The Role of Milk Protein and Whey Permeate in Lipid-based Nutrient Supplements on the Growth and Development of Stunted Children in Uganda: A Randomized Trial Protocol (MAGNUS)

Stunting is associated with cognitive impairment and later chronic disease. Previous trials to prevent stunting have had little effect, and no trials seem to have provided larger amounts of energy and high-quality proteins to already stunted children. We aimed to assess the effects of milk protein (MP) and whey permeate (WP) in large-quantity lipid-based nutrient supplements (LNS-LQ), among stunted children, on linear growth and child development. This was a randomized, double-blind, 2-by-2 factorial trial. Stunted children aged 12-59 mo from eastern Uganda (n = 750) were randomly assigned to receive 100 g LNS-LQ with or without MP and WP (n = 4 × 150) or no supplement (n = 150) for 3 mo. The primary outcomes were change in knee-heel and total length. Secondary outcomes included child development, body composition, anthropometry, and hemoglobin. Micronutrient status, intestinal function, and microbiota were also assessed. Our findings will contribute to an understanding of the role of milk ingredients and LNS in linear catch-up growth. This trial was registered at www.isrctn.com as ISRCTN13093195.

The reporting of this protocol followed the Standard Protocol Items: Recommendations for International Trials (SPIRIT) 2013 checklist. The MAGNUS study (Milk Affecting Growth, Cognition, and the Gut in Child Stunting) was a randomized, double-blind, 2-by-2 factorial trial testing the effects of MP and WP in LNS-LQ. An unsupplemented group was included as a reference. For a 12-wk period between February and December 2020, 750 Ugandan children classified as stunted received 1 of 4 formulations of LNS-LQ as a daily supplement (n = 4 × 150) or continued with the family diet (n = 150) (see Figure 1). All caregivers received individual nutrition counseling at baseline. All participants were followed up at the same intervals throughout the intervention period (see Figure 2). This design will allow us to assess the individual and combined effects of MP and WP among the 600 children allocated to LNS, based on the factorial design: If the effects are independent, then we can compare the 300 given LNS with MP to the 300 given LNS without milk. And likewise, we can compare the 300 given LNS with WP to the 300 given LNS without WP. If the effects are not independent, then we will compare each of the 4 combinations pairwise. In addition, we will be able to assess the effect of LNS by comparing the 600 given LNS, irrespective of milk ingredients, to the 150 given no supplements. The primary analysis will compare LNS-LQ with and without MP and WP in a 2-by-2 factorial design with 150 participants in each given combination. Secondary analysis will compare all LNS-LQ interventions (n = 600) with the reference group (family diet, n = 150). LNS-LQ, large-quantity lipid-based nutrient supplements; MP, milk protein; WP, whey permeate. MAGNUS data collection time points and visits. Timeline and visit overview for participants enrolled in the study. Phone follow-up and home visits were carried out as required and thus were unfixed time points. All participants were invited to the same follow-up visits. Time points were considered valid if taken within ±7 d of baseline, week 2 and week 4, and ±14 d from all other time points. Hemoglobin was a secondary outcome. LNS-LQ, large-quantity lipid-based nutrient supplement; MAGNUS, Milk Affecting Growth, Cognition, and the Gut in Child Stunting; MUAC, midupper arm circumference; S, screening in village for referral; T, week of visit from baseline (T0) to discharge (T12); WASH, water, sanitation, and hygiene. LNS are fortified lipid-based pastes that are well adapted for use in resource-limited settings; they are produced to a high safety standard, do not require refrigeration or preparation, and are packaged in standard portion sizes. Our LNS-LQ supplements, manufactured by Nutriset (Malaunay, France), varied with respect to the incorporation of WP and MP isolate (MPI). The MPI contained casein and whey proteins in the same proportions as milk; the lactose and mineral components were removed so that the MPI was close to 90% protein by weight with a Digestible Indispensable Amino Acid Score (DIAAS) of 120 (47). As a comparator to MPI, soy protein isolate, a high-quality plant protein with a DIAAS of 84, was used (47). WP contained 80–85% lactose and minerals (potassium, phosphorus, magnesium, calcium, sodium, and to a lesser extent zinc). As a comparator to WP, maltodextrin, a standard ingredient used in LNS products was used. All formulations were standardized to contain similar proportions of energy, protein, and carbohydrates. The supplements contained a mineral and vitamin mix to improve micronutrient content, and in 2 of the formulations the milk minerals provided by WP were in addition to the standard amount provided in all formulations (Table 1). Nutrient composition for 4 formulations of LNS-LQ supplied to 1- to 5-y-old stunted children1 The 600 participants randomly assigned to LNS-LQ received one 100-g sachet (530–535 kcal)/d for 12 wk, distributed every 14 d. Those randomly assigned to the family diet received laundry soap at each visit. The study was conducted from 2 local community health