Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Computer-optimized weaning food blends

R. E. Hayes, J. M. Mwale, P. Tembo, and J. I. Wadsworth


This computerized linear programming study was performed to formulate low-cost, commercially processed, blended weaning foods, and even less expensive, home-prepared, blended wearing foods, with excellent energy and protein value for use in areas of Lusaka, Zambia, where the risk of malnutrition is high. Ingredient input information included proximate and essential amino acid analyses, into-the-blend costs, and factors for gastrointestinal absorption of proteins and total energy. For most computed mixtures, the minimum cost at selected levels of absorbed protein quality was optimized. Formulations were determined, with calculated cost and nutritional parameters, using only mother-favoured ingredients or, alternatively, selections from among all listed market ingredients; fat versus fat plus sugar as concentrated energy; for different seasons (home-prepared); and, for commercial blends, both with and without puridies amino acid(s) that do or do not include vitamin-antioxidant-mineral mix. The methodology should be widely applicable elsewhere.


This study describes the development of both home prepared and commercially processed weaning blend formulations that incorporate indigenous food come modifies. The blended foods were specifically designed for use in areas of high population density in Lusaka, the capital city of Zambia. The same set of procedures can be used to develop weaning food formulations for any country. The ingredient types and proportions were optimized by computer [1]. using linear programming, in most cases to obtain mixtures of minimum cost and at selected protein quality levels as determined by absorbed amino acid score. In a few instances of commodity sets with substandard protein quality, computer optimization was directed to specify the mixture that would most closely approach the standard. These optimizations were carried out in conformity with certain compositional, nutritional, and acceptability criteria together with considerations of commodity availability.

Blended nutritious foods would benefit children at any age but are especially targeted toward those of weaning age. This usually begins at about four months of age and occasionally extends up to two years or more in the Lusaka area. The weaning period begins when the first food is used to complement breast milk and ends when a child is fed completely with foods other than breast milk.

The need was to develop weaning foods for Zambia that would be practical and affordable by minimizing imported components in commercially processed blends. Equally important was a set of formulations that could be prepared at home, as an alternative to purchased products. For both types of weaning foods, computer optimization, which was developed at the Southern Regional Research Center, US Department of Agriculture [1, 2], is especially applicable.

A very important first step in developing weaning foods is to assess accurately the food-consumption pattern of the target population, together with the influential cultural, social, and economic factors [3, 4]. Therefore, this study was preceded by a survey of weaning practices and foods in areas of Lusaka with high population density [5]. The computed for formulations would also be useful in other geographical areas with similarly available and similarly priced components.

Materials and methods

Home- and village-prepared weaning blends, known as the HOVIPREP type [4], are made of homegrown foods or those available in local markets. Such blends are considered to have greater utility than either commercially processed or donated food aid mixtures [4, 6]. Although HOVIPREP blends may not be as nutritionally complete or as carefully controlled from the standpoint of food safety, they have the advantage of being much more affordable by families with limited income.

Major commodities available in substantial quantities throughout much of the year are considered as ingredients for commercially processed blends, and often vitamin and mineral supplements are added. These industrially produced foods are intended primarily for use in supplementary and school feeding programmes, and in hospitals and other institutions. They are usually purchased by the government, organizations concerned with health, and church groups, and are also offered for public sale to those who can afford them.

Home-prepared ingredients

Since protein-energy malnutrition is an important problem in Zambia, a survey was made of the principal markets in Lusaka to determine available protein- and energy-providing foods. This list was narrowed to a group of items that were known to be the most widely available and commonly used. Prices were determined for these market commodities, as was seasonal availability.

The processing strategy was the same as that used for sarbottam pitho (Super Flour), a successful homeprepared weaning food developed in Nepal [7; M.E. Krantz, unpublished observations, 1986]. It uses local foods and simple processing techniques, and produces quick-cooking, nutritious blends.

