Risk of malaria in young children after periconceptional iron supplementation

Abstract This study in Burkina Faso investigated whether offspring of young mothers who had received weekly periconceptional iron supplementation in a randomised controlled trial were at increased risk of malaria. A child safety survey was undertaken in the peak month of malaria transmission towards the end of the trial to assess child iron biomarkers, nutritional status, anaemia and malaria outcomes. Antenatal iron biomarkers, preterm birth, fetal growth restriction and placental pathology for malaria and chorioamnionitis were assessed. Data were available for 180 babies surviving to the time of the survey when their median age was 9 months. Prevalence of maternal iron deficiency in the last trimester based on low body iron stores was 16%. Prevalence of active placental malaria infection was 24.8%, past infection 59% and chorioamnionitis 55.6%. Babies of iron supplemented women had lower median gestational age. Four out of five children ≥ 6 months were iron deficient, and 98% were anaemic. At 4 months malaria prevalence was 45%. Child iron biomarkers, anaemia and malaria outcomes did not differ by trial arm. Factors associated with childhood parasitaemia were third trimester C‐reactive protein level (OR 2.1; 95% CI 1.1–3.9), active placental malaria (OR 5.8; 1.0–32.5, P = 0.042) and child body iron stores (OR 1.13; 1.04–1.23, P = 0.002). Chorioamnionitis was associated with reduced risk of child parasitaemia (OR 0.4; 0.1–1.0, P = 0.038). Periconceptional iron supplementation of young women did not alter body iron stores of their children. Higher child body iron stores and placental malaria increased risk of childhood parasitaemia.

placental iron homeostasis. Placental adaptation, allowing optimal transfer of iron from the maternal circulation to the fetus (Sangkhae et al., 2020;Scholl, 2011), may be impaired by malaria. Malaria in pregnancy is also reported as a predictor of infant haemoglobin (Hb) levels Le Cessie et al., 2002).
Cohort studies from Benin (Moya-Alvarez et al., 2017), Malawi (Jonker et al., 2012), Tanzania (Gwamaka et al., 2012), Zambia (Barffour et al., 2017), Kenya and Uganda (Muriuki et al., 2019) have reported that better iron status in young children predicted increased future malaria risk, with iron deficiency significantly decreasing odds of subsequent parasitaemia. A twofold increased risk in young children of malaria parasitaemia and clinical malaria was also seen in mothers experiencing P. falciparum infections during pregnancy (Park et al., 2020), with several studies identifying an association of increased child malaria with placental malaria in their mothers Asante et al., 2013;Awine et al., 2016;Bardají et al., 2011;Le Port et al., 2011;Schwarz et al., 2008;Sylvester et al., 2018;Sylvester et al., 2016). Given the protective effect of iron deficiency during pregnancy for malaria Kabyemela, Fried, Kurtis, Mutabingwa, & Duffy, 2008;Moya-Alvarez et al., 2015;Senga, Harper, Koshy, Kazembe, & Brabin, 2011), it follows that young children born to iron replete mothers, enhanced by routine periconceptional or antenatal iron supplementation, could be at higher risk of malaria, although the issue is complex as immunological mechanisms (Brickley et al., 2015;Broen, Brustoski, Engelmann, & Luty, 2007;Dechavanne et al., 2017;Feeney, 2020;Hviid, 2009;Hviid & Staalsoe, 2004;Park et al., 2019;Sylvester et al., 2018) and maternal factors such as breast feeding and use of antimalarial drugs are relevant (Hawking, 1954;Kakuru, Staedke, Dorsey, Rogerson, & Chandramohan, 2019;. To determine childhood malaria risk in relation to maternal and offspring iron status a cohort of children born to mothers who participated in a randomised double blind controlled safety trial of periconceptional weekly iron supplementation  was assessed during the peak month of malaria transmission towards the end of the trial. A child safety survey became necessary after publication in 2012 (post trial commencement) of two cohort studies that reported iron repletion predicted increased malaria risk in pre-school children in Malawi (Jonker et al., 2012) and Tanzania (Gwamaka et al., 2012). The objectives of this cross-sectional child safety study were firstly, in children under 2 years of age, to determine prespecified malaria-related outcomes and anaemia and iron biomarkers by trial arm, and secondly, to characterise maternal and child iron status and related factors associated with child malaria.