centers in Walukuba and Buwenge. All participants were recruited from within the surrounding district of Jinja, in the Busoga Subregion, eastern Uganda. Here, the prevalence of child stunting is estimated to be 29%, similar to the national average (48). To identify stunted children, communities within the district of Jinja were mobilized by Village Health Teams for an initial screening for referral. Study staff screened children in the community for age, stunting, and severe acute malnutrition (SAM). All children identified as having SAM were referred for appropriate treatment; others who met the inclusion criteria for stunting and age were invited to one of the study sites for eligibility screening. At the study sites, children were considered eligible if they were aged between 12 and 59 mo and had an HAZ of less than −2, according to the WHO growth standards (49). Children <12 mo old were not eligible to avoid interfering with breastfeeding. Caregivers had to be living in the catchment area and willing to return for follow-up visits, and able to provide written informed consent and agree to both phone follow-up (if a phone contact was available) and home visits. Children were excluded if they were identified with SAM according to the WHO classification (50), had medical complications requiring hospitalization, a history of allergy to peanuts or milk, obvious disability that impeded eating capacity, or a disability that impeded the measurement of length or height. Children were also excluded if they were participating in another study, if the family planned to move away from the catchment area within 6 mo, if previously enrolled in the MAGNUS study, or if another child from the same household was already included. If all eligibility criteria were met, trained staff took the caregiver through the informed-consent information individually, using the most appropriate of 3 commonly spoken languages in the region (English, Lusoga, or Luganda). The same information was given verbally and in writing. Caregivers were also taken through a short verbal questionnaire to ensure that the information provided was adequately understood. After necessary clarifications were given, the caregiver consented on behalf of the participant. If illiterate, a literate witness was present during the informed-consent process. Consenting caregivers were asked for permission to store 1–2 mL of blood and stool samples from the participant for future use; this was independent of trial consent. The caregiver could opt to withdraw consent at any time. The sachets of LNS-LQ were labeled with a unique 3-letter code that corresponded to the different formulations. Two unique codes were given to each of the 4 formulations and a further 2 codes were created for the reference group so that 10 unique codes were used in the allocation sequence list. Only the manufacturer (Nutriset) had access to the blinding code. Two allocation sequence lists, one for each study site, were computer generated using R (R Foundation for Statistical Computing). These were generated and sealed by a member of staff at the University of Copenhagen, Denmark, who was otherwise not involved in the study. Site-stratified, block randomization, with variable block sizes of 10 and 20 were used to allocate the sequential list of ID numbers to the 10 unique codes. Upon inclusion, administrative staff allocated a unique ID from a sequentially ordered list. After completion of baseline activities, the study pharmacist allocated the intervention according to a hard-copy random allocation list. Only the pharmacist had access to the allocation list, which was checked for each participant, at each visit. Using QR codes, the pharmacist recorded the code of what was distributed in a spreadsheet, which was regularly monitored by an independent assessor in Copenhagen. Hard copies of the allocation lists were kept securely in sealed envelopes at the University of Copenhagen. Outcome assessors and data analysts were blinded both with respect to the allocation of the intervention and to the type of ingredients contained in differently coded LNS sachets. Caregivers were blinded with respect to the type of LNS allocated, since the taste, smell, and appearance of all 4 products were indistinguishable. Caregivers were not, however, blinded with respect to receiving LNS or not. Only the Data Safety Monitoring Board, which operated independently of the study, could choose to break the blinding in order to monitor safety parameters. The LNS was distributed in packs of 14 sachets. To counteract the likelihood of sharing, an additional pack of the same LNS product code was distributed every 2 wk to caregivers with other children aged between 6 and 59 mo living in the same household. The additional stock provided to the household increased the likelihood that the participating child had access to the required daily quota. When collecting new sachets, the caregiver was requested to return any empty and unused sachets from the previous 2-wk supply, including those from the additional pack. The primary outcomes were changes in knee-heel length (mm) and total length/height (cm) from baseline to 12 wk. All secondary outcomes were measured over time from baseline to 12 wk. Child development was assessed at baseline and at discharge using a locally adapted version of the Malawi Development