Contents - Previous - Next

Compilation of energy intakes published before and after 1980

Whitehead et al (1981) compiled energy intakes of infants from the literature predating 1940 and up to 1980. The work represented 9046 data points during infancy, weighted to account for sample size. Analysis of the energy intake data revealed a highly significant curvilinear relation between energy intake per body weight in kg and age in months:

Energy intake (kcal/kg/d) = 120 - 10.4 age + 0.76 age²
r² = 0.41 (1)

The quadratic term was significant (P = 0.001). No differences were seen between sexes. The authors attributed the sharp fall in energy intake from O to 6 months of age to the rapidly decelerating velocity of growth, a reduction in the rate of fat storage, and a decrease in energy needed for maintenance per kg body weight. The rise in energy intake from 6 to 12 months of age was ascribed to the increase in physical activity as infants begin to crawl and then walk.

Because of possible secular trends in infant feeding practices, we examined energy intakes of presumably well-nourished infants reported after 1980. An analysis was performed on the mean energy intakes from 19 longitudinal or cross-sectional studies comprising 3574 data points (Table 2, Figures 1 and 2). As noted in Table 2, dietary methodology varied across studies.

Table 1 Energy requirements of infants from birth to l year (FAO/ WHO/UNU 1985)

Weighed dietary records, dietary recall methods, or the test-weighing method for breast milk intake were used. Food intakes were converted to metabolizable energy intakes using food composition tables, or macro nutrients were analyzed and converted to gross or metabolizable energy using Atwater factors. Bomb calorimetry was used to measure the gross energy content of breast milk and formula in a few studies. Mean energy intakes as reported were used in the present analysis. Mean total energy intakes (inclusive of solids) of breast fed and formula-fed infants were weighted by sample size at each monthly interval yielding 107 weighted mean values used in the regression analysis (BMDP1R: Dixon, 1990). The multiple regressions of energy intake per kg body weight on age and age² are summarized below.

All: Energy intake (kcal/kg/d)

= 119 - 9.9 age + 0.82 age²
r² = 0.29; n = 107 (2)

BF: Energy intake (kcal/kg/d)

= 118 - 12.8 age + 0.89 age²
r² = 0.66; n = 59 (3)

FF: Energy intake (kcal/kg/d)

= 122 - 8.5 age + 0.73 age²
r² = 0.36; n = 48 (4)

Each of these three equations was tested against the earlier curvilinear equation published by Whitehead et al (1981). We do not have evidence for a strong secular trend in energy intakes of infants before and after 1980, since the regression coefficients did not differ significantly between the Whitehead and present databases. The ~ test for equality of the regression lines across feeding groups was significant, indicating differences in the relationship of energy intake and age between breast-fed and formula-fed infants (P = 0.001) (Figure 3).

Equations (2), (3) and (4) were derived from energy intake data as reported. Two technical problems with reported data arise in the case of the breast-fed infants. Breast milk intakes measured by the test-weighing method were corrected for insensible water loss (IWL) during the course of the measurement in a few studies only (Heinig et al, 1993; Michaelsen et al, 1994). The systematic negative bias caused by not correcting for IWL during, the test-weighing is well recognized: the difficulty has been to determine the magnitude of correction necessary to fairly represent the ranges of metabolic rates, ambient temperatures, humidities, and air circulation rates likely to be encountered. Rates of IWL measured by a number of investigators were as follows: 1.5g/kg/h, Levine et al (1929); 0.83g/kg/h, Kajtar et al (1976); 0.4-0.6 g/kg/h, Doyle & Sinclair (1982); 2.5 g/kg/h, Orr-Ewing & Heywood (1982); 1.9g/kg/h, Hendrikson et al (1985); 1.14 g/kg/h, Butte et al (1990b); 3 g/kg/h, Dewey et al (1991). Most of the measurements were performed under thermoneutral conditions. Levine et al (1929) noted that rates of weight loss may increase threefold above basal levels in temperatures sufficiently high to induce visible perspiration.

Table 2 Energy intakes of infants reported in the first year of life (kcal/kg/d). Mean ± s.d. (N)

^a Abbreviations: Social economic status (SES); food composition tables (FCT); metabolizable energy (ME); gross energy (GE).

To correct test-weighing values for IWL, the number and duration of breastfeedings also must be known. The systematic bias caused by IWL may be estimated for 1-4 month-old (Butte et al 1985) and 12 month-old breast-fed infants (Dewey et al 1991). Based upon the published weights, milk intakes, number of feedings, and duration of feedings (20 min was assumed for the Dewey report), and an estimated average rate of IWL of 2 g/kg/d, IWL would cause a 4 and 6% underestimation of intake in the 1-4 month-old and 12 month-old breast-fed infants, respectively.

