Menu
Menu
Menu
Menu
Background Innovative strategies are needed to enhance the nutritional impact of agriculture. Value chain approaches, which use supply chains to add value (usually economic) to products as they move from producers to consumers, can be used to increase access to nutritious foods and improve nutritional status. This study tested whether a dairy value chain could be used to distribute a micronutrient-fortified yoghurt (MNFY) (conditional upon the producer supplying a minimum amount of cow milk/day) to improve hemoglobin and reduce anemia among preschool children in a remote area in Northern Senegal. Methods A cluster randomized control trial was used to compare 204 children (24 to 59 months of age at baseline) from households who received the MNFY coupled to a behavior change communication (BCC) campaign focusing on anemia prevention to 245 children from a control group (receiving BCC only) after one year. Randomization was done at the level of the family concession (households from the same family) (n = 321). Eligible households had a child of the target age and were willing to deliver milk to the dairy factory. Changes in anemia and hemoglobin between groups were assessed using mixed regression models. Key findings Anemia prevalence was very high at baseline (80%) and dropped to close to 60% at endline, with no differences between intervention groups. Hemoglobin increased by 0.55 g/dL, 95%CI (0.27; 0.84) more in the intervention compared to the control group after one year, in models that controlled for potentially confounding factors. The impact was greater (0.72 g/dL, 95%CI (0.34; 1.12)) for boys, compared to girls (0.38 g/dL, 95%CI (-0.03; 0.80)). Conclusion The dairy value chain was a successful strategy to distribute MNFY among pastoralists in Northern Senegal, and increase Hb concentrations among their children. This study is one of the first proofs of concept showing that a nutrition-sensitive agriculture value chain approach can contribute to improved child nutrition in a remote pastoralist population.
Richard Toll is located in the northern Saint-Louis region of Senegal, bordering Mauritania and the Senegal River. The district extends into a vast arid area where access to health centers is limited. In the region of Saint-Louis, the average distance to reach a health post is 7 km. Only 19 health posts and one hospital were available at the time of the study in the district of Richard Toll [9]. BCC activities on nutrition and health at the community level are part of the curriculum of the Cellule de Lutte Contre la Malnutrition (CLM), the agency in charge of technical assistance in defining and implementing the national policy in nutrition, called the Programme de Renforcement de la Nutrition (PRN) (Nutrition Strengthening Program). However, BCC activities were not yet implemented in the study area in 2012. The local dairy factory (LDB) was set up in Richard Toll in 2006, and based its activity on a system of milk collection from semi-nomadic pastoralists. Pastoralists belong to the semi-nomadic ethnic-group of Pulaar (Fulanis) who live in concessions (extended family unit) and migrate with their herds every year during the dry season (from November to June) when local pastures are no-longer sufficient for animal grazing. Due to insufficient water and pasture, milk production is at its lowest during the dry season. For the Pulaar, gender roles in milk production are clearly established at a young age, with women being in charge of milk production, and men of herd management [10]. The LDB collects milk from each producer twice a day through a network of 94 collection points close to their home. Milk collection trucks follow four 50 km routes throughout the area, collect the milk, and bring it back to the factory for processing and production of non-fortified yoghurt usually sold in urban centers. For this study, a formal contract was signed between the LDB and the pastoralists, whereby producers committed to supply 0.5 L of milk per lactating cow per day, 5 days per week. Households in the intervention group who fulfilled the contract in the previous week received one sachet of MNFY per child 24–59 months of age per day for seven consecutive days, in addition to the normal payment of 200 FCFA per liter of milk delivered. Households in the control group who fulfilled the contract received the normal payment for each liter of milk delivered but no MNFY. MNFY sachets were delivered daily to collection points easily accessible to women by the same truck that collected milk from the suppliers, thereby saving on transport/delivery costs. The MNFY was produced specifically for the study, using the milk collected from the pastoralists. The 80 g sachet of yoghurt was mixed with grains of millet (recipe of a traditional Senegalese yoghurt) and fortified with 2.1 mg of iron-EDTA (in addition to 2.25 mg of zinc, 24 μg of Iodine and 120 μg of Vitamin A). Implementation of the program was conducted by the Gret, an international NGO, working closely with the milk factory and its milk suppliers. The target age range for children in the intervention (24 to 59 months) was agreed upon among all actors involved in the intervention; children beyond the first 1,000 days were selected because of the dairy factory’s’ concerns regarding authorizations and infrastructures needed to produce complementary foods for infants and young children less than 24 months of age. The micronutrient formulation of the MNFY was therefore tailored to the needs of children 24–59 months of age. In addition, starting 3 months after the beginning of the intervention and continuing for 9 months, the CLM launched a BCC intervention implemented by two local non-governmental organizations (NGO) in both intervention and control areas. The BCC strategy included group sessions organized once a month in each village, home visits for households located in remote hamlets, social mobilization activities (events such as theater plays presented in big villages twice during the one-year study), and nine radio spots. Messaging focused on essential nutrition actions (ENA) [11] including optimal infant and young child nutrition during their first 24 months of life, the importance of micronutrients, the role of dietary diversity, the importance of consuming iron-rich, or fortified foods to prevent anemia, and the identification of symptoms of anemia. The BCC activities did not provide any information on the MNFY itself, because they were part of the standard PRN program