Type of infant formula increases early weight gain and impacts energy balance: a randomized controlled trial

ABSTRACT

Background

Millions of infants are fed breast milk substitutes, and the type of infant formula can impact weight gain patterns.

Objective

We conducted a randomized controlled trial to determine the direct impact of 2 types of infant formula (cow milk formula, CMF; extensively protein hydrolyzed formula, EHF) on growth and energy balance.

Design

A racially diverse group of formula-fed infants (n = 113) were randomly assigned to either CMF or EHF from the age of 0.75 to 12.5 mo. At each monthly visit, anthropometric measures were obtained to determine growth z scores and weight gain velocity, and to categorize early weight gain patterns as rapid or nonrapid. Also, diet records were collected to determine energy from formula and other sources. Comprehensive assessments of energy balance (intake, expenditure, loss) were made at 0.75, 3.5, and 12.5 mo.

Results

Beginning 3 wk after randomization, CMF infants had significantly higher weight, but not length, z scores than did EHF infants, and this persisted after solid foods complemented the formula diet. On average, weight gain velocity from 0.75 to 4.5 mo was within the range of typically growing infants for both groups, yet velocity was 3.9 g/d greater for CMF infants (P = 0.002), who were more likely to be classified as an early rapid weight gainer, than EHF infants (46% compared with 18%; P = 0.007). Early differences in energy intake and fecal loss, yielding greater energy available for deposition among CMF infants, contributed to the differential weight gain patterns. There were no significant differences between the formula treatment groups in total energy expenditure or sleeping energy expenditure.

Conclusions

Among healthy infants, the type of formula impacted on early rapid weight gain patterns owing to energy intake and loss mechanisms. Research is needed to identify the macronutrients and other compositional constituents in EHF and breast milk that promote satiation and healthy weight gain during sensitive periods of development. This trial was registered at clinicaltrials.gov as: NCT01700205.

INTRODUCTION

The infant diet is unique in that it typically consists of a sole source of nutrition—breast milk or breast-milk substitutes (hereafter referred to as infant formulas). However, formula-fed infants should not be treated as a homogeneous group because infant formulas can differ in macronutrient composition and such compositional differences impact flavor, satiation, and trajectories of weight gain in the infants who are fed them (1).

Cow-milk formula (CMF), the formula given to most healthy infants (2), differs slightly in the amount of protein and greatly in the form of protein from an isocaloric yet extensively protein-hydrolyzed formula (EHF) (3), which is fed to infants who have cow-milk protein allergy or intolerance to intact proteins. Unlike the intact proteins found in CMF, the proteins in EHF are treated with enzymes to break down peptide bonds, resulting in higher concentrations of small peptides and free amino acids (FAAs). As a reference, the FAA concentration is lower in CMF than in breast milk and is lower in breast milk than in EHF (4). On the other hand, while total protein concentration of CMF is lower than EHF, both are higher than that found in mature breast milk (5). Such compositional differences between these 2 types of formulas provide an experimental model system to study the impact of early diet on flavor preferences, satiation, development, and growth (1, 6, 7).

Differences in macronutrient composition can lead to distinct flavor profiles of formulas (8). The greater abundance of FAAs in EHF, many of which taste sour, bitter, and savory by binding to chemoreceptors in the mouth (9, 10), contributes to its unpalatability, especially to those naive to its taste. But palatability is personal, and is dependent on both age and experience. While infants <4 mo old readily accept its taste, beginning at 4 mo and continuing thereafter, EHF will be rejected unless the child has been fed EHF at <4 mo of age. Most remarkably, feeding EHF at <4 mo of age programs infants to like its taste (11, 12), and this preference persists several years later (13–15), a finding consistent with clinical reports of children who are fed other hydrolysate-based formulas [e.g., children with phenylketonuria (16, 17)].