Assessment Tool (MDAT). Anthropometric indices HAZ, WAZ, and WHZ were assessed as well as weight (g), MUAC (cm), and head circumference (mm). Body composition was assessed using bioimpedance and the triceps and subscapular skinfold thicknesses (mm). The raw data from bioimpedance were used to calculate the fat mass (FM) (kg), fat-free mass (FFM) (kg), fat mass index (kg/m2), and fat-free mass index (kg/m2). Hemoglobin concentration was assessed from blood samples collected at baseline and 12 wk. Blood and stool samples were collected at baseline and at week 12. Blood samples will be analyzed for growth factors (IGF-I and insulin), markers of micronutrient status [i.e., iron (ferritin, soluble transferrin receptor), folate (serum folate), vitamin B-12 (cobalamin, methylmalonic acid), and vitamin A (retinol binding protein)], markers of systemic inflammation [C-reactive protein and α1-acid glycoprotein (AGP)], and markers of intestinal function (citrulline), together with other amino acids. Stool samples will be analyzed for markers of intestinal inflammation [myeloperoxidase (MPO), neopterin (NEO)] and function (α1-antitrypsin (AAT)] and the gut microbiota. Data will be reported on the proportion of children who, during the intervention period, deteriorated to moderate acute malnutrition or SAM according to the WHO classifications (49). The proportion of parti-cipants who died during the study period will be reported, as well as the number of morbidity episodes including the duration and severity of the illness. Finally, the number of children who were lost to follow-up will be reported. Caregivers were called with reminders to attend upcoming or missed appointments. Loss to follow-up was defined as those who had not returned for the 12-wk follow-up visit by 14 wk post-inclusion. Additional information collected at baseline included demographics, a dietary intake assessment, and a WASH assessment taken at the initial home visit. Time points for each measurement are shown in Figure 2. Knee-heel length was measured using a digital caliper with a resolution of 0.01 mm (Mitutoyo) mounted with knee and heel caps, cast in hard plastic. The distance between the knee (from the lateral condyle) and the heel (calcaneus) was measured 5 times consecutively on the left leg while the child was seated with both legs hanging over the edge of a table or the caregiver's lap. All other anthropometric measurements were repeated in triplicate. Participant length and height measurements were taken using a wooden Shorrboard (Weight and Measure), ensuring 4 points of contact with repositioning between measurements. Maternal height was measured using a fixed wall stadiometer (SECA 206). The weights of the mother and participant were measured using an electronic double-weighing scale (SECA 876). Head circumference, MUAC, and skinfold thickness were measured using a windowed, nonelastic head circumference tape (SECA 212); a nonelastic MUAC tape (UNICEF SD); and a Harpenden skinfold caliper (Baty International), respectively. Height or length, weight, MUAC, and head circumference were measured according to accepted international standards for anthropometric measurement (51). Skinfold thicknesses were measured on the left side, according to the manufacturer's instructions. For referral and inclusion, z scores were calculated using the WHO field growth charts. The WHO Anthro program will be used to calculate z scores for data analysis (52). Bioimpedance was measured using the Bodystat 500 (50 kHz) and in accordance with the manufacturer's instructions (Bodystat Ltd.). Measurements were taken while the child was lying on his/her back, with limbs spread apart, preferably at rest and with removal of wet or soiled diapers. A measure was repeated a minimum of 2 times but up to 3 times if the child's positioning or movement rated poorly. Measurements for impedance, resistance, reactance, and phase angle were recorded. Using an equation, the raw data will be used to calculate FM and FFM. Child development officers, trained in use of the MDAT (Manual V06, March 2018), took the participant through a series of activities adapted for the Ugandan context. The activities were related to 4 domains of development: gross motor, fine motor, language, and social development (53). The participant was graded as to whether or not he/she could complete each task successfully. The assessment continued until the child had failed to complete 6 tasks consecutively. At baseline, an interviewer-administered questionnaire was also used to gather information from the caregiver about household and family indicators for the support of child development (54). A thorough clinical examination was carried out at baseline. It included rapid tests for HIV and malaria, a thorough medical history with questions related to signs and symptoms of wasting or hospitalization due to SAM, and assessment of vital signs (pulse, blood pressure, and respiratory rate). At follow-up visits, a short review was conducted, assessing the most recent medical history, milk intake, and where applicable, monitoring of adverse events. To maintain blinding, the pharmacist distributing LNS inquired about adherence and if the caregiver had experienced problems with the LNS. Stool samples were collected at 2 time points and stored for later analysis of markers of gut function and microbiota. If