Contents - Previous - Next

Table 6 Revised estimates of average requirements for Table 32 in the 1985 FAO/WHO/UNU report on Energy and Protein Requirements

^a Based on growth data in Table 4.

It is possible to compare mean intakes of breastfed infants with the safe levels in Table 7 (although theoretically, group mean intake should be somewhat greater than the 'safe level' of intake; see section 1). However, such a comparison should be made only on the basis of total nitrogen intake (not estimated protein intake), as the factorial model used for Tables 6 and 7 already takes into account an efficiency of utilization of about 70%, both for the growth component (an assumption of the model) and for maintenance (inherent in the estimate of 90 mg N/kg/d, because the assumed slope was 0.73). In fact, the utilization of nitrogen from human milk is likely to be higher than 70% (see section 2.2.8). Nonetheless, if the safe levels for nitrogen intake during the first six months of life are compared with the average nitrogen intakes shown in Table 1, it is clear that the values are quite similar. This comparison is best made using a probability approach, as illustrated in the following section.

2.2.8. Validation of the revised factorial model using an epidemiological approach. In 1988, Beaton and Chery published a paper challenging the protein requirement estimates of the 1985 report. In that paper, they gathered together fragmentary information relevant to the modelling of distributions of intakes and requirements and asked the questions 'Could requirements be as high as had been estimated?' and if not 'How high could they be before there was major inconsistency between the predicted prevalence of inadequate intakes and the starting assumption that breast fed infants at 3-4 months of age were adequately nourished?' They concluded that the 1985 report had seriously overestimated protein requirements of young infants. The principles of the approach that Beaton and Chery applied have been described in greater detail in a National Research Council report (Subcommittee on Criteria for Dietary Evaluation, 1986) and embody what has become known as the 'probability approach' to assessment of observed intakes - an approach that recognizes that two distributions exist - a distribution of intakes and a distribution of requirements.

Table 7 Revised estimates for safe level of protein intake for infants (Table 33 in the 1985 FAO/WHO/UNU report on Energy and protein Requirements)

a. where CV_M = CV for maintenance, taken as 12.5%, CV_G = CV for growth, taken from Table 5 based on 3-month increments.

For the analyses published in 1988, Beaton and Chery could not identify data bases that included in one source all of the variables they needed. They finally approached the problem by assembling information about the likely distribution of protein: energy intake ratios among breastfed infants and then modelling requirements in the same form. This was not their preferred approach but it was workable with the data available at that time. Beaton and Chery suggested that 1.1 g/kg/d was a much more reasonable estimate of the average protein requirement at 34 months than was the 1985 estimate of 1.47 g/kg/d.

In retrospect there were many flaws in the assumptions and estimates that Beaton and Chery were compelled to use. For example, they used the then-existing estimates of energy needs (103 kcal/kg), now known to be considerably higher than the actual intakes of breastfed infants at 3 months. Given that there are now databases which include all of the required information, it was considered worthwhile to reattempt the type of analyses done by Beaton and Chery For this report, data from the Davis Area Research on Lactation, Infant Nutrition and Growth (DARLING) study (Dewey et al, 1991; Heinig et al, 1993) and a subsequent study in the same location (Dewey et al, 1994) were utilized. The dataset included 104 exclusively breastfed infants at 3 months of age. Table 8 shows descriptive data for this cohort. While this cohort is not necessarily representative of all breastfed infants, the dataset allows for a more direct approach to the questions raised in the 1988 paper. Beaton agreed to work with the first author to again ask the question 'How high could infant protein requirements be, given the premise that at three months of age, the breast fed infant is adequately nourished with regard to protein?' In the end, this analysis became an epidemiologic testing of the factorial model presented in Tables 6 and 7.

Table 8 Summary statistics for cohort of 104 exclusively breastfed infants used to test factorial models of protein requirements^a (Mean ± s.d.)

^a Heinig et al (1993); Dewey et al (1994).

Several components of the factorial model were varied to test the impact of changes in the individual components on the predicted prevalence of inadequate intakes among the 104 infants. These trials made use of the individual infants' observed energy and nitrogen intakes, body weights at 3 months and actual growth rates estimated by linear regression using weights at 2, 3 and 4 months. Each factorial model developed was applied to the known body size of each infant to compute a 'requirement' for that infant. This information was used to ask 'for what proportion of the infants was the observed intake below the estimated requirement?'- a measure of the expected prevalence of inadequacy. Given the starting premise, the expected answer would be 0 - all of the infants should have had intakes above the modelled requirements if the components of the model were correct. It became clear from these modelling exercises that anything affecting estimated mean requirement (e.g. assumed efficiency of utilization) had distinct effects on the estimated prevalence of inadequacy. Changes in the estimates of variability had much less impact.

