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

Kenneth H. Brown, Hilary Creed-Kanashiro, and Kathryn G. Dewey


Optimal nutritional care of young children requires application of feeding guidelines based on scientific knowledge of children's nutrient requirements and the ability of breastmilk and other foods to satisfy these nutritional needs. This paper reviews recent information on the appropriate duration of exclusive breastfeeding and timing of introduction of complementary foods; the relationship between frequency of feeding, dietary energy density, and total daily energy consumption; and the importance of nutrient composition and selected organoleptic characteristics of complementary foods as determinants of dietary intake. The role of child appetite is also discussed Finally, programmatic options for the promotion of enhanced complementary feeding and relevant practical experiences in Peru are reviewed.


Exclusive breastfeeding is widely recognized as the optimal means of feeding and caring for young infants during the first few months of life. Nevertheless, at the stage in the growing infant's development when breastmilk alone is no longer able to satisfy the increased physiological requirements for energy and specific macro- and micronutrients, additional foods or nutrient supplements must be provided to prevent nutritional deficiencies and secondary functional impairment. The introduction of "complementary foods" (that is, foods consumed in addition to breastmilk to meet nutrient needs) is not, however, without risk. Recent evidence indicates that these foods may displace breastmilk, thereby producing a greater likelihood of nutrient deficiencies if, as is common in many low-income countries, the density and bioavailability of nutrients in the complementary foods are not equal to or greater than those of breastmilk. By the same token, these foods may interfere with the absorption of nutrients in breastmilk. Moreover, the frequent occurrence of microbial contamination of non-breastmilk foods in many developing countries can substantially increase the risk of diarrhoea and other infectious diseases, possibly resulting in a net negative impact of these foods on nutrition status and other adverse health outcomes. Thus, the presumed benefits of initiation of non-breastmilk food sources must be balanced against these possible risks.

For these reasons, the optimal timing of introduction of complementary foods has been controversial, and numerous articles have been written on the so-called weanlings' dilemma [1-5]. The present paper will briefly re-examine these issues in the light of newly available information, and will provide specific feeding recommendations based on current, albeit incomplete, scientific knowledge. Specific feeding guidelines will be emphasized, because we assume that caregivers require detailed, quantitative information on particular aspects of child-feeding regimens and dietary components to permit them to establish caregiving behaviours that will ensure adequate child nutrition.

Issues such as the timing of introduction of complementary foods, their composition and organoleptic characteristics, and the frequency of feeding will be considered. Much of this discussion will focus on factors that influence total daily energy intake. It must be recognized, though, that the content and bioavailability of micronutrients in complementary foods may be equally or more important determinants of total energy consumption and growth in some situations.

However, a full presentation of these aspects of "dietary quality" is beyond the scope of this paper. Following the examination of specific feeding behaviours and composition of complementary foods, the microbiological quality of these foods will be discussed briefly. Finally, programmatic experience in promoting improved complementary feeding practices in Peru will be presented.

Timing of introduction of complementary foods

Full-term, normal-birth-weight infants

Until recently, our concept of the appropriate timing for introduction of complementary foods was based on a comparison of the theoretical energy requirements of young infants and their energy intake from breastmilk at different ages. Thus, it was assumed that when the average energy intake from breastmilk falls below these theoretical requirements, additional energy sources need to be offered. This concept, however, ignored the fact that our understanding of the theoretical energy requirements of breastfed infants was limited and was based mostly on observed intakes by nonbreastfed infants who were fed infant formula in feeding bottles [6].