not collected on site, a sample collection kit was given to caregivers along with specific instruction on stool sample collection at home. Collection vials contained StayRNA (A&A Biotechnology), allowing samples to be stored at room temperature for up to 5 d after collection. A maximum of 6.0 mL of venous blood was collected on site at 2 time points. A small amount was used for rapid tests: HIV status, malaria, and hemoglobin status. The remaining sample was processed and stored within hours for later analysis of selected markers. All biological samples were stored at −20°C until delivery to the main storage site in Kampala where they were stored at −80°C until they were shipped on dry ice to the University of Copenhagen, Denmark, for additional analyses. At baseline, the child's age and birth weight were recorded, wherever possible, using a birth information card. Information on sociodemographic characteristics, breastfeeding status, food frequency, and diet diversity was collected via interviewer-administered questionnaires. At the baseline home visit, GPS coordinates of the home site location were collected to facilitate later follow-up. In addition, trained staff conducted a short assessment of observed household WASH characteristics, including water source, access to basic sanitation, and the use of soap. To minimize response bias, local study staff with a good knowledge of the language and the culture were trained in asking questions to get as clear and precise answers as possible. If visits were missed and phone contact was unsuccessful, attempts were made to visit the caregiver's home. To facilitate attendance, a travel reimbursement was provided at each visit to cover the cost of return transport and food while at the clinic visit. Any participants requiring hospital attention were referred for treatment. If a caregiver requested for their child to stop receiving LNS, this was permitted; however, all included participants continued to be followed up for the remainder of the 12-wk intervention period. In case of participant withdrawal, all available data up to the point of withdrawal were used in data analysis. Participant data were collected in a paper case report form and were double entered using Epidata software (https://www.epidata.dk/) with inbuilt range checks. The secure electronic data collection platform REDCap (Open Source; Vanderbilt University) was used to monitor participant registration and visits but not for primary data collection. All source data will be kept securely on file for a minimum of 5 y after completion of the study. Adverse and serious adverse events occurring during the intervention period were recorded and reported to the sponsor and the institutional review board. Events occurring after a subject was discontinued from the study were not reported unless the investigator suspected that the event was related to the LNS-LQ intervention. To detect a 0.35-SD or greater difference between any 2 groups, with 5% significance and 80% power, 129 children were required in each group. To allow for 10% loss to follow-up, 150 children were included in each group, based on the 4 combinations of MP and WP. If there were no interactions between the 2 experimental interventions, 2 groups of 300 children could be compared, enabling differences of 0.24 SD to be detected. In the Treatfood trial (18), the SD of knee‐heel length at baseline was 18.1 mm (18), so that a 0.24-SD difference corresponded to 4.3 mm. In secondary analysis, to assess the effect of LNS, 600 supplemented children were compared with 150 unsupplemented children, with the ability to detect a 0.27-SD difference, corresponding to 4.9 mm. Primary and secondary outcomes will be analyzed using linear mixed models that account for the correlation between repeated measurements from the same participant, whereas tertiary outcomes will be analyzed using ordinary ANCOVA models. In all of these ANCOVA models, the baseline value will be included as a covariate. Additional covariates may be included as appropriate. Results will be reported as estimated differences with corresponding 95% CIs and P values. A statistical analysis plan was prepared before unblinding of the trial and uploaded to the ISRCTN registry. This is an effectiveness trial. Therefore, the primary statistical analysis will be carried out as intention to treat. In subsequent per-protocol statistical analysis, participants with major protocol deviations or violations are excluded. The study was conducted in accordance with the ethical principles set forth in the current version of the Declaration of Helsinki and all applicable local regulatory requirements. The study was approved by the School of Medicine Research Ethics Committee at Makerere University and The Ugandan National Council of Science and Technology. The study also received consultative approval from the Danish National Committee on Biomedical Research Ethics. The study was initiated only after approval was given by all aforementioned authorities. Written informed consent was obtained from all caregivers who consented to study participation of the child in their care. The rights, safety, and well-being of the children involved in the study prevailed over science and society. Before participant recruitment, the study was registered at www.isrctn.com as ISRCTN13093195.

N/A

• Dima Health