| MATERIALS AND METHODS
Written, informed consents were given by all individuals, with additional guardian consents provided for minors. The work described was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). The primary trial outcome was malaria prevalence at the first antenatal visit, but women were followed till delivery and gestational age, preterm birth, birthweight, placental malaria and neonatal deaths have been reported . The survey in children was introduced as a safety amendment to the registered trial protocol with a prespecified secondary outcome of malaria parasite prevalence. The main results of the study were communicated to communities at the end of the study.

| Trial procedures
The main trial was undertaken between April 2011 and January 2014 in the rural area of Nanoro in Burkina Faso where malaria is hyperendemic with seasonal transmission. The study participants were enrolled within the Nanoro Health and Demographic Surveillance catchment area which had a population of approximately 55000 inhabitants. HIV prevalence in this population was 1.2% in women aged 15-49 years and 0.76% among pregnant women (Institut National de la Statistique et de la Démographie [INSD] et ICF International, 2012). Background and published data on the trial, randomisation and design (Brabin, Gies, et al., 2019;Diallo et al., 2020;Gies et al., 2018) are summarised in Appendix S1.

| Pregnancy assessments
Nulliparous participants had been individually randomised to receive, directly observed, either a weekly capsule containing ferrous gluconate (60 mg) and folic acid (2.8 mg) (n = 980), or an identical capsule containing folic acid alone (2.8 mg) (n = 979) for up to 18 months, or till pregnancy occurred ; Appendix S1). Iron status was assessed at recruitment. For women who became pregnant, a venepuncture sample was collected for measurement of iron status and malaria infection at a study antenatal visit scheduled at around 13-16 weeks gestation (ANC1) and at a second scheduled study visit between 33 and 36 weeks gestation (ANC2). At delivery the placenta was analysed for malaria parasites. From ANC1 onwards women received standard antenatal care including daily iron and folic acid

Key messages
• Iron deficiency prevalence in young mothers in Burkina Faso was low.
• Periconceptional iron supplements did not alter placental or child malaria risk.
• Malaria parasitaemia risk was higher in children with better iron status.
• Placental malaria increased risk of child malaria.
• Intensified efforts are needed to reduce risk of placental malaria which should, in turn, reduce risk of malaria and anaemia in young children.
supplements until delivery (60-mg iron, 400-μg folic acid daily). All women received a first dose of intermittent preventive treatment with sulphadoxine pyrimethamine (IPTp) at ANC1 if gestational age was >13 weeks. Women ≤ 13 weeks gestation, if positive for malaria by rapid diagnostic test (RDT) (Bioline SD, Malaria Antigen Pf detecting P. falciparum histidine-rich protein 2), were treated with oral quinine. A second scheduled IPTp-SP dose was provided through routine antenatal care. Gestational age was estimated by ultrasound examination at ANC1 with a FF Sonic UF-4100 (Fukuda Denshi) scanner, with measurement of crown rump length in the first trimester and by biparietal diameter, femur length and abdominal circumference afterward.

| Newborn assessments
At delivery study nurses examined babies within 24-48 h of delivery, and recorded birthweight, using an electronic scale to within 10 g (SECA 384, Hamburg, Germany, precision ± 5 g for weights < 5000 g, ±10 g above 5000 g). Delayed umbilical cord clamping was routinely practised. Following hospital delivery, placental biopsies (2.5 × 1 cm) were excised from fetal and maternal sites at mid-distance between umbilical cord insertion and the placental border and placed in 10% neutral buffered formalin. Rural health centres were also trained and equipped to take placenta samples. Placental biopsies were not available from home deliveries.