Except for the electric grinding mill, the items of equipment are commonly available in households of the areas for which the approach was designed. It was assumed that a type of mill designed by the Adaptive Technology Unit, School of Engineering, University of Zambia, could be installed at each clinic in these areas so that mothers bringing their roasted grains, oilseeds, and nonoilseed legumes could have them ground after mixing them in correct proportions. Specified non-roasted ingredients, such as dried leaves, milk (fresh or dried), raw egg, sugar, and cooking oil, are added when the porridge is constituted and cooked.

Seeds were cleaned (separated from foreign materials) in a winnowing basket that also separated hulls when necessary. After winnowing, rice was spread on a clean, flat surface to remove stones by hand. Before winnowing, hulls were loosened in a wooden mortar with a wooden pestle. The mortar and pestle were also used for pulverizing other commodities such as dried leaves. One of the main reasons hulls were removed from certain seeds was to control fibre content, which should be kept low for good digestibility and caloric density in weaning foods [8]. Congealed materials from pounded groundnuts, leaf veins from pounded leaves, and scales and some unpulverized bones from pounded kapenta (Limnothrissa and Stolothrissa species of small, dried fish) were removed with a wire mesh sieve.

Grains, oilseeds, and non-oilseed legumes were roasted by placing two cupfuls (total volume 560 ml) in an aluminium pot with a fitted lid. The pot was placed over a prepared charcoal brazier and, to provide even roasting, the mixture was stirred periodically. Roasting gelatinizes starch, which reduces cooking time when the flour is later made into a gruel, and inactivates antiphysiological factors in legumes [4]. In the case of sarbottam pitho, roasting and grinding make the whole grains and pulses highly digestible and increase their acceptability by imparting a nutty flavor to the food [7].

Roasted grains and legumes have been used successfully to prepare weaning foods in India [9, 10] and Thailand [11]. In Zambia groundnuts generally are roasted in a pan; therefore this approach was ape plied in preparing the weaning food because of its cultural suitability.

Three approaches were taken to determine the length of roasting required under the conditions described. Viscosity measurements were made on two minute cooked gruels prepared from certain cereal grains that had been roasted for different periods of time. This was done to assess the degree of starch gelatinization achieved, which in turn would influence the length of cooking time required to prepare a blended gruel. These cooked gruels were also tested for grittiness, another index of the degree of doneness. Besides the tests on cereal grains, certain legumes, each roasted for different lengths of time, were prepared as two-minute cooked gruels and tested for grittiness. Softness in cooked legumes is considered important for good digestibility in feeding to young children [4]. The amount of urease inactivation in soy indicated the degree of inactivation of antibiological factors such as trypsin inhibitor [4,12, 13].

Before conducting viscosity and softness tests, the roasted seeds were ground through a 0.3-mm screen, the smallest available (8 Lab Mill, 8,000 rpm; Christy and Norris Ltd., Chelmsford, England). Sieve analyses were conducted on ground maize and ground rice samples to compare particle size with that of white maize breakfast meal (65% extraction rate), which is used in Lusaka for porridge and nshima. (Nshima is a thick dumpling usually made by continuously stirring cereal meal into boiling water to obtain the desired consistency.)

Porridges were made to various thinner consistencies. All maize samples throughout this study were the white maize variety. In conducting these analyses, three screens of 3.36, 0.600, and 0.250 mm were positioned over the pan, in order of opening size, with the one with the largest opening at the top. A 50-g sample was poured evenly over the top screen, a cover was placed over the screen, and the set was shaken for exactly 15 minutes on a mechanical sieve shaker with tapper [14, 15]. The procedure was repeated using a 0.150-mm screen.

Viscosities were determined with a Brookfield model RVT viscometer. At 50 rpm, a no. 5 spindle was used for gruels prepared from cassava flour, breakfast maize meal, and maize meal, and rice flour made from roasted grain. Also at 50 rpm, a no. 7 spindle was used for gruels prepared from roasted sorghum, which were more viscous than the other pore ridges. Measurements were made on both cooked and uncooked porridges made from the commodities that had been roasted for different lengths of time [16].

