3

Quality Factors for Infant Formula: Protein Quality

EVALUATION OF THE PROTEIN EFFICIENCY RATIO

The committee was asked to examine the state of the science on methodologies to assess the biological quality of infant formula. The U.S. Food and Drug Administration (FDA) further included information in an open session with the committee that the committee should consider benefits and limitations of the protein efficiency ratio (PER) design as described in 21 CFR § 106.96(f) (see Chapter 2, Box 2-1). This chapter describes the regulatory historical context for using PER to assess protein quality, evidence on PER identified from the scoping review, and challenges with the PER methodology.

Historical Context

The Infant Formula Act of 1980 (P.L. 96-359) amended the Food, Drug, and Cosmetic Act (FD&C Act) to include section 412 (21 U.S.C. § 350a), which provides requirements relating to infant formula nutrient content, quantity, and quality control, among other requirements for new infant formula submission to FDA. The law was amended in 1986 to require FDA to establish quality factors for infant formula (U.S.C § 350a). During its rulemaking to implement the law, FDA evolved its regulatory definition of quality factors (see Table 3-1). Table 3-2 shows a timeline for the regulatory actions related to quality factors.

TABLE 3-1 Regulatory Definition of Quality Factors

SOURCES: 61 FR 36154, September 23, 1996; 79 FR 7934, February 10, 2014; 79 FR 33037, June 10, 2014.

Consistent across definitions was that the quality factors

• applied to the infant formulas as prepared for market,
• were to demonstrate safety of the infant formula,
• were to demonstrate the bioavailability of its nutrients when fed as a sole source of nutrition, and
• were focused on healthy infant growth when fed as a sole source of nutrition.

In 1996, FDA’s proposed regulations pertaining to quality factors, one of which was the “biologic quality of protein” (61 FR 36154, July 9, 1996). Citing Senator Howard Metzenbaum’s 1986 testimony on the Infant Formula Act, FDA wrote that “although the testimony to the Senate does not specify the identity of the nutrient for which there was a basis for a quality factor, the quality factor was the PER used for assessing protein quality” (61 FR 36154 at 36180).

In the interim final rule published in 2014, FDA clarified that the PER was not the quality factor. “The quality factor is the biological quality of the protein, and the PER is a method used to assure such quality” (21 CFR § 106.96(f)). The final regulation reads, “An infant formula shall meet the quality factor of sufficient biological quality of protein” (21 CFR § 106.96(e)).

TABLE 3-2 Timeline of Legislative and Regulatory Actions on Quality Factors

NOTES: CFR = Code of Federal Regulations; FR = Federal Register; IFA = Infant Formula Act; IFR = Interim Final Rule; P.L. = Public Law; QF = quality factors; STAT = statute. In 2002 and 2003, FDA also held Food Advisory Committee meetings, 67 FR 12571, March 19, 2002; 67 FR 63933, October 16, 2002; and 68 FR 8299, February 20, 2003, and reopened the comment period to solicit comments on quality factors among other things, 66 FR 20589, April 28, 2003.

SOURCES: P.L. 96-359—September 26, 1980; P.L. 100 STAT. 3207-38; October 27, 1986; 61 FR 36154, September 23, 1996; 79 FR 7934, February 10, 2014; 79 FR 33037, June 10, 2014.

The manner by which the formula is shown to meet the quality factor is specified in 21 CFR § 106.96(f):

The final rule also included a provision that an alternate method to the PER could be used if:

Defining the quality factor as the “biological” quality of the protein added emphasis that it could not be established via chemical analysis only. As FDA explained in the final rule:

FDA selected PER as the method to be used to demonstrate the biological quality of the protein as a sufficient, if indirect (using weight gain to indicate utilization of protein) measure of bioavailability of the protein in the formula as processed for market. “Bioavailability” is defined by FDA as “the degree to which a nutrient is absorbed or otherwise becomes available to the body” (79 FR 7934 at 7945). FDA defended the use of PER, stating it “is not aware of any other available method to assess protein bioavailability” (79 FR 7934 at 8022). PER takes into account the amino acid (AA) composition of the protein in the test infant formula, digestibility of the protein, and absorption and utilization of the limiting AA for growth of the weanling rat. FDA noted that certain food products can influence the results of a PER assay (Harris et al., 1988; Steinke, 1977), and since infant formula has high fat content, low protein content, and high lactose concentration, PER needed appropriate modifications, which were specified in the regulations (79 FR 7934). As noted, in 2023, FDA proposed Draft Guidance for Industry on the conduct of the PER bioassay; as of November 2024, this draft guidance had not been finalized (FDA, 2023). Growth monitoring studies (GMS) provide evidence of utilization of AAs

___________________

¹ The FDA proposal that set out nutrient specifications for infant formula (49 FR 14402, April 11, 1984) included a specification that the biological quality of protein be not less than 70 percent casein; the reference to casein implies PER, although it was not named in the rulemaking (61 FR 36154 at 36180).

from dietary protein, but animal models and other control diets than those used in the PER could be used for this purpose.

