1 Murphy, Beaton & Calloway (1992).
² Basal: amount of nutrient needed to prevent clinically detectable signs of functional impairment.
³ Normative: in addition to basal criteria the amount to generate and maintain reserve storage of a nutrient.

All infants were breast fed on demand during the day and night. The estimated duration of breast feeding was 17 months, with nearly 100% of the children weaned completely by 18 months. Despite the introduction of supplemental feedings between the 2nd and 3rd months of life, breast feeding was maintained on demand throughout the day and night. Breast feeding frequency and time spent per feeding were measured, but not the volume of breast milk ingested.

3.3. Study cohorts and duration

Cohorts were comprised of the following age groups and studied for the designated duration: Infants (n = 130): from birth to 6 months of age. Toddlers (n = 120): from 18 to 30 months. Schoolchildren (n = 138): from 7 to 9 years; a subsample of schoolchildren was restudied at 12 to 14 years of age. Adults (n = 247): mothers and fathers (biologic parents): during 1 to 2 years. Pregnant women (n = 138): from pre-pregnancy or the 1st trimester of pregnancy to 6 months post partum. Non-target groups: all household members not included in the cohorts were measured on at least two occasions during the course of the study.

3.4. Methods

Identical or very similar study design and methods were used by the three NCRSP projects. Adaptations were made to local conditions as indicated. The methodology used in Kenya is described in detail in the two Kenya Project Final Reports (Neumann & Bwibo, 1987; Neumann, Bwibo & Sigman, 1992).

3.4.1. Anthropometry Height or length and weight measurements were made independently by two individuals on each subject once a month and the results averaged.

To assess growth, reference data from the National Center for Health Statistics (NCHS), adopted by WHO, were used (Hamill et al., 1979). Studies in well-off Kenyan families from various tribes, both rural and urban, showed that these populations reach or exceed the median values of the NCHS curves, thus justifying the use of NCHS data as a reference (Stephenson, Latham & Jansen, 1983).

Table 2. Mean stature and Z scores for length/height in Embu family members of different ages

3.4.2. Food intake Food intake was measured quantitatively on two successive days per month, rotating the days of the week. Cooked dishes and raw ingredients were weighed and/or volumes measured. Food intake between 6 pm and 7 am was assessed by recall, with use of standard recipes based on weighed ingredients. Nutrient values for the foods were derived from biochemical analyses of 45 commonly eaten ingredients and mixed dishes and the most appropriate food composition data from the USDA, FAO or other African food tables (Murphy et al., 1990). An International Mini-list developed by Murphy & Calloway (1991) was also used. To assess diet adequacy, measured intakes were compared with FAD/WHO/UNU (1985) reference values, and a probability method of assessing dietary deficiency was used, particularly for minerals (Murphy, Beaton & Calloway, 1992).

3.4.3. Morbidity Information on illness was collected weekly by trained observers using recall and observation under the close supervision of the project nurses and physicians. Diagnostic coding was based on pre-set criteria, and severe or potentially serious illness was verified by the physician and/or nurses. Life-threatening or serious illnesses were treated and patients referred whenever necessary.

3.5. Results

3.5.1. Stature and growth rate Mean stature and Z scores for length/height in Embu family members of different ages are presented in Table 2. Growth in length/height for the three child cohorts is shown in Fig. 1, in comparison with the NCHS reference. Table 3 shows the growth velocities, expressed as centiles of the Roche & Hines (1980) reference.

3.5.1.1. Infants The most striking finding was that linear growth faltering started early in life. Infants were born with normal or near normal birth weights, but lengths were at a lower mean Z score relative to weight. Mean birth weight was 3.1 ±0.5 (SD) kg and mean birth length was 49.3 ±2.3 (SD) cm. Eleven percent of neonates were of low birth weight, 76% of which were small-for-gestational-age (SGA). Fourteen percent of newborns weighed between 2501 and 2800 g. Gestational age was determined using the Dubowitz Examination Scale (Dubowitz & Dubowitz, 1977), and preterm infants were omitted from growth analyses.

Fig. 1. Comparison of infant and toddler mean length and schoolchild mean height by month of age with NCHS growth curves (Hamill et al., 1979).

Roche & Hines (1980).

Declines in the rate of weight gain and linear growth were noted in the infants as early as 3 to 4 months of age, with linear growth declining more steeply than weight. By 6 months of age weight-for-age (WA) and length-for-age (LA) Z scores declined further, with incremental growth rates showing corresponding declines (Roche & Hines, 1980). The mean Z score at 6 months of age was -1.8 ±1.0 (SD) for LA and 0.7 ±1.1 for WA.

