The Effects of 1 Egg per Day on Iron and Anemia Status among Young Malawian Children: A Secondary Analysis of a Randomized Controlled Trial

Background: Young children with diets lacking diversity with low consumption of animal source foods are at risk of iron deficiency anemia (IDA). Objectives: Our objectives were to determine the impact of supplementing diets with 1 egg/d on 1) plasma ferritin, soluble transferrin receptor (sTfR), body iron index (BII), and hemoglobin concentrations and 2) the prevalence of iron deficiency (ID), anemia, and IDA. Methods: Malawian 6-9-mo-old infants in the Mazira trial (clinicaltrials.gov; NCT03385252) were individually randomly assigned to receive 1 egg/d for 6 mo (n = 331) or continue their usual diet (n = 329). In this secondary analysis, hemoglobin, plasma ferritin, sTfR, C-reactive protein (CRP), and α-1-acid glycoprotein (AGP) were measured at enrollment and 6-mo follow-up. Iron biomarkers were corrected for inflammation. Ferritin, sTfR, BII, and hemoglobin were compared between groups using linear regression. Prevalence ratios (PRs) for anemia (hemoglobin <11 g/dL) and ID (ferritin <12 μg/L, sTfR >8.3 mg/L, or BII <0 mg/kg) between groups were compared using log binomial or modified Poisson regression. Results: A total of 585 children were included in this analysis (Egg: n = 286; Control: n = 299). At enrollment, the total prevalence of anemia was 61% and did not differ between groups. At 6-mo follow-up, groups did not differ in geometric mean concentration of hemoglobin [mean (95% CI); Egg: 10.9 (10.7, 11.1) g/dL; Control: 11.1 (10.9, 11.2) g/dL] and inflammation-adjusted ferritin [Egg: 6.52 (5.98, 7.10) μg/L; Control: 6.82 (6.27, 7.42) μg/L], sTfR [Egg: 11.34 (10.92, 11.78) mg/L; Control: 11.46 (11.04, 11.89) mg/L] or BII [Egg: 0.07 (0.06, 0.09) mg/kg; Control: 0.07 (0.05, 0.08) mg/kg]. There were also no group differences in anemia [Egg: 46%; Control 40%; PR: 1.15 (95% CI: 0.96, 1.38)], ID [PR: 0.99 (0.94, 1.05)], or IDA [PR: 1.12 (0.92, 1.36)]. Conclusions: Providing eggs daily for 6 mo did not affect iron status or anemia prevalence in this context. Other interventions are needed to address the high prevalence of ID and anemia among young Malawian children. This trial is registered at http://www.clinicaltrials.gov as NCT03385252. The Mazira Project was conducted between February 2018 and January 2019 in the Mangochi District of Malawi (clinicaltrials.gov registry {"type":"clinical-trial","attrs":{"text":"NCT03385252","term_id":"NCT03385252"}}NCT03385252). Children were randomly assigned to an intervention group, receiving 1 egg/d for 6 mo, or a control group that did not receive additional eggs. Details of the study design have been reported previously (20, 21). The study was promoted through community outreach events and study participants were recruited by home visits from household listings. Children were eligible if they were between the ages of 6.0 and 9.9 mo, were of singleton birth, and planned to reside in the catchment areas of the Lungwena or Malindi health centers for the study duration. Children were excluded based on wasting (midupper arm circumference ≤12.5 cm), severe anemia (hemoglobin ≤5 g/dL), bipedal edema, acute illness warranting hospital referral, history of egg allergy, congenital defects, or other morbidities that may impede growth or development. Children were referred to a health center if they presented with signs of severe dehydration or screened positive for wasting, bipedal edema, malaria, or severe anemia during any study visits. Caregivers were oriented to the clinic facilities, activities, purpose, and procedures of the research study and had opportunities to ask questions and discuss concerns in a group setting and privately with staff members. They provided written informed consent at enrollment by signature or thumbprint to confirm their study participation, consent to future use of collected blood samples, and right to withdraw at any time. This study followed principles of ethical conduct approved by the Institutional Review Board at the University of California, Davis and the Research Ethics Committee at the University of Malawi College of Medicine. The target sample size for the main trial was 662 children, based on the desire to detect a 0.25 SD difference between groups in the primary outcome measure of length-for-age z-score with α = 0.05 and 80% power. Children were block-randomized in groups of 10 and allocated to the egg intervention or non-intervention group in a 1:1 ratio after enrolling and completing baseline assessments. From the current block, caregivers randomly selected 1 opaque, unmarked envelope containing a card with a unique randomization code to reveal their group assignment. Study staff conducting assessments were masked to group assignments. A full description of the intervention groups has been published elsewhere (20). Briefly, each week caregivers in the egg intervention group received 7 eggs to feed the enrolled child 1 egg/d, plus 7 additional eggs to share with other household members. Study staff delivered eggs to intervention households twice per week and conducted recalls on the most recent egg feeding. The control group continued their usual diet, and their households were visited twice per week to report on the child's most recent meal. They received wash tubs, buckets, and plastic bins as participation incentives during the study and a mixed basket of foods, including eggs, at the completion of the study. This package of goods was selected to be of equal value to that of the eggs provided to the intervention group. All study