Study setting and population
The study was conducted in the Madarounfa rural health district in south-central Niger, in the Sahel region, a region affected by seasonal fluctuations in food availability, infectious illnesses and high rates of acute malnutrition among young children. The study population included 2412 children enrolled in a randomized trial of routine amoxicillin in the treatment of uncomplicated SAM conducted from 2012 to 2014 (ClinicalTrials.gov Identifier, NCT01613547) [16]. In brief, children were enrolled in the parent trial if they presented to health centers for outpatient treatment of uncomplicated SAM, resided within 15 km of study health centers, were available for 12 weeks of follow up and had no clinical complications requiring antibiotic treatment at admission. Children also could not have been admitted to a nutrition treatment program in the preceding 3 months or have congenital abnormalities. The inclusion criteria for outpatient SAM treatment were age 6 to 59 months, WHZ < − 3 based on World Health Organization (WHO) growth standards, or mid-upper arm circumference (MUAC) < 115 mm. Other inclusion criteria included sufficient appetite for successful oral nutrition and absence of bipedal edema or clinical complications requiring hospitalization [16].
Study design and interventions
The parent study was a randomized, double-blind placebo-controlled trial. Children were randomized to receive routine amoxicillin or placebo at admission, but otherwise received standard clinical care and ready-to-use therapeutic food for the outpatient treatment of SAM in line with guidelines from Médecins Sans Frontières (MSF) and the Ministry of Health of Niger [16]. Detailed descriptions of the study design and methods are provided elsewhere [16].
Study procedures and follow-up
At admission, study nurses interviewed caregivers and collected socio-demographic information, including child feeding practices, immunization status, child illnesses and medical history, as well as maternal characteristics and household socioeconomic status. Hemoglobin status (HemoCue Hb 301, HemoCue, Angelholm, Sweden) and malaria infection (SD Bioline Malaria Antigen P.f, Standard Diagnostics Inc., Republic of Korea) were assessed in all children at baseline.
Follow-up was conducted by study staff at health centers on a weekly basis until discharge from the nutritional program (minimum stay 3 weeks for recovery), and at 4, 8 and 12 weeks after enrollment. At each visit, anthropometric measurements were taken. Weight was measured to the nearest 100 g using a hanging Salter scale (≥ 15 kg) or SECA scale (< 15 kg), and length (standing height for children ≥24 months) using a wooden length/height measuring board. MUAC was measured to the nearest 0.1 cm using a non-stretchable MUAC tape. Study physicians conducted a physical exam and completed a medical history that included health events, medical consultations and treatments for children in the last 7 days. When a child missed a scheduled visit, home visits by a study nurse and community health agent were conducted and anthropometry and clinical condition were assessed.
Dietary diversity
Diet diversity (DD) has been recognized as a tool for evaluating dietary intake and quality for children in low income settings where gold standard tools such as 7 day diet records are not feasible or are limited by cost [17]. While individual foods are diverse and vary by culture, food groups are limited in number and thus are amenable to study and comparison across contexts. Dietary intake and diversity, the primary exposures of interest, were assessed at baseline using a food frequency questionnaire (FFQ) which asked caregivers if the child had eaten a list of 45 foods (in 12 food categories) in the previous 7 days, and if yes, for how many of these 7 days. As proposed by Dewey et al. [18], 8 food groups were created: grains, roots and tubers; legumes and nuts; dairy products; flesh foods; eggs; vitamin A-rich fruits and vegetables; other fruits and vegetables; fats and oils. A diet diversity score (DDS) can be created by awarding one point for each food group consumed over the past 24 h and 0 otherwise, using a 10 g cutoff per food group for the first seven groups, and a 1 g cutoff for fats and oils. In this analysis, which used FFQ not 24-h recall data, a revised scoring algorithm was developed which awarded one point for each food group consumed at least once per day and scaled, partial points for all less frequent consumption (e.g. if eaten on 1 day in the 7-day recall, 1/7 point was awarded). DDS was computed as the sum of food groups consumed. This 8-food group DDS was selected as it has been shown to be predictive of diet quality of complementary foods for children 6–12 months of age as indicated by correlations with mean nutrient density adequacy for selected micronutrients including iron and vitamin A [18]. We also considered other measures of child dietary diversity including the WHO index to assess infant and young child feeding (2007), a 7-food group measure that included grains, roots and tubers; legumes and nuts; dairy products; flesh foods; eggs; vitamin-A rich fruits and vegetables; and other fruits and vegetables [19]. We excluded 22 children from the analysis that were exclusively breastfed at the time of admission.