Table 9 Sample of the testing of factorial models of nitrogen requirements at 3 months of age. Tested with data from 104 exclusively breast fed California^a infants

^a Heinig et al (1993); Dewey et al (1994).
^b Need for growth (mg N kg/d) = weight gain (g/kg/d) × 0.106/6.25 × 1000.
^c Total requirement as utilized N = average basal need + average growth need. Standard deviation estimated as the square root of the weighted
sum of variances and CV calculated as s.d./total requirement (see footnote to Table 7). These estimates were identical for all infants in each model.
^d Milk N requirement = Utilized N requirement/Efficiency of utilization.

The factorial components used in the final models, and the associated predicted prevalences of inadequacy, are shown in Table 9. In these models, a basal requirement of 60 or 66 mg N/kg/d was chosen. The lower value (60) corresponds to the lower end of the range calculated from the data of Fomon et al (1965), as explained in section 2.2.2. The higher value (66) is the estimate that would be consistent with the maintenance estimate used in Tables 6 and 7 (90 mg N/kg/d), assuming a slope of 0.73. Two estimates for the efficiency of utilization of dietary nitrogen (for both maintenance and growth) were used: 73% or 80%. The CV for the maintenance requirement was assumed to be 12.5% (as assumed in the 1985 report). The proportion of weight gain attributable to protein accretion was estimated to be 10.6%, based on the data in Table 4 (the average proportion for 2-3 and 3-4 months). The CV for rate of weight gain (24.7%) was based on the actual data (over a 2-month interval).

These estimates of requirements were compared with observed intakes of total nitrogen. Although human milk contains about 25% non-protein nitrogen, not all of which is utilizable (see section 2.1.3), the evidence to date indicates that the percentage utilization of total nitrogen from cow's milk-based formulas and human milk is similar (75-79% in preterm infants fed human milk or formula: Stack et al, 1989; 72-74% in term infants fed human milk: Fomon and May, 1958). The resulting estimated prevalences of inadequate intakes were 1.6-5.1 % when basal requirement was set at 60 mg N/kg/d (Options A and B) and 3.0-8.1% when basal requirement was set at 66 mg N/kg/d (Options C and D).

An important feature to note from these analyses is that the factorial model closest to that used in Tables 6 and 7 (Option C) leads to a predicted prevalence of 'inadequacy' (8.1%) that is higher than expected. Unless one rejects the starting assumption that virtually all breastfed infants meet their protein requirements, this would suggest that one or more of the components in the factorial model is still being overestimated. It is not likely to be the estimates of variability, as they have minimal impact on the prevalence estimate. The most likely components are the estimation of the basal requirement and the efficiency of utilization of dietary nitrogen. Option D illustrates that when the latter is assumed to be 80%, the prevalence of 'inadequacy' is 3.0%, and Option B illustrates that when both of these components are modified, the prevalence of 'inadequacy' is only 1.6%.

Two other potential explanations for the non-zero prevalence of 'inadequacy' were considered but set aside as improbable. The first related to an explicit assumption of the probability approach - the assumption that protein intakes and protein requirements are independent of one another except insofar as both relate to body size. Protein intake is related to energy intake, and it is reasonable to assume that this relates to body size. However, because protein requirements per unit body size were being modelled, this association would not be relevant. Beaton and Chery (1988) considered the possible effect of a correlation mediated through energy requirement (and hence protein intake) for the growth component of protein requirement. This was again tested in the present models and found to have negligible impact.

The second potential explanation relates to within subject (day-to-day) variation in protein intake. The individual intake data used in these analyses, although based on 3-4 days of measurement (average = 3.7 days), would still have a residual 'random' error in relation to 'true' usual intakes. This would act to inflate the distribution of intakes, thereby increasing the estimated prevalence of both low and high intakes. To examine this effect, data for the 71 infants in the DARLING sample were further analyzed. The results indicated that the within-subject component of variance had a CV of about 11.1% and the between-subject variation in 'true intake' had a CV of about 15.1% (actual variance ratio 0.54); this is a substantially lower day-to-day variation than is customarily seen in intakes of older children and adults. With three day means, the remaining random error component would be about 6% while the between subject variation would continue at about 15%, yielding a variance ratio of about 0.18. These estimates were considered in theoretical distribution models, but the impact was too small to explain the non-zero prevalences of 'inadequacy'.