3.5.1.2. Toddlers By 18 months of age, average length was below the 3rd NCHS centile (Fig. 1). Wasting was not seen, as assessed by weight-for-height (WH), Z scores being above 1.0. Z scores for LA remained lower than those for WA, with no further decline. The incremental growth rates for length from 18 to 30 months of age were consistently low (Roche & Hines, 1980) (Table 3).

3.5.1.3. Schoolchildren Schoolchildren were found to be underweight for age and stunted. They showed very little catch-up, with only slight improvement in Z scores of attained height and in incremental growth centiles (Fig. 1 and Table 3).

3.5.1.4. Teenagers In 12 to 14 year olds, mean Z scores, particularly for HA, but also WA, showed a slight further decline (Fig. 1 and Table 3). This was in part due to a delay in the onset of the pubescent growth spurt compared to the earlier onset in the NCHS sample of age-matched North American children (Hamill et al., 1979). Such a delay in the onset of the growth spurt until the late teenage years has been noted in other Kenyan teenagers (Kulin et al., 1982).

3.5.1.5. Adults The parents exhibited slightly improved Z scores for height (mean: -1.5) (Najjar & Rowland, 1987).

3.5.1.6. Summary A major deceleration of linear growth was seen in infants as early as 3 months of age with a steady decline until 18 months. From 18 months onward the children appeared to grow at a low but normal incremental rate, but never caught up to their neonatal length status of less than 1 Z score below the median reference values. Slight catch-up was seen in late puberty. Foregone growth in stature was considerable, one year of growth by 30 months, which was never regained when compared to NCHS reference values.

3.5.2. Determinants of linear growth

3.5.2.1. Effects of maternal parameters on infant growth To better understand why growth failure developed so early in life, it is important to consider the mothers' food intake and changes in body weight and fat during prepregnancy, pregnancy and lactation.

At the beginning of pregnancy mothers had a mean weight of 54 ±7.6 (SD) kg and BMI of 21.2 ±2.7 (SD). Equally distributed throughout the second and third trimester, mothers gained a mean total of 6.2 ±3.4 (SD) kg, about half of the weight gain of women in the United Kingdom or the USA (Institute of Medicine, 1990). No weight gain was seen in 4% of women, and 3% lost weight during pregnancy. The mean net post-partum weight gain, represented largely by fat stores (Institute of Medicine), was only 1.2 ±3.7 (SD) kg compared to 3 to 4 kg in better fed women.

Mothers progressively lowered their food intake during pregnancy, from a daily intake of 1700 to 1800 kcal/d in prepregnancy to 1400 kcal during the 3rd trimester. The main compensation for this diminished energy intake during pregnancy was the loss of body fat. With time allocation studies, a doubling of inactive time was documented (Neumann, Bwibo & Sigman, 1992). Resting energy expediture per kg showed no compensatory change in relation to decreasing energy intake (Neumann et al., 1990).

Maternal mean daily intake of energy, protein and fat during pregnancy related positively to infant length from 0 to 6 months of age, the strongest association being with fat intake. The mean daily intake during lactation also related positively to attained infant lengths at 2 through 5 months, with fat intake giving the strongest correlations. Pearson's 2-tailed correlations during lactation between maternal intake and infant length ranged from r = 0.29 to 0.36, P < 0.01 to 0.0001 (n = 118).

The mother's size at the beginning of pregnancy determined the infant's size at birth and its subsequent growth. Maternal height, prepregnancy weight, weight in all trimesters and in lactation, as well as height and changes in body weight all consistently and positively correlated with infant length. By regression analysis, maternal height and weight accounted for 17% of the variance in infants' length at 6 months of age.

All parameters of maternal fat stores (sum of 6 fatfolds, percent body fat) during pregnancy and lactation also related positively to length of the infant at 6 months of age, with Pearson's correlations ranging from 0.30 to 0.37, P < 0.001.

The results of multivariate regression analysis for predicting infant length and velocity of linear growth are shown in Table 4. For attained length at 6 months of age, birth weight was the strongest predictor, followed by maternal pregnancy weight and fat gain, and weight gain during lactation. Energy intake from supplemental feeding in the first 3 months had a negative impact on the infants' length. The above variables account for 27% of the variance in weight and for 41% in height at 6 months. The less mature and smaller the infant (lower gestational age), the more rapid was the rate of linear growth.

Maternal fat gain in pregnancy and lactation and maternal energy intake in lactation promoted the velocity of linear growth. This is most probably due to the positive effect of energy intake on milk production, although this was not directly measured. The above factors account for 24% of the variance in rate of linear growth.