| Child assessments
In the wet season in October 2013, mothers who had delivered and remained in the study area were invited to have their offspring assessed at the Clinical Research Unit of Nanoro (CRUN) Health Clinic. The follow-up child assessment was performed by experienced paediatric nursing research staff and included: medical and vaccination history (BCG and polio at birth; pentavalent diphtheria, tetanus, pertussis, hepatitis B, Haemophilus influenzae type B and polio at 2, 3 and 4 months; measles at 9 months; yellow fever at 12 months). Health cards were examined and reports of recent illness, fever and drug intake within the last 2 weeks were checked.
Anthropometric measurements were performed in duplicate by two independent observers and mean values computed. Children were weighed to the nearest 10 g with an electronic scale (SECA 384, Hamburg, Germany, precision ± 5 g for weights < 5000 g, ±10 g above 5000 g). Length was measured to the nearest 5 mm using a measuring mat (SECA 210, Germany) and mid-upper arm circumference to the nearest mm using a circumference tape (SECA 201, Germany). Clinical examination was completed with temperature. A venepuncture blood sample was collected. The volume of blood sampled differed by weight, 1 ml in EDTA and 4 ml in a dry tube for infants ≥ 5600 g and 0.5 and 3.0 ml, respectively, for lower weights. This was used for malaria microscopy and RDT for malaria, Hb, serum iron biomarkers (ferritin, transferrin receptor and zinc protoporphyrin), C-reactive protein (CRP) and a reserve filter paper blood spot. Children diagnosed with malaria, or any other concurrent health problem, received free treatment according to National Guidelines, with appropriate follow-up.

| Laboratory procedures
Blood samples were transported within 3 h from the clinic to the central project laboratory at CRUN. Sera aliquots were stored at −80 C.
Hb was measured (Sysmex automated analyser) and ZPP by fluorometry (Aviv Biomedical) on fresh whole blood. Anaemia in children ≥ 6 months was defined as Hb < 11 g dl −1 . Plasma ferritin and TfR were measured using mean values from duplicate ELISA samples RAMCO Inc.) and CRP by ELISA (EU59131IBL, GmbH). Intra-assay coefficients of variation (CVs) were all <10%. Ranges for normal controls were ferritin, 69.1-114.7 μg L −1 ; sTfR, 4.2-5.9 mg L −1 ; CRP, 5-8 mg L −1 . Definitions of iron status were based on (1) adjusted ferritin using the internal regression correction approach , allowing for inflammation as described by Mei et al. (2017), or (2) the ratio of sTfR (mg L −1 ) to log 10 ferritin (μg L −1 ), which assesses both stored and functional iron and is possibly less affected by inflammation.
Values > 5.6 in children derive from the cut-offs sTfR > 8.3 μg ml −1 and ferritin < 30 μg L −1 . This best predicted iron deficient bone marrow stores using the same assay as in the present study, in an area of high malaria transmission (Phiri et al., 2009). Body iron stores (BIS) (mg kg −1 ) using the regression-adjusted ferritin estimate were calculated using the equation derived by Cook, Flowers, and Skikne (2003): body iron (mg kg −1 ) = −[log 10 (1000 × sTfR/ferritin) − 2.8229]/0.1207 [39]. Low BIS was defined as <0 mg kg −1 . Plasma hepcidin was measured by competitive ELISA at an International Reference Laboratory (Kroot et al., 2010). A malaria RDT was performed. Malaria parasite density was obtained from the mean count of two independent readers counting the number of asexual parasites per 200 white blood cells in a thick blood film stained with 3% Giemsa, assuming a white cell count of 8000/μl. For discrepant findings (positive/negative; more than twofold difference for parasite densities ≥ 400/μl; >log 10 if <400/μl), a third independent reading was made, with the mean of the two closest observations accepted as the true value.

