A cluster-randomized, controlled trial of nutritional supplementation and promotion of responsive parenting in Madagascar: the MAHAY study design and rationale

Hundreds of millions of the world’s children suffer from poor nutrition, and as a consequence they experience delays in physical and mental health and cognitive development [1]. Linear growth retardation (stunting) is associated with poor cognitive, language and behavioral development [2], high rates of morbidity and mortality, and negative long term outcomes including compromised labor market participation [3, 4]. Outcomes are even worse for severely stunted children [3, 5‐7]. Despite evidence of the potential effectiveness of interventions in early childhood, policy makers and planners are challenged by designing interventions that sustainably improve growth and child development, particularly in the context of extreme poverty. Furthermore, since linear growth retardation is largely irreversible after the age of two or three [8], there is an urgent need to design interventions that begin early in life.

Madagascar is a very low-income country (gross national annual income per capita of 430 USD [9]), with a prevalence of stunting of almost 50 % [1]. Children in Madagascar living in poverty show deficits in developmental outcomes, such as language and cognition [10]. In order to address the severity and prevalence of poor nutrition in Madagascar, the National Office of Nutrition (ONN) initiated a large-scale community-based nutrition intervention in 1999, which is still operational throughout the country. The community-based program includes monthly growth-monitoring activities for infants and young children, cooking demonstrations, community mobilization, and nutrition and hygiene education of their primary caregivers. A rigorous multi-year evaluation of the long term effects of the existing program on weight-for-age showed small, significant effects of the program for young children, but no effects of the program on height-for-age [11]. To address the lack of effectiveness of Madagascar’s current program on linear growth faltering, the interventions described here were developed in collaboration with the research team and Madagascar’s Office of Nutrition.

The proposed evaluation has as its goal to identify the most effective, cost-effective and scalable interventions to address linear growth faltering and promote child development in Madagascar. Our study uses a multi-arm randomized-controlled trial (RCT) to test the effects and cost-effectiveness of interventions designed to address chronic malnutrition and poor child development, and has been designed based on the existing evidence [3, 12, 13]. The questions we are addressing through this research are described in detail below.

Common strategies to improve nutritional status include the promotion of behavior change to increase dietary quality; the fortification of food to improve micronutrient content of staple foods; or the use of nutritional supplements for vulnerable population groups (e.g. very young children) [13‐15]. Our study aims to test the added value of variations of these strategies over and above the existing, community-based, growth monitoring. Growth monitoring and education strategies are only successful when they involve intensive counseling and strongly emphasize dietary diversity and the promotion and consumption of animal source foods; [14] the quality of counseling and its intensity are believed to be a necessary condition for growth monitoring programs to be effective [16]. Two existing programs (one in Peru [17], and the other in China [18]) have demonstrated that primary caregivers are able to act on the advice of community nutrition workers to provide higher quality, micronutrient rich foods. Similarly, home visits in Malawi including nutritional information from birth to the first year of life showed significant improvements in the consumption of protein rich foods and in measures of height for age [19]. The findings are in line with evidence from behavioral economics on the importance of the way the messages are framed, perceived and internalized [20].

In spite of these encouraging studies, intensive counseling and behavioral change may not be sufficient to address severe nutritional deficiencies in settings with high food insecurity and intense poverty. Thus, nutritional supplementation may be a more effective option and can come in the form of micronutrient tablets or drops, multiple micronutrient powders (e.g. “Sprinkles”) mixed into the child’s typical complementary foods, or lipid-based nutrient supplements (LNS). While multiple micronutrient powders appear to be effective for improving hemoglobin, iron status and some cognitive outcomes in children, they have shown limited effects on child growth [21, 22]. Multiple micronutrient supplements in pregnancy have been found to have a small impact on birth weight [23], though the effects appear to be minimal in women of low BMI in early pregnancy. For LNS findings, evidence is emerging from a number of efficacy trials in Africa and Asia that suggest there are important benefits of LNS products to infant growth and development [24]. For example, in Malawi, LNS supplementation reduced severe stunting to 3.5 % or less as compared to 12.5 % in the control group [25], an effect that was maintained 12 months after the intervention period ceased [26]. In Ghana, LNS supplementation has been associated with higher length-for-age z-score when provided to infants from 6–12 months of age compared to a micronutrient tablet or powder [27]. In Burkina Faso, mothers starting LNS in early pregnancy [28] showed an increase in birth length; effects were biggest for those women with low early pregnancy BMI, and in those whose late pregnancy stage occurred during the rainy/lean season [29]. In a separate trial in Burkina Faso, LNS supplementation for children 9–18 mo together with morbidity monitoring and treatment, was associated with a 10 percentage point reduction in stunting [30]. Thus far, few studies have examined developmental outcomes after LNS supplementation and the data available are conflicting. In Malawi, children supplemented from 6–18 months of age with LNS had no difference in age of achievement of motor or language developmental milestones compared to controls [31]. In contrast, in Ghana, the group receiving LNS from 6–12 months of age had a greater proportion of children who could walk independently at 12 months (49 %) compared to a non-intervention group (25 %) [27].