We now know that energy intakes of breastfed infants are less than those of formula-fed infants [79], and the patterns of weight gain of the two sets of children differ considerably [10]. Although the current level of recommended energy intake for infants from 6 to 12 months of age is 98 kcal/kg/d [6], breastfed infants typically consume somewhat less than this, even in affluent populations [9]. In such populations, the lower energy intake by breastfed compared with formula-fed infants is not due to inadequate milk production by their mothers or to inadequate amounts of complementary foods. Rather, it appears that breastfed infants voluntarily self-regulate their energy intake at about 80-90 kcal/kg/d [11, 12],

Despite the different levels of energy intake by breastfed and formula-fed infants, there is no evidence of any functional impairment associated with the lower energy intakes and weight gains of breastfed infants. Indeed, a comparative study of matched breastfed and formula-fed infants in California found that the breastfed infants were at least as physically active as those receiving infant formula, and the breastfed infants had less morbidity from infections [13, 14]. Thus, it seems unlikely that breastfed infants are "underfed," and it is perhaps more plausible that formula-fed infants may be "overfed."

An alternative, experimental approach to deciding when to introduce complementary foods is to randomly assign exclusively breastfed infants to receive additional foods at different ages. If the complementary foods add significantly to the infants' total energy intake and promote increased growth, it might be assumed that breastmilk alone was not meeting the children's physiological needs. If, on the other hand, these additional foods simply replace breastmilk without contributing to total energy intake and growth, it might be assumed that the infants' energy requirements were being satisfied with breastmilk alone. Because of the high frequency of microbial contamination of complementary foods in many developing countries, it would be concluded in the latter case that delayed introduction of these foods is preferable.

A study has recently been completed in Honduras using this experimental design [15]. Exclusively breastfed infants were randomly assigned to receive complementary foods at four months of age or to continue being exclusively breastfed until six months of age. The foods offered to the former group were designed to satisfy the requirements for all essential nutrients and were commercially prepared, precooked, and packaged to ensure that they were of adequate nutritional quality and free from microbial contamination. This approach guaranteed that these factors would not adversely affect the intake and growth of the children assigned to earlier introduction of complementary foods. Notably, the children who received the complementary foods at four months consumed significantly less breastmilk at five and six months, thereby counterbalancing the energy consumed from the non-breastmilk sources (fig. 1).

There were no differences between groups in total energy intake (fig. 1) or patterns of growth (fig. 2). Rates of infectious morbidity were similar across groups, presumably because of the ideal conditions under which the complementary foods were provided. It is unlikely that this would be true under usual field conditions, where introduction of non-breastmilk foods and liquids is consistently associated with increased rates of illness [16, 17]. Of further importance, there were no differences between study groups in the children's subsequent acceptance of non-breastmilk foods or growth beyond six months of age [18]. Thus, introducing the complementary foods before six months appeared to offer no advantage for the children's overall growth performance.

The results of a descriptive study of infant feeding practices and growth of poor Peruvian infants provide further support for these conclusions. This study found that infant growth during the first six months was positively related to the proportion of energy consumed from breastmilk [19]. Moreover, the patterns of growth of predominantly breastfed, low-income Peruvian infants did not differ from those of predominantly breastfed Californian infants of relatively affluent mothers [20].

The Peruvian infants' growth may have been adequate for several reasons in addition to the generally excellent nutritional quality of breastmilk. Not only did the predominantly breastfed Peruvian infants have considerably less diarrhoea and other illnesses than their mixed-fed and non-breastfed counterparts [16], but the breastfed infants tended to maintain their usual energy intake even when they had diarrhoea [21].

Interestingly, whereas non-breastfed infants demonstrated the expected negative relationship between diarrhoea! prevalence and growth, these illnesses were not associated with reduced growth rates among young breastfed infants [19]. A similar reduction of the adverse nutritional consequences of diarrhoea in association with breastfeeding has also been reported among West African children 1221.

FIG. 1. Consumption of energy from breastmilk and complementary foods by age and study group, Honduran infants.