| Placental histology
After the delivery of the placenta, it was placed in a receptacle fetal side upwards. Using scissors, a biopsy of 2.5 cm × 1 cm was excised at mid-distance between the insertion of the umbilical cord and the placenta border and placed into a prefilled specimen container with 10% neutral buffered formalin (CellStor Pot, CellPath Ltd. Newtown SY16 4LE, UK). A 1 cm cross-section from the umbilical cord was cut with scissors at about 5 cm from the insertion and placed into the same receptacle. After turning the placenta in order to expose the maternal side upwards, a second biopsy of about the same size was excised at half distance between the centre and the border of the placenta and placed into a second container with formalin. A 10 cm × 10 cm piece of the membrane was cut with scissors and placed into the same container.
Formalin fixed specimens were stored for up to 3 months at room temperature in an air-conditioned room (20 C) at the CRUN laboratory. After transport to the department of pathology at the National University Hospital Yalgado Ouedraogo in Ouagadougou, tissue samples were processed by experienced technicians according to standard histopathological procedures.
Samples were embedded in paraffin wax following standard methods. Membranes were rolled to cylinders in order to obtain sufficient material to be cut and embedded. For each set of samples, paraffin sections 3 to 5 μm thick were placed on two slides, the fetal side of the placenta together with the cord on one slide, the maternal side of the placenta together with the membrane on another slide. Slides

| Statistical analysis
The sample size was determined from formal power calculations for the malaria trial endpoints . The number of children available for assessment per study arm is shown in Figure S1 (see Appendix S2). The primary analyses presented here are comparisons of prespecified child malaria-related outcomes and placental malaria in babies by trial arm on an intention to treat basis. Iron and inflammation biomarkers and placental chorioamnionitis were prespecified exploratory outcomes. Preterm was defined as a live birth or stillbirth that took place at least 20 but before 37 completed weeks. Fetal growth restriction (SGA) was defined as birthweight below the 10th centile for gestation and gender, indicated by standard reference data (Villar et al., 2014). Clinical malaria was defined as fever or history of fever (≥37.5 C) in the previous 48 h with parasitaemia. Outcomes were summarised by median (interquartile range) or N (%) and compared between treatment arms using ordinary or logistic regression models adjusting for child sex and age at assessment. Age was fitted with a cubic spline function with 5 degrees of freedom, with sensitivity analyses confirming that this was a sufficient representation of the non-linear age relationships. Results are presented as odds ratios for categorical outcomes or differences between arms for continuous variables with 95% CI. Where appropriate, outcomes were log transformed and the presented effect size can be interpreted as a ratio between arms.
Malaria prevalence by age in various subgroups was visualised using the fitted probabilities from a logistic regression model using a cubic spline representation with the degrees of freedom selected to capture the main features of the age trends without spurious artefacts.
Shading around the fitted lines to indicate ±1SE, and rugplots showing the location of positive and negative values (tick marks along top and bottom axes) were added where these did not obscure the presentation.
The associations between infant/maternal factors and infant malaria were assessed using logistic regression models for malaria outcomes against the relevant factors, adjusting for age and sex as above.
The associations between iron biomarkers and malaria were visualised as scatterplots of the relationship between the biomarker and age for the children with and without malaria, with cubic spline regression lines added for each group. All analyses were performed in the R statistical environment version 3.6.

| Participants
During the trial 478 pregnancies occurred, with 348 known deliveries occurring before the survey period ( Figure S1 in Appendix S2). Following exclusions due to out-migration, stillbirths and neonatal and infant deaths, there were 262 babies eligible for this study. Of these, 180 (69% of those eligible) were contacted, assessed and included in the primary analyses presented here.