The existing data and evidence relating to LNS interventions suggest that lipid-based supplements have the potential to improve size at birth and growth through infancy, but the optimal timing and composition of the supplements is not clear. It is possible that if supplementation were to cover both mothers and infants during critical periods of growth and development, the effect sizes could be larger than supplementing at one time point alone. Although the evidence from such combined interventions is not yet available, there are some trials of LNS for pregnant women and/or children underway in other settings [32]. Testing the added value of supplementation of pregnant and lactating women over and above supplementation of children during a critical age window of 6–18 months of age will be very informative for understanding the importance of pre- and post-natal nutrition as a strategy for stunting prevention. The biological and economic literature suggests that there are dynamic complementarities between early and later investments in human capital.

Our study will test the cost-effectiveness of LNS targeted to children, as well as pregnant and lactating women, as tools to prevent child stunting and promote optimal child development. While the results from the pregnancy and child LNS supplementation studies are promising, there are important considerations about the cost-effectiveness of these interventions. Given the high prevalence of stunting, and the particularly severe stunting in Madagascar, the rationale for this intervention is strong, particularly if it can be combined with an improved behavior change intervention [33, 34]. The quantities of LNS are intended to be preventative, and thus are less likely to be shared with household members other than the targeted child or pregnant women. Thus, there should be little dilution of the benefits due to food sharing within the household or substitution of existing food intake. While we think that there are likely to be important nutritional benefits of supplementing both mothers and their children, it is important to consider that this approach will represent a two year period of supplementation per target child (one year to mothers plus one year of supplementation to their children). Thus, the costs of this intervention will need to be weighed carefully compared to one year of direct supplementation to the child only.

The program under study is already at scale, covering 6500 communities across Madagascar. The community-based program is implemented by the Government and is a key component of the country's nutrition and food security strategy. There are undeniable strengths of the existing program, which builds on an existing large scale infrastructure of service delivery, a strong sense of ownership, and many people intensely dedicated to making the program work. Each of the alternative program designs will compare estimated impacts in terms of reduced stunting and improved cognitive development against the marginal cost of each additional intervention. The preventive new lipid-based supplements are going to be imported from abroad for the purpose of this study. The implementing agency (PNNC/ONN) has already begun discussion for local production of the supplements, however, which would make it a more sustainable approach in the future. The new projected local unit cost of the intervention will be used to assess the benefit/cost ratios of the different packages that include LNS.

There has been a call for large scale programs that integrate health, nutrition and the promotion of child development [12], but there are limited data on the feasibility, advantages, or disadvantages of integrating such interventions. Existing studies consist primarily of small, efficacy trials, and have shown little evidence of synergistic interaction between nutrition and stimulation on child development [35]. There are clear benefits to integrating interventions, because health and nutrition sectors are often the only services reaching children under three years. Thus, delivering support for child development through these mechanisms may be a key way to reach young children, particularly those at risk of poor development. It is also possible that integration can provide additive benefits for children at a lower cost than for a stand-alone program.

The objective of the MAHAY (“smart” in Malagasy) study is to determine whether any of the specified intervention packages can significantly alter the pattern of severe growth faltering and delays in child development among young children in Madagascar. The design will not only allow us to measure the impact of each treatment arm compared to the status quo, but also to estimate differential effects and costs of each incremental layer.

Our primary objectives are:

Our secondary objective is to measure intermediate indicators such as knowledge and hygiene, feeding and stimulation practices that can help us model the pathways through which the different interventions might have an effect on child growth and development.