3.5.2.2. Impact of supplemental feeding on infant growth By three months of age 90% of infants received supplemental feeding. Of those supplemented, the mean daily intake was 188 kcal/d at 3 months, 228 kcal/d at 6 months. Supplemental feeding consisted of non-commercial preparations of boiled cow's milk and maize or millet gruel, occasionally fermented. These foods did supply some zinc and vitamin B₁₂, which exerted positive effects on linear growth (see below). Overall, supplemental feeding had a weak but statistically significant negative effect on infant linear growth (Table 4). A small increase in febrile but not in diarrheal illness was also found in supplemented infants, in direct relationship to the quantity of supplemental energy received.

Table 4. Multiple regression models for the determinants of growth in Kenyan infants: 0 to 6 months of age

* Other significant variables predicting infant length at 6 months: prepregnancy, pregnancy, lactation height, weight, BMI; maternal intake of animal protein in pregnancy; infant illness: in % days of fever, malaria, diarrhoea (negative association).

Although breast milk intake was not measured, no differences were noted in the duration or frequency of breast feeding in supplemented vs non-supplemented infants. The most frequently stated reason for supplemental feeding was the mothers' perception that their infants were not satisfied solely by their breast milk.

Regardless of the type of feedings, illness episodes had strong negative associations with length and rate of linear growth, weight gain and fat stores in the infants. Such illnesses commonly included fever, clinical malaria, respiratory infection and diarrheal episodes between 4 and 6 months of age (Table 5).

3.5.2.3. Determinants of growth in toddlers Household and familial factors, such as parental size and socioeconomic score (SES), predicted height and weight of the 18 month toddler upon entry into the study (Table 6). The strongest predictor of attained length of the toddler at 24 and 30 months was the length at 18 months (Table 7). After controlling for initial length, parental size alone accounted for 26% of the variation in length, and the quality and quantity of the diet (daily energy, animal protein and fat intake) for 30% of the variation in length.

Other important determinants of linear growth of toddlers, as shown in multiple regression analysis, were SES, season, morbidity and maternal fat intake. The cold season coincided with the pre-harvest months and food scarcity, as well as an increase in diarrheal and other severe illnesses. These factors had negative impacts on the slope of linear growth from 18 to 30 months, as well as attained size at 30 months of age (Table 6). Anemia also had a negative impact on linear growth.

Table 5. Relation to infant illness to attained size and rate of growth from 0 to 6 months (Kenya)*

* Pearson's correlations (2-tailed); all correlations presented are significant: P < 0.001 to 0.05.

Table 6. Household and familial determinants of toddler length at 18 months (multiple regression model)

* Other significant predictors of outcome: child's intake: protein, fat; paternal literacy; number of children in household < 6 years of age.

** Male = 1; Female = 2.

Throughout the toddler years, from 18 to 30 months, energy intake remained low, 75% of the RDA (FAO/WHO/UNU, 1985). Although this was sufficient to maintain growth velocity at a steady level but below the 5th centile, it did not permit any catch-up growth in length. Also, as shown below, diet quality and micronutrient intake remained poor.

3.5.2.4. Determinants of growth in schoolchildren In examining the determinants of growth in school-age children, SES, household sanitation, dietary components and parental size emerge as important factors. From regression analysis, the proportion of the variance in height explained by maternal height was 40%. Diet quality, particularly the intake of animal fat and protein, each accounted for 30% and 18% of the variation in height respectively, even when controlling for energy intake.

Multiple regression analysis was used to predict height at 8 years and the velocity of linear growth from 7 to 9 years of age. For height at 8 years, SES, maternal height and the child's energy intake had a significant effect (Table 8). Fiber intake had a slightly negative effect on height. For slope of linear growth the same variables entered the equation except that energy intake was replaced by animal protein.

3.5.2.5. Micronutrient intake and linear growth In addition to low energy intake, low intakes of iodine, zinc, vitamin B₁₂ and calcium were found to be present in varying degrees in the study population (Table 1). The dietary intake data were supported by clinical and biochemical evidence of iodine deficiency, and hematologic and biochemical evidence (breast milk levels) of low B₁₂. These nutrients were each found to be related to linear growth.

Table 7. Multiple regression models for attained toddler length at 30 months and velocity of linear growth from 18 to 30 months

* Other significant predictors of outcome: Length at 30 months: maternal height; SES; child's daily protein intake. Length velocity 18-30 most: parental height; maternal fat intake.

Table 8. Multiple regression models for attained height at 8 yr. in Kenyan children

Sex: 1 = Male, 2 = Female.

