Trial of ready-to-use supplemental food and corn-soy blend in pregnant Malawian women with moderate malnutrition: A randomized controlled clinical trial

Background: Malnutrition during pregnancy in sub-Saharan Africa is associated with poor birth outcomes. Objective: This study compared maternal and offspring anthropometry for moderately malnourished pregnant women receiving ready-to-use supplemental food (RUSF), a fortified corn-soy blend (CSB+) with a daily multiple micronutrient antenatal supplement [United Nations International Multiple Micronutrient Preparation (UNIMMAP)], or standard of care comprising CSB+ and iron and folic acid (IFA). Design: A single-blind randomized controlled clinical trial was conducted in southern Malawi among 1828 pregnant women with moderate malnutrition, defined as a midupper arm circumference (MUAC) $20.6 and #23.0 cm. Women received 1 of 3 dietary treatment regimens that provided w900 kcal/d and 33–36 g protein/d. Maternal and infant anthropometry were followed until the child was 3 mo old. Results: Newborns had a mean length-for-age z score of 21.3 6 1.2 and 22% were stunted at birth. Mothers receiving RUSF had the highest weight gain during supplementation (3.4 6 2.6, 3.0 6 2.2, and 3.2 6 2.4 kg for the RUSF, CSB+ with UNIMMAP, and CSB+ with IFA groups, respectively; P = 0.03). Newborn birth weights and lengths were similar across intervention groups, but the incidence of newborns with a birth weight,2.4 kg (weight-for-age z score,22) was higher in the CSB+ with UNIMMAP group than the other groups (17%, 18%, and 24% for the CSB+ with IFA, RUSF, and CSB+ with UNIMMAP groups, respectively; P = 0.02). At birth, HIV-exposed newborns had a similar length and weight as newborns without HIV exposure, but their head circumference was smaller (34.0 6 1.5 and 34.3 6 1.6 cm, respectively; P = 0.02). At 3 mo of age, HIV-exposed infants had smaller weights, lengths, and head and arm circumferences than infants without HIV exposure. Conclusions: RUSF improved maternal weight gain compared with CSB+ with UNIMMAP. The large amount of food given and the modest effect on linear growth in newborns suggests that stunting in utero is unlikely to be reduced by supplemental food alone. This trial was registered at clinicaltrials.gov as NCT02120599. This assessor-blinded randomized controlled clinical trial was conducted in southern Malawi and compared maternal and newborn outcomes when one of 3 supplementation regimens was given to moderately malnourished pregnant women. Moderate malnutrition was defined as a midupper arm circumference (MUAC) ≥20.6 and ≤23.0 cm (9). The primary outcomes at study onset were as follows: 1) proportion of women who reached an MUAC ≥23.1 cm for 2 consecutive visits, 2) change in MUAC, 3) newborn birth weight, 4) newborn birth length, and 5) proportion of infants born prematurely. Secondary outcomes were as follows: 1) change in maternal weight from treatment onset until the final weight measurement, 2) change in maternal hemoglobin concentration from treatment onset until 10 wk later, 3) duration of treatment, and 4) infant anthropometry and survival at 6 and 12 wk. The sample size goal was estimated to be 1800 moderately malnourished pregnant women, divided equally among the 2 treatment groups (RUSF and CSB+ with UNIMMAP) and the control group (CSB+ with IFA). This size allowed for 15% attrition, yielding a final sample size of 1530 (n = 510/group) With a 2-tailed significance of 0.05, there was 80% power to detect a difference of 50 g in birth weight between groups, a mean difference of 0.1 cm in maternal MUAC between groups, a mean difference in 0.3 cm in newborn birth length, an 8% difference in the proportion of women reaching an MUAC >23 cm, and a 5% difference in the proportion of children born prematurely. All participants were randomly assigned to receive RUSF, CSB+ with UNIMMAP, or CSB+ with IFA using a random number generator that prospectively assigned participant identification numbers to a treatment group in blocks of 60. Women chose a sealed envelope that contained a study number that was linked to the previously randomly assigned treatment. Because the supplements were visually distinct, the study subjects were not blinded. To minimize study personnel awareness of a mother’s assigned treatment regimen, the study aide who dispensed the food looked up each participant’s study number, selected the assigned food, and placed the food in an opaque bucket. Data entry and analysis was conducted completely blinded to intervention group. Pregnant women >18 y of age presenting at 15 antenatal clinics in southern Malawi with moderate malnutrition (defined as an MUAC ≥20.6 and ≤23.0 cm) were recruited between March 2014 and December 2015. Enrollment criteria were as follows: a fundal height (FH) <35 cm, willingness to attend the antenatal clinic every 2 wk during pregnancy, and plans to remain in the area for delivery and until 3 mo postpartum. Women who had pregnancy complications, such as severe malnutrition (MUAC <20.6 cm), gestational diabetes, preeclampsia, hypertension, or severe anemia (blood hemoglobin concentration <70 g/L) at enrollment, were excluded from the study but were provided the standard of care nutritional treatment and were referred to a health facility clinician for medical management. After 296 women were enrolled, the enrollment criteria were expanded to include female individuals 16 or 17 y of age to extend participation to a substantial number of otherwise eligible subjects. Most of the subjects were subsistence farmers, who grew corn, tubers, and legumes on small plots of land, resided in mud huts without electricity, and collected water from boreholes or wells. This trial was registered at clinicaltrials.gov as {"type":"clinical-trial","attrs":{"text":"NCT02120599","term_id":"NCT02120599"}}NCT02120599 and was approved by the Institutional Review Boards of Washington University in St. Louis, California Polytechnic State University, San Luis Obispo, and the University of Malawi College of Medicine. Participants gave verbal and written informed consent to be in the study for themselves and their infants. On enrollment, participants were interviewed and demographic and health information was recorded. Participants also completed the Household