Impact of reduced dose of ready-to-use therapeutic foods in children with uncomplicated severe acute malnutrition: A randomised non-inferiority trial in Burkina Faso

Background: Children with uncomplicated severe acute malnutrition (SAM) are treated at home with ready-to-use therapeutic foods (RUTFs). The current RUTF dose is prescribed according to the weight of the child to fulfil 100% of their nutritional needs until discharge. However, there is doubt concerning the dose, as it seems to be shared, resulting in suboptimal cost-efficiency of SAM treatment. We investigated the efficacy of a reduced RUTF dose in community- based treatment of uncomplicated SAM. Methods and findings: We undertook a randomised trial testing the non-inferiority of weight gain velocity of children with SAM receiving (a) a standard RUTF dose for two weeks, followed by a reduced dose thereafter (reduced), compared with (b) a standard RUTF dose throughout the treatment (standard). A mean difference of 0.0 g/kg/day was expected, with a non-inferiority margin fixed at -0.5 g/kg/day. Linear and logistic mixed regression analyses were performed, with study site and team as random effects. Between October 2016 and July 2018, 801 children with uncomplicated SAM aged 6-59 months were enrolled from 10 community health centres in Burkina Faso. At admission, the mean age (± standard deviation [SD]) was 13.4 months (±8.7), 49% were male, and the mean weight was 6.2 kg (±1.3). The mean weight gain velocity from admission to discharge was 3.4 g/kg/day and did not differ between study arms (Δ 0.0 g/kg/day; 95% CI -0.4 to 0.4; p = 0.92) confirming non-inferiority (p = 0.013). However, after two weeks, the weight gain velocity was significantly lower in the reduced dose with a mean of 2.3 g/kg/day compared with 2.7 g/kg/day in the standard dose (Δ -0.4 g/kg/day; 95% CI -0.8 to -0.02; p = 0.041). The length of stay (LoS) was not different (p = 0.73) between groups with a median of 56 days (interquartile range [IQR] 35-91) in both arms. No differences were found between reduced and standard arm in recovery (52.7% and 55.4%; p = 0.45), referral (19.2% and 20.1%; p = 0.80), defaulter (12.2% and 8.5%; p = 0.088), non-response (12.7% and 12.5%; p = 0.95), and relapse (2.4% and 1.8%; p = 0.69) rates, respectively. However, the reduced RUTF dose had a small 0.2 mm/week (95% CI 0.04 to 0.4; p = 0.015) negative effect on height gain velocity with a mean height gain of 2.6 mm/week with reduced and 2.8 mm/week with standard RUTF dose. The impact was more pronounced in children under 12 months of age (interaction, p = 0.019) who gained 2.8 mm/ week with reduced and 3.1 mm/week with standard dose (Δ -0.4 mm/week; 95% CI -0.6 to -0.2; p < 0.001). Limitations include not blinding participants to the RUTF dose received and excluding all children with negative appetite test. The results are generalisable for relatively food secure contexts with a young SAM population. Conclusions: Reducing the RUTF dose provided to children with SAM after two weeks of treatment did not reduce overall weight or mid-upper arm circumference (MUAC) gain velocity nor affect recovery or lengthen treatment time. However, it led to a small but significant negative effect on linear growth, especially among the youngest. The potential effect of reducing the RUTF dose in a routine program on treatment outcomes should be evaluated before scaling up. The study was performed in accordance with the principles in the Declaration of Helsinki. The research protocol obtained ethical clearance from the national ethics committee (Comité d'éthique pour la recherche en santé [CERS]) and the clinical trials board (Direction Générale de la Pharmacie, du Médicament et des Laboratoires [DGPML]) of Burkina Faso. An independent Data Safety Monitoring Board composed of one paediatrician and one statistician was responsible for monitoring serious adverse events and conducted five complete data reviews during the course of the study. Caregivers provided verbal and written consent prior to enrolment and were made aware of their right to withdraw from the study at any time. Caregivers in both arms were given an instant photo of their child at the end of the treatment period and a bucket with soap at the end of the 3-month post-discharge follow-up period to compensate for the time spent on study procedures. We conducted a randomised controlled clinical trial (called MANGO) comparing the efficacy of a reduced RUTF dose to a standard RUTF dose in the management of uncomplicated SAM in children 6–59 months of age in a non-inferiority design. The study was conducted in the Fada N’Gourma health district located in the Eastern region of Burkina Faso. Malaria is endemic, with 69.3% of children presenting a positive rapid test [29]. HIV prevalence is 1.0% among 15–49-year-olds. In 2016, the prevalence of severe wasting (weight-for-height z-score [WHZ] <−3) and moderate wasting (WHZ between −3 and −2) was 2.4% and 8.6%, respectively [30]. There were 42 health centres in the district in 2015, all run by the Ministry of Health and supported by Action