The World Health Organization (WHO) recommends, as standard therapy for recovery from severe malnutrition, the use of the F100 catch-up diet, which is a nutrient-dense milk-based food formulated specifically to fulfill the high nutritional needs of severely malnourished children (1). In particular, to support rapid catch-up growth, its contents of micronutrients, especially potassium, magnesium, and zinc, are high (2). Alternative therapeutic foods are ready-to-use-therapeutic foods (RUTF). These solid fortified foods have a nutritional composition similar to that of F100 and have been shown to support recovery from severe malnutrition (3). For safe community-based rehabilitation, they have the additional advantage of resistance to bacterial contamination (4). RUTF is recommended for community-based nutritional rehabilitation because, in contrast to F100, bacteria do not grow in it when it is accidentally contaminated (5).
Both F100 and RUTF are relatively expensive, and they are not available in many settings where severe malnutrition is present. Such limitations make the use of home-prepared rehabilitation diets, which are based on nutrient-dense locally available foods, appealing because of their distinct advantages in terms of cost, logistics of procurement, distribution, and sustainability (6,7). Nevertheless, a recent study showed that in disadvantaged environments, it is difficult to formulate a diet that will ensure an infant's nutrient needs are met by the use of only locally available foods (8). In comparison with adequately nourished infants, the nutrient needs of severely malnourished infant are high to support rapid rates of catch-up growth (2). Consequently, before home-prepared rehabilitation diets based only on locally available foods can be recommended, it is essential to establish whether they are likely to meet a severely malnourished infant's high nutrient needs. Because the nutrient content of F100 was specifically formulated to meet these high nutrient needs, the aim of the present study was to assess whether home-prepared rehabilitation diets for severe malnutrition would achieve the nutrient density levels of F100, taking as examples 6- to 24-month-old children living in Bangladesh, Ghana, and Latin America. A secondary aim was to identify the key limiting nutrients in each region that may require supplementation to ensure rapid recovery from severe malnutrition.
MATERIALS AND METHODS
Initially, the nutrient densities of individual foods were calculated to explore, for severely malnourished children living in Ghana, Bangladesh, and Latin America, potential limiting nutrients and candidate foods for home-prepared rehabilitation diets. Afterward, by use of linear programming analyses, the feasibility of formulating home-prepared rehabilitation diets that achieved the nutrient density of F100 was assessed.
Food List and Food Composition
All of the analyses were carried out by use of a recently published list of foods with reported maximum food portion sizes (n = 59 foods) for 6- to 15-month-old non-breast-fed children living in Bangladesh, Ghana, and Latin America (ie, Peru, Guatemala, and Honduras) (8) (Table 1). To increase the probability of achieving the desired nutrient density levels, anchovies were added to this list, whereas to avoid inadvertently encouraging mothers to replace breast milk with other types of milk after recovery from severe malnutrition, infant formula and cow's milk were excluded from the main analysis. In a separate set of analyses, the impact of excluding these 2 foods was assessed by allowing, in the optimized linear programming solutions, as much as 1000 g/d of cow's milk or 140 g/d of dry fortified milk-based infant formula (9).
The energy and nutrient contents of most foods, in these analyses, were obtained from the World Food Dietary assessment system food composition database (10). However, when values disagreed with similar types of foods, in the World Food Dietary assessment system food composition database or with values published elsewhere (11,12), the published values were substituted. In total, 24 substitutions were made for 7 nutrients in 14 of the 59 foods modeled (ie, 6 substitutions each for thiamin and vitamin E, 4 substitutions each for niacin and copper, 2 substitutions for riboflavin, and 1 substitution each for calcium and vitamin B12).
Initially, the nutrient densities of individual foods were calculated to identify those foods from the list of 59 foods that achieved or exceeded the nutrient densities of F100. To do this, the nutrient densities of each food were calculated (ie, nutrient content per 1000 kcal) and expressed as a percentage of their nutrient densities in F100. The number of foods for each nutrient that equaled or exceeded 100% of its density in F100 was then summed and expressed as a percentage of the 59 foods evaluated.