Scoping Review Results: Protein Efficiency Ratio

The scoping review identified only three articles published between 2000 and 2024 that measured PER in infant formulas; none of these articles provided evidence that PER reflects the protein quality of the formulas fed to rats (Hoskin, 2023; Sarwar, 2001) or human infants (Panjkota Krbavčić et al., 2009). Supplementation of soy-based infant formula with tryptophan had no effect on growth, as determined by no difference in PER, despite increased blood and brain tryptophan and the metabolites, serotonin and 5-hydroxyindoleacetic acid (5HIAA), compared with rats fed unsupplemented diets (Sarwar, 2001). The unsupplemented soy formulas had substantially lower tryptophan concentrations than human milk, suggesting that the rat has a lower requirement for tryptophan for growth than to support non-growth metabolic functions (Sarwar, 2001). This finding raises the possibility that human milk amino acids concentration may be a better guide than using PER to determine protein quality for infants. Hoskin (2023) found that PER casein control diets formulated with the mineral mixes used in infant formula manufacturing provided insufficient sulfur in rats; rats fed control diets with the same mineral mix as the test infant formula had compromised growth. These results suggest that sulfur might be added to the list of nutritional variables that must be matched between test and control diets in PER studies is needed.

Challenges with Protein Efficiency Ratio

Challenges with PER include its lack of global recognition (only used in the United States and Canada) and commercial availability, unclear predictive ability, and criticism from international expert scientific bodies, such as the Food and Agriculture Organization (FAO) and World Health Organization (WHO), which have recommended against it as a method to determine protein quality (FAO, 1991, 2013; Wallingford, 2023). That same conclusion was reached by the Life Sciences Research Office (LSRO) panel contracted by FDA, which stated: “The Expert Panel recommended that the assessment of protein quality be based on an AA score with human milk as the reference protein, thus eliminating the PER as an index of protein quality for infant formulas” (Raiten et al., 1998). Regulatory bodies outside North America no longer use PER. In 2007, Codex Alimentarius (Codex) set the standard for protein quality to be the human milk amino acid pattern (HMAA) (FAO/WHO, 2023) which has been adopted by regulatory bodies in Europe (European Commission, 2006) and Australia/-

New Zealand (FSANZ, 2024). Furthermore, while PER is the most common method for assessing protein quality in infant formula within North America, its specifics differ in the United States compared to Canada (which performs the assay on a food after fat extraction).

A key attribute of the PER assay is that it measures the biological quality of the protein in the product as processed. Infant formula is manufactured primarily through a “wet-blend” process, where ingredients are mixed in an aqueous environment. This allows for chemical reactions among nutrients. Similarly, heat used in the processing may encourage chemical reactions. Some reaction products could render nutrients unavailable to the body. An example is the Maillard reaction product of the epsilon amino group of lysine and lactose, producing lactosyl lysine, which is not absorbed (Aalaei et al., 2019a,b; Erbersdobler et al., 1991; Finot et al., 1981; Henle et al., 2000; Langhendries et al., 1992; Pischetsrieder and Henle, 2012; Rutherfurd and Moughan, 2007). Assays that measure only free lysine but not its reaction products in feces would overestimate absorbed lysine. Other “matrix effects” involving indispensable² (essential) AAs are possible and could potentially be detected in a growth assay (See Box 3-1).

However, commercial infant formulas produced in a variety of matrices and processing methods support normal physical growth, suggesting that matrix effects are physiologically important. The potential effects of processing would only become detectable if, because of creation of unavailable AA products, the free AA became limiting. The formation of lactosyl lysine,

___________________

² The term indispensable is equivalent to essential amino acid.

for example, affects 5 percent or less of the lysine in milk (Rutherfurd and Moughan, 2007). The sensitivity of any assay for bioavailable AAs must be considered relative to the extent of any processing effects. It is unlikely that PER could detect minor differences in available AAs (Wallingford, 2023).

Conclusion 1: The protein efficiency ratio (PER) is not the preferred method for assessing protein quality of infant formula.

Recommendation 1: The Food and Drug Administration should not use the protein efficiency ratio (PER) as the method for establishing the biological quality of protein of new infant formulas and should reconsider the need for the existing draft guidance on PER.

ALTERNATIVE METHODS FOR DEMONSTRATING PROTEIN QUALITY OF INFANT FORMULAS

This section of the report describes several methodologies for demonstrating protein quality of infant formula that are used by various authoritative bodies around the world, including the European Food Safety Authority (EFSA), FAO, WHO, Food Standards Australia New Zealand (FSANZ), and Codex Alimentarius (Codex).

Amino Acid Composition Compared to Human Milk

The committee was asked to consider quality factor study designs that may be in common use in other global regions. Therefore, it looked to these regions’ regulatory bodies, such as EFSA and Codex, which use the AA composition compared to human milk. This section of the report describes the historical and regulatory context for these regulatory bodies.

Background

The most recent EFSA scientific opinion on the topic of compositional requirements for infant formulas was issued in 2014, when the EFSA Panel on Dietetic Products, Nutrition and Allergies (NDA) reviewed the current European Commission’s directives (EFSA NDA Panel, 2014). The updated report was largely consistent with the original directive in 2006 (European Commission, 2006), with general rules on the composition of infant formula being recommended in 2009 (European Commission, 2009).

A few methodological considerations for determining infant formula quality were laid out in EFSA’s 2014 scientific opinion. These include that all formulas

In addition, the rules state that the composition of human milk from “healthy, well-nourished mothers” can provide guidance for the composition of infant formulas, although it was acknowledged that “compositional similarity to human milk is not the only appropriate determinant or indicator of safety and nutritional suitability of such formulae” (EFSA NDA Panel, 2014). For a food that is the sole source of energy and nutrients, such as infant formulas, EFSA’s 2014 scientific opinion states that infant formula compositional requirements can be “based on energy and nutrient needs of the targeted population” and “set by specifying minimum and maximum content of nutrients, and the minimum content of a nutrient” (EFSA NDA Panel, 2014). It further states that the required nutrient content of infant formula can be “derived from the intake levels considered adequate for the majority of infants in the first half of the first year of life”; this intake was set as 500 kcal per day (EFSA NDA Panel, 2013). The value of 500 kcal per day was based on “the Average Requirements for energy of boys and girls aged 3 to less than 4 months (479 kcal/day) and rounding up” (EFSA NDA Panel, 2014).