Three groups of Honduran infants were randomly assigned to receive: 1) exclusive breastfeeding (EBF) until six months of age (N = 50), 2) ad libitum breastfeeding and complementary foods (CF) at four months of age (N = 44), or 3) breastfeeding at maintained frequency and complementary foods (CF-M) at four months of age (N = 47). Significant increase in total energy intake by age (p < .01); no significant effect of study group or interaction of study group by age. Data from ref. 15[22]

FIG. 2. Weight-for-age z-scores from 0 to 6 months Honduras vs. USA breastfed infants [15]. See figure 1 for abbreviations

There is also a need to consider the adequacy of micronutrient status of infants who are exclusively breastfed for varying lengths of time. In the Honduras study, exclusively breastfed infants had higher rates of anaemia at six months of age, especially in the subgroup with low birth weight [23]. Nevertheless, for the reasons noted above, it would probably be ill-advised to recommend earlier complementary feeding of these infants to prevent nutritional anaemia. Moreover, even among those infants who received the iron-fortified complementary foods provided by the project, nearly one-fourth developed anaemia by six months of age.

Thus, it may be preferable to focus on enhancing maternal health and nutrition status to reduce rates of low birth weight, and on promoting obstetrical practices such as late clamping of the umbilical cord to ensure maximal transfer of blood and associated nutrients to the newborn infant. Furthermore, in those situations where earlier supplementation of infants is necessary to prevent nutritional anaemia, this might be accomplished by providing specific nutrient supplements instead of potentially contaminated food.

Low-birth-weight infants

It is still uncertain whether this same recommendation for withholding complementary foods until six months is appropriate for low-birth-weight infants ( <2,500 g), particularly those who are small for gestational age. Because these infants are often born to undernourished women, concern has been raised that their mothers may not produce an adequate quantity and quality of breastmilk to maintain expected growth rates through the first six months of life. However, the relationship between maternal nutrition status and milk quantity and composition remains controversial, because few appropriately designed studies have explored these relationships across the full spectrum of maternal nutrition status [24]. Evidence to date indicates that milk production is unlikely to be impaired unless maternal nutrition status is quite severely compromised.

The Honduras study described above had too few low-birth-weight infants to draw definitive conclusions regarding the effects of different ages of initiation of complementary foods. Nevertheless, the subgroup of low-birth-weight children seemed to respond to early supplementation in the same way as those who were of normal birth weight, suggesting that the same conclusions might hold for low-birth-weight infants. Additional studies are needed to gather more empirical data on this subgroup of children.

Optimal feeding behaviours

The following discussion of "feeding behaviours" refers specifically to those caregiver behaviours that may influence children's consumption of non-breastmilk foods once they have been introduced. For example, the caregiver has primary control over the frequency with which foods are offered to the young child, the amount served, the technique with which they are provided (that is, by spoon, hand, feeding bottle, etc.), the composition of these foods, and the level of encouragement or forcefulness used to feed them.

Methodological Issues

The relationships between caregivers' feeding behaviours and children's total energy intakes have not been studied very extensively, so it is worth digressing for a moment to consider some critical methodological features that must be recognized for the proper interpretation of existing studies and planning of future ones. There are two primary methods of studying the relationships between caregiver feeding behaviours and children's energy intakes, neither of which is perfect. Probably both methodologic approaches will be needed to understand these relationships fully.

One study method is simply to observe current feeding practices and to analyse subsequently the relationships between the caregiver behaviours and the children's food consumption. This approach is extremely valuable because it provides descriptive information on current practices in the home environment. However, because there is no control of the number of meals offered, timing of meals, types and preparation of foods, etc., the enormous number of variables makes the statistical modelling extremely complex. Moreover, these descriptive studies only permit an examination of associations between the independent variables (e.g., feeding behaviours, food composition) and total food (energy) consumption.

Because of the nature of the research design, it is impossible to draw definitive conclusions regarding the causal direction of any observed associations or the reasons for these associations. For example, caregivers may learn over time how much individual children are likely to eat at a particular meal. Thus, positive associations between the amount served and the amount consumed do not necessarily imply a causal relationship. Likewise, positive relationships between the frequency of meals and total daily energy consumption may indicate either that caregivers who feed more frequently are able to encourage children to eat more or that children with greater appetites demand more frequent meals.