The instant or uncooked gruel was patterned in concentration after instant corn (maize)-soy milk [13]. It was prepared as follows, using the temperatures and viscosity measurements suggested by Jansen et al. [16]. Twenty-five grams of the ground sample was stirred into 100 g of water at 70C in a 400-ml beaker Any lumps were mashed, and the viscosity was measured at 45C to 50C as described above. The concentration and the general procedure chosen for the cooked gruel were determined from viscosity results on maize [16]. The sample and water weights were chosen to result in a gruel concentration in the range producing high viscosity for processed, cooked maize |16].

The cooking time was modified to conform to that used in preparing corn (maize)-soy milk for consistency measurements [13]. In the cooking procedure, 17.6 g of sample was placed in a 400-ml beaker, and 5 g of cold water was added and mixed to make a smooth paste. Boiling water, 90 g weighed into an appropriate container, was added to the paste and the mixture was brought to a boil for exactly two minutes. The beaker was then removed from the heat and cooled to 50C. Compensation was made for evaporation of some of the water by adding more water at 50C to make up the mixture's total original weight. The viscosity was measured as described above.

A simple technique of pressing porridge between microscope slides was used as an index of softness of cooked cereals and legumes [17]. In general, the method of gruel preparation for measuring consistency of corn (maize)soy milk [13] was followed. The concentration was similar to that used for the previously described viscosity measurement, with slightly different sample and water quantities, and allowance was made for some evaporative water loss in the softness test.

The technique was standardized as follows. Water, 100 ml, was brought to a boil in a 400-ml beaker. Immediately on boiling, 17.8 g of the ground commodity was stirred into the boiling water. The mixture was reboiled and stirred for exactly two minutes. Then the beaker was removed from the heat and allowed to cool for exactly three minutes. A level teaspoonful (5 ml) of the cooked gruel was spread as evenly as possible on a microscope slide. Another microscope slide of identical size was placed on top of the porridge. A 5-kg weight was centered on the top slide and allowed to remain there for exactly two minutes. After this the weight was removed from the top slide, and the pressed porridge on the bottom slide was examined for grittiness and general consistency. Using the same procedure, softness tests were also conducted on some representative potential blend ingredients that were not roasted.

The method of the American Association of Cereal Chemists (AACC) [18] was used to determine urease in duplicate soya bean samples, ground as described. Tests were made on samples roasted for 10 and 15 minutes.

To calculate the cost of each ingredient that would go directly into a weaning blend, processing losses were determined for commodities requiring modification from the "as purchased" condition in the market. For example, some losses occur due to removal of dirt, stones, weed seeds, and other debris; evaporation of water during roasting; and removal of hulls, leaf veins, fish scales and bones, and congealed material after pounding. In the case of some potential ingredients such as cooking oil, sugar, and dried yeast, for which there was no processing loss, the market price required no modification.

Commercially processed ingredients

These items were selected on the basis of protein and/or energy value, consistent availability throughout much of the year, and organoleptic and cultural acceptability. Consideration was given to incorporating foods that had been previously investigated for use in such mixtures by the National Council for Scientific Research in Zambia.

Wholesale prices were obtained for the potential ingredients of commercially processed blends. One option provides for not using any imported chemicals, with a resultant lower protein quality. Other options-provide for these blends to use, as required, one or more imported purified amino acids with and without imported vitamins, minerals, and antioxidants. Selection was made from among the four essential amino acids most likely to be deficient in weaning foods: L-lysine, L-methionine, L-tryptophan, and L-threonine. Import duty is not levied in Zambia on such chemicals used in weaning foods.