Combination of Foods Selected Using Linear Programming Analysis
Linear programming analysis is a rigorous mathematical technique that can be used to formulate nutrient-dense diets from a list of foods (13–15). In this study, for each region separately and for all regions combined (n = 4 models), it was used to select an energy-dense combination of foods that achieved or exceeded the nutrient densities of F100. The number of eligible foods modeled per region was 31, 27, 34, and 59 foods in Ghana, Bangladesh, Latin America, and all regions combined, respectively (Table 1). The objective function, in each model, minimized the total gram weight while respecting the nutritional and palatability constraints (Tables 1 and 2). The nutritional constraints ensured that the optimized model solutions provided 1000 kcal of energy, achieved the energy and nutrient densities of F100 (16), and had a phytate:zinc molar ratio of ≤15 (Table 2). The last-named nutritional constraint, which was expressed in a linear form as described elsewhere (17), ensured moderate zinc bioavailability (ie, the estimated zinc bioavailability of F100) (18). For each food modeled, the palatability constraints established an upper gram intake level to ensure realistic maximum food portion sizes (Table 1). These upper gram intake levels were obtained from published maximum portions for infants living in Bangladesh, Ghana, and Latin America (8). Except for eggs and food items consumed in small amounts, these published portions were generally 90% of the observed maximum amounts consumed in any of the 5 different countries. For eggs and food items with a maximum portion below 20 g/d, the maximum portions were set at 50 and 20 g/d, respectively (8). For anchovies, the published maximum portion for fish was used (8).
Linear programming analysis results in either a feasible (optimal) or an unfeasible (ie, at least 1 model constraint is unachievable) solution. For example, a model solution is unfeasible when the highest level of a nutrient that is achievable is less than the model constraint (eg, a model solution's vitamin E density = 20 mg/1000 kcal, which is below the model constraint of 37 mg/1000 kcal). Such a result occurs when nutrient densities (eg, vitamin E) of local foods are low in relation to that of F100. In the context of this study, an unfeasible model solution indicates that it is impossible to formulate a home-prepared rehabilitation diet that achieves the nutrient density levels of F100, using only locally available foods. In such cases, its solution will respect as many of the nutritional constraints as possible; for those it cannot achieve, the nutrients will be provided at optimal levels. An unfeasible linear programming solution, therefore, also defines the key limiting nutrients that may require supplementation. All of the linear programming models were run by use of the standard Microsoft Excel Solver (Frontline Systems, Incline Village, NV).
For 4 nutrients (vitamin E, riboflavin, zinc, and copper), fewer than 5 of the 59 foods achieved (or exceeded) the nutrient density of F100 (Figure 1). Specifically, only 2 foods achieved the desired nutrient densities for vitamin E (spinach and taro leaves) and zinc (spinach and chicken livers), and only 4 foods achieved the desired nutrient density for riboflavin (spinach, taro leaves, chicken liver, chicken eggs) and copper (spinach, taro leaves, tomatoes, and soybeans). Five foods provided the desired levels of vitamin B12, and 10 foods provided the desired levels of calcium. For all of the other nutrients, more than 15 of the 59 foods provided the desired amounts.
Combination of Foods Selected Using Linear Programming Analysis
All 4 linear programming model solutions were unfeasible. These results indicate that realistic combinations of local foods, in each region individually and in all regions combined, will not achieve the nutrient density levels of F100. In all 3 regions, the minimum constraints on vitamin E, riboflavin, niacin, zinc, and copper were not met. In 2 regions (Latin America and Bangladesh), the minimum constraint on calcium was not met, and in Bangladesh, the minimum constraint on thiamin was not met (Figure 2). All of the other constraints, including those on energy density and the phytate:zinc ratio, were fulfilled. The foods selected, in each unfeasible solution, are shown by region in Table 3. They include between 2 (Ghana) and 4 (Latin America) animal-source foods (liver, eggs, anchovies, cheese), soybean oil (only Bangladesh and Latin America), legumes, roots, and a wide variety of fruits and vegetables. Despite the inclusion of these nutrient-dense foods, the densities of vitamin E, riboflavin, zinc, and copper in the optimized model solutions were consistently below 50% of the F100 levels (Fig. 2). Likewise, even when as much as 1000 g/d of cow's milk or 140 g/d of dry fortified milk-based infant formula was allowed as eligible food variables, all 4 linear programming models were unfeasible, and several nutrients remained below 50% of the F100 levels (22%–25% for zinc, 12%–16% for vitamin E, 35%–37% for riboflavin, and 42%–51% for copper). Thus, the exclusion of these foods, in the main analysis, did not have an important influence on the study conclusions.
