Contents - Previous - Next

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

Optimal complementary feeding practices to prevent childhood malnutrition in developing countries

Methods to reduce viscosity

Several methods have been described to reduce the viscosity of cereal mixtures, such as the addition of amylase [39, 40], fermentation [41], and extrusion cooking [42]. Of these, amylase has the greatest and most dramatic ability to reduce the viscosity of a cereal porridge rapidly. The use of this enzyme is particularly attractive because it can be produced at the household level in developing countries by germinating local grains and producing malt flour, which is rich in amylase.

However, the impact of viscosity reduction on total energy intake is still being debated. Whereas some investigators have reported a positive impact of adding amylase or malt flour to children's diets on their total energy intake [39, 43], others have not [33, 40, 44]. A recent review of this topic sponsored by the WHO concluded that amylase- (or malt-) treated diets failed to produce a consistently positive impact on children's total energy intakes [45].

Furthermore, a number of concerns have been expressed regarding the safety of this technology [41, 45]. Possible risks include the increased likelihood of microbial contamination of foods inoculated with home-produced malt flours after cooking, and the possibility of cyanide poisoning from improperly processed germinated grains. Because of the uncertain benefit of amylase-treated foods and the possible aforementioned risks, there does not presently seem to be sufficiently compelling evidence to warrant extensive promotion of this technique, except under controlled conditions.

Other organoleptic characteristics

Other organoleptic features, especially flavour, may affect children's intakes of complementary foods. A study of formula-fed infants in the United States, for example, found that they increased their consumption when sugars with a higher sweetness index were added to their formulas [46]. By contrast, preliminary results from a study of semi-solid diets offered to recovering malnourished Peruvian children indicate that replacing starch with sugar did not affect intakes. However, the children consumed about 10% more when a non-caloric sweetener, saccharine, was added to the high-starch diet [47]. Unlike sweetening agents, the addition of small amounts of salt to blended foods does not seem to affect consumption by young infants [48].

Nutrient content and bioavailabillty

A full discussion of the optimal content and bioavailability of specific nutrients in complementary foods is beyond the scope of this article. Nevertheless, it is important to recognize that these factors may be more important determinants of physical growth and development than is energy consumption per se [49-51]. For example, recent studies found that the growth of rural Mexican children was associated not with their total energy consumption, but with the proportion of their energy intake derived from animal sources [49].

Earlier intervention trials in New Guinea [52] and Sudan [53] have also shown that the quality of the diet, not just the amount of energy offered, influences children's growth. In these studies children who received milk supplements demonstrated greater linear growth than did those who received similar amounts of energy from either beans, tubers, or isolated fat sources. Even in the more affluent setting of western Europe, the physical growth and psychomotor development of vegetarian (macrobiotic) children lag behind those of children who receive animal products in the diet [54].

Although suggestions are available on the appropriate nutrient content of complementary foods [55, 56], better guidelines are urgently needed on suitable ways of achieving these recommended levels of nutrients and assuring their bioavailability, especially when the complementary feeding mixtures are composed of local foods derived exclusively or almost exclusively from plant sources. These issues are of particular concern for the poorest families, because their economic conditions and educational and time constraints may limit their ability to improve the quality of the diet.

Case-studies of dietary and behavioural determinants of energy intake by infants and young children (examples from Peru and Nigeria)

We reanalysed previously collected information on dietary intakes by Peruvian infants in a low-income, peri-urban community and by Nigerian infants and young children in several rural villages. The major purpose of these analyses was to provide descriptive information on the children's meal patterns and selected characteristics of their complementary foods.

Additionally, we analysed factors that we hypothesized might have influenced the children's total daily energy intakes. The results of these analyses are still preliminary. Nevertheless, they provide interesting insights into specific dietary factors that influenced energy intake and how these varied from one location to another. These findings have important implications for the design of interventions to improve complementary feeding practices.

The data from the Peruvian infants were collected during longitudinal studies in Huascar, a low-income, pert-urban community on the eastern limits of Lima [57, 58]. A total of 153 singleton newborns weighing 22,500 g at birth were enrolled in the study for a period of one year. Dietary intake was measured in a subgroup of 131 infants on one or two days each month by means of direct observation and weighing of all foods and breastmilk consumed during 12 daytime hours, as described in detail previously [31]. Nighttime intakes were obtained by recall history and by extrapolation of the amount of breastmilk consumed during the daytime observations.