| Maternal and child characteristics by trial arm
Maternal characteristics at delivery and offspring characteristics by trial arm are shown in Table 1. Maternal characteristics of children surveyed did not differ from those of children lost to follow-up (see Table S1 in Appendix S2). Maternal iron biomarkers did not differ by trial arm and prevalence of maternal iron deficiency based on low BIS at ANC2 was 16%. Prevalence of placental malaria (active and past infection) was 86% in supplemented women and 81% in controls. Babies born to iron supplemented women had shorter median gestational age and lower mean birthweight (Brabin, Gies, et al., 2019). The median age of children surveyed was 9.1 months (range 1-22 months), 20% were undernourished (Z score ≤ 2SD), 2.2% were referred for severe malnutrition (Z score ≤ 3SD) and three children had received recent haematinics. Malaria and iron biomarker outcomes are outlined in Table 2. In children ≥ 6 months, 69% were iron deficient (based on a sTfR/log ferritin ratio > 5.6), 33% had low BIS and 98% were anaemic. In children ≥ 6 months, low BIS prevalence was higher than in younger children < 6 months (33% vs. 6%, P < 0.001).
Mean Hb concentration, iron biomarkers levels or anaemia prevalence in children did not differ between trial arms, nor were there significant differences for any malaria-related outcomes. Median CRP concentration increased from 2.7 mg L −1 in children ≤ 6 months to 6.3 mg L −1 for older children (P < 0.001), but values did not differ by trial arm.
GIES ET AL. 7 months. Prevalence of all malaria infection parameters progressively increased with age ( Figure 1).
The age-specific pattern of child iron biomarkers is shown in Figure 2 in relation to malaria parasitaemia. Malaria infection was associated at all ages with lower Hb, higher CRP concentration and higher values for adjusted ferritin (P < 0.001) and BIS (P < 0.001). Similar iron biomarker differences were seen in relation to parasitaemia and/or recent malaria treatment (see Figure S2). Parasitaemia risk was higher with evidence of childhood iron repletion ( The absence of a difference in child malaria outcomes or iron status by trial arm is consistent with the findings that periconceptional iron supplementation did not improve maternal iron status. There was good adherence to weekly supplementation (79% iron; 80% control; Gies et al., 2018), and the lack of effect of maternal iron supplementation was attributed to poor maternal iron absorption due to the high prevalence (>40%) of chronic untreated asymptomatic malaria parasitaemias . In principle screening for iron deficiency, prevalence would be implemented before introduction of routine supplementation programmes. WHO in 2020 suggested that a prevalence of 5-19.9% iron deficiency, based on ferritin concentrations, might be considered a mild public health problem (World Health Organisation, 2020). Based on adjusted ferritin alone (<15 μg L −1 ), prevalence at ANC1 in our study was 12.8% . Using BIS as an indicator, 16% of mothers in the last  providing periconceptional iron routinely to these women, most of whom were not iron deficient, also increased risk of preterm birth and on this basis should not be recommended (Table 1; Brabin, Gies, et al., 2019).
Younger babies would be more likely to have recrudescences from congenital malaria which, in this area, affected 10% of newborns (Natama et al., 2017). Malaria parasite prevalence (with or without prior treatment) and RDT positivity in children were, with the exception of ZPP, strongly associated with biomarkers of better iron status and iron repletion at all ages (Table 3). ZPP concentration was higher in children with malaria, but its specificity in children, as well as in pregnant women, is reduced secondary to the anaemia of inflammation (Asobayire, Adou, Davidson, Cook, & Hurrell, 2001;Senga, Koshy, & Brabin, 2012;Stoltzfus et al., 2000).
Preterm delivery and low birthweight lead to higher iron requirements for growth and are factors likely to contribute to the high prevalence of iron deficiency and anaemia in children older than 6 months in this area (Brabin, Gies, et al., 2019;Domellöf, 2017).
Malaria contributes an added risk for anaemia and infection had been experienced by approximately 80% of children reaching 6 months of age. Comparable high iron deficiency and anaemia prevalence has been reported in the Eastern region of Burkina Faso in children 6-12 months of age (Bliznashka, Arsenault, Becquey, Ruel, & Olney, 2020).
