 7; V8, visit 8; V9, visit 9; V10, visit 10; V11, visit 11; V12, visit 12; wks, weeks. Primary outcome measured at V4 or V7, indicated by the red bar connecting both visits. Dotted lines represent follow-up visits not included in the main trial. Figure prepared using BioRender.

Trial design

REVAMP-TT is a two-arm open-label individual-randomized controlled trial in pregnant women with moderate or severe anemia (capillary Hb<10g/dL) during their third trimester (27-35 weeks gestation). Participants will be randomized to receive either parenteral iron – in the form of FCM– or standard-of-care oral iron. Women are recruited from health centers across Zomba district in Southern Malawi (Figure 1). As REVAMP-TT is an effectiveness trial, the standard-of-care oral iron arm will be provided to participants following the established recommendations issued by the Malawi Ministry of Health. Screening of hemoglobin for eligibility and administration of the intervention (i.e., provision of the intravenous iron) takes place in government health centres. Mothers and their babies will be followed up to 12 months postpartum. The trial was registered with the Australia and New Zealand Clinical Trial Registry (ANZCTR 12621001239853) prior to the start of recruitment and followed the SPIRIT guidelines for randomized trials²¹ throughout the design and reporting of this protocol.

Study site and population

The trial is based at the Training and Research Unit of Excellence (TRUE) center at Zomba Central Hospital in Southern Malawi. The trial has eight government health centers across Zomba district available for screening, recruitment and provision of the intravenous drug as part of REVAMP-TT, namely: Likangala, Bimbi, Lambulira, Domasi, Naisi, Matawale, Sadzi and City clinic (Figure 1). These facilities receive close to 1500 antenatal visits per month. All subsequent study visits will be done at TRUE center in Zomba.

Participants are eligible for the trial if they have a confirmed singleton pregnancy 27- to 35-weeks’ gestation either using last menstrual period or fundal height, have a capillary Hb concentration of less than 10 g/dL, measured by HemoCue 301+, and have negative malaria RDT. Additionally, the participants must be afebrile with no evidence of septicemia, reside and plan to give birth within the study catchment area of Zomba district and be able to provide written informed consent (or assent if <18 years old).

Women are excluded if they were previously enrolled in the REVAMP trial^16,18 or are actively participating in another intervention trial. Women with known hypersensitivity to any of the study drugs, clinical symptoms of malaria or other infection, any condition requiring hospitalization, known history of sickle cell or sickle-hemoglobin C anemia, clinically unstable with a low hemoglobin level requiring a blood transfusion (usually Hb <5g/dL), or pre-eclampsia are excluded from participating in this trial. HIV positivity is not an exclusion.

Randomization, treatment allocation and blinding

Using a randomization schedule of randomly permuted blocks stratified by site to achieve balance between the arms within each site, participants are randomly allocated 1:1 to one of the two treatment arms. An independent statistician computer-generated the randomization list, and sealed, opaque envelopes were used to randomly assign participants. Despite the trial being open-label, laboratory scientists measuring hemoglobin concentration, midwives collecting birth outcome data, and investigators and researchers at the Walter and Eliza Hall Institute, Melbourne, Australia (including data managers and statisticians) are all blinded to the treatment allocation throughout the trial duration until the database is locked and ready for unblinding.