In summary, the inferences drawn from the above analyses seem robust and provide strong support for a recommendation that the estimates of protein requirements for young infants be substantially lower than those published in the 1985 report. The requirement estimate in Table 6 for infants 3-4 months of age is 170 mg N/kg/d, which is similar to Option C in Table 9 but somewhat higher than Options A, B and D, which yielded more reasonable prevalences of 'inadequacy'. This discrepancy arises because Table 6 was designed to present estimates that would be appropriate for a range of ages and feeding modes, and therefore was based on relatively conservative estimates for maintenance requirement and for efficiency of utilization of dietary nitrogen.

2.3. Using a direct experimental approach

A direct experimental approach to estimating protein needs during infancy has been used in several studies of formula-fed infants. Many of these studies have utilized plasma amino acid profiles to help understand protein metabolism and requirements. Postprandial concentrations of amino acids in plasma are reflective of protein intake and synthesis as well as turnover, amino acid metabolism (catabolism, synthesis and excretion) and tissue utilization. Therefore, amino acid concentrations can be viewed as a rough indicator of the balance between intake and utilization, and have been used for assessing deficient or excessive intakes (Heird, 1989). An underlying assumption is that the protein and amino acid intake and the plasma aminogram of exclusively breast-fed infants is optimal. Serum urea nitrogen can be used as an additional parameter in that excessive protein intake will result in elevated serum urea nitrogen levels; however, it should be considered that human milk as well as some infant formulas are high in urea (Donovan and Lönnerdal, 1989b) and that absorbed urea may contribute to the serum urea nitrogen level.

The case for using plasma amino acid and serum urea nitrogen levels to assess protein status is based in part on the potential risks associated with profiles that differ from those of the breastfed infant. For example, it has been cautioned that the high metabolic activity of the liver and the kidney necessary to catabolize and excrete high levels of plasma amino acids and urea nitrogen could cause undue stress to immature organs and potentially lead to long-term consequences (Herin and Zetterström, 1987; Axelsson et al, 1988). Evidence for such consequences later in life is lacking for infants fed high protein formulas, although some animal studies of high protein diets have resulted in tissue damage. However, most of the animal studies used protein levels considerably higher (relative to requirements) than those given to formula-fed infants. It must be emphasized that studies looking at long-term consequences, such as kidney function in adulthood as related to protein intake during early life, are usually not available, making claims of the safety of high protein formulas relatively weak. In terms of specific amino acids, high concentrations of branched-chain amino acids have raised some concern in that the higher than normal transport of these across the blood-brain barrier may interfere with the transport of other essential amino acids (Ginsburg et al, 1985; Scott et al, 1985). The high levels of threonine observed in infants fed high protein whey-predominant formula has also evoked some concern (Rigo and Senterre, 1980; Grugan et al, 1988), although it has been argued that no negative consequences have been associated with hyperthreoninemia. Amino acids that tend to be lower in plasma of formula-fed than breastfed infants include tryptophan and phenylalanine (Janas et al, 1985, 1987; Lonnerdal and Chen, 1990; Hanning et al, 1992). In general, low tryptophan levels are observed in infants fed low-protein whey-predominant (60: 40) formula, while lower phenylalanine levels are observed in infants fed casein-predominant formula. In both cases, hypothetical scenarios can be made for impaired neurotransmitter synthesis, although evidence is lacking. Differences in sleep patterns have been reported between breast-fed and formula-fed infants (Butte et al, 1992); plasma tryptophan levels have been implicated as one possible explanation for this difference (Fernstrom and Wurtman, 1971). However, difficulties arise when judging what constitutes a 'normal' sleeping pattern.

Most experimental studies on the protein requirement of infants have been on formula-fed infants during the first three months of life. Several of these studies have shown that infants fed formula with a true protein level of 15 g/1 have significantly higher plasma levels of several amino acids and urea nitrogen than those found in breast-fed infants (Järvenpää et al, 1982a,b; Janas et al, 1985, 1987; Räihä et al, 1986a,b; Hanning et al, 1992). A true protein level of 13 g/1 has been shown to result in a plasma amino acid profile similar to that of breast-fed infants (Lönnerdal and Chen, 1990), although a few minor differences were observed. One study by Räihä et al (1986a,b) indicated that feeding a formula protein level of 13 g/1 resulted in lower serum urea nitrogen and plasma concentrations of some essential amino acids than found in breast-fed infants. However, the formula used had an unusually high level of urea and the true protein level was actually 11 g/1 (Lonnerdal and Chen, 1990). The low levels of some amino acids were only observed during the first few weeks of life and it is difficult to evaluate if they were truly inadequate even if they were significantly lower than in breast-fed infants. It should be noted that as bovine casein has a different amino acid composition compared to that of human casein, and bovine whey is quite different from human whey (Picone et al, 1989), it is virtually impossible to create an amino acid profile identical to that found in human milk. Even if the amino acid composition of the formula could be made very similar to that of human milk, digestibility and absorption of amino acids and peptides from such a formula would be quite different from that of breast milk, thus resulting in different plasma amino acid profiles.