The compositional differences in infant formula can impact satiation within a meal and weight gain over the long term. In the short term, within-subject experimental studies consistently reveal that infants <4 mo old satiate on smaller volumes when fed EHF than when fed CMF (18–20), perhaps because, relative to the intact proteins in CMF, the hydrolyzed proteins in EHF are transported and absorbed more rapidly and sensed by chemoreceptors in the gut (21, 22). In the longer term, the prospective, randomized German Infant Nutrition Intervention trial revealed that infants with a hereditary risk for atopy who were randomly assigned at birth to EHF (but whose mothers were encouraged to breastfeed) had lower BMI gains during the first year of life than did those randomly assigned to CMF (23). The findings of normative weight gain patterning among infants feeding on EHF and accelerated weight gain patterning among infants feeding on CMF were later confirmed in an 8.5-mo randomized controlled trial (RCT) on exclusively formula-fed infants who had no hereditary risk for atopy and who began feeding on the formulas within the first weeks of life (24).

Because the energy balance mechanisms underlying differences in weight gain owing to the type of infant formula remain unknown, the present trial randomly assigned healthy formula-fed infants to CMF or EHF and measured both growth and energy balance during infancy, a time when the energy requirement for growth is a greater component of total energy requirements than later in life. We characterized the direct impact of these formulas on early weight gain, which, if rapid, increases later risks of obesity and other comorbidities (25–31). Our study population was racially diverse, and the majority were black Americans, who are more likely to formula feed (32) and are at particular risk of later obesity (33).

METHODS

Participants

Newly parturient women were recruited through advertisements in regional newspapers, websites, and hospitals in the greater Philadelphia area. Inclusion criteria included mothers whose decision not to breastfeed was firmly established and infants born at full term whose birth weight was between 2500 and 4500 g; whose gestational age was ≥37 wk but ≤42 wk, and whose prior feeding history was CMF. Exclusion criteria included major congenital malformations, documented systemic or congenital infections, family history of atopy, or maternal gestational diabetes. Mothers and study personnel were blind to the hypotheses and group assignment.

The primary anthropometry outcomes were weight-for-agez score (WAZ), length-for-agez score (LAZ), and weight-for-length z score (WLZ) over time, and the proposed hypotheses were based on findings from our prior RCT that randomly assigned 0.5-mo-old healthy infants to be fed either CMF or EHF for 7 mo and that found significant group differences in patterns of gains in weight but not length (24). For this reason, power analyses were based on the group difference in WLZ scores at 4.5 mo; z scores were normally distributed within each group, with a mean SD of 1.3 (24). We powered the study to determine a group difference of 0.67 in WLZ because a difference of this magnitude defines rapid weight gain (28). A priori power analyses determined that a sample size of 94 is required to detect a difference of 0.67 with 90% power and a 5% type 1 error rate. To account for a 20% attrition rate, the target enrollment was 112 infants. The Office of Regulatory Affairs at the University of Pennsylvania approved all procedures, and written informed consent was obtained from each mother prior to study entry.

Study formulas

Powder infant formulas—CMF (Enfamil, Mead Johnson Nutrition) and EHF (Nutramigen, Mead Johnson Nutrition)—contained no added pre- or probiotics, and were provided to mothers as part of the study design. The manufacturer provided the formulas gratis in 350 g containers, each yielding ∼80 oz (2365 mL) of reconstituted formula. Formula containers were labeled either Formula A (white label) or Formula B (blue label) so that mothers and study staff would be blind to the type of formula. Throughout the trial, the formulas were manufactured in 2 batches and released following the quality standards required for commercial product release, which included analysis of the micronutrient and macronutrient concentrations (34).

The formulas were identical in caloric density (66.7 kcal/100 mL), percentage of energy provided from fat (48%), and type of fat blend (i.e., palm olein, soy, coconut, high–oleic sunflower oils); the percentages of energy from carbohydrate (EHF: 41%, CMF: 44%) and protein (EHF: 11%; 2.8 g protein/100 kcal; CMF: 8%; 2.1 g protein/100 kcal) differed slightly (35). While both infant formulas met the US Food and Drug Administration (FDA) criteria for minimum and maximum macro- and micronutrient concentrations (36), the major differentiator in the 2 formulas is the form of protein. While CMF contains a mixture of intact casein and whey proteins, EHF is a casein hydrolysate, containing small-molecular-weight peptides (37) and a substantially higher FAA content (80,375 µmol/L) than CMF (864 µmol/L), including the FAA glutamate (CMF compared with EHF: 7479 compared with 109 µmol/L) (3).