Because of the widespread use of extrusion cookers in large-scale production of weaning blends [19, 20], and because of the recent introduction of this equipment into Zambia, it was assumed that this processing approach would likely be used. Although some potential ingredients (e.g., vegetable oil, sugar) can be used directly without modification, others must be processed before they are put into an extrusion cooker. It may be necessary to remove extraneous debris, dehull or deskin to reduce fibre and tannins, and grind some items into flour to ensure uniform blending.

Initial cleaning, removal of hulls and skins, and grinding are generally done by industrial machines. However, these operations were conducted manually in our laboratory for purposes of estimating processing losses up to the point at which the ingredient would go into an extrusion cooker. Samples were ground for chemical analyses through a 0.3mm screen in a Christy mill, as described earlier.

The computer formulation used in this study takes into account the prices of individual ingredients in a blend. The actual cost of most blend ingredients per kilogram as they enter the extruder cooker is greater than the wholesale costs by factors that account for actual losses of material (e.g., in hull removal) and also the expense of the precooker processing (e.g., for operating the machinery to remove the hulls). Estimates of such processing costs were not readily available in Zambia. Therefore, approximations on the basis of percentage raw material cost were obtained from industrial and trade association sources in the United States.

Chemical analyses

Proximate analyses for all potential ingredients and essential amino acid analyses for protein and other amino acid-providing ingredients are required as part of the database for the described computer formulation of weaning blends.

Proximate analyses on most of the study ingredients were performed in the laboratories of the National Council for Scientific Research. Duplicate determinations for ash, crude protein, fat, and crude fibre were done by AACC procedures [18]. Vacuum oven moistures were measured by methods of the Association of Official Analytical Chemists (AOAC) [21]. The proximate analysis values are stated to be representative, because it was sometimes necessary to repeat particular analyses on other batches of the same commodity. In such instances, values were adjusted for differences in moisture content between the original and the new batch. Values in the literature were used for raw egg and fresh milk [22], sugar [23], and the vitamin-antioxidantmineral premixes [1]. An industrial company supplied proximate analysis information for the four amino acids considered for the commercially processed blends.

Adjustments had to be made in the conventional crude protein (N x 6.25) determination for yeast and mushrooms because substantial proportions of Kjeldahl nitrogen are nonprotein in origin. It was assumed that four-fifths of the nitrogen in yeast and two-thirds of that in mushrooms is protein nitrogen [22]. In accordance also with published recommendations [22], total nitrogenous matter, rather than just protein nitrogen, was used to calculate the food energy for these foods. So that the proximate analyses would total 100% for yeast and mushroom, the remaining non-protein component was arbitrarily included in the carbohydrate by difference category of the proximate analysis.

Samples of ingredients in both home-prepared and commercially processed weaning blends, prepared as described, were sent by air to the United States for amino acid analysis. Total nitrogen was determined by a macro Kjeldahl procedure (Armour Pharmaceutical Company, method 28). Except for tryptophan and the sulphur amino acids, the essential amino acids were determined by ionexchange chromatography using an amino acid analyser as specified by Pellett and Young [24]. For tryptophan, a basic hydrolysis procedure [24] was followed by precolumn derivatization with phenylisothiocyanate and separation by high-performance liquid chromatography (HPLC) [25, 26]

For cysteine, cystine, and methionine, the performic acid pre-oxidation and acid hydrolysis [24] were followed by phenylisothiocyanate derivatization and separation by HPLC [25, 26]. Literature values were used for essential amino acid contents of whole and dried milk, and raw eggs [27]. The amino acid come position of baker's yeast, based on direct determination of protein rather than on N x 6.25, was supplied by an industrial company.

Roasted home-prepared ingredients were evaluated for available lysine using 1-fluoro-2,4-dinitrobenzene [24] and for available sulphur amino acids through initial reduction with 2-mercaptoethanol, followed by performic acid oxidation and separation by ion-exchange chromatography [28; J. W. Finley, personal communication, 1991].


Contents - Previous - Next