In this study, all of the analyses (ie, individual foods analyses and the 4 linear programming models) indicate that even optimal home-prepared rehabilitation diets in Ghana, Bangladesh, and Latin America are unlikely to achieve the nutrient density levels of F100, especially for vitamin E, riboflavin, zinc, and copper (ie, generally <35% of their desired levels). Instead, to achieve the nutrient density levels of F100, our results suggest that additional supplemented nutrients are required at least for these 4 key limiting nutrients; and perhaps also for thiamin, niacin, and calcium.
In all of the analyses, the models aimed to achieve the nutrient density levels of F100. These levels were chosen because F100 is known to support rapid catch-up growth after severe malnutrition (2), it is endorsed by the WHO, and this product or nutritionally similar products are widely used in the treatment of severe malnutrition (1,2). In general, to correct preexisting nutrient deficiencies and to support rapid weight gains that are often 10 to 20 times higher than those of well-nourished children of the same age (1), the nutrient densities of F100 are high compared with those of “natural” foods. In addition, the nutrient requirements of severely malnourished children differ from those of well-nourished children because the types of tissue laid down during recovery from severe malnutrition differ (19). Such differences most notably influence their requirements for potassium, magnesium, zinc, and a range of vitamins to support muscle synthesis. Thus, it is not surprising that the nutrient density levels of F100, as a therapeutic food formulated to ensure that the unique nutrient requirements of severely malnourished children are met, are difficult to achieve in a rehabilitation diet based only on locally available foods.
Nevertheless, the use of F100 as the reference diet for our analyses has its limitations. First, its formulation was based on results from only a limited number of studies in a few settings showing rapid weight gain with its use (2). Consequently, in other settings with different underlying nutrient deficiencies, similar weight gains would perhaps occur with fortification levels lower than those of F100 (6,20,21). Second, and more important, the clinical consequence of rapid weight gain may be less important than restoration of normal body composition and normal body function, so that lower weight gain may also result in satisfactory recovery. For these reasons, our results (ie, unfeasible linear programming model solutions) do not contraindicate the use of home-prepared rehabilitation diets as therapy for severe malnutrition. Instead, they raise an important cautionary note against their blind acceptance, particularly when one considers the generally poor growth responses observed in the 1980s and early 1990s, when fortified foods or micronutrient supplements were not in common use (7). Therefore, when a home-prepared rehabilitation diet is promoted as the only therapy for recovery from severe malnutrition, it is essential to critically evaluate its nutrient density and to seriously consider supplementing key limiting nutrients, especially when recovery rates are poor in small-scale feeding trials.
Also of note was that the linear programming model solutions contained generous amounts of nutrient-dense foods (Table 3), which are likely unrealistic for poor families. For example, the linear programming solutions contained between 2 and 4 animal source food portions per 1000 kcal, contained seasonal fruits such as mangoes and avocados, and excluded key staple foods such as rice, tortillas, and corn dough products. Moreover, even when milk or fortified infant formula was allowed in the optimized diets, the key limiting nutrients still generally remained below 50% of desired levels. Clearly, it will be difficult for poor families in different regions of the world to prepare home-prepared rehabilitation diets that achieve the high nutrient density levels of F100, using only locally available foods.
Even though linear programming is a rigorous mathematical approach, its validity relies on model constraints and, in this study, on the comprehensiveness of its food lists and maximum food portion size estimates, especially for locally available nutrient-dense foods. In this study the maximum food portion sizes were obtained from 6- to 15-month old children who were not severely malnourished. The appetite of a severely malnourished child is substantially increased during the catch-up growth phase; therefore, their ability to consume large amounts of food may exceed that of other children. Notwithstanding, the maximum food portion sizes used in this study were generous: they were set at 90% of the maximum portions consumed by any child in the studies reported (8). They were also well above the average portion sizes reported for 6- to 24-month-old American children (22). Slight errors in portion size estimates also would not modify the study's conclusions because for some nutrients, their densities in local foods were well below their corresponding levels in F100 (Fig. 2).