Minimum and Maximum Energy and Protein Composition

The EFSA requirements for energy and protein composition of infant formula were determined based on human milk composition, using the above-described methodology.

Energy

The energy density³ of human milk has been shown by several studies to be an average of 65kcal/100ml (European Commission, 2006). While energy content of human milk can vary widely within a feed and change over time, infant formula, in contrast, has a stable composition. To ensure that the growth and development of formula-fed infants are similar to those of infants who are exclusively fed human milk, the minimum energy content for infant formula in the European Union was set at 60kcal/100ml, with a maximum of 70kcal/100ml (European Commission, 2006) (see Box 3-2). EFSA recommended that infant formula have an energy content that is closer to the lower bounds, to limit excess weight gain.

___________________

³ Gross energy is the reporting standard for human studies.

Protein

Protein concentrations of human milk change rapidly during the first few days of life, but mature human milk (>14 days) was reported to be within a smaller range, 0.8–2.1g/100ml (average 1.3g/100mL) or 1.2–3.2g/100kcal (average 2.0g/100kcal) (European Commission, 2006). The compositional requirements for minimum and maximum protein in infant formula were set for the permitted sources of protein—cow’s milk, goat’s milk, isolated soy, and hydrolysates—based on the range of protein concentration in human milk (see Table 3-3).

Protein Quality

Based on the assumption that intake of protein from a healthy, well-nourished mother satisfies the AA requirements for an infant exclusively fed human milk for the first 6 months of life, EFSA considered HMAA to be the best reference for a substitute product for human milk in infants. The evidence considered for AA composition was based on the 2003 European Commission report on the revision of requirements for infant formula by the Scientific Committee on Food (SCF, 2003).

TABLE 3-3 EFSA Recommended Minimum and Maximum Protein Composition for Infant Formula

NOTES: EFSA = European Food Safety Authority; g = gram; kcal = kilocalories.

^a EFSA NDA Panel (2014) recommended that this value for soy protein be revised to 2.25g/100kcal.

^b EFSA NDA Panel (2014) recommended that for protein hydrolysates, a minimum protein content cannot be proposed and should instead be clinically evaluated.

^c EFSA NDA Panel (2014) concluded that the 3g/100kcal maximum does not have a physiological basis and should be reduced to 2.5g/100kcal for infant formula based on cow’s milk and goat’s milk protein and 2.8g/100kcal for isolated soy protein and protein hydrolysates.

SOURCE: European Commission, 2006.

In addition, as infant formulas are considered human milk substitutes, EFSA considered that infant formula “should provide indispensable and conditionally indispensable amino acids in amounts on an energy basis at least equal to the reference protein” (i.e., human milk), “irrespective of the protein source” (EFSA NDA Panel, 2014). No correction for ileal digestibility of AAs is proposed by EFSA or Codex.

As a comparison, Table 3-4 provides the AA pattern recommendations for the indispensable AAs and the conditionally indispensable AAs, cysteine and tyrosine, from the 2007 WHO/FAO/UN University (UNU) and FAO 2013 reports.

Regulatory Usage of Amino Acid Composition of Infant Formula Compared with Human Milk

As discussed previously, EFSA’s 2014 report concluded that the composition of human milk from a healthy, well-nourished mother can provide guidance for the composition of infant formula. A 2018 analysis identified the change of pattern of AAs over the first 6 months of lactation (van Sadelhoff et al., 2018). Table 3-5 shows the protein content (g/L), total nitrogen, N, (g/L) and concentration of nine AAs during months 1–6. The composition of individual AAs on a mg/g N basis are relatively consistent over the first 6 months. Davis et al. (1994) and Zhang et al. (2013) presented similar findings of a consistent amino acid composition across stages of lactation. These findings confirm the validity and stability of the approach to use as reference for an average HMAA profile during the first 6 months of lactation. In its review of evidence, the committee considered total AAs of human milk, which include the sum of those derived from the hydrolyzed protein fractions and the free AAs present in the non-protein fraction of human milk. Thus, matching the AA pattern to human milk AA composition provides AAs being utilized for protein synthesis, growth, and non-protein pathways. As necessary, protein quality in infant formula may be improved by addition of free AAs (European Commission, 2016; FAO/WHO, 2023; FSANZ, 2024).

Strengths in using the HMAA include that strong data on compositional amino acid pattern of human milk exists and a published amino acid scoring pattern is available in Codex, EFSA, the European Union, and FSANZ which would contribute to international regulatory harmonization.

Analysis of human milk composition by Moughan et al. (2024) highlighted the importance of standardizing human milk AA analysis methodology. The current standard is to use fixed time of acid hydrolysis (20–24h) to release all AAs into free AAs before quantification. However, some AAs require longer hydrolysis time to be released from human milk

TABLE 3-4 Proposed Minimum Requirement for Indispensable and Conditionally Indispensable Amino Acids for Infant Formula to Align with Human Milk Reference

NOTES: Methionine and cysteine are provided as sulfur amino acids. Phenylalanine and tyrosine are provided as total aromatic amino acids. EFSA = European Food Safety Authority; FAO = Food and Agriculture Organization; g = gram; mg = milligram; kcal = kilocalories; N = nitrogen. The nitrogen-to-protein conversion factor of 6.25 was used to calculate the amounts of an amino acid per g of protein.