An alternative research method is to implement an experimental design whereby complementary foods of defined composition are offered a specified number of times per day according to a fixed protocol, either in a clinical unit or in the child's home. The total amount of food consumed can then be assessed in relation to the number of meals offered and their composition. The person feeding the child may either be his or her usual home caregiver or a study aide specially trained to offer food in a consistent manner from one meal to the next. The advantage of this study design is the ability to control both the composition of the diet (to examine the effects of specific food components or organoleptic characteristics on total intake) and the frequency of feeding. By manipulating specific feeding behaviours or components of the diet in a controlled fashion, the experimental design permits causal inferences regarding the effects of these factors on the children's intakes. The disadvantage of this design is the uncertainty regarding the applicability of these results to "natural conditions" in the home.

Regardless of the experimental approach, the research methods are complex, tedious, and costly because they are very labour-intensive. Not only is quantitative information regarding food intake needed, but breastmilk consumption must also be monitored to assess the impact of different complementary feeding regimens on consumption of milk by infants who are still breastfeeding. Preliminary results of studies using the experimental approach are described in the next section. The results of other descriptive studies of current home feeding behaviours in selected settings are described in the case-studies below.

Frequency of meals

We are currently conducting a series of hospital based studies of the effects of different frequencies of feeding and composition of meals on total daily energy intake by fully weaned, recovering malnourished children. The ultimate objectives of these studies are to develop specific feeding recommendations and to design appropriate foods for young children. As indicated in the foregoing paragraph, these studies have the advantage of the exquisite degree of control of feeding practices and food composition that is possible in the metabolic ward, but the results must be interpreted with caution due to the nature of the study subjects and the fact that they are no longer receiving any breastmilk. Because the study subjects were recovering from severe malnutrition, their energy intakes were substantially greater than would be expected for normal infants. Thus, it is conceivable that these results may not be fully generalizable to non-malnourished children.

The first set of studies examined the effects of three meal frequencies and varied energy densities of semi-solid mixed diets composed of rice, milk, sugar, vegetable oil, and supplemental vitamins and minerals [25]. Colouring, thickening, and flavouring agents were added to the diets so that they were generally indistinguishable in taste and appearance. The preparations were fed ad libitum three, four, or five times per day to 18 children from 6 to 18 months of age whose total daily amounts of food and energy consumption were measured by weighing the feeding bowls before and after each meal.

The effects of varied energy density on the amounts of food and energy consumed are reviewed below. Controlling for the level of energy density, the total daily amount consumed was approximately 16% more when the number of meals was increased from three to four per day (p < .001) and 7% more when the feeding frequency rose from four to five meals per day (p = .0.05). The proportionately greater increase in intake with the change from three to four meals per day versus the change from four to five meals per day was statistically significant. Therefore, there may be diminishing returns with still further increases in meal frequency. Importantly, approximately 15 to 20 minutes were required per meal before the children reached satiety, regardless of the meal frequency. Thus, there is an obvious tradeoff between meal frequency and the amount of time required to provide additional meals. This time cost may be a severe constraint to greater meal frequency, especially when the caregivers have multiple competing responsibilities, including other child-care tasks. The total amount of time required to feed the children each day was related only to the number of meals served (p < .001), and not to energy density.

It is important to recognize that the clinical studies described in this section were conducted on fully weaned children. We have not yet completed similar studies in breastfed infants. Recommendations regarding the frequency of feeding complementary foods often fail to consider the potentially adverse effects of frequent meals on breastmilk intake. Even if a mother continues to offer the breast often, it is difficult to stimulate a child who has been satiated by other foods to breastfeed.

Thus, recommendations aimed at increasing the intake of complementary foods should take this into account, especially if the energy and nutrient densities of these foods are lower than those of breastmilk. Obviously, at some age the child will be completely weaned from the breast, but the speed at which this occurs may be influenced by the amount and frequency of other foods offered. Additional studies of the effects of different complementary feeding practices on both total energy and nutrient intakes and on intakes from breastmilk are urgently needed.