The total amount of food intake was converted to energy and nutrient intakes using Peruvian and other regional food composition tables and direct measurement of the proximate components of samples of breastmilk. Information is available for 1,626 days of observation of these 131 infants. The index children were also visited in their homes three times weekly to inquire about the presence of specific symptoms of illness during the period since the previous visit [58, 59]. Body weight and recumbent length were measured monthly.

The dietary data were originally analysed as total energy and nutrient intakes per day. Data were also available from most children on intakes from individual recipes, mode of feeding, time of day at which the food or recipe was provided, and amounts offered and consumed. These latter pieces of information provided the opportunity to examine the relationships among feeding frequency, energy density, feeding mode, amounts offered, and source of energy (independent variables), and the total energy consumed during the day of observation and total energy consumed from different food sources (dependent variables). To compare data for children of similar ages from both data sets (i.e., Peru and Nigeria), a total of 720 days of observation of 117 Peruvian children from 6 to 11 months of age were included in the present analyses.

The data for the Nigerian children were obtained during longitudinal studies conducted as part of the Dietary Management of Diarrhoea project in three villages near Ilorin, Kwara State [26, 32]. A total of 214 children from 6 to 28 months of age were included in daily surveillance for diarrhoea! disease and monthly anthropometric assessments. Children were enrolled in dietary studies only after the start of an episode of diarrhoea. Quantitative observations of dietary intake were completed for at least 12 hours on one or more days during diarrhoea, early convalescence, and after full recovery.

All foods prepared, all portions served, and any leftovers were weighed. The amounts of each ingredient consumed were converted into nutrient intakes, using African food composition tables. Breastmilk intakes were estimated from 12-hour test weighings. As with the Peruvian study, information was available on the number of meals, energy density, serving size and amounts consumed for each food and recipe item. Only data from 197 days of observation of 53 children from 6 to 11 months of age are presented in this report.

TABLE 3. Characteristics of study subjects 6-11 months of age and selected food consumption variables in Huascar (Lima), Peru, and Kwara State, Nigeria

  Study site
Variable Limaa Nigeriab
Age (mo) 8.7 1.8 8.8 1.6
Weight (kg) 8.3 1.1 7.0 1.0
Length (cm) 69.1 3.0 67.6 3.3
Total energy intake (kcal/kg/d) 80 26 92 25
Breastmilk energy intake (kcal/kg/d) 43 21 66 21
Non-breastmilk energy intake (kcal/kg/d) 36 38 26 20
Breastfeedings (N/12 h) 4.0 2.0 6.4 1.8
Meals (N/12 h) 3.8 2.5 4.4 1.3
Energy density of meals (kcal/100 g) 67 42 26 18

a N = 117 children, 720 days of observation.
b N = 53 children 197 days of observation

Description of analytic methods

The energy intake data from the Peruvian and Nigerian children were analysed using multivariate analysis of covariance. Response variables were total energy intake, energy intake from breastmilk, and energy intake from other foods, all measured in kilocalories per kilogram body weight per day. Explanatory variables included country (Peru vs. Nigeria), presence of fever, presence of diarrhoea, body weight, age, number of breastfeeds, number of meals, average amount of food offered per kilogram body weight per meal, and average energy density of food offered. The original model included all of the above variables as well as a random subject effect (nested within the country effect) and all two-way interactions of country with the continuous variables. Variables that were not significant (p < .05) in the MANCOVA were removed in a stepwise fashion. Analyses were done with PC-SAS Release 6.04.

The children's weight, length, total energy intakes, and selected characteristics of their dietary patterns are shown for both study sites in table 3. Although the children's energy intakes per kilogram body weight were somewhat less than current WHO recommendations [6], they were similar to observed intakes by breastfed infants of relatively well-off families in California [9].

Nevertheless, because the Peruvian and Nigerian children were smaller than expected for age (mean weight-for-age SD score was -0.51 z for the Peruvian children and -1.94 z for the Nigerian children), their total energy intakes (647 + 201 and 633 + 167 kcal/d, respectively) were considerably less than the amounts currently recommended by WHO (720 to 1,050 kcal/d, depending on age and sex).