Study interventions and procedures

This trial takes place in primary health centers spread across Zomba district in Malawi, with several of the health centers considered to be remote – Likangala, Bimbi and Lambulira. (Figure 1). As REVAMP-TT is delivered through government health centers using the available government health centers’ infrastructure, we used a co-design approach to develop interventions to support its implementation (REVAMP-IS study¹⁷). Co-design is a two-staged approach to health systems and service improvements that typically involves an information-gathering stage followed by a second stage in which end users jointly develop the strategies with health service providers and the project leaders²². Where the workload permits, government nurses directly administer the intervention and, in all clinics, government nurses received training and are encouraged to assist the trial team. Women in the intervention groups receive intravenous iron in the form of ferric carboxymaltose (Vifor Pharma) at a dosage of 20 mg/Kg body weight if <50 kg and a maximum of 1000 mg if body weight ≥50 kg given over 15 minutes in 250 mL of normal saline, once on recruitment day. The participant is observed for a minimum of 30 minutes before being discharged home. Women in the control group receive oral iron-200 mg ferrous sulphate (approx. 65 mg elemental iron) twice daily for the remainder of pregnancy. Importantly, the oral iron is given under real-life health service delivery conditions where the participant is given three months of oral iron at presentation to the antenatal clinic. This strategy is being employed as it is a component of the key hypothesis of this trial that there is a reduced effect of oral iron on maternal anemia due to poor adherence to the full course of treatment. In addition, all participants receive IPTp with SP, 1500 mg sulfadoxine and 75 mg pyrimethamine (three fixed tablets of SP strength at 500 mg/25 mg) as recommended in the national guidelines during study follow up visits²³. As part of the safety assessment, IPTp-SP post-randomization is directly observed by study staff.

All study activities at each study visit are summarized in Table 1.

Data collection and management

Data are recorded in digital form with REDCap using electronic tablets and backed up nightly to a local backup server at TRUE, Blantyre, Malawi, with a de-identified fortnightly backup to the servers at the Walter and Eliza Hall Institute, Melbourne, Australia. REDCap is hosted on infrastructure belonging to the trial’s organizational team in Malawi and is subject to the same security and backup regimen as other systems (e.g., the network file servers).

Primary outcome

The primary outcome is the proportion of women with anemia (defined as a venous blood hemoglobin < 11 g/dL) at 36 weeks gestation or delivery, whichever comes first. Gestational age in this setting is evaluated by last menstrual date and/or measurement of fundal height. These were chosen due to their common use in the government primary care centers where there is no availability of ultrasound, thus ensuring the generalizability of the results.

Secondary outcome(s)

Maternal secondary outcomes include hemoglobin, ferritin, anemia, iron deficiency, iron-deficiency anemia collected at 36 weeks gestation, delivery, and at one-month postpartum. Iron-deficiency is determined by measuring levels of serum ferritin and setting thresholds at ferritin<15ug/L or ferritin<30ug/L if C-reactive protein>5mg/L. Iron deficiency anemia indicates Hb<11g/dL and serum ferritin<15ug/L or ferritin<30ug/L if C-reactive protein>5mg/L²⁴.

Neonatal outcomes include birth weight and low birth weight (birth weight <2,500g) within 24 hours of delivery and infant length-for-age z-score, weight-for-age z-score, and weight-for-length z-score as well as iron and hematological parameters collected at one-month postpartum.

Safety outcomes

Safety outcomes for this trial include maternal and infant adverse events and severe adverse effects, hypophosphatemia (defined as Phosphate (PO4 <0.80 mmol/L), and inflammation status (threshold for inflammation set at C-reactive protein>5mg/L), measured at the same timepoints as specified for primary and secondary outcomes detailed above.

Data Monitoring Committee

An independent Data Monitoring Committee (DMC) oversees the trial and is comprised of three international experts in clinical trials, obstetrics, epidemiology and statistics. The DMC recommends to the sponsor and investigators whether to continue, modify or terminate the trial on ethical grounds. The DMC evaluated the results of the sample size re-estimation (see sample size estimation and power calculation) and recommended that no sample size change was necessary.

Statistical analysis principles / overview

Efficacy analyses will be performed according to the participants randomized treatment group and safety analyses will be performed according to the participants actual treatment group. Frequencies and percentages will be reported for categorical variables and mean and standard deviation or median and quartiles (25^th and 75^th percentile) for continuous variables. Treatment effects will be presented as point estimates with two-sided 95% confidence intervals (CIs). A two-sided P value <0.05 will be used to indicate statistical significance for the primary outcome and birth weight. The Holm procedure²⁷ will be used to control the family-wise error rate at 0.05 for maternal outcomes (hemoglobin concentration, and serum ferritin concentration at 36 weeks gestation or delivery (whichever comes first) and at one-month postpartum) and neonatal outcomes (hemoglobin concentration, serum ferritin concentration and weight at one-month of age) separately. Specifically, the Holm procedure is a stepwise multiple test method that controls the probability of one or more type 1 errors to be at most 5%, whereby null hypotheses are tested step-down using adjusted individual significance levels until no further rejections can be made²⁶. Analyses will be conducted using Stata/SE (StataCorp. 2019. College Station, TX: StataCorp LLC). The open-access software R can be used for sample size calculations where Stata/SE is unavailable.