SOURCES: EFSA NDA Panel, 2014; FAO/WHO, 2023; WHO/FAO/UNU, 2007.

TABLE 3-5 Pattern of Change in Indispensable Amino Acids Over the First 6 Months of Lactation

NOTES: AA = amino acid; g = gram; L = liter; mg = milligram; mo = months; N = nitrogen. The values are the result of calculations using the data for protein (g/L) and AAs (µmole/L). The AA pattern as given in mg/g nitrogen is relatively constant throughout lactation. Total protein includes non-protein N, whereby N multiplied by 6.25 equals total protein. Protein (g/L) was converted to nitrogen (g/L) by dividing it by 6.25. The AAs were converted to mg/L by multiplying with their respective molecular weights and then converted into mg/g nitrogen by dividing this result by the respective (calculated) nitrogen concentration (g/L).

SOURCE: Van Sadelhoff et al., 2018.

proteins, while others oxidize/are destroyed with such longer hydrolysis (Darragh and Moughan, 2005). An alternate method is proposed that quantitates individual AAs based on different hydrolysis times; the resulting composition pattern is different (Hodgkinson et al., 2023; Moughan et al., 2024). Furthermore, based on newer piglet studies (Charton et al., 2023) and the use of ileal AA digestibility values, a correction factor is being recommended to the final HMAA (Moughan et al., 2024). This pattern results in 16 percent higher concentrations for leucine, lysine, and threonine and 30 percent higher concentrations for histidine and tryptophan. FAO/WHO should assess this new proposed pattern. Challenges with this approach include variability in measures related to ethnicity, diet, and circadian rhythms. Since the values presented by Moughan et al. (2024) have not been confirmed in the published literature, they were not included in Table 3-4.

Conclusion 2: Based on its assessment of the evidence related to the lack of use of PER to assess protein quality and review of relevant regulations in countries outside of the United States, the preferred method of assessment of protein quality in infant formula is to match the indispensable and conditionally indispensable amino acid composition of human milk. An advantage of this approach is that a published amino acid scoring pattern is available in Codex, European Food Safety Authority, the Food Standards Australia New Zealand, and the European Union, which would contribute to international regulatory harmonization.

Recommendation 2: The Food and Drug Administration should adopt the human milk amino acid pattern as the reference pattern to assess the protein quality of infant formula.

Alternative Approaches to Measuring Protein Quality

The following section describes alternative approaches to measuring protein quality that, as of 2024, are not used by a regulatory body but could be considered. For each approach, potential challenges are also described.

Assessment of Digestibility and Availability

FDA’s rulemaking and Draft Guidance for Industry on assessment of protein quality emphasizes the bioavailability of AAs from protein as distinct from the quality of protein:

As noted, FDA defined “bioavailability” as “the degree to which a nutrient is absorbed or otherwise becomes available to the body” (79 FR 7934 at 7945). A commonly used definition of protein quality is the age-appropriate AA score of a candidate protein relative to a reference, corrected for digestibility. Digestibility was historically measured by the difference of dietary protein and fecal protein, the latter representing unabsorbed dietary protein, and an estimated amount of endogenous protein secreted into the gastrointestinal tract. The digestible protein represents what was absorbed or otherwise available to the body. Digestibility and bioavailability are often used interchangeably (Paoletti et al., 2024; Tome, 2024) because, after AA composition, digestibility is another impactful variable that can determine protein quality (Sarwar Gilani et al., 2012). Sophisticated in vivo methods directly measure the AAs that become bioaccessible during digestion (i.e., dialyzable). Such methods presume that bioaccessible AAs are utilizable for metabolism and body protein synthesis.

Imbalances among absorbed AAs or interaction between absorbed AAs and other absorbed molecules could affect AA utilization (e.g., growth). However, the required nutritional composition of infant formula, the AA pattern of formula protein sources, and history of use of formulas support the conclusion that absorbed AAs are utilized.

One alternative approach to measurement of protein quality is to assess the digestibility and availability of protein using animal bioassays, which are then applied to a scoring method that takes into account the protein and AA requirement of the relevant life stage. Swine and rodent models have been used to estimate protein digestibility in humans, including infants.

Fecal Digestibility

The Protein Digestibility-Corrected Amino Acid Score (PDCAAS) methodology evaluates the quality of protein based on AA requirements and digestibility (FAO/WHO, 1991; see Box 3-3). This method compares the indispensable AA composition of the test protein to the reference indispensable AA pattern for the target population and corrects this chemical AA score for the digestibility of the protein measured at the end of the gastrointestinal tract (in the feces). The indispensable AA with the lowest score is considered the first limiting AA and determines the PDCAAS value for the test protein. Proteins with a score greater than 1.0 are truncated to 1.0. Although PDCAAS has been used extensively to evaluate the protein quality of food protein ingredients and food

products, the scoping review did not identify the use of any animal models to estimate PDCAAS specifically in infant formulas.

Limitations of the PDCAAS method have been identified (Mansilla et al., 2020):

• Truncation of scores to 1.0 leads to the underestimation of protein with high protein digestibility (Wallingford, 2023);
• A single value of protein digestibility is used, although the digestibility of individual AAs in the protein differs (Wallingford, 2023);
• It is assumed that AAs are digested and absorbed throughout the gastrointestinal tract, although significant absorption does not occur in the large intestine; and
• Dietary AAs in the digesta that enter the large intestine can be modified by colonic microbiota.