In conclusion, our results suggest that home-prepared rehabilitation diets prepared from locally available foods for severely malnourished children in Ghana, Bangladesh, and Latin America will not achieve the nutrient density levels of F100 without additional supplemented nutrients. In all regions, vitamin E, riboflavin, zinc, copper, and niacin were the key limiting nutrients, which will likely require supplementation, and in some regions calcium and thiamin were also limiting. Finally, before home-prepared rehabilitation diets can be promoted as sound community-based therapy for recovery from severe malnutrition, our results point to a need for well-designed clinical trials to establish their efficacy and feasibility, in terms of acceptability, practicability, sustainability, cost, nutrient bioavailability, and rates of child weight gain.
1. World Health Organisation. Management of Severe Malnutrition: A Manual for Physicians and Other Senior Health Workers. Geneva: World Health Organisation; 1999.
2. Waterlow JC. Protein Energy Malnutrition. London: Edward Arnold; 1992.
3. Ciliberto MA, Sandige H, Ndekha MJ, et al. Comparison of home-based therapy with ready-to-use therapeutic food with standard therapy in the treatment of malnourished Malawian children: a controlled, clinical effectiveness trial. Am J Clin Nutr 2005; 81:864–870.
4. Briend A. Treatment of severe malnutrition with a therapeutic spread. Field Exchange 1997; 2:15.
5. Prudhon C, Prinzo ZW, Briend A, et al. Proceedings of the WHO, UNICEF, and SNC Informal Consultation on Community-Based Management of Severe Malnutrition in Children. Meeting report. Food Nutr Bull 2006; 27:S99–S104.
6. Khanum S, Ashworth A, Huttly SR. Controlled trial of three approaches to the treatment of severe malnutrition. Lancet 1994; 344:1728–1732.
7. Ashworth A. Efficacy and effectiveness of community-based treatment of severe malnutrition. Food Nutr Bull 2006; 27:S24–S28.
8. Dewey KG, Cohen RJ, Rollins NC. WHO technical background paper: feeding of nonbreastfed children from 6 to 24 months of age in developing countries. Informal Working Group on Feeding of Nonbreastfed Children. Food Nutr Bull 2004; 25:377–402.
12. Ferguson EL, Gibson RS, Opare-Obisaw C, et al. The zinc, calcium, copper, manganese, nonstarch polysaccharides and phytate content of seventy-eight locally grown and prepared African foods. J Food Comp Anal 1993; 6:87–99.
13. Briend A, Darmon N. Determining limiting nutrients by linear programming: a new approach to predict insufficient intakes from complementary foods. Pediatrics 2000; 5:1288–1290.
14. Briend A, Ferguson E, Darmon N. Local food price analysis by linear programming: a new approach to assess the economic value of fortified food supplements. Food Nutr Bull 2001; 22:184–189.
15. Briend A, Darmon N, Ferguson E, et al. Linear programming: a mathematical tool for analyzing and optimizing children's diets during the complementary feeding period. J Pediatr Gastroenterol Nutr 2003; 36:12–22.
16. UNDP-IAPSO. Emergency Relief Items: Compendium of Generic Specifications, Vol 1. Copenhagen: UNDP, 1995:107.
17. Darmon N, Ferguson E, Briend A. Linear and nonlinear programming to optimize the nutrient density of a population's diet: an example based on diets of preschool children in rural Malawi. Am J Clin Nutr 2002; 75:245–253.
18. WHO/FAO/IAEA. Trace Elements in Human Nutrition and Health. Geneva: World Health Organisation, 1996.
19. Ashworth A, Millward DJ. Catch-up growth in children. Nutr Rev 1986; 44:157–163.
20. Greco L, Balungi J, Amono K, et al. Effect of a low-cost food on the recovery and death rate of malnourished children. J Pediatr Gastroenterol Nutr 2006; 43:512–517.
21. Ashraf H, Ahmed T, Hossain MI, et al. Day-care management of children with severe malnutrition in an urban health clinic in Dhaka, Bangladesh. J Trop Pediatr 2007; 53:171–178.
22. Fox MK, Reidy K, Karwe V, et al. Average portions of foods commonly eaten by infants and toddlers in the United States. J Am Diet Assoc 2006; 106:S66–76.
© 2008 Lippincott Williams & Wilkins, Inc.