Main analysis of the primary outcome

A log-binomial regression model will be used to examine maternal anemia at the primary timepoint – 36 weeks’ gestation or delivery, whichever comes first – with study participants as a random intercept to account for multiple timepoints. A treatment and treatment-by-study-visit interaction and an adjustment for the stratification variable site (used during randomization) will be included in the model. The reference group for the model will be the standard-of-care oral iron treatment group. The prevalence ratio of IV iron versus standard-of-care oral iron will be estimated from this model as the treatment effect. In the event of non-convergence, we will use alternative models.

Analysis of secondary outcome(s)

Identical to the analysis of the primary outcome repeated time-point binary outcomes (anemia, moderate/severe anemia, iron deficiency, iron deficiency anemia) will be analyzed. A likelihood-based longitudinal data analysis model²⁸ will be used to examine secondary repeated time-point continuous outcomes (hemoglobin concentration and ferritin concentration). Unstructured variance-covariance among the repeated measurements and a common baseline mean will be assumed by the model for the two treatment arms. The main effects of the model will be the stratification factor (site), the treatment and treatment by study visit interaction, and the study visit (timepoint) as a categorical variable. We will consider alternative structures (first-order autoregressive, Toeplitz, compound symmetry) in the event of non-convergence. The mean change between IV iron and standard-of-care oral iron will be used to estimate the treatment effect using this model. Before analysis, ferritin (μg/L) will be log_e transformed, and the treatment effect will be expressed as a geometric mean ratio. A linear regression model, adjusting for the stratification factor (site) will be used to analyze birth weight. The absolute difference in mean birth weight between IV iron and standard-of-care oral iron will be used to estimate the treatment effect based on this model.

Using the same analysis as used for birth weight, birth length within 24 hours of birth and hemoglobin concentration, ferritin concentration, length-for-age z-score, weight-for-age z-score, and weight-for-length z-score at one-month of age will be analyzed. If the variables are considered skewed, appropriate transformations may be applied to them prior to fitting the model. A log-binomial regression model, adjusting for the stratification factor (site), will be used to examine low birth weight and stillbirth. The risk ratio of IV iron versus standard-of-care oral iron will be estimated from this model as the treatment effect. In the event of non-convergence, we will use alternative models.

Analysis of safety outcome(s)

Analyses of maternal and neonate adverse event outcomes, and neonate safety biomarkers (hypophosphatemia and malaria RDT positive) will be analyzed similarly to analyses of secondary single time-point binary outcomes detailed above. Maternal safety biomarkers (hypophosphatemia, inflammation and malaria RDT positive) will be analyzed using the same analysis as the primary maternal outcome (anemia).

Additional analyses

Using a model similar to the main analysis model of the primary outcome described above, the treatment effect at 36 weeks’ gestation will be examined for the primary outcome. Adjusted analyses for pre-specified covariates and analysis excluding participants with protocol violations (e.g., multiple births) will be included in additional analyses of primary, key secondary and other secondary outcomes. To determine the heterogeneity of the differences between IV iron and standard-of-care oral iron six pre-specified subgroup analyses (gravidity (primigravida vs. multigravida), baseline HIV status (positive vs negative), baseline severe anaemia status (yes vs no severe anaemia), baseline iron-deficient status (yes vs no ID), baseline iron-deficient anaemia status (yes vs no IDA), baseline inflammation status (yes vs no elevated CRP)) will be performed for primary and key secondary maternal and neonate outcomes. The missing-at-random assumption will be used to fit the model for the primary maternal outcome (anemia) and the missing-completely-at-random assumption will be used to fit the model for the primary neonate outcome (birth weight). Pattern-mixture models will be conducted to assess sensitivity of the results to these assumptions.