Iodine deficiency

The presence of iodine deficiency was supported by low iodine concentrations in drinking water and the presence of goiter in 18% of non-pregnant women, 24% of pregnant women and 9% of pre-pubertal school age children (Neumann, Bwibo & Sigman, 1992; Neumann et al., 1990). Low thyroid hormone levels (T3, T4, free T4) were found in 11% and elevated TSH in 10% of children (Neumann, Bwibo & Sigman, 1992; Neumann et al., 1991).

In toddlers household use of iodized salt related positively to height (P = 0.04). In the schoolchildren, serum free T4 levels also correlated positively with height (P = 0.04). These relationships remained significant after controlling for SES in toddlers, and for maternal and paternal size in addition to SES in the schoolchildren.

Zinc deficiency

Zinc intake was low in all age groups. Moreover, given the high phytate and fiber content of the diet, the bioavailability of dietary zinc was reduced. An algorithm was used for the calculation of available zinc (Murphy, Beaton & Calloway, 1992) which was then examined in relation to linear growth.

Infants: Maternal pregnancy intake of zinc related positively to the newborn's length. The amount of available zinc consumed by the 3 to 6 month old infant from supplemental feeding (6.7 ±0.5 (SD) mg/d), predicted attained length at 6 months, even after controlling for maternal height but not for SES. The phytate: zinc molar ratio related negatively to length.

Toddlers: In the 18 to 30 month old toddlers, available zinc intake correlated positively with length, as did mean daily energy and animal protein intake. The phytate: zinc molar ratio was negatively related to height at 24 and 30 months of age. Partial correlations between available zinc and length remained positive and statistically significant after taking SES, iodized salt intake and maternal height into consideration (Table 9).

Schoolchildren: Zinc, available zinc and animal protein intake related positively to height in the 8 year olds. The phytate:zinc molar ratio, as expected, showed a negative relationship to linear growth. Partial correlations between available zinc and height lost statistical significance after controlling for SES and maternal height, pointing to their more potent influence than low dietary zinc intake (Table 9).

Vitamin B₁₂ deficiency

Vitamin B₁₂ intake in all cohorts was found to be below 2/3 of the RDA and was associated with very low intakes of animal protein (Table 1). Megaloblastic anemia was seen in 4 to 8% of children and women, and breast milk B₁₂ levels were found to be markedly low (Neumann & Bwibo, 1987; Neumann, Bwibo & Sigman, 1992).

Table 9. Relation between mean daily intake of available zinc during the past year and attained length in toddlers and schoolchildren

* Pearson's correlations (2-tailed).
** Partial correlations, controlling for confounding variables.

The relationship between maternal vitamin B₁₂ intake during lactation and linear growth in infants was positive and statistically significant, even after controlling for maternal energy intake. During lactation, maternal B₁₂ intake and breast milk B₁₂ levels were strongly and positively correlated. Vitamin B₁₂ from supplemental feedings (0.21 ±0.26 (SD) mg daily) did predict length at 6 months even after controlling for maternal energy intake.

In the toddlers, regression analyses indicated that B₁₂ and animal protein intake each significantly predicted height at 30 months. These associations remained significant even after controlling for SES, iodized salt, and maternal height. Animal protein intake correlated highly with B₁₂ intake (r = 0.60, P < 0.001).

In the schoolchildren, although B₁₂ intake and animal protein were positively and significantly correlated with height (n = 43, r = 0.36, P < 0.02); these correlations lose significance after controlling for SES, maternal height and iodized salt use, as was seen for zinc intake in this age-group.

3.5.2.6. Food patterns and linear growth An analysis of food patterns in relation to growth is particularly useful for planning practical interventions. Intake of foods with high content of energy, particularly fat and sugar, animal protein and calcium were positively associated with the toddlers' and schoolchildren's attained length and velocity of linear growth (Table 10) (Calloway et al., 1992).

Two main food patterns were identified by factor analysis (Calloway et al., 1992). The first included milk, cereal grains, fat, potatoes and sugar. This pattern correlated-positively and significantly not only with SES, but with length and with the slope of linear growth in toddlers and schoolchildren. The second pattern included maize, sorghum, green leafy vegetables and beans. The maize component, in particular, was negatively associated with length. This latter pattern contains foods with high phytate and fiber content which reduce diet quality by decreasing the bioavailability of iron and zinc. The first pattern is positively, the second negatively correlated with SES. Poor households cannot afford animal protein foods and rely heavily on maize as the main diet component.

Table 10. Food type predictors of attained size and slope of growth in toddlers and schoolchildren

*Based on factor analyses by Murphy, Calloway & Lien (1992).
** Pearson's correlations (2-tailed), shown only if P < 0.05.

Contents - Previous - Next