Although there are no empirical data from controlled studies of the effects of varied feeding frequency and energy density of complementary foods on total energy intakes by breastfed infants, the amounts of energy that might be required from complementary foods at different ages have been estimated [26]. These estimates, revised as shown in table 1, were calculated as the difference between current WHO recommendations for total energy intake [6] and the average amounts of breastmilk con gumption reported from studies in a number of developing countries [15, 27-33].

These calculations provide only very crude estimates of the average amounts of energy that must be offered from complementary foods, because there is tremendous variability in the amounts of breastmilk consumption that were reported from the different studies, possibly because of different complementary feeding practices, as well as substantial inter-child differences within studies. Also, it must be remembered that currently recommended levels of energy intake very likely overestimate the true needs of breastfed infants. Nevertheless, these estimates provide some idea of the general amounts of energy that might be necessary from complementary foods at different ages.

TABLE 1. Estimated amount of energy required from complementary foods by age group

  Age group (mo)
  6-8 9-11 12-23
Recommended energy intake (kcal/day)a 783 948 1,170
Amount of breastmilk consumed (g,/24 h)b 673 592 538
Energy intake from breastmilk (kcal/d)c 437 337 350
Energy required from complementary foods (kcal/d)d 346 561 820

a. Recommended energy intake from WHO, 1985.
b. Data from ref. 15, 27-33.
c. Assumes 65 kcal/100 g breastmilk.
d Difference between recommended energy intake and estimated energy intake from breastmilk

Appropriate composition of complementary foods

In addition to the particular ways in which foods are presented to children, the composition of the meals can independently affect intake. Children, like adults, respond to the organoleptic, or sensory, properties of food, such as aroma, flavour, "mouth-feel," colour, and appearance. Moreover, because of the relatively high energy requirements of young children and their limited gastric capacity, the energy density or "bulkiness" of the diet can also influence total consumption. Finally, because the neuromuscular mechanisms of chewing and swallowing may not be fully developed in the young infant, the viscosity or consistency of the food may also influence consumption. Each of these issues will be discussed briefly in the following section.

Energy density

Despite the critical importance of adequate complementary foods in maintaining the nutrition and health of young children, there is surprisingly little recent scientific information on the optimal formulation of these foods in terms of their energy density and nutrient content and bioavailability. With regard to energy density, the classic manual on feeding infants and young children by Cameron and Hofvander [34] states only that the energy density should be between 100 and 200 kcal/100 g of food, but no information is provided on the scientific basis for this recommendation.

To develop guidelines for appropriate energy density of semi-solid food mixtures, we varied the range of densities offered during the same clinical studies described above [25]. Diets of four different densities (40, 70,100, or 150 kcal/100 g) were provided in random sequence during each of the dietary periods when three, four, or five meals were offered per day. The mean amounts of the diet consumed (g/kg/d) were significantly greater with successive reductions in the energy density of the diet (p < .001, fig. 3). Nevertheless, the total daily energy intakes (kcal/ kg/d) increased significantly with the more concentrated diets (p < .001).

Surprisingly, there was no clear evidence of a threshold of energy density after which further increases in total energy consumption ceased. Moreover, there were no significant interactions between energy density and feeding frequency. In other words, greater energy intakes occurred with each added meal regardless of the energy density of the diet, and these intakes increased with each higher level of energy density for all meal frequencies.

These results indicate that the children were able to adjust the amounts they consumed to compensate partially for the different energy densities of the diets. However, they still ingested greater total energy from the more concentrated diets. Similar conclusions were drawn in earlier studies of younger infants [35], older preschool children [36], and other recovering malnourished children [37].