Trial design

The trial was a randomized, controlled, double-blind study that consisted of 14 study visits, which began when infants were 0.5 mo old and ended when they were 12.5 mo old. To ensure balanced randomization, groups were stratified for sex and race or ethnicity. A statistician generated a permuted block randomization and placed assignments in serially numbered, opaque, sealed envelopes for study staff. At the first study visit, mothers came with their 2-wk-old infants to the Monell testing facility to learn about the study and to be weighed and measured, after which mothers received a 1-wk supply of CMF to feed the infants. After 1 wk (0.75 mo), baseline sleeping energy expenditure (SEE) was measured and dyads were randomized to 1 of the 2 formula treatment groups (CMF, EHF).

At the end of the 0.75-mo visit and every monthly visit thereafter, each mother was provided with the same number of containers (∼12) of their randomized formula (labeled Formula A or B). At each of these visits, infants were fed their assigned formula using established methodologies to ensure adherence to the protocol, and mothers were queried about the number of times they formula fed their infant daily (24). Mothers received instructions on how to prepare the formula and to return all unused containers, or, if needed, to contact the research team if they needed more formula before the month's end. Mothers were not instructed by study personnel on how or how much to feed their infants, or when or how to introduce solid foods.

Anthropometrics

For each of the 14 study visits, weight and length were measured in triplicate by personnel who were trained in standard anthropometric techniques and certified by one of us (JIS) using calibrated equipment; the pediatric scale (Scale Tronix) and stadiometer (Harpenden Infantometer 702) were accurate to 0.1 kg and 0.1 cm, respectively. From these data, anthropometric measures were normalized to z scores using WHO growth standards (38) to yield the primary anthropometry outcomes measures of WLZ, WAZ, and LAZ. We chose to express the data as z scores because of the advantage that the score is standardized to a reference population based on the age and sex of the infant, thus permitting the evaluation of growth patterns between the infant formula treatment groups over time. To further describe changes over time, we also calculated body mass index z scores (BMIZ) and velocities of gains in both weight and length, defined as change of weight in grams or length in centimeters, respectively, divided by change of age in days (39). At the individual level, we determined whether each infant was an “early rapid weight gainer,” defined as an increase in WLZ of >0.67 SD, or an “early nonrapid weight gainer,” defined as a change in WLZ of ≤0.67 SD between baseline (0.75 mo) and 4.5 mo (28).

Energy balance

At 3 study visits [0.75 (baseline), 3.5, and 12.5 mo of infant age], comprehensive measures of each component of energy balance (intake, expenditure, loss) were assessed. At the 0.75-mo visit, postprandial SEE was measured prior to randomization, whereas the 3-d weighed energy intake (EI), stool collections [energy loss (EL)], and diaper collections [total energy expenditure (TEE)] occurred during the first days the infants were fed the randomized study formula.

For each of these 3 visits, we determined EI (calories per day) by 3-d weighed bottle intake; we standardized the number and size of the bottles, because this can affect both formula intake and weight gain (40). For 3 d, mothers recorded their infant’s intake of any liquid or food other than formula. For the remaining 11 visits, the EI of formula and other sources was determined from written records in which mothers reported all formula and foods consumed by infants for a 24-h period. Diet data were analyzed using Nutrient Data System for Research (University of Minnesota) to determine the available EI (41). All records were reviewed by registered dietitians (NP, JCT), and food records with physiologically implausible EI, either based on estimated energy needs (42) or because incomplete, were eliminated.

Postprandial SEE (calories per day), a proxy for resting energy expenditure in infants (43), was measured for ≥30 min by open-circuit, indirect calorimetry using a metabolic cart (Sensor Medic 2900 Z; Sensor Medics) with a canopy hood, in a quiet, thermally neutral room. Infants rested in a supine position under a large, clear, ventilated hood. Expiratory gases were sampled and analyzed every second, and 1-min averages were recorded. The first 10 min were devoted to environmental acclimation of the infant and not used in the calculations. SEE was calculated from oxygen consumption and carbon dioxide production by the Weir equation (44).