participants received fabric cloth, sugar, and soap tablets after completing each visit. During the initial clinic visit and 6-mo follow-up, children were assessed for growth, development, and dietary intake. Child recumbent length and weight were converted to z-scores using the sex- and age-specific WHO Growth Standards (22). Enrollment surveys and initial household visits assessed demographic characteristics of the study child and household members, including household assets and food insecurity using the Household Food Insecurity Access Scale (HFIAS) (23) and home environment using the Home Observation for Measurement of the Environment (HOME) (24) indicator. At clinic visits, trained nurses collected venous blood samples to measure hemoglobin concentration using a portable spectrophotometer (Hemocue Hb 201; HemoCue, Inc.) and presence of malaria antigens using a rapid diagnostic test kit (SD Bioline Malaria Ag P.f/Pan; Abbott Diagnostics) with >85% sensitivity and ≥90% specificity for Plasmodium falciparum (25, 26). Blood samples were collected in lithium heparin tubes, immediately placed in a cooler with ice, centrifuged at ambient temperature for 15 min at 1040 × g, and placed in aliquots on site. Plasma samples were temporarily stored at –20°C and transported in coolers at the end of each day to a storage freezer maintained at –80°C. Aliquots were shipped on dry ice to laboratories completing plasma analyses. Plasma ferritin, sTfR, C-reactive protein (CRP), α-1-acid glycoprotein (AGP), and retinol binding protein (RBP) were measured by combined sandwich techniques with ELISA methods by the VitMin Lab (27). All analytes were measured from a single well containing 50–75 µL plasma for all children who provided a minimum of 450 µL plasma sample from blood draws. A 10% subset of samples was reanalyzed for quality assurance. Replicates of pooled plasma samples were run with each tray, and the CV for each indicator was calculated as the following: ferritin (2.3%), sTfR (3.6%), CRP (5.8%), and AGP (8.1%). A detailed statistical analysis plan was developed and posted (https://osf.io/vfrg7) prior to analysis and analyst unblinding. All data cleaning, management, and analyses were performed using de-identified data in Stata (version 15; StataCorp LLC) (28). Indices above the upper limit of detection were replaced by the maximum observed values, and indices below the lower limit of detection were replaced by zeroes and converted to half the limit of detection as needed for analytical models performed on the log scale. Ferritin and sTfR were corrected for subclinical inflammation on the log-transformed scale using the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) regression approach. This method adjusts for elevated CRP and AGP above the lowest decile set by an external reference group of preschool-aged children (29, 30). Body iron index (BII) was calculated according to Cook’s formula (31) by applying constants to the ratio of sTfR:ferritin using inflammation-adjusted values, such that the quantitative estimates of iron stores are indicated by positive values and the magnitude of iron deficit is depicted with negative values (32). Dichotomous variables were created for anemia (hemoglobin <11 g/dL), ID (ferritin <12 µg/L, sTfR >8.3 mg/L, or BII <0 mg/kg), and IDA (both anemia and ID) (33–35). Descriptive statistics were calculated for demographic characteristics, iron indices, and inflammation (CRP >5 mg/L or AGP >1 g/L) (36) at enrollment by group assignment. Linear regression models assessed groupwise differences in mean concentrations of hemoglobin and inflammation-corrected ferritin, sTfR, and BII. The prevalences of anemia, ID, and IDA were compared by group assignment using prevalence ratios estimated using logistic regression with a logarithmic link function and prevalence differences estimated using linear probability models with heteroscedasticity-consistent SEs (37). Modified Poisson models were used when log binomial models failed to converge (38). Our primary inferences were drawn from minimally adjusted models that controlled for baseline values of the outcome variable. For fully adjusted models, covariates were selected based on a bivariate association with the outcome variable (P < 0.1) from the following set of a priori identified variables: child age, child sex, maternal education, household asset index, number of children under 5 in the household, month of assessment, blood processing time, and inflammation-adjusted retinol binding protein (RBP). Malaria was examined for inclusion as a covariate based on a bivariate association (P < 0.1) with hemoglobin and anemia but not for ferritin, sTfR, BII, ID, or IDA, since these indicators included malaria in the correction for inflammation. Linear regression models were used to impute missing baseline values, which affected 11% of hemoglobin and 20% of ferritin, sTfR, CRP, and AGP covariates included in analytical models. Demographic variables were evaluated for bivariate associations with each biomarker and were used for imputation of missing baseline measures when they retained significance (P < 0.1) in multivariable linear regression models. Additional sensitivity analyses were conducted excluding children missing baseline data and imputing with the mean value. Participant characteristics of children lost to follow-up or missing outcome measures were compared to children with complete measures. We used an inverse probability of censoring-weighted approach to reweight the analytic sample to match the enrolled sample and then compared these results with those from the principal models.