The children were fed quite frequently during the 12 hours of observation in their homes. Indeed, it was often difficult to decide exactly when one meal ended and another began, so a "feeding episode" was arbitrarily defined as any intake of non-breastmilk food separated from other occurrences of food consumption by at least 10 minutes. The Peruvian children received an average of four breastfeeds and nearly four additional meals during the observation period. The Nigerian children received more than six breastfeeds and four other meals. The energy densities of the non-breastmilk foods differed considerably in the two populations. Whereas the Peruvian children received mostly non-human milks and soups as complementary foods, with an average energy density of 67 kcal/100 g, the Nigerian children received mostly watery cereal paps, which had an average energy density of only 26 kcal/100g.

The results of the statistical analyses examining those factors that predicted total energy intake and intake from breastmilk and non-breastmilk sources are shown in table 4. The selected results shown in the table are regression coefficients for the number of feeding episodes of non-breastmilk foods and the daily average energy density of these foods. The models control for the children's age, presence of fever, number of breastfeeds per 12 hours, and amount of food offered per meal, all of which were significantly associated with the energy intake variables.

Interestingly, the magnitude of association of the two independent variables of primary interest (i.e., the number of meals and the energy density of complementary foods) differed for the two sets of children. In Peru, each additional meal was associated with an increase in total energy intake of 6.5 kcal/ kg/d, but each kilocalorie augmentation in energy density was associated with an increase in total energy intake of only 0.11 kcal/kg/d. Thus, if all other factors are held constant, an increased consumption of 10 kcal/kg/d would be expected to occur if the meal frequency were increased by 1.5 meals per day or the energy density were increased by 91 keel/ 100g.

In contrast with the findings in Peru, each additional feeding episode in Nigeria was associated with an increase of only 1.8 kcal/kg/d, whereas each kilocalorie increase in energy density was associated with 0.68 kcal/kg/d greater total energy intake. Thus, to achieve an increase in intake of 10 kcal/kg/d in Nigeria would require either 5.6 more meals per day or an increment in energy density of only 15 kcal/100 g. These results must be accepted with all the caution described above, because the original studies were observational and not experimental. Nevertheless, the results imply that different approaches to enhancing energy intake might be appropriate in these two settings.

Greater meal frequency might be expected to yield a more positive impact in Peru than increased energy density of complementary foods, especially considering the difficulty in achieving the densities that would be required. On the other hand, enhanced energy density would seem to be a more effective approach in Nigeria than an increased frequency of feeding, particularly in view of the already high combined frequency of breastfeeding and other meals. These results also highlight the potential value of quantitative studies of dietary intake and the importance of defining current feeding practices before initiating programmatic interventions.

TABLE 4. Relationship between feeding frequency, energy density of non-breastmilk meals, and total daily energy intake by infants 6-11 months of age in Huascar (Lima), Peru, and Kwara State, Nigeria.

  Regression coefficients
Dependent variables Independent variables Lima Nigeria
Total energy intake (kcal/kg/d) No. of meals/12 h 6 5 1.8
Energy density of meals (kcal/100 g) 0.11 0.68
Energy intake from breastmilk No. of meals/12 h (kcal/kg/d) -1.4 -2.5
Energy density of meals (kcal/100 g) -0.1 -0.2
Energy intake from non-breastmilk No. of meals/12 h (kcal/kg/d) 7.9 4.3
Energy density of meals (kcal/100 g) 0.1 0.9

a. Model controls for age, presence of fever, number of breastfeedings/12 h, and amount of food offered per meal.

Importance of child appetite

Despite the important relationships described above, it should be obvious that children's dietary intakes are not only a function of caregiver feeding behaviours and the composition of the diet. Other factors intrinsic to the child, such as the presence of diarrhoea, fever, and parasitic infections [21, 32, 33, 6066], micronutrient status [67, 68], and interactions with the caregiver [69, 70], often determine whether foods that are offered will be accepted. To understand the frequency with which child-related factors interfere with eating patterns, we have recently completed studies of the epidemiology of reported poor appetite in a low-income community on the outskirts of Lima, Peru [71].