Given these limitations, estimating digestibility from fecal AAs can lead to overestimating the bioavailability of AAs in proteins that are poorly digested in the small intestine. The weaned, growing rat has been considered as an acceptable model for determining protein digestibility using PDCAAS (Deglaire and Moughan, 2012). Its limitations include coprophagy, physiological stage/age in comparison to the human infant, and species differences in AA requirements.

In 1989, an FAO expert panel evaluated alternative methods for PER and concluded that PDCAAS is “the most suitable regulatory method for evaluating protein quality of foods and infant formulas” (FAO/WHO,

1991). WHO/FAO further concluded in 2007 that infant AA requirements should be used when scoring PDCAAS for infant foods (WHO/FAO/UNU, 2007). PDCAAS was adopted by FDA in 1991 (58 FR 2079) for assessing protein quality in all foods except foods for infants and young children. It was also used in a 2005 Dietary Reference Intake report for evaluating the relative nutritional quality of different protein sources (IOM, 2005).

Ileal Digestibility

DIAAS methodology requires direct measurement of the digestibility of each indispensable AA in the protein at the end of the small intestine (the ileum; see Box 3-3; FAO, 2013). The digestible AAs are then compared to the indispensable AA requirements of the reference population, which for infants is HMAA pattern (FAO, 2013). The lowest value, as a percentage, is the DIAAS for that protein. Scores are not truncated, so proteins with a high biological value can have values greater than 100 (Mansilla et al., 2020).

The primary limitation of the method is that it is an invasive procedure (Mansilla et al., 2020). Because ethical considerations and challenges limit access to ileal digesta in humans, the growing pig has been used to determine DIAAS for many proteins by placing a cannula at the end of the ileum to allow collection of digesta. Studies indicate that the true ileal digestibility of indispensable AAs for protein sources in the human and pig are highly correlated (Hodgkinson et al., 2022). Because the anatomy, physiology, and metabolism of infants and piglets are similar, the young, growing pig is considered an optimal model for studies of protein digestion in the human infant (Darragh and Moughan, 1995). Another limitation of DIAAS is AA secretion into the small intestine, which can vary with diet composition, thus impacting the indispensable AAs collected in the ileal digesta. However, endogenous losses can be accounted for using methods such as a protein-free diet to estimate true ileal digestibility. DIAAS also can overestimate the availability of some indispensable AAs, such as lysine, due to their reaction with lactose in the presence of heat to form Maillard reaction products that may reduce bioavailability. However, only a small percentage of the lysine in infant formulas is blocked from utilization due to the formation of reaction products, and the remainder exceeds the lysine requirement (Mehta and Deeth, 2015; Rutherfurd and Moughan, 2007). It is acknowledged that beyond lysine, other amino acids can also contribute to Maillard reaction products; in soy-based infant formulas, arginine contributes significantly toward N ε- (carboxyethyl)lysine (CEL), and several dietary advanced glycation end products (dAGE) (Xie et al., 2023).

DIAAS is considered to be a more accurate measure of protein quality than PDCAAS and was recommended by FAO in 2011 to evaluate protein quality of foods (FAO, 2013). A study in pigs demonstrated that the cow’s

and goat’s milk proteins used in infant formulas are highly digestible, with an ileal true digestibility of 90 percent or greater, whereas that for most plant-based proteins is much lower (Mathai et al., 2017). However, DIAAS values were determined only for food proteins and not after inclusion in an infant formula matrix in that study (Mathai et al., 2017). Rutherfurd et al. (2006) also showed true ileal digestibility of greater than 90 percent in both cow’s and goat’s milk–based infant formulas. Thus, little or no correction may be needed to adjust for AA digestibility with cow’s milk proteins (Wallingford, 2023), but correction may be needed for proteins from other sources.

The results of the scoping review identified only two studies (Calvez et al., 2024; Charton et al., 2023) between 2000 and 2024 using animal models to estimate DIAAS in infant formulas. Studies in a rat model suggest high but variable digestible indispensable AA ratios ranging from 0.80 (for aromatic AAs) to 1.97 (for sulfur AAs) for bovine milk–based formulas (Calvez et al., 2024). Comparison of an infant formula and human milk in a piglet model showed that both are highly digestible, with a true digestibility of 90 percent or greater (Charton et al., 2023). The digestible indispensable AA ratios (corrected for AA losses during chemical hydrolyses and endogenous basal losses and expressed mg/g true protein with 6.38 as the N-to-protein correction factor and total nitrogen corrected for non-protein nitrogen) for all AAs in the infant formula and human milk were greater than 99 except for the aromatic AAs (83). As the DIAAS value is the lowest ratio among the indispensable AAs for the test protein, scores of 101 for human milk and 83 for infant formula reflect the digestible AA score for the first limiting AA (Charton et al., 2023). Thus, a higher protein level is needed for infant formulas to compensate for the unbalanced AA profile that is due largely to the limiting aromatic AAs compared to the profile for human milk.

Human Studies for the Assessment of Protein Quality

The direct measurement of protein quality in humans is difficult due to the complexities of obtaining ileal digesta following a test protein intake. In humans, mostly adults, direct (oro-ileal digestibility) and indirect (dual isotope tracer method and indicator AA oxidation (IAAO)) methods have been developed to measure protein and AA digestibility (FAO and IAEA, 2024).