TEE (calories per day) was measured over 7 d using the doubly labeled water method and established protocols for infants (45, 46). After obtaining a baseline urine sample, infants were dosed based on body weight with 0.3 g H₂¹⁸O and 0.15 g ²H₂O/kg estimated total body water, and mothers were given special study diapers for collecting morning urine samples each day thereafter for 7 d; samples were stored frozen until analysis. The abundance of ²H and ¹⁸O in the dose water and urine samples (45) was determined by isotope ratio mass spectrometry. Isotope dilution spaces (kilograms) were calculated based on the elimination rates of ¹⁸O and ²H (45). Carbon dioxide production rate was computed from the difference in elimination rate of the isotopes (45, 47). TEE was calculated using the modified Weir equation (44) and an assumed respiratory quotient of 0.86 (46).

The free fat mass (FFM; kilograms) was calculated for each subject by dividing the total body water derived from ²H and ¹⁸O dilution by age-specific hydration constants (48, 49). When FFM measures by isotope dilution were not successful, FFM was determined by multiple linear regression imputation. Weight was regressed onto measured FFM using data from similarly aged subjects with successful FFM measures; the resultant equation was used to impute FFM for those with missing FFM values.

Stool EL (calories per day) was determined from 3-d stool collection by bomb calorimetry (Covance Laboratories) (50). The energy available for deposition was calculated for each infant as available EI minus TEE (51).

Statistical analyses

The treatment codes were unblinded after the data analyses were complete. Descriptive statistics were used to examine the distribution of infant sex, race or ethnicity, and category of early weight gain (rapid, nonrapid), as well as maternal age (years), BMI, education, and income. Fisher's exact analyses for categorical variables and ANOVAs for continuous variables were used to demonstrate that the randomized procedures resulted in no differences in these measures between the groups (CMF, EHF) at baseline. Results for continuous variables are presented as means ± SEs, and percentages are used for categorical variables.

The RCT design, which minimizes selection biases, enables the determination of the effects of treatment (CMF, EHF) while other variables are kept constant. For the intention-to-treat analyses, we focused on all 113 randomly assigned infants regardless of study withdrawal. We conducted generalized estimating equations (GEEs) that included treatment group (CMF, EHF), time (age of infants at the time of measurements), and group × time interaction on anthropometric outcomes. The GEE approach accounts for the repeated measurements of outcomes over time for each infant and examines whether the slopes of the line created differ between treatment groups. Visualization of the correlation structure for inclusion within the GEE model was conducted using spaghetti plots. For LAZ, an autoregressive correlation structure was identified. For all other growth z scores, a fit exchangeable correlation structure, where within-subject measurements were equally correlated over time, was found. For energy balance measures (TEE, SEE, EL, EI), separate repeated-measures ANOVAs were conducted with group as the between-subjects factor; SEE analysis was conducted with and without adjustment for FFM. Statistical significance level was set at P < 0.05 (2-tailed) for unadjusted P values. Data were analyzed using Stata version 12 (StataCorp LP, USA) and Statistica version 13.1 (Dell, Inc.).

RESULTS

Characteristics of participants and study outcomes

Of the 113 dyads randomly assigned, 83 (73%) completed the study (Supplemental Figure 1). An additional 28 women consented but were not randomly assigned to the study formulas because they changed their minds shortly after consenting, owing to the perceived burden of study participation. The study cohort was diverse in race or ethnicity, family income, and maternal education (Table 1), all of which reflect the urban setting in which they live, the county of Philadelphia (52). Groups did not differ in any of these measures, including the sex ratio and baseline anthropometric measures of the infants (Table 1), completion of primary outcome assessments of energy balance (Supplemental Table 1), and completion of the study (Supplemental Figure 1). During each of the 14 study visits, both groups of infants were fed their assigned formulas to satiation, and mothers answered questions which revealed that they perceived that their infants enjoyed the taste of the formula in the bottle. There were no differences in the baseline anthropometric measures between those dyads who completed the 1-y trial and those who did not (Supplemental Table 2). However, the mothers who did not complete the trial were more likely to have less education, and tended to be younger and of lower income.

TABLE 1.