Against Hunger; 10 were chosen as study sites based on criteria on minimum SAM caseload (>7 new SAM admissions/month), accessibility, and a suitable schedule to couple study visit days with routine growth monitoring days. Between October 2016 and July 2018, study participants were selected from children presenting with SAM at the 10 participating health centres for curative and preventive activities. Study staff checked admission criteria: WHZ <−3 and/or mid-upper arm circumference (MUAC) <115 mm, positive appetite test (performed as per the national protocol [31]), no oedema or medical complications, and between 6 and 59 months of age. Exclusion criteria included having received treatment for SAM within 6 months, caregiver planning to travel or unable to comply with the weekly checkup schedule, peanut or milk allergy, or disability affecting food intake. Children with any grade of oedema or medical complications, as defined by the Burkina national protocol for CMAM [31], at any time during the study were referred to inpatient care. Randomisation was stratified by health centre using varying block sizes from 2 to 8. Randomisation lists were generated using the website www.randomization.com. After confirming eligibility and obtaining consent from the caregiver, children were given a unique study identifier (ID) by a team supervisor and assigned to a treatment group. Only the RUTF distributors had access to the randomisation lists, while staff involved in assessing the eligibility and study outcomes of the child were blinded to the trial arm. Participants could not be blinded to the RUTF dose received. Investigators remained blinded to treatment groups until the final analysis stage. Upon admission, the child’s caregiver was interviewed regarding household socioeconomic characteristics, care practices, and recent morbidity of the child and encouraged to adhere to weekly visits until recovery. Anthropometric measurements and a clinical examination were performed at each visit from admission to discharge. As per national SAM treatment protocol, seven key messages were delivered to caregivers in both groups, including advice to continue breastfeeding and to offer family foods in addition to RUTF if needed. Anthropometrics were measured in duplicate at each visit: weight using an electronic scale (SECA 876, SECA, Hamburg, Germany) to the nearest 100 g, height (recumbent for <24 months of age; standing for ≥24 months of age) using a wooden measuring board (locally made) to the nearest 1 mm, and MUAC using a non-stretchable colourless measuring tape to the nearest 1 mm. Using WHO field tables, WHZ was determined and used for admission and discharge. In later analysis, WHZ was calculated using the package ‘zscore06’ [32] in STATA 15 (StataCorp, College Station, TX). Children were followed up until recovery. Children missing their study visit were contacted either directly by telephone or via a community health worker and encouraged to return. Children referred did not return to trial after inpatient phase, as referral was considered a trial endpoint. Recovered children were followed up fortnightly for 12 weeks and relapses recorded. A supplementary feeding program accompanied the post-discharge follow-up, providing ready-to-use supplementary foods when available. Treatment followed the Burkina national CMAM guidelines in all aspects except the RUTF dose. Half of the children received a reduced dose from the third treatment week onwards (Table 1). Medical treatment included 7 days of amoxicillin for all children at admission (50–100 mg/kg/day), albendazole at the second treatment visit for children ≥12 months (200 mg to 12–23-month-olds; 400 mg to ≥24-month-olds) and catch-up doses for missed routine vaccinations or vitamin A supplementation (100,000 IU to 6–11-month-olds; 200,000 IU to 12–59-month-olds, every 6 months) at admission. Any illness, such as malaria, respiratory tract infections, or diarrhoea, diagnosed during the study was treated according to national protocol. See S2 Text for the full protocol for the study. Abbreviation: RUTF, ready-to-use therapeutic foods. Two study teams were comprised of one nurse, three measurers, one food distributor, and one supervisor per team. All team members were trained on research ethics and processes; standard operating procedures were defined, tested, and applied. Data were collected via tablets using the Open Data Kit (ODK1 software), and continuous data monitoring and cleaning were performed by a data manager under the supervision of the principal investigator. Electronic data were password protected, and field registries were kept in a locked office. Data were de-identified prior to analysis. The primary outcome was weight gain velocity (g/kg/day) from admission to discharge. Other outcomes included weight gain velocity after two weeks, length of stay (LoS), discharge anthropometrics, linear and MUAC growth, treatment outcome, morbidity, and relapses. Weight gain velocity from admission to discharge was calculated by dividing the weight gain (weight at discharge − weight at admission) in