Oro-Ileal Digestibility Studies

To directly measure ileal digestibility in humans, naso-ileal intubation is required. The tube is introduced through the nose and allowed to migrate to the terminal ileum (radiography confirms the final position). It consists of three lumens: to infuse a

non-absorbable marker to measure flow rates, help with the peristaltic movement of the tube, and sample the digesta. Once test protein intakes have been provided orally, collection of ileal contents occurs. These contents are tested for nitrogen and AA concentrations (Bandyopadhyay et al., 2022).

Healthy adult humans have been tested using this method for digestibility of a variety of foods including milk, milk protein isolates, casein, eggs, meat, pea isolate, soy isolate, and some foods with low digestibility, such as sunflower seed and zein. Some studies also used foods that have intrinsically labeled stable isotopes (15N, 2H, 13C), which allow protein digestive kinetics to be assessed (FAO and IAEA, 2024). The strength of the method is that it provides a direct quantitative estimate of digestibility of protein (nitrogen) and AAs and is considered the “reference method” (see Table 3-6). However, the procedures are highly invasive and not adaptable to other life stages, including infants and children. Furthermore, the high cost of such studies limits the number of foods that are tested.

Oro-Ileal Digestibility, Ileostomates

Individuals who have a permanent ileostomy have been used to test ileal digestibility of foods, and the method was compared to ileal digestibility in pigs (Hodgkinson et al., 2022; Moughan et al., 2005). These studies are possible because the ileal contents are readily available, and some foods over the range of digestibility have been tested (FAO and IAEA, 2024). However, concerns have been raised as to the suitability of these individuals, who may have other comorbidities, as study participants and whether digestive disturbances (e.g., microbial differences) may interfere with the results. These studies are noninvasive, since access to the samples is easy. However, they are not applicable to infants and children.

Dual Isotope Tracer Studies

The dual isotope tracer method is conducted using foods that have been intrinsically labeled with one isotope (15N or 2H) and a standard reference labeled with another isotope (2H or 13C) of known digestibility. Test proteins are fed in small frequent meals and under steady conditions, and plasma samples are collected. The ratio of plasma appearance of AAs from the test protein to the reference protein, in comparison to the meal administered and corrected for the AA digestibility of the reference protein, allows for determining the true ileal AA digestibility. As the results are based on plasma appearance rather than an ileal sample measure, this is considered an indirect method. Since the first publication using this method (Devi et al., 2018), a variety of foods, including milk, eggs, legumes/pulses, and cereals, have been tested in healthy adult volunteers (FAO and IAEA, 2024; Bandyopadhyay

TABLE 3-6 Overview of Protein Quality Assessment Methods in Humans

SOURCE: Committee adapted from FAO and IAEA (2024) and Daniel Tome presentation at the June 9, 2024, open session workshop.

et al., 2022). The method has also been applied in healthy young children (Shivakumar et al., 2019) and children with stunting (Devi et al., 2020). An advantage of the method is its applicability in different age groups and the ability to test proteins after food processing. However, several assumptions used in the data analysis, such as whether the food versus reference isotope proteins have differential digestion, have not yet been validated. Additional challenges relate to the costs associated with production of the foods with intrinsic isotope labels, expertise required to conduct these complex studies, and partial invasiveness of the plasma sampling (see Table 3-6).

Indicator Amino Acid Oxidation

The concept of limiting AA is that the protein quality of the food source is driven by the limiting AA, and thus determining the whole-body utilization of it gives an estimate of the protein quality. The IAAO method is based on the principle that a single indispensable AA being limited in the diet would lead to an overall decline in body protein synthesis and all other AAs, including the indicator AA (another indispensable AA, which has a 1-13C-label), would be oxidized to 13CO2 measured in breath samples (Elango et al., 2012). The response to changes in 13CO2 in breath samples to intake of a limiting indispensable AA from a protein in food (for example, soy protein) is compared with that obtained using a free AA (in crystalline form). The relative difference in 13CO2 is a measure of whole-body relative bioavailability of the AA and reflects the effects of its digestion, absorption, and utilization of the test amino acid. That measure is a key difference of IAAO compared to the other human methods. It has been termed “metabolic availability,” since the results reflect the metabolic utilization of the limiting test indispensable AA (Paoletti et al., 2024).

The method was originally developed in pigs, adapted to healthy adult humans, and applied in children (Caballero et al., 2020). Foods that have been tested include casein, soy protein isolate, rice, and legumes and pulses, with the focus being on the limiting AA in each protein, such as lysine in rice and methionine in soy. Furthermore, the IAAO method has been used to test the effect of food processing (e.g., impact of heat treatment on lysine bioavailability) (Prolla et al., 2013) and the effect of protein complementation (e.g., mixing a cereal and legume to provide a complete protein) (Rafii et al., 2022). Bandyopadhyay et al. (2020) used the IAAO method to test the metabolic availability of lysine in spray-dried milk powder and heat-treated spray-dried milk powder in healthy young men. Heat treatment reduced lysine metabolic availability by 22 percent in the milk powder and shows the utility of the method to detect processing impacts on protein quality. Lysine is well known to react with lactose during processing (Xie et al., 2023) to form Maillard reaction compounds.