Despite the children's apparent attempts to adjust intakes in response to differing energy density, their total energy intakes still varied according to energy concentration. At the lower densities, it is likely that the children were physically unable to consume enough to satisfy their physiological needs because of limitations in gastric capacity, which has been estimated at 30 to 40 ml/kg body weight [37]. The absence of a plateau in intake at the higher densities may occur either because the children had not satisfied their requirements for post-malnutrition compensatory growth, even with the highest density diet, or because the adaptive decrease in intake with a high-density diet is not perfectly regulated. For reasons discussed at greater length elsewhere [25], the latter explanation seems more likely.

As indicated above, specific guidelines are needed on the minimal energy density that will satisfy children's theoretical energy requirements. These guidelines obviously will depend on the assumptions employed regarding the children's recommended level of energy intake. If we decide that the minimum energy density should be sufficient to ensure that nearly all fully weaned children can satisfy their theoretical energy needs from a mixed diet, the target level of energy intake should be set at 2 SD above the assumed mean energy requirement for all children of a particular age.

The current recommended daily energy intake for children 6 to 24 months of age ranges from 95 to 108 kcal/kg, depending on age and sex, and the coefficient of variation in energy requirements is estimated to be 12.5% [6]. Thus, the energy density of the diet of 24-month-old girls, whose mean theoretical energy requirement is 108 kcal/kg/d, should be sufficient to permit any one of them to consume as much as 135 kcal/kg/d, which is 25% greater than the average requirement for all children in this subgroup.

The estimated minimum levels of energy density that would have permitted the average child in the study described herein to consume different amounts of energy are shown in table 2. According to this table, for children in this study to consume 135 kcal/kg/d in a total of just three meals, a minimum energy density of 121 kcal/100 g of diet was necessary. The most conservative approach to establishing a single recommendation for energy density for all children within the age range of 6 to 24 months would be to select the subgroup with the highest theoretical energy requirements and then estimate the energy density that would be needed to permit that level of consumption.

According to the 1985 FAD/WHO recommendations, 24-month-old girls, whose energy requirement is noted above, have the highest energy requirement (per kilogram body weight) within this age range. As indicated in table 1, this implies that an energy density of at least 120 kcal/100 g of diet should be promoted to ensure that this level of energy intake can be achieved by most children, unless more than three meals per day can be provided. Again, it is uncertain whether these same recommendations are appropriate for breastfed infants, and more experimental data are needed.

FIG. 3. Intake of diets of varied energy density by frequency of feeding (N = 18) [25]

Methods to increase energy density

TABLE 2. Estimated energy density (kcal/100g) required to attain different levels of energy intake at various feeding frequencies

  Frequency of meals (per day)
Level of intake (kcal/kg/d) 3 4 5
71 46


76 49


81 52


95 61 48 43
102 65 52 47
108 71 56 51
119 92 63 59
127 106 69 64
135 121 83 69

Unfortunately, it may not be possible to achieve increased energy density of the diet without sacrificing other critical features. For example, reduction of the water content to achieve greater energy concentration is accompanied by an increase in the thickness or viscosity of the diet. Cereal starches become extremely viscous because of their high water-holding capacity once they are gelatinized by cooking. Methods to reduce the viscosity of starches are described below. Alternative means of augmenting energy density are to add non-gelatinous carbohydrates, such as simple sugars or fats. However, each of these approaches implies the addition of empty calories without accompanying protein or micronutrients. Thus, it is extremely important that the entire nutrient profile of the final mixed diet be analysed before glibly recommending the addition of sugar or oil to complementary foods.

For example, adding one teaspoonful of vegetable oil to 100 g of a typical West African pap would increase its energy density from approximately 30 to 70 kcal/100 g but decrease the percentage of energy as protein from about 9% to about 4% [38]. If consumed at a level sufficient to satisfy children's energy needs, the oil-supplemented pap would not meet their protein requirements. A similar effect on micronutrient density (in relation to energy content) could have undesirable nutritional consequences if those micronutrients are ordinarily present in marginally adequate concentrations.

Click here to continue

Contents - Previous - Next