Subject characteristics¹

		Treatment group
Characteristics	All2 (n = 113)	CMF (n = 59)	EHF (n = 54)	P 3
Infants
Age at enrollment, mo	0.41 ± 0.04	0.41 ± 0.0	0.42 ± 0.0	0.66
Female, n (%)	57 (50)	28 (47)	29 (54)	0.63
Race or ethnicity, n (%)
Black	70 (62)	35 (59)	35 (65)	0.14
White	25 (22)	17 (29)	8 (15)
More than one race or ethnicity	18 (16)	7 (12)	11 (20)
Anthropometry at enrollment
WAZ	−0.30 ± 0.08	−0.36 ± 0.11	−0.25 ± 0.11	0.49
LAZ	−0.48 ± 0.10	−0.49 ± 0.13	−0.46 ± 0.15	0.91
WLZ	−0.19 ± 0.09	−0.26 ± 0.11	−0.12 ± 0.14	0.43
BMIZ	−0.05 ± 0.08	−0.14 ± 0.11	0.05 ± 0.11	0.22
Mothers
Age, y	27.1 ± 0.5	27.1 ± 0.9	27.0 ± 0.7	0.89
Household income,5n (%)
<$35,000	81 (73)	44 (76)	37 (70)	0.70
$35,000–75,000	12 (11)	5 (9)	7 (13)
>$75,000	18 (16)	9 (15)	9 (17)
Education level, n (%)
Primary school	17 (15)	12 (20)	5 (9)	0.21
High school or TA certificate	68 (60)	32 (54)	36 (67)
College degree or higher	28 (25)	15 (26)	13 (24)
BMI at enrollment, kg/m2	30.9 (7.9)	30.7 (7.4)	31.2 (8.4)	0.75
Gestational weight gain, 5 kg	13.4 ± 0.9	13.0 ± 1.2	13.8 ± 1.3	0.65

¹BMIZ, body mass index z score; CMF, cow-milk formula; EHF, extensively protein-hydrolyzed formula; ITT, intention to treat; LAZ, length-for-age z score; TA, technical school; WAZ, weight-for-age z score based on WHO growth reference data; WLZ, weight-for-length z score.

²ITT study sample (N = 113).

³ Statistical Analyses. P values for main effect of treatment group, obtained from chi-square test or one-way ANOVAs with group as the between-subject factor.

⁴Mean ± SEM (all such values).

⁵ n = 111. Some values do not sum to the total because of missing data.

Anthropometric measures by treatment group

At baseline and throughout the study, z score measures were within 1 SD of the WHO growth standards for both groups. However, there were significant group × time interactions for each of the primary weight anthropometric measures, WAZ (P = 0.001) and WLZ (P = 0.047), as well as BMIZ (P = 0.035), but no group difference in the length measure, LAZ (P = 0.319). Figure 1 depicts anthropometric z scores of the 2 treatment groups. Infants randomly assigned to CMF had significantly higher WLZ, WAZ, and BMIZ than infants randomly assigned to EHF. The differences in weight z scores were first apparent at 1.5 mo, a few weeks after feeding started with the randomized formula, and reached a maximal difference with a WLZ of 0.67 at 4.5 mo, a time when formula provided 95% ± 1% of the EI.

FIGURE 1.

Anthropometric measures based on WHO growth standards by treatment group over time (0.5–12.5 mo): infant z score trajectories for WAZ (A), LAZ (B), weight-for-length (WLZ; C), and BMIZ (D), from age 0.5 to 12.5 mo. Data are least square means ± SEMs. The arrow (↓) at 0.75 mo indicates when infants began feeding on CMF (ITT CMF; circles; n = 59) or EHF (ITT EHF; triangles; n = 54). *Significant difference between groups at P < 0.05 (generalized estimating equations). ^∧Trend at P = 0.06. BMIZ, body mass index z score; CMF, cow-milk formula; EHF, extensively protein-hydrolyzed formula; ITT, intention to treat; LAZ, length-for-age z score; WAZ, weight-for-age z score; WLZ, weight-for-length z score.

Mean WAZ, WLZ, and BMIZ scores at 4.5 mo for the CMF group were between the 50^th and 75^th percentiles, which represented an upward crossing of one growth channel from where they began at 0.75 mo, whereas these weight z scores for the EHF group were in the 25^th and ∼50^th percentiles, which represented the same growth channel that they began at 0.75 mo, the age of randomization to the study formulas. In other words, over this time period, the infants’ weights, on average, remained within the same growth channel (25^th–50^th percentile lines) for the EHF group and increased by one growth channel (to the 50^th–75^th percentile lines) for the CMF group, both of which are within the normal reference range of WHO standards (38). While formula treatment group differences in WLZ were significant until 11.5 mo, this was only a trend at 12.5 mo (P = 0.06), a time when formula provided, on average, 31.6% ± 3.2% of EI and when EHF infants consumed more calories from other sources than CMF infants (Table 2). WAZ and BMIZ remained significantly higher in the CMF group than in the EHF group until 12.5 mo (Figure 2).