Published intakes of breast-fed infants are in terms of metabolizable energy in some reports, and gross energy in others. Gross energy intake may be converted to metabolizable energy intake using Atwater factors (Watt & Merrill, 1963). Application of the Atwater factors to human milk components (Butte et al, 1984), indicates that human milk would be 96.4% metabolizable. The applicability of the Atwater factors to infants has been questioned, since the original studies were performed on adults (Schulz & Decombaz, 1987). Balance data on ten breast-fed infants fed unpasteurized human milk are available from one study (Southgate & Barrett, 1966). Metabolizable energy averaged 92%.

If not already corrected, the energy intakes of breast fed infants presented in Table 2 were corrected uniformly for IWL and metabolizable energy. A 5% correction was applied to compensate for IWL, and metabolizable energy was assumed to be 94% of gross energy intake. The energy intakes reported by Martinez et al (1985) differed substantially from those of the other formula-fed infants. These six mean values were eliminated from the database.

All: Energy intake (kcal/kg/d)
= 121 - 10.2 age + 0.72 age²
r² = 0.43; n = 101 (2a)

BF: Energy intake (kcal/kg/d)
= 116 - 12.3 age + 0.83 age²
r² = 0.66; n = 59 (3a)

FF: Energy intake (kcal/kg/d)
= 125 - 9.3 age + 0.64 age²
r² = 0.67; n = 42 (4a)

The curvilinearity of the equation of energy intake on age has important ramifications for energy requirements during infancy. The above analysis confirms White head's earlier observations of decreasing need in the first half of infancy, followed by increasing need in the latter half of infancy. However, the above analysis may be misleading because of a mathematical artifact. Energy intake standardized by body weight was regressed on age, which was highly correlated with weight (r_{age, weight} = 0.97). By dividing the ordinate (energy intake) by the abscissa value (age) or in this case a proxy (weight) for the abscissa, a curvilinear relation is created mathematically with this quadratic equation, irrespective of the actual data (Tanner, 1949). It is misleading to describe the relationship of energy intake on age, with energy intake divided by weight.

To circumvent this artifact, another model relating energy intake (kcal/d) to age with weight as a covariate was developed. Mode refers to breast-fed (coded 0) or formula-fed (coded 1). Data were weighted for sample size.

All: Energy intake (kcal/d)
= 100 - 57.7 age + 3.3 age² + 92.8 weight + 43.6 mode + 13.8 age × mode
r² = 0.81; n = 101 (5)

BF: Energy intake (kcal/d)
= 581 - 21.7 age + 1.1 age² + 24.8 weight
r² = 0.63; n = 59 (6)

FF: Energy intake (kcal/d)
= 11.8 - 71.8 age + 4.0 age² + 130 weight
r² = 0.94; n = 42 (7)

In the regression model of all cases there was both a negative linear term (age) and a positive quadratic term (age²) (P = 0.001). A significant interaction between age and feeding mode was encountered (P = 0.006). Splitting on feeding mode, the age² terms for breast-fed and formula-fed infants were significant (P = 0.04 and 0.001, respectively). A curvilinear trend in energy intake was evident. Further analysis revealed that the curvature could be explained by a significant interaction between age and weight. Energy intake (kcal/d) can best be described by the following regression equations weighted by sample size:

All: Energy intake (kcal/d)
= 210 - 59.2 age + 37.2 mode + 63.1 weight + 14.0 age × mode + 5.6 age × weight
r² = 0.80; n = 101 (8)

BF: Energy intake (kcal/d)
= 640 + 25.6 age-40.1 weight + 1.7 age × weight
r² = 0.62; n = 59 (9)

FF: Energy intake (kcal/d)
= 101 - 89.6 age + 105 weight + 7.7 age × weight
r² = 0.87; n = 42, (10)

In the overall model, weight (P = 0.001) and the interactions of age × mode and age × weight were significant (P = 0.01 and 0.002). The older the infant the greater the positive contribution of age × weight term to energy intake becomes. Energy intake of infants across the 1st year of life is best described in this multiple regression, with weight treated as a covariate.

Energy requirements of infants have been estimated from dietary intake using equations (2a), (3a), (4a) and (8)-(10) (Table 3). NCHS median weights were used to calculate energy requirements. For the estimation of the energy requirements of all infants, it was assumed that half the infants were breast-fed and half were formula fed. The current FAO/WHO/UNU energy requirements for infants are 2-15% higher than these estimates based on energy intakes recorded after 1980. The discrepancy is partially due to the 5% increment added to the 1985 FAO/WHO/UNU energy requirements to compensate for assumed underestimation of energy intakes.

Contents - Previous - Next