In Banyopadhyay et al. (2020), furosine, an early Maillard reaction product, more than doubled in concentration due to heat treatment compared to non-heat-treated spray-dried cow milk powder; lysinoalanine, a late Maillard reaction product, was absent in the non-heat-treated milk powder and present in the heat-treated milk powder in large amounts. Maillard reaction products and their impact on infant formula to reduce protein quality could be assessed using IAAO. Due to the noninvasive nature of the study (breath sampling), the method can be applied across life stages and in individuals with various diseases. Since the method only studies one AA at a time, the throughput is low, and its complexity requires expertise and training and thus increases the cost. None of the articles identified by the scoping review utilized this method.

In Vitro Approaches

Another alternative approach to measuring protein quality is in vitro models of digestibility estimation, which includes both static and dynamic methods (see Box 3-4). The 1991 FAO report on protein quality recommended that further research was needed to “perfect and evaluate” in vitro procedures for determining protein digestibility (FAO, 1991). The FAO 2013 report further expressed the need to develop, standardize, and independently validate in vitro methods for predicting protein and AA digestibility in humans (FAO, 2013).

Static Digestibility Methods

Static in vitro assays treat suspensions of food with a mixture of digestive enzymes (Krul et al., 2024). They are research tools to study food structure and digestibility, determine nutrient bioavailability, and provide protein digestibility coefficients (Krul et al., 2024). Static digestibility methods include pH drop, pH-stat, and commercially available assay kits.

The pH drop method is based on the principle that protons are released when peptides are cleaved during enzymatic digestion, causing a drop in pH over a specified time (Krul et al., 2024). The pH-stat method maintains the pH over a specified time by titration with sodium hydroxide (Abrahamse et al., 2012; Pedersen and Eggum, 1983). The static INFOGEST method measures protein digestion during the oral, gastric, and intestinal phases using standardized protocols (Brodkorb et al., 2019; Egger et al., 2017; Krul et al., 2024). It was developed through a research network with the aim to harmonize the various static digestibility assays.⁴ The protocol was created (Minekus et al., 2014) and modified (Brodkorb et al., 2019) to standardize experimental conditions and reagents for the

___________________

⁴ https://infogest.hub.inrae.fr/ (accessed December 1, 2024).

procedure, including digestive enzymes, bile, electrolytes, pH, and time of digestion, across a range of foods. It has been tested in foods with different digestibilities and the results compared with results from in vivo studies (Sousa et al., 2023). The INFOGEST method is currently undergoing the International Organization for Standardization application process for dairy products (ISO, 2024).

The strengths of these static in vitro methods of digestibility include ease of implementation across different lab settings, high throughput, and less cost compared to in vivo methods. PDCAAS values have been reported for a number of protein foods based on static in vitro methods of digestibility (House et al., 2019; Krul et al., 2024; Nosworthy et al., 2018; Tavano et al., 2016). Although these methods cannot fully reproduce all factors involved with in vivo digestibility, measures of in vitro and in vivo digestibility generally agree across various protein sources (House et al., 2019; Nosworthy et al., 2018; Tavano et al., 2016). However, in vitro static methods have limited validation for individual AA digestibility (Krul et al., 2024).

The scoping review revealed 14 studies using in vitro static models to examine the effect of infant formula composition and processing on digestion characteristics (Bavaro et al., 2021; Byrne et al., 2023; Cattaneo et al., 2017; Chitchumroonchokchai et al., 2023; de Figueiredo Furtado et al., 2022; Feng et al., 2022; Halabi et al., 2020; Huang et al., 2022; Le Roux et al., 2020a,b; Nguyen et al., 2016; Su et al., 2017; Zenker et al., 2020;

Zhu et al., 2021). However, comparison to human milk was not included in the experimental design. A recent study using the INFOGEST model that was modified to mimic in vivo infant digestive conditions (Komatsu et al., 2024) showed higher protein digestion of human milk than a bovine protein–containing infant formula during gastric digestion but similar digestion by the end of the intestinal digestion phase.

Dynamic Digestibility Assays

Dynamic digestibility assays are designed to simulate gastrointestinal digestion phases and nutrient bioaccessibility using sophisticated, computer-controlled, temperature-regulated digestion chambers (Krul et al., 2024). They have been used to measure AA and nitrogen digestibility (Krul et al., 2024). An example of a commercially available dynamic digestibility system is the TIM-1, which simulates the stomach, duodenum, jejunum, and ileum compartments of the gastrointestinal system. Semi-dynamic in vitro digestion methods simulate changes in the digestive process, especially in the gastric phase, including gradual changes in pH, digestive enzymes, and gastric fluid.

Strengths of a dynamic digestibility system include that it can be readily tailored to specific life stages, such as through changes in temperature or residency time, and allows for integrative study of nutrient digestion. It also allows a better simulation of physiological conditions than static in vitro methods and can mimic gradual changes of pH and enzyme activity and fluids, and thus, performance of kinetic studies. Challenges include the high cost of system acquisition, service contracts, and operation (Krul et al., 2024) as well as limited suppliers and sample throughput. Additional challenges with this approach include the various conformations that are available with different characteristics, e.g., to mimic gastric and peristaltic movements or fluid additions, that make harmonization between laboratories more difficult.