of community based BCC implemented nationwide, and thus it was intended to increase demand for any fortified product. Given the slight delay in rolling out the BCC strategy and the consequences of seasonality and migration on milk production, the intervention was implemented slightly differently during three periods: 1) February-March 2013 (dry season): MNFY was distributed, as planned, in the intervention group conditional on households providing 0.5L of milk/per cow, with no BCC in either intervention or control group; 2) April-June 2013 (peak dry season): MNFY was distributed in the intervention group and BCC in both groups; however starting in mid-May, the conditionality for milk provided was decreased purposively to 0.3 L of milk per lactating cow per day, 5 days per week, because implementers realized that the cut-off used previously was too difficult to achieve, especially during the dry season; 3) July-December 2013 (rainy season until November): MNFY was distributed in intervention group conditional on households providing 0.3L of milk/per cow with BCC implemented in both intervention and control group. The evaluation used a cluster randomized control trial to compare the Hb concentration and anemia prevalence of children aged 24 to 59 months old in two groups: intervention and control. Concessions, which are units of 3–7 households usually related to each other, were randomly assigned to intervention and control group. In December 2012, 16 public gatherings were organized along the four milk collections routes by the implementing partner Gret to inform suppliers who were selling milk to the milk factory at the time, or those who had in the past and were willing to re-engage, about the objectives of the study, the milk production contract, the conditions of the MNFY delivery, the biological and other data collected, and the possibility to withdraw the study at any time. An information sheet was also distributed. The same day, public lotteries were performed to assign concessions to the intervention or control group. Suppliers were invited to meet at several points of the milking routes. Taking into consideration the context of the project area and for logistical reasons, we stratified the randomization of concessions (n = 321) at the lottery level per milking route level, in order to guarantee that, within each stratum or milking route, each of the intervention arms was represented equally. All the suppliers’ names were written on small pieces of paper, stapled by concession, collected into a barrel, and selected at random (publicly) to allocate concessions to intervention and control group. Eligibility criteria for households of active milk suppliers were that: 1) they agreed to participate in the study and 2) they had at least one child within the target age of 24–59 months. Households who accepted to participate in the study and declared having children aged 24–59 months old living in their household were asked to sign a contract. To avoid sharing of the MNFY with non-target children, the project provided MNFY to all children 24–59 mo old in the target concessions, even if they did not have a contract with the milk factory. Households from the intervention and control groups were paid 200 CFA/liter monthly for the milk they delivered to the dairy factory. We estimated that daily consumption of MNFY by target children would reduce the prevalence of anemia by 15 percentage points (pp) at endline from an initial level of 75%. With a power of 80%, 321 clusters randomized, an anemia intra cluster correlation (ICC) set at 0.3, a one-sided test, and equal cluster sizes assumed, we estimated, using sampsi/sampclus command in Stata, that a total of 133 children aged 24–59 months per group were needed to detect a 15 pp difference between interventions groups at endline. Taking into account a potential 20% loss of follow-up and 20% of households not meeting the criteria for receiving the incentive, we estimated that we would need to enroll 186 children per group. We estimated that 160 clusters per group would allow to detect a difference of 0.43 g/dL in hemoglobin concentration from an initial hemoglobin concentration at 11.5 g/dL (SD = 0.17)[3]. The main criteria for household enrolment was to have at least one child 24–59 months of age living in the household at the time of the baseline survey, and having committed to fulfil the contract of milk delivery. Written informed consent was obtained from all survey respondents (household head and spouse). A contact sheet was left with each household as well as copies of informed consent sheets which were also stored by the survey team. The baseline survey was conducted in January 2013 and the endline survey in January 2014, with two additional follow-up surveys in April 2013 (follow-up 1, F1) and in September 2013 (follow-up 2, F2). At endline, children followed-up were 36–71 months. Data were collected using questionnaires at the household, maternal and child level, during the baseline and endline surveys. Maternal questionnaires were administered on all mothers in a household with children 24–59 months of age at baseline, which given polygamy and household structure meant that some households had more than one mother questionnaire administered. At baseline, heads of household were asked about household characteristics such as household composition, cattle ownership, ownership of assets, and household food security using the HFIAS score[12]. Maternal knowledge was assessed through a questionnaire asking mothers to list iron-rich foods and the consequences of anemia in children. Mothers were considered knowledgeable if they were able to mention at least 2 out of 3 iron-rich foods (i.e. liver, fish or meat) and at least 2 out of 5 anemia consequences (i.e. difficulty at school, impaired development, slow growth, weak immunity defenses, and fatigue). Data on availability of the MNFY at the household level were collected at the two follow up surveys (F1 at 3 months; and F2 at 8 months following baseline) and at endline, through maternal recall. Mothers were asked if children from their household had consumed the MNFY, in the 3 days prior to the interview. Data on anemia (the study’s primary outcome) was collected at all four data points (baseline, F1, F2 and endline) using Hb concentration measured through a finger prick using a Hemocue device 201+ (HemoCue Ltd, United Kingdom). HemoCue devices 201+ were cleaned and checked for accuracy prior each data collection, using blood samples