The evaluation is a cluster RCT with assignment at the community level. The sample of eligible program communities is assigned to either the current status quo or to one of four arms, based on the randomized design. The study is longitudinal, with sampled children and their households interviewed at three points in time, with a baseline survey administered right before the interventions begin, and two follow-up surveys (midline and endline) administered approximately one and two years after the baseline survey.

Our sampling frame is the universe of eligible project sites in the 5 target regions of the program. From this universe of project communities, a sample of 25 program sites is randomly assigned to one of five groups. The comparison group for our study is the program as currently designed (as opposed to "no program" in most of the evaluation literature). In the four randomized arms, we sequentially add increasing levels of intensity and complexity to the current intervention, starting with the lowest cost option and incrementally adding layers of intensity (and cost) to test the value added of each layer in terms of its ability to reduce growth faltering and promote child development. We will then compare the added value of each layer with its added cost, and will use this information to  assess the cost-effectiveness of alternative modes for service delivery.

In order to identify the variants in program design that could be added to the status quo program, we examined interventions that have shown effectiveness in previous studies. Most of the evidence stems from efficacy trials or small scale interventions, so our objective is to test the implementation within an existing large scale program. Furthermore, we considered interventions that can reasonably be incorporated into the existing program in Madagascar, which is already at scale, will be acceptable to the local population, and that the government would consider adopting as a new package of services.

The interventions are summarized briefly here and then described in more detail below (see also Table 1):

The five regions included in the study have some of the highest poverty rates in Madagascar and are highly vulnerable to extreme weather shocks (e.g., flooding and drought). The regions are Amoron’i Mania, Androy, Atsimo Atsinanana, Haute Matsiatra, and Vatovavy-Fitovinany. Four of these regions are geographically located in Fianarantsoa province on the south-east coast of Madagascar, and the fifth is located at the very southern tip of the country. Sites eligible for inclusion in the RCT were all active government nutrition program sites in these 5 regions as of January 2014. Sites opened after 2013 are excluded. Each site covers an average of 2–3 small communities of about 1000–2000 inhabitants each, who rely economically on agriculture or fishing, with limited diversification into commerce and trading. Most of the sites have access to a health center and a primary school and many have access to a secondary school. Supervision of government program activities is operated by local Malagasy non-profit organizations (NGOs), which operate uniquely at the regional level.

All pregnant women and women with age-eligible children living in the catchment area of a project site are eligible to participate in the standard growth monitoring and nutritional education that occurs in a group setting in a community center in all sites. The LNS supplement is available to all pregnant/lactating women and age-eligible children in the participating communities assigned to T₃ or T₂, respectively. Although no nutritional screening criteria are required for eligibility into the intervention program, severely malnourished children >6 months old will be referred for treatment according to national guidelines. In the T₄ intervention communities, only women in households selected to be interviewed for the baseline survey are eligible to receive the home visiting and early stimulation component of the program. All other children in the T4 communities who are not part of the baseline survey are eligible for the T₀ status quo program.

Primary nutritional outcomes include length/height-for-age and weight-for-length/height (Table 3). Measures of child development are language, cognitive, motor and social development. Secondary indicators include weight-for-age, anemia, and iron status. Intermediate indicators include measures of nutrition and development, and intermediate indicators including dietary diversity, household food security, maternal knowledge of child care and feeding practices, and home stimulation practices. A household survey will be administered with the following questionnaires:

The prevalence of anemia, an important risk factor for poor child development, will be measured at baseline, midline and endline using portable Hemocue analyzers. Anemia will be defined as hemoglobin concentration <110 g/L [42, 43]. In serum samples collected at endline, ferritin and transferrin receptor as markers of iron status, retinol binding protein as a marker of vitamin A status, and c-reactive protein (CRP) and alpha-1 acid glycoprotein (AGP) as markers of inflammation will be measured. Iron and vitamin A biomarkers will be adjusted for inflammation using the method described by Thurnham et al. [44].

The endline survey will include a validation study of the maternal-report ASQ-I instrument to compare it with a direct assessment of child development (the Bayley Scales of Infant Development - BSID). The BSID has been well-validated and provides a strong indication of how children are developing when the test is administered [45]. The BSID requires a trained assessor, and is costly to purchase, which is why we are including it for only a subset of children [46]. The BSID provides scores for both a Mental Development Index and a Motor Development Index, as well as an indicator of socio-emotional development.