An early peak in malaria prevalence occurred at around 4 months of age followed by uniformly rising prevalence with increasing age. An early peak in prevalence was similarly associated with placental infection in a cohort of Cameroonian children (Le Hesran et al., 1997).
However in a cross-sectional survey an early peak may be an artefact since age reflects intensity and cumulative exposure to malaria. Children at 4 months may have had more exposure (almost a full season at the time of the survey), compared with younger children who have had less exposure, and older ones who have developed some acquired immunity from a previous season's exposure . This would occur in children of mothers with and without placental malaria, so does not explain the higher malaria prevalence in those with placental malaria. A recent metaanalysis of 11 studies found an overall malaria risk in young children (adjusted hazard ratio 1.46, 95% CI 1.07-2.0, P < 0.001) associated with malaria in pregnancy but via indeterminate mechanisms (Park et al., 2020). Placental malaria may influence primarily risk for the first infant malaria episode (Bouaziz et al., 2018;Le Hesran et al., 1997;Le Port et al., 2011), and better child iron status could increase subsequent malaria risk (Georgiadou et al., 2019) and possibly risk of nonmalarial infections Rachas et al., 2012).
Malaria risk in young children is also influenced by other maternal factors, although this study found no association with gesta-  (Natama, Moncunill, et al., 2018). Immunological interactions are especially relevant in primigravidae who are at increased risk of malaria in high transmission areas. In the present study, all women were primigravidae and more than 90% were adolescent . The lower risk of child malaria observed in this study with chorioamnionitis may be attributable to treatment of related maternal symptoms of vaginal discharge from genital infection with metronidazole, which has antimalarial effects and hence could reduce placental parasite load (Pallangyo, Minjas, & Sarda, 1986).
Metronidazole for lower genital infections with vaginal discharge was available free of charge both before and during pregnancy and 8.2% reported a discharge at least once during pregnancy (Brabin et al., 2017; background data, Appendix S1). Strengths of the present study include assessment of gestational age by ultrasound, lack of confounding due to maternal parity or age, inclusion of key determinants of maternal and child malaria and completion of this cross-sectional survey during the peak month of malaria transmission . The findings should be generalizable to comparable areas with high malaria transmission. A limitation was the reduced sample size for the child survey. This was predicated by the primary safety trial endpoints in mothers and the requirement to complete the survey during the peak malaria transmission month. The sample size for the child safety survey was reduced due to perinatal losses , as well as migration outside the study area which contributed to failure to attend for delivery, or led to postnatal migration and failure to attend for the child survey. Mobility was high among girls as they married and moved to join their husbands in nonstudy areas (Campaoré, Gies, Brabin, Tinto, & Brabin, 2018). Village deliveries and those occurring en route to the hospital reduced placental biopsy samples. Attrition and child nutritional and age profiles were equivalent between trial arms.

| CONCLUSIONS
High prevalence of placental and child malaria was associated with better iron status in young mothers and children. Intensified efforts are needed to reduce risk of placental malaria which should, in turn, reduce risk of malaria and anaemia in young children.

ACKNOWLEDGMENTS
The study on which these data are based was a collaborative effort of many individuals involved in field work, data management, laboratory tests and data monitoring . We also acknowledge the community field workers who worked for the

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.

CONTRIBUTIONS
BB and SG designed the research. BB was the Principal Investigator for the main randomised controlled trial and recipient of the grant.
SG, SD, BB and HT conducted the field research. SD and OL conducted and supervised the laboratory research. SAR analysed the data. BB and SR wrote the paper. All authors reviewed and approved the final manuscript.

DATA AVAILABILITY STATEMENT
Until placed in a public repository, study data can be requested from the corresponding author and made available following an end user data agreement and sponsor approval.