Food Insecurity Access Scale (10) to identify the level of household food insecurity. Participants’ HIV status was recorded for those who had testing results available. For participants not already tested, HIV status was established by a rapid test after routine counseling by clinic voluntary counseling and testing counselors. Participants’ weight, height, MUAC, and triceps skinfold thickness were measured. Weight was measured with a Seca 803 Precision for Health scale. Height was measured with a Seca stadiometer. MUAC was measured twice on the left arm in centimeters to the nearest 10th of a centimeter with a flexible measuring tape (TALC), according to standard procedures; if the measurements differed by >1 mm, a third measurement was taken and the 2 closest measurements were recorded and averaged. FH was measured in the supine position with a nonelastic tape, measured to the nearest 0.5 cm (11). Study nurses were trained by a certified nurse trainer to measure FH, and inter-rater reliability was within 1.0 cm. Enrolled women returned to the clinic every 2 wk for anthropometric measurements (weight, MUAC, and triceps skinfold thickness) and health checks (blood pressure, interim illness questions, and FH measurement) and to receive their 2-wk supply of treatment food or supplements. When their MUAC reached ≥23.1 cm for 2 consecutive visits, the treatment food was no longer provided and the women “graduated” from the study; however, IFA supplements were continued for the remainder of their pregnancy. After graduation, women were asked to visit the clinic every 4 wk and were assessed for relapse of malnutrition. If a relapse (MUAC ≤23.0 cm) occurred, mothers resumed consumption of their assigned treatment food or supplements. A trained, dedicated birth anthropometry team frequently communicated with women for whom delivery was thought to occur within the next 2 wk. A team was dispatched to measure the infant within a few hours of being notified of delivery. Recumbent length was measured (Seca 417 length board) in triplicate to the nearest millimeter and averaged for analysis. Birth weight was measured in duplicate (MTB20 digital scale; Adam Equipment) to the nearest 10 g and averaged. If the 2 measurements differed by >10 g, a third measurement was taken and the outlier was eliminated. Head circumference was measured (Seca head circumference measuring band 212) to the nearest millimeter in duplicate and averaged. Efforts were made to minimize interobserver bias by periodic interobserver comparison and standardized training and techniques. At the 6-wk and 3-mo postpartum visits, maternal weight and MUAC were measured and the participants’ health history was taken. Infant weight, length, head circumference, and MUAC were also measured. In addition, infant hemoglobin was measured at the 3-mo visit. Two interventions (RUSF and CSB+ with UNIMMAP) were compared with the standard of care, which consisted of CSB+ with IFA. The food supplements were provided in biweekly rations of ten 250-g bottles of RUSF or 5 kg of CSB+. The RUSF provided 920 kcal/d, 36 g of protein/d, and ∼200% of the Recommended Dietary Allowance for most micronutrients during pregnancy. The RUSF energy content was designed to provide 450 kcal/d to support the increased energy needs during the third trimester of pregnancy plus an additional 470 kcal/d to support recovery from moderate malnutrition. The CSB+ with UNIMMAP treatment ration had amounts of energy, protein, and micronutrients similar to the RUSF treatment (Table 1). The antenatal UNIMMAP micronutrient tablet contained 15 micronutrients (Supplemental Table 1). The CSB+ with IFA treatment provided the same quantity of CSB+ as the CSB+ with UNIMMAP ration, but with a daily iron (60 mg) and folic acid (400 μg) tablet instead of a daily UNIMMAP tablet. The CSB+ formulation was “super cereal: CSB with sugar” (12). Nutrient content of supplemental foods1 Data were double entered and discrepancies were resolved after examination of the data collection cards. Characteristics and outcomes were tabulated. Women for whom newborn outcomes were not collected were considered to have defaulted. Data were analyzed with JMP Pro software (version 12.1.0; SAS Institute). Anthropometric z scores were calculated with the WHO R macro (WHO Anthro, version 3.2.2, January 2011; http://www.who.int/childgrowth/software/en/). Only singleton pregnancies were included in the newborn comparisons. An intention-to-treat analysis was performed to compare all outcomes using ANOVA for continuous parameters and a chi-square test for categorical outcomes. Wilcoxon’s rank-sum tests were performed if the normality requirement was not met, and a Fisher’s exact test was used if the expected cell count was <5 for categorical outcomes. For pairwise comparisons, connecting letter plots were created using Tukey-Kramer honestly significance difference tests. Some women delivered before receiving their second treatment ration, raising the concern that they had not been treated long enough to observe a clinical effect. In a second analysis, the intention-to-treat analysis was repeated with these women excluded. Women who delivered before receiving their second ration were also compared as a group to women who received ≥2 treatment rations. In addition, a subgroup analysis was conducted for women with HIV who were treated with 1 of the 3 treatment regimens, and another analysis was conducted to compare women with HIV to women without HIV. The effect of the different treatment regimens or duration of treatment on total maternal weight gain was analyzed with multivariable linear regression. We centered the numeric predictor (length of treatment) to reduce collinearity in the model. A generalized linear model with binomial error and logit link and the same predictor parameterization was used to determine whether duration of treatment or treatment group influenced the categorical response (stunting at birth). Partial F tests or likelihood ratio tests (chi-square tests) were first calculated to determine whether treatment group or duration of treatment had any association with the response variable through either the main effect or the interaction term. Statistical significance was set at P < 0.05, and all statistical tests were 2 sided.