Data collection is being carried out by a survey firm with long term experience in collecting health and nutrition data. The team in the survey firm participated in the planning and data collection of regular Demographic Health Survey (DHS) in Madagascar as well as the Anthropometrics and Child Development Surveys (ADE 1997, 2004, 2007, 2011). Through these surveys, the team has developed key expertise in integrating child development tests within a household survey, as well as experience in fielding longitudinal surveys with tracking protocols and data entry in the field [10, 47].

We collect data with Computer Assisted Personalized Interviews (CAPI) using Android tablets, using a free platform developed by the World Bank (solutions.​worldbank.​org). All documentation regarding the participants, including the laboratory samples, source data and the questionnaires, will be identified using appropriate participant codes. All records will be kept in a secure server and will be accessible only to approved survey firm personnel and research team members. Clinical information will not be released without written permission of the subject, except as necessary for clinical monitoring. Data from these registers may be given only to authorized members of the research team and authorities guiding health research in Madagascar.

In addition, we collect monthly administrative data on all the treatment arms for the added community nutrition worker, in addition to the regular weighing data from the growth monitoring activities. In addition, midline and endline survey data will have a participation module with a detailed account of participation in all program activities.

Unit costs for the “status quo” program are available from existing project data and are approximately 15 USD per target beneficiary. The total cost includes the wages of the community worker, the training cost (refresher) of the community worker, the wages, training and supervision cost (NGO and district supervisor), the unit cost of the cooking demonstration, the information, education and communication (IEC) materials, the equipment and registries at the community level, and the health card of the child.

The additional cost of the intervention included in the intensive counseling arm (T1) will add additional salary, training and supervision of the added community worker.

For the purpose of this study, the LNS supplements are produced internationally and imported from overseas, which results in a substantial addition to the unit cost in the supplementation arms. The actual costing for the cost-effectiveness analysis (and for informing scale up of the study) will include the scenario of local production around which the ONN has already begun discussions.

During the implementation of this study, districts and NGO supervisors across all intervention arms will be provided with costing data sheets to record actual expenditures incurred at regular points during implementation. A framework is being developed to monitor these costs over time, which will feed into the cost benefit analysis work.

Last, we plan to collect information to incorporate private costs to mothers. We include a small time use section to be administered to the primary caregiver, with questions related to the time needed to participate in each program activity. We will also have wage rates for agriculture and casual labor in the community questionnaire, which are likely to represent an upper bound on the opportunity cost of time. The information will be triangulated with the history of seasonal wage rates available from the commune census collected by Cornell and Oxford University (http://​www.​ilo.​cornell.​edu/​ilo/​data.​html). The opportunity costs will be highly dependent on the season, so we want to make sure that the full accounting of private costs will not be a function of the time of the interviews.

The study will refer participants for treatment if we observe either of the two following outcomes: soy or nut allergies that would make it potentially dangerous to consume LNS supplementation, and/or acute malnutrition.

The number of communities was selected based on testing the null hypothesis that the treatment effect for any pair-wise comparison of treatment arms is equal to zero at a conventional 95 % confidence level (alpha = 0.05) to achieve a statistical power of 80 %.

An effect size of 0.25 to 0.30 SD for the supplementation arms is expected to be seen based on the review on lipid-based complementary feeding interventions in developing countries [14] and was chosen for the mid-point of the sample size calculation.

The intra-cluster correlation was drawn from the 2011 Anthropometrics Survey for height-for-age z-scores (HAZ). With an effect size of 0.30 SD and an estimated intra-cluster correlation of 0.10, the number of communities to be included in each pair-wise comparison of treatment arms should be 50 communities (25 per treatment arm) with 30 children sampled from within each community.

No external data safety monitoring board is pre-planned, but quality control during data collection is ensured by the Madagascar Office of National Nutrition using their standard operating procedures and regular collection of the monitoring data system. We do not anticipate serious risk to our participants from participating in the study. However, there may be adverse consequences to participation that were unintended or unexpected. In the absence of a local or external monitoring board, we rely on local internal monitoring for reports of adverse events, as well as preliminary analyses following the midline study, to stop any intervention arm if clear evidence of harm is found from program participation. In the event that the monitoring reports or midline analyses indicate evidence of substantial early benefit from one of the intervention arms, we will continue the trial, with the Government’s approval, in order to capture differential effects of the intervention arms over time and by age cohort of the children.