In a series of studies. Räihä Axelsson and co-workers have evaluated intake and growth during the period from 4 to 6 months, when the study infants were gradually increasing their intake of solid foods (Axelsson et al, 1988; Räihä and Axelsson, 1991). Breastfed infants were compared with those fed experimental formulas varying in protein content. Weight and length gain were significantly higher in infants fed high-protein formulas (ranging from 18 to 27 g/1) than in those who were breastfed or fed a low-protein formula (13 g/1). Nitrogen excretion was much higher in the infants fed the high protein formulas, which the authors interpreted as indicating excessive nitrogen intake. Their conclusion was that a lower-protein formula results in anthropometric and biochemical indices more similar to those of breastfed infants.

A recently completed study by Fomon et al (1995) provides additional information that is useful for understanding protein requirements. By design, one group of male infants (Experimental Group) was provided graded intakes of protein that were intended to match, as closely as possible, the protein requirements for normal male infants estimated by Fomon (1991). Another group of male infants (Control Group) was fed a commercially available formula with a protein concentration of 15 g/1. The source of protein in all formulas was whey-predominant bovine milk proteins. Observed protein intakes are indicated in Table 10 for each 28-d period of the study. Protein intake was calculated assuming that 14% of total nitrogen in the formulas was non-protein nitrogen (Donovan and Lonnerdal, 1989b) and that all of the non-urea portion and 17% of the urea portion (assumed to be 5.6% of total nitrogen) were fully utilizable; this calculation is equivalent to 95.4% of total nitrogen × 6.25. Gains in weight and length by the Experimental Group were slightly less than those by the Control Group, but the differences were not statistically significant. However, in comparison to a larger reference group (n = 380) of formula-fed male infants (Nelson et al, 1989), gain in weight of the Experimental Group was significantly lower for the age intervals 56-112 days, and 8112 days, and gain in length was significantly lower for the intervals 5-56 and 56-112 days. Growth of the Experimental Group was similar to that of male breastfed infants (Nelson et al, 1989). Concentrations of serum albumin of the Experimental Group were normal, but concentrations of serum urea were extremely low. In each age interval, serum urea concentrations were significantly lower than in the Control Group and in either the formula-fed or the breastfed reference group. It was concluded that the lower protein intake had a mild but detectable effect on growth of infants in the Experimental Group. Protein intakes by infants in the Control Group were clearly not limiting for growth.

Table 10 Protein and energy intakes of infants in the experimental (n = 15) and control groups (n = 13); values are means (s.d.) (Fomon et al, 1995)

^a Adapted from Fomon et al (1995), using 6.25 to convert from nitrogen to protein, and adjusting for NPN by multiplying by 0.954 (see text).

The results of Fomon et al suggest that the amount of protein consumed by the Experimental Group was somewhat less than the group mean intake needed to ensure adequacy among essentially all formula-fed infants (theoretically, the latter is somewhat greater than the 'safe' level of intake defined as mean requirement + 2 s.d. of the requirement; see section 1). In comparison with the average adjusted protein intake of breastfed infants (Table 1), the amounts consumed by the Experimental Group were lower from 0 to 1 month, but generally similar thereafter.

Protein requirements of formula-fed infants may be greater than those of breastfed infants if there are differences in the efficiency of utilization of formula vs human milk. Even after accounting for the non-protein nitrogen content of the milk consumed, there may be differences in utilization of the protein and non-protein fractions. As pointed out above, it is virtually impossible to create a formula that matches the digestibility and amino acid composition of human milk. Waterlow et al (1960) compared nitrogen absorption and retention in infants fed either expressed breast milk or a cow's milk formula with a similar concentration of protein (1.1%). Nitrogen absorption and weight gain did not differ significantly between diet periods, but nitrogen retention was 17% higher on human milk than on the formula (P = 0.01). Further studies that include measurements of the composition of weight gain and use infant formulas developed more recently are needed to resolve this issue.