TABLE 2.

Energy balance measurements¹

			Treatment group,3 kcal/d
	Infants’ age,2 mo		CMF	EHF	P 4
Energy in
Available energy intake from formula and other sources4	0.75	Formula	466 ± 14	383 ± 15*	<0.001
		Other	0 ± 0	0 ± 0
	3.5	Formula	628 ± 21	596 ± 21	0.34
		Other	11 ± 5	10 ± 6
	12.5	Formula	425 ± 50	284 ± 56	0.03
		Other	712 ± 51	863 ± 57*
Energy out5
SEE	0.75		206 ± 6	201 ± 6	0.72
	3.5		329 ± 8	321 ± 8
	12.5		507 ± 11	509 ± 11
TEE	0.75		283 ± 18	243 ± 19	0.72
	3.5		488 ± 24	444 ± 26
	12.5		859 ± 42	858 ± 45
Energy loss in stool6	0.75		22.3 ± 3.9	34.9 ± 4.1*	0.02
	3.5		20.7 ± 1.8	17.0 ± 1.9
	12.5		27.0 ± 3.2	33.8 ± 3.4

¹*Significant difference between treatment groups, P ≤ 0.05. CMF, cow-milk formula; EHF, extensively protein-hydrolyzed formula; ITT, intention to treat; SEE, sleeping energy expenditure; TEE, total energy expenditure.

²The ITT study sample size was 113 at 0.75 mo of age, 101 at 3.5 mo of age, and 83 at 12.5 mo of age. Baseline measures at 0.75 mo were obtained either immediately before (SEE) or during the first days (TEE, stool energy loss) that infants were fed randomized formula.

³Values are means ± SEMs.

⁴Energy intake (available) from formula and other food sources by age. Because there was no variation in other food intake at age 0.75 mo, each time point was analyzed separately for group effects.

⁵ P values for effect of treatment × time, repeated-measures ANOVA. Statistical analyses for SEE conducted with and without adjustments for free fat mass; results remained unchanged.

⁶After accounting for difference between groups in fecal energy loss at 0.75 mo (see Methods), the adjusted available energy intake for EHF is 370 ± 15, which remains significantly different from CMF.

FIGURE 2.

Available EI (kcal/d; means ± SEMs) from formula (left axis: solid bars) and other sources (left axis: hatched bars) from 0.5 to 12.5 mo of age by treatment group (ITT CMF, n = 59, or ITT EHF, n = 54) and percentage of energy derived from formula (right axis: CMF, circles; EHF, triangles). The arrow (↓) at 0.75 mo indicates when infants began receiving the randomized formula. *Significant difference between groups at P < 0.05 (repeated-measures ANOVA). ^∧Trend at P = 0.068. CMF, cow-milk formula; EI, energy intake; EHF, extensively protein-hydrolyzed formula; ITT, intention to treat.

Weight gain velocity and early rapid weight gainers by treatment group

There was a significant group × time interaction in velocities of weight gain (P = 0.01). Figure 3 shows that the difference between groups was significant only for velocities of weight gain from 0.75 to 4.5 mo (P = 0.002). While there were no group differences in gains in length, for gains in weight the EHF group gained a mean of 25.1 ± 0.9 g/d, whereas the CMF group gained 29.0 ± 1.0 g/d. Although, on average, the weight gain velocities of both groups are regarded by WHO growth standards as within the range of typically growing infants (53), the group difference of 3.9 g/d is statistically significant (P = 0.002). After 4.5 mo, when the weight z scores of the 2 groups were increasing in parallel but remained higher in the CMF group (Figure 1), the velocities of gains in weight and length were not statistically different between the groups.

FIGURE 3.