grams by the weight at admission in kilograms and the LoS in days. Weight gain velocity after two weeks was measured as follows: (weight at discharge − weight at visit 3 [in g]) ÷ (weight at admission [in kg]) ÷ (LoS − 14 [in days]). Missing weights at visit 3 (60 in reduced and 58 in standard arm) were imputed using mean weekly weight gained between an earlier visit (1 or 2) and later visit (4 or 5). The length of the stay was calculated as the number of days spent from admission to either recovery, referral, nonresponse, false discharge, or last visit before defaulting, lost to follow-up, or death. Linear and MUAC growth were defined as gains in millimetres (exit measure − admission measure)/week (LoS/7). A minimum acceptable mean rate of weight gain of 3.0 g/kg/day was defined at the protocol stage as a quality cutoff for evaluating general program performance. Nutritional recovery was defined as reaching a WHZ of ≥−2 for those admitted with a WHZ <−3 only, or MUAC ≥125 mm for those admitted with a MUAC <115 mm only, or both WHZ ≥−2 and MUAC ≥125 mm for those admitted with both WHZ <−3 and MUAC <115 mm upon two consecutive visits and absence of any illness. Referrals included children referred to inpatient care as a result of medical complications, >5% weight loss within three weeks, or ≤100 g weight gain over four weeks in the absence of apparent illness. Nonresponse included children not reaching anthropometric discharge criteria by 16 weeks of treatment who were referred to inpatient care for further examinations. Defaulters were defined as having missed three consecutive visits, but the child was confirmed to be alive. Transfers to health centres not involved in the study were categorised as defaulters. ‘Lost to follow-up’ was defined as having missed three consecutive visits without a known status of the child. False discharges included children who were erroneously discharged as recovered or referred, but upon analysis did not meet the criteria. Relapses were recorded over 12 weeks following recovery and were defined as presenting a WHZ <−3 and/or a MUAC <115 mm, or any grade of bilateral oedema. We assumed an expected mean difference in weight gain velocity between the two groups of 0.0 g/kg/day and a standard deviation (SD) of 2.6 g/kg/day with a non-inferiority margin of 0.5 g/kg/day. Assuming a power of 80% and a 5% significance level for a one-sided test, 335 children were needed in each group to demonstrate non-inferiority. To allow for dropout, the total target sample size was 800 children. Applying a 0.22 SD difference as could be observed in the main outcome with the calculated sample would allow us to detect a difference of 12% in recovery, seven days in LoS, 0.4 mm/week in MUAC gain velocity, and 0.3 mm/week in height gain velocity. Baseline characteristics of the study population were summarised as percentages and means (±SDs). Linear mixed models were used to compare primary and secondary outcomes of weight, linear and MUAC growth velocities, LoS, and anthropometric endpoints. Results were reported as differences of reduced dose from standard dose (reduced minus standard), with positive values meaning greater estimates among reduced dose. For programmatic outcomes, logistic mixed models were used to compare groups. Time to recovery was analysed using the Cox proportional hazards model. Study sites and research teams were included in the mixed models as random effects. Unadjusted models and models adjusted for sex, age, admission measure of weight, height, MUAC, WHZ, wealth, LoS in treatment, and month of admission were fitted. Adjustments were defined in the statistical analysis plan development stage prior to data analysis (S3 Text). Model checking was based on residual plots and normal probability plots, when applicable. All analyses were performed using STATA 15 (StataCorp). Both intention to treat (ITT) and per protocol (PP) analyses were carried out for the main outcome and key secondary outcomes. ITT analysis included all children admitted to the study for whom an endpoint observation was available. PP analysis included children without missed visits who, according to maternal recall, consumed >50% of the daily dose at all times and excluded those who had received a wrong treatment dose or had been falsely discharged. Interactions were only investigated in ITT analyses. Interactions between treatment and age group (<12 months versus ≥12 months), sex, MUAC category (<115 versus ≥115 mm), WHZ category (<−3 versus ≥−3), and stunting (height-for-age z-score [HAZ] < −2 versus HAZ ≥ −2) at admission were evaluated for the main outcome of weight gain velocity and the key secondary outcomes of recovery, LoS and height gain velocity, by means of likelihood ratio tests. Only significant interaction terms led to subgroup analyses. ‘Urban’ was defined as those living ≤30 minutes’ return trip from the regional capital city. Low birth weight (<2,500 g) was confirmed from an official birth certificate or health card. Household Food Insecurity Access Scale (HFIAS) was constructed according to FANTA indicator guide [33].