Table 11 Comparison the amino acid composition of whole protein in immature mammals of different species

Values are mg amino acid/g total amino acid excluding cysteine and tryptophan for which few reliable data are available.
Notes:
1. Species are arranged in order of the rate constant of postnatal growth.
2. For source references see Davis et al (1993).
3. Note that as the postnatal growth rate increases (i.e. from the human infant to the rat), the glycine and proline contents decrease, and hence the contribution of collagen to protein mass is higher in the human infant than in the rat pup.

All of the experimental studies described above were conducted with formula-fed infants, because manipulating the protein intake of breastfed infants is of course much more difficult.. However, data are now available from an intervention study in Honduras in which infants were randomly assigned to be exclusively breastfed for the first 6 months (n = 50), or to receive pre-prepared solid foods (with egg yolk as the main source of protein) in addition to breast milk beginning at 4 months (n = 91) (Cohen et al, 1994; Dewey et al, in press). Neither weight gain nor length gain from 4 to 6 months differed between groups despite a 20% higher protein intake in the latter group. The 20 infants with the highest protein intakes in that group were matched to 20 exclusively breastfed infants on the basis of energy intake; protein intake was 33% higher in the solid foods subgroup, but growth rate did not differ between groups. Similarly, the 20 infants with the lowest protein intakes in the exclusively breastfed group were matched (by energy intake) to 20 infants given solid foods; protein intake was very low in the former compared to the latter (0.81 ± 0.13 vs 1.04 ± 0.20 g/kg/d; P < 0.001), yet there was still no difference in growth. Infant morbidity was relatively low and did not influence the results. These analyses indicate that protein intake is not likely to be a limiting factor with regard to growth of breastfed infants from 4 to 6 months of age.

2.4. Using an 'operational' approach

Waterlow (1990) has suggested that the most satisfactory operational approach to assessing protein requirements of infants is by using protein-energy ratios (P: E). Although protein-energy ratios have been used previously (FAO/WHO/UNU, 1985), they required estimating the ratio of the safe level for protein to the mean requirement for energy at each age, with all of the difficulties and assumptions inherent in those calculations (Town et al, 1992). Waterlow proposes instead that one starts with the assumption that breast milk provides sufficient protein for virtually all infants up to about 4 months of age, provided that they consume enough to satisfy their energy needs. The protein-energy ratio of human milk is approximately 8-8.5% at 3-4 months post partum. This estimate is based on (a) a 'protein' concentration of 9.6-100.0 g/l, which includes 46% utilization of the NPN fraction, (b) a gross energy density of 670 kcal/l and (c) a conversion factor of 5.65 kcal per g protein. The question then becomes, what is the appropriate P: E ratio for older infants? To calculate this, one can determine the ratio of requirements at older ages to those of the infant at 3-4 months. For example, the ratio of protein requirements (per kg) at 12 months vs 4 months is approximately 0.7-0.8. Waterlow argues that the 'safe' P: E ratio of the diet of the 12 month infant would thus be (8-8.5) × (0.7-0.8), or about 6%, presuming that the quality of the food is the same as that of breast milk. Adjustments can then be made for the digestibility and protein quality of the diet.

While this approach is attractive in terms of its theoretical simplicity and ease of application, it bypasses the sticky issue of defining minimum protein requirements. This is both a virtue and a limitation. Like the approach based on observed intakes of breastfed infants, it can be used to set 'safe' levels of intake, but the mean requirement level would be at some (unknown) level below the 'safe" level.

2.5. Amino acid requirements

Amino acid needs can be visualized as stemming from two independent pathways of utilization: growth and maintenance. The accretion of some amino acids (glutamine) and their metabolites (e.g. creatine, heme, glutathione and taurine) within the lean body mass should be considered as part of growth, but in the context of amino acid needs, protein deposition is the dominating influence. The basal need for a given amino acid in support of protein deposition is simply the product of the rate of protein deposition and the contribution of the amino acid in question to the proteins being deposited.

Table 11 shows data on the amino acid composition of the body protein of various mammalian 'infants'. With regard to the essential (indispensable) amino acids, the composition of body protein is virtually identical. Thus the relative needs of each essential amino acid (ma amino acid/g total amino acid) for growth will be the same across species. Table 12 compares current definitions (by nitrogen balance) of the amino acid needs for growth of human infants, pigs and rats. To control for differences in protein 'requirements' these are expressed in terms of total essential amino acids (i.e. mg amino acid/mg total essential amino acids). When adjusted on this basis, the patterns are very similar across species. Thus it seems reasonable to conclude that the obligatory or minimum needs for essential amino acids to support growth (protein deposition) are a close function of the growth rate of the individual. It follows that defining amino acid requirements will depend on what is regarded as an appropriate rate of protein deposition (plus the needs for maintenance).

Continued

Contents - Previous - Next