Weight-gain velocity (A) and length-gain velocity (B), defined as change of weight in grams or length in centimeters divided by change of age in days, during intervals from randomization (0.75 mo) to 4.5 mo of age, from 4.5 to 8.5 mo of age, and from 8.5 to 12.5 mo of age for CMF (n = 48) and EHF (n = 44) groups. *Significant difference between treatment groups at P = 0.002 (repeated-measures ANOVA). CMF, cow-milk formula; EHF, extensively protein-hydrolyzed formula.

We next focused on individual infants rather than groups and determined whether each infant was an early rapid weight gainer from 0.75 to 4.5 mo, defined as an increase in WLZ of >0.67 SD (28). At 4.5 mo, almost half of the CMF infants (46%, 22 of 48) but only 18% of EHF infants (8 of 44; P = 0.007) were categorized as early rapid weight gainers. Infants who gained weight rapidly had significantly greater velocities of weight gain (32.8 ± 0.9 g/d) compared with nonrapid weight gainers (24.3 ± 0.6 g/d, P < 0.0001) during this time period.

Supplemental Table 3 includes the means, CIs, and statistical analyses of the 12.5-mo z scores based on infant formula treatment group (CMF, EHF) or on whether infants were rapid compared with nonrapid early weight gainers. Given that 4 z-score outcomes (WAZ, LAZ, WLZ, BMIZ) were examined and that statistical significance was determined independently for each GEE model, we applied a predetermined false discovery rate (FDR) of 0.10 to control for multiplicity (54, 55). We selected such a conservative FDR rate to balance type I and type II errors, and to account for the relatively small sample size and small number of comparisons, and we provide both unadjusted and adjusted Pvalues in Supplemental Table 3. WAZ, WLZ, and BMIZ scores at 12.5 mo were significantly greater in the CMF group than in the EHF group, as well as among rapid compared with nonrapid early weight gainers (P < 0.05 for all). The findings remained significant after FDR adjustments.

Energy balance by treatment group

Energy-in and energy-out mechanisms explained the group differences in energy deposition. At age 0.75 mo, formula was the sole source of energy for all but one infant. From ages 0.75 to 2.5 mo, EHF-fed infants ingested fewer calories per day (P = 0.003; Figure 2), fewer calories per kilogram of body weight per day (P = 0.01), and less formula per feed (P = 0.03) than did CMF-fed infants. As shown in Figure 2, EI from formula peaked at 6.5 mo and remained the predominant source of nutrition until 10.5 mo for both groups. At 12.5 mo, the groups were consuming similar calories from formula, but the EHF group consumed more calories from other foods than the CMF group (P = 0.03). Groups did not differ in the age at which infants began eating fruit (CMF compared with EHF: 5.8 ± 0.2 compared with 5.7 ± 0.2 mo; P = 0.78), vegetables (5.8 ± 0.2 compared with 6.0 ± 0.2 mo; P = 0.49), or cereal (6.4 ± 0.3 compared with 6.2 ± 0.3 mo; P = 0.58).

There was a significant treatment group × time interaction (P = 0.02) in the amount of energy lost in stools (Table 2). Fecal EL was greater during the first days of feeding on EHF compared with CMF, but did not differ at 3.5 or 12.5 mo of age. There were no significant group × time interactions or group effects for TEE or SEE, with and without adjustments for FFM. The means of imputed FFM (21%; n = 61/297) compared with actual FFM values did not differ significantly at any of the time points (P > 0.40 for all).

To determine the energy available for deposition, the group differences in the digestibility of the energy substrates, as evidenced by significantly greater fecal EL in EHF-fed infants during the first few days of feeding the assigned formula (0.75 mo), had to be accounted for. First, we calculated the percentage fecal EL for each treatment group:

(1)

Treatment groups significantly differed in percentage fecal EL (CMF compared with EHF: 4.5% compared with 7.9%; P = 0.0002). Using the CMF group as the reference, we then adjusted the available EI of EHF-fed infants by reducing it for the difference in percentage fecal EL between treatment groups as follows:

(2)

At the 0.75-mo time point, CMF-fed infants had significantly more energy available for deposition (190 ± 20 kcal/d) than did the EHF group (118 ± 23 kcal/d; P = 0.02), and this difference remained significant without adjustment of available EI for the EHF group. At 3.5 or 12.5 mo of age, the formula treatment groups did not differ in energy available for deposition.