to check for repeatability. Anemia was defined as Hb below 11 g/dL and severe anemia, as Hb below 7g/dL. Children who were severely anemic and/or detected as suffering from severe acute malnutrition at any of the data collection points were referred to the closest health center. Mothers were given a small fee for transportation. Basic treatment (iron treatment, antiparasitic treatment, antibiotics) was provided at no cost to the mother or the health center, conditional on presenting the project’s referral letter. Health centers were reimbursed by the survey managers for treatment costs after the surveys and data on children were collected. Data were collected using computer assisted personal interviews (CAPI) using CSPro software. In the initial study protocol, the intervention was planned to last 6 months, using a cross-sectional design, with a baseline survey in January 2013, an endline survey at 6 months and a post-intervention assessment at 12 months. Before the start of the program, a census was organized in the study area to interview the 700 milk suppliers of the LDB on their willingness to participate in the program and to verify the number of children of target age living in their households. At the time of the census, all supplier households reported having at least one child between 24 and 59 months of age with an average of two children in this age range, suggesting a total number of eligible children of 1400. However, when enumerators assessed the children’s exact age during the baseline survey, only 310 of these households, from 247 concessions (instead of 320), were found eligible (with a total of 462 children aged 24 to 59 months old). With 247 concessions, the sample needed to detect a 15 pp difference of anemia prevalence between interventions groups at endline with a power at 80% was 346 children in total (~ 173 per arm). Adding 20% of potential loss of follow-up and 20% of households not meeting the criteria for receiving the incentive, we needed 484 children in total (~ 242 per arm), slightly more than the estimated number of eligible children in the population sampled (462 children in total (~ 231 per arm)). Given this slightly lower sample size than expected, the study team decided to move from a cross-sectional design to a longitudinal design, which followed children over time from baseline to endline. The program also faced some delays in implementation, as noted above. The MNFY distribution started in February (instead of January) and the BCC campaign in April. Given these delays, the duration of the program was extended by 6 months, until December 2013 (instead of June), and measurements were taken at baseline (January 2013), follow-up 1 (F1) in April 2013, follow up 2 (F2) in September 2013, and at endline in January 2014. The initial protocol for this study and supporting CONSORT checklist are available as supporting information; see S1 Consort Checklist and S1 Protocol. Comparison of baseline characteristics between intervention and control group was conducted to assess the quality of randomization in bivariate analysis, linear mixed-effect regression models using random effect at the concession level (cluster) for continuous variables and for binary variables (linear probability model) and using clustered Chi-squared tests for categorical variables (stata command clchi2). Comparison of anemia and severe anemia prevalence between intervention groups and boys and girls, percentage of children surveyed at different time-point by intervention group, maternal knowledge and household level of MNFY consumption between boys and girls were also done using linear mixed-effect regression using random effect at the concession level (cluster). The impact analysis was performed using intent to treat estimates, taking into account the cluster design and a level of statistical significance of 5% for a 1-sided test. We compared main outcomes by intervention group using linear mixed-effects regression models with random effects at two levels—random intercept and random slope at the child level and random intercept at the cluster level of concession. We found that adding time as a random slope improved the fit of the model significantly using a likelihood ratio test. We used an unstructured covariance for the random effects at the child level, after comparing models with different covariance structures using the Akaike information criterion. Models for Hb and anemia included all children having a baseline measure of Hb, regardless of whether or not they were present at any follow up measurement, mixed models allowing missing data for repeated measures. Interactions terms (Intervention X Time) were used to assess the effect of the intervention on Hb and anemia (Intervention vs control) between baseline and different time points (F1, F2, endline) (baseline serving as reference). Models were run separately by gender to take into account potential differences between boys and girls related to health or migration. Models controlled for the following potential confounding factors: wealth index, size of the household, number of lactating cows (as a wealth proxy), maternal age, children’s age, sex, and iron treatment for severe anemia at baseline (effective visit at the health center). A wealth index was established for each household, using multiple component analysis (mca stata command). The creation of the wealth index was based on the ownership of 27 assets (household assets: benches, tables, seats, beds, jars/calabashes, pestles/mortars, mattresses, radios, watches, chairs/Living-room, TV, mobile phones, fridges, carts, bicycles, motorcycles, cars; agriculture equipment: sickles, seed drills, spades, traditional hoes, other hoes, machetes/axe; animals: poultries, lambs, goats, donkeys). The two first principal components explained 78.71% of the total variance and were used in the analysis. All data analysis was performed using STATA 14.0 software. Ethical approval for this survey was obtained from the National Committee for Ethics and Research for Development for one year, under the protocol number SEN12/57 and from the International Food Policy Research Institute Institutional Review Board (IFPRI IRB). An amendment of the protocol was submitted to both the National Committee of Senegal and the IFPRI IRB to include an additional measure of hemoglobin in April 2013. The protocol was registered in the clinical trial.gov website ({“type”:”clinical-trial”,”attrs”:{“text”:”NCT02079961″,”term_id”:”NCT02079961″}}NCT02079961).