To assess the validity of maternal reports of poor infant appetite, we first compared these histories with measured energy consumption on days of observation. Mean + SD total energy intakes on days with reported anorexia were 81 + 21 kcal/kg in infants from 1 to 6 months of age and 71 + 22 kcal/kg in infants greater than 6 months old compared with 94 + 22 and 82 + 21 kcal/kg in the respective age groups when appetites were reportedly normal (p < .001). Energy intake from non-breastmilk sources was more affected than energy intake from breastmilk.

The epidemiology of poor appetite was then assessed in 153 infants who were monitored longitudinally during their first year of life. The prevalence of reported anorexia increased progressively from 2.2% of days of observation at less than 1 month of age to 31.7% at 11 months of age (fig. 4). The presence of reduced appetite was positively associated with infant age and with the presence of fever, diarrhoea, and respiratory illnesses and negatively associated with the consumption of breastmilk.

Previous analyses of data from a subset of these same children showed that the prevalence of poor appetite was negatively associated with growth increments from 6 to 12 months of age [19]. Because of the high prevalence of reported anorexia and its relationship with impaired growth, it is likely that poor appetite, rather than lack of food or improper child feeding behaviours, may explain some proportion of the low energy intake and growth faltering of infants in this and similar communities.

Prevention of microbial contamination

Several studies have demonstrated that foods provided to young children in developing countries are frequently colonized with high levels of faecal indicator organisms [59, 72, 73]. Because organisms transmitted in stool are potentially capable of causing enteric infections, optimal child care practices must include prevention of faecal contamination of their foods. Several studies have found that specific foodrelated practices that influence the likelihood of contamination are the methods of preparation, processing, and serving, as well as the amount of time elapsed between cooking and serving, and the method of storage. These issues are discussed briefly in the following section.

Food preparation and method of serving

Because of the high rates of contamination of food and water in many low-income settings, the best way to ensure that these items are free of pathogens is to heat them to a sufficiently high temperature (>70C) immediately prior to serving [74]. Despite the simplicity of this recommendation, the limitations of available time and cooking fuel often make this impractical for many child caregivers.

In household studies in Peru, we found that food rarely harboured faecal organisms immediately after cooking [59]. However, the degree of subsequent contamination varied according to the type of serving utensil. For example, teas and herbal infusions served with a cup and spoon almost never had high levels of faecal organisms. By contrast, these same preparations given in feeding bottles had considerably greater rates of contamination. The results of subsequent studies indicated that the bottles themselves and the rubber nipples were frequently contaminated before any liquid was put into them. These findings support current recommendations to discourage bottle-feeding.

In some countries hand-feeding of young children is the norm [75]. Because the caregiver's hands may also be a source of potential pathogens, handwashing before feeding episodes is crucial.

FIG. 4. Prevalence of reported anorexia, by age with 99% confidence interval [71]

Food storage

In the aforementioned studies from Peru, the same foods that were free from contamination after cooking had progressively greater concentrations of bacteria with increased duration of storage, presumably because of multiplication of microbes introduced by contaminated utensils or hands at the time of initial serving [59]. The bacterial counts were especially high during the warmer months of the year. Because a minimum dose of bacteria is required to induce some types of enteric infections, the multiplication of bacteria that occurs during food storage may be instrumental in the transmission of infection. Under typical conditions where refrigeration is not available, the only method to reduce this risk is to reheat the foods to a sufficiently high temperature long enough to kill bacteria before serving.

Use of food additives and processing techniques to prevent microbial contamination

A wide range of simple food preservation techniques, such as drying, salting, and fermentation, is practiced commonly at the household level in traditional societies. On the other hand, there is little experience with low-cost food additives that might be used to prevent contamination of food once it is prepared. There is some suggestion that fermented foods may be more resistant to such contamination, possibly because of their low pH or other antimicrobial factors produced during fermentation [76-78], but this body of knowledge is still limited.

One recent study from Tanzania found that children in one village who received fermented cereal drinks had a reduced diarrhoeal incidence compared with children from another village and ethnic group who did not receive fermented foods [79]. These results must be viewed with caution, however, because the study design did not control for the possibly confounding effects of village, ethnic group, and other factors that may have independently influenced the rates of diarrhoea.

Click here to continue


Contents - Previous - Next