Study outcomes
The primary outcomes of the analysis were nutritional recovery, transfer to hospital and death at program discharge and 12 weeks of follow up. Secondary outcomes included change in child weight (g/kg/d) and height (mm/d) from baseline to weeks 1 and 2, program discharge and 12 weeks of follow up, among the children that recovered. Nutritional recovery in the program was determined at 3 weeks or beyond and was defined as WHZ ≥ − 2 on two consecutive visits and MUAC ≥115 mm. Recovered children had no acute complications or edema for a minimum of 7 days and had completed all antibiotic and anti-malaria treatment at program discharge. Default from the program was defined as 3 or more missed weekly visits. Non-response was defined as when a child did not meet criteria for recovery at 8 weeks. Children with weight loss greater than 5% between two consecutive visits, lack of weight gain after 2 weeks or clinical complications requiring inpatient management were transferred to hospital for inpatient management. Death was considered as all-cause mortality.
Statistical analysis
We evaluated the socio-economic and demographic characteristics of the study children. Chi-squared tests for dichotomous variables and the Wilcoxon rank sum test for continuous variables were used to assess significant differences between levels of dietary diversity (±4 food groups/day). We described intake of individual food groups and the overall DDS, and compared differences between seasons, using chi-squared tests and the Wilcoxon rank sum test.
We calculated DDS at admission, and evaluated the associations between individual food group intake and DDS with the risk of nutritional recovery, transfer to hospital or death at program discharge and 12 weeks. Separate binomial regression models were used for individual food groups or DDS. Among recovered children, we also evaluated associations of food group intake and DDS with weight and height change using linear regression.
We adjusted for amoxicillin/placebo trial regimen, as well as potential confounders assessed at admission and selected based on univariate significance with the outcome at the level of P < 0.20. Covariates considered included: child characteristics (age: 6–11, 12–23 vs. 24–59 months), sex: male vs. female), breastfeeding (yes vs. no), child anthropometric status (wasting: none, moderate vs. severe, stunting: none, moderate, severe and MUAC: < 115 vs. ≥115 mm), child health status (anemia: yes vs. no, respiratory rate, and body temperature), presence of morbidity (yes/no: malaria, tachypnea, upper respiratory infection, cough, nasal discharge, diarrhea, vomiting, and ear pain), child receiving ready to use supplementary foods in the past 7 days (yes/no), vaccination status: yes/no (measles, pertussis, tetanus, and polio), previous consultation for child’s health status (yes/no). We also considered household characteristics (number of children < 5 years, household size, wealth index, household food security index (HFIAS score), bednet use(yes/no)), maternal and paternal characteristics (maternal and paternal age, maternal education (none, koranic school, primary school complete, CEG complete), use of health services (yes/no) and season(lean/non-lean). Confounders were measured at admission.
Finally, to more broadly inform the identification of potential boosters of nutritional recovery, we assessed independent predictors of treatment recovery in this setting. Potential predictors identified using logistic regression models with univariate significance at P < 0.20 were included in a final multivariate model. All analyses were conducted using SAS version 9.4.
Ethics
Ethical approval for the parent study was obtained from Comité Consultatif National d’Éthique, Niger and the Comité de Protection des Personnes, Île-de-France XI, Paris. Written informed consent was obtained from the parent or legal guardian of all participating children.