The scoping review commissioned by the committee revealed six studies that used a two-step semi-dynamic in vitro model designed to mimic infant digestion (Abrahamse et al., 2022; Bourlieu et al., 2015; Chen et al., 2022; de Almeida et al., 2021; He et al., 2022; Lambers et al., 2023) and five studies that used a dynamic in vitro model (Chauvet et al., 2023; Le Roux et al., 2020b; Maathuis et al., 2017; Mudgil et al., 2022; Song et al., 2024). Most of these compared the effects of processing on digestive qualities of infant formula, without reference to human milk. He et al. (2022) and Abrahamse et al. (2022) reported that the physiochemical behavior of infant formula containing bovine milk proteins differed from that of human milk. Maathuis et al. (2017), using an in vitro gastrointestinal multicompartmental dynamic model simulating infant conditions (tiny-TIM), reported that although the initial digestion of a cow’s milk–based infant formula was slower than that of goat’s milk–based infant formula and

human milk, the true ileal digestibility and estimated DIAAS in both formulas were similar to those of human milk.

Considerations for In Vitro Approaches

Other considerations arise with in vitro assays to assess protein quality. These include using animal-derived enzymes, which have less ethical implications, as these animals are destined for the meat sector. Additional data are needed to validate the in vitro methods and determine an appropriate reference pattern for infants. An advantage of in vitro approaches is that they do not require Institutional Review Board approval, which is needed for animal or human studies.

SUMMARY OF METHODOLOGY TO ASSESS PROTEIN QUALITY

The recognized method in the United States to evaluate the protein quality of infant formulas involves a rodent bioassay to determine PER. Based on the review of studies on measures of protein quality of infant formula and methods advocated in relevant regulations globally, the committee summarized alternative methods for the assessment of protein quality. These methods, including assessing digestibility and availability, measuring plasma AAs and urea, human and in vitro models, and measuring AA scores based on requirement assays (see Table 3-7).

The type of evidence needed to establish protein quality depends on the extent of information available about the protein source, including its processing, the processing of the formula, and the potential for interactions between the protein and other components of the formula. Digestibility is a suitable surrogate for bioavailability given evidence to support the absorption, metabolism, and utilization of AAs from the digesta of the protein source. No methods, including PER, have been validated by accepted organizations to assess protein quality in the intact infant formula. Thus, all preclinical methods of assessment are preliminary to evidence of protein utilization by infants, demonstrated as the quality factor of normal physical growth. Table 3-7 lists various conditions that may affect protein quality, ordered by relative risk, and the respective, appropriate methods of assessment of protein quality. The variable determining the assessment method must have adequate sensitivity to detect a nutritionally significant change.

Conclusion 3: Changes in protein composition or processing of infant formula may require evidence of digestion, absorption, and utilization of the amino acids in the protein source of the formula. Both in vivo and in vitro alternative approaches to the PER are available to demonstrate

digestion (see Table 3-7). Specific protein quality measures targeted for application in different scenarios of manufacturing processes and formulation changes may be warranted (see Table 3-8).

Recommendation 3: The Food and Drug Administration should provide guidance to manufacturers of infant formula when there is need for evidence of digestibility and bioavailability and on acceptable methods of assuring protein quality that reflect the types of changes in the composition or processing of the formula.

TABLE 3-7 Assumptions and Possible Regulatory Applications Related to Protein Quality Assessments

NOTES: No assessment methods have been validated specifically for infant formula. AA = amino acid; DIAAS = Digestible Indispensable Amino Acid Score; EU = European Union; FAO = Food and Agriculture Organization; FSANZ = Food Standards Australia New Zealand; IAA = indispensable AAs; IAAO = indicator AA oxidation; N/A = not available; N = nitrogen; PDCAAS = protein digestibility–corrected amino acid score; PER = protein efficiency ratio.

^a Recommended to be discontinued by FAO/WHO (1991) and Raiten et. al. (1998). Discontinued by Codex in 2007 (FAO/WHO, 2023), European Commission (2006).

^b Adopted by Codex, EU, and FSANZ and recommended by LSRO.

^c Pigs are preferred models.

^d Can use formulas and human milk as controls.

TABLE 3-8 Decision Matrix for the Assessment of Protein Quality in Infant Formula

NOTES: DIAAS = Digestible Indispensable Amino Acid Score; HMAA = human milk amino acid pattern.

^a Major changes (21 CFR § 106.3) that involve protein include any infant formula produced by a manufacturer that is entering the U.S. market or having a significant revision, addition, or substitution of a macronutrient (i.e., protein, fat, or carbohydrate), with which the manufacturer has not had previous experience. The committee considered new protein to include (a) a new species of protein, (b) a new mixture or large quantitative difference of existing proteins (American Academy of Pediatrics described as 20 percent), or a quantitative level at or below the minimum in regulation (where Codex and European regulations trigger a clinical evaluation).

^b Major changes that involve other ingredients include any infant formula having a significant revision, addition, or substitution of a macronutrient (i.e., protein, fat, or carbohydrate), with which the manufacturer has not had previous experience and any infant formula manufactured containing a new constituent not listed in section 412(i) of the Federal Food, Drug, and Cosmetic Act (21 U.S.C. 350a(i)), such as taurine or L-carnitine.

^c Major processing change is defined as any infant formula powder processed and distributed by a manufacturer that previously only produced liquids (or vice versa); any infant formula manufactured on a new processing line or in a new plant; any infant formula processed by a manufacturer on new equipment that uses a new technology or principle (e.g., from terminal sterilization to aseptic processing); or an infant formula that has a fundamental change in the type of packaging used (e.g., changing from metal cans to plastic pouches).

^d DIAAS; the preferred in vivo method for assessing digestion and accessibility of the products of digestion. HMAA; provides evidence of balance among and amount of indispensable amino acids to meet requirements. Animal growth provides evidence of use, and piglets are the preferred model.

REFERENCES