DISCUSSION

Healthy infants whose mothers decided to formula feed them and who were randomly assigned to CMF had increased velocities of weight gain compared with infants randomly assigned to EHF and double the incidence of early rapid weight gain from 0.75 to 4.5 mo of age, a time in life when formula is the predominant source of nutrition for the vast majority of formula-fed infants (56). Such patterns of weight gain during early sensitive periods is clinically relevant, because gaining weight rapidly during early life is regarded as a modifiable determinant of risk for many chronic diseases (25–31).

The infant formulas used in this RCT were commercially available products that have been sold in the United States for decades. Before marketing these formulas, the manufacturer was required by law to provide the FDA with documentation that each formula meets FDA nutrient requirements and provides adequate nutrition for infants to thrive. Thus, as expected, both groups exhibited age-appropriate growth, as evidenced by WAZ, WLZ, LAZ, and BMIZ scores within 1 SD of the WHO growth standards (38) and weight gain velocities within the range of typically growing infants, as defined by WHO standards (53). However, the treatment groups differed significantly in each of the 3 weight z scores over time, as well as in their velocities of early weight gain.

One month of feeding CMF resulted in significant increases in WLZ, WAZ, and BMIZ relative to the EHF group that persisted and reached a maximum at age 4.5 mo, when 100% of energy intake was from formula for all but 18 infants. From age 0.75 to 4.5 mo, the EHF group gained on average 25.1 g/d, whereas the CMF group gained 29.0 g/d. This between-group difference of 3.9 g/d in weight gain velocity is considered nutritionally significant according to the US FDA Infant Formula Act (36). Furthermore, when we focus on the individual infants, nearly half (46%) of the CMF group but less than one-fifth (18%) of the EHF group experienced early rapid weight gain from randomization until they reached 4.5 mo of age. Early rapid weight gainers gained on average 33 g/d during this early time period and had higher weight zscores at 12.5 mo of age compared with those who did not gain weight rapidly from age 0.75 to 4.5 mo. From age 4.5 to 12.5 mo, the weight gain velocities and the patterning of the weight curves did not differ between the formula treatment groups, but weight z scores remained significantly higher for the CMF group than for the EHF group until 11.5 mo of age. However, the treatment group differences in WLZ attenuated at 12.5 mo of age, when the percentage of energy from nonformula sources increased and the percentage of energy from formula decreased to one-third of the infants’ daily caloric intake.

The treatment groups did not differ in linear growth, and length z scores increased in both groups at rates typical of healthy infants during the first year of life (38). Thus, not only did the infants have sufficient energy for linear growth, but also the differences in early weight gain cannot be attributed to differences in length. Nor can the group differences be attributed to the taste or aroma of the formulas. Infants were introduced to EHF at age 0.75 mo, which is during the early sensitive period of taste acceptance (1). During each of the 14 monthly compliance checks, they were fed their assigned formulas to satiation and were perceived by their mothers as enjoying the taste of the formula. There were no group differences in the mothers’ willingness to feed their infants the formula, compliance to protocol, or study completion rates. Furthermore, mothers and study staff were blind to the treatment group assignment and hypotheses.

Comprehensive measurements of each component of energy balance revealed that the differences in weight gain and weight z scores between the groups were owing to early differences in EI of formula and in EL, which led to greater energy available for tissue deposition in the CMF group than in the EHF group. CMF infants consumed more formula and thus more calories from age 0.75 to 2.5 mo than EHF infants, consistent with experimental studies showing that CMF is less satiating to infants than EHF (18–20). CMF infants also lost significantly less energy in their stools during the first days of feeding on the formula than did EHF infants, perhaps because CMF transits the gastrointestinal tract more slowly than EHF (57, 58). Group differences in fecal EL were no longer observed at the 3.5-mo visit, so whether the formula-induced change in fecal EL was transient or persisted for a few months remains unknown. Of interest, prior research has demonstrated that when infants are switched from CMF to EHF, the consistency of the stool changes, resembling the watery stools characteristic of breastfed infants (59), and the number of stools per day decreases, remaining relatively stable for the first month of feeding on EHF (60). That there were no group differences in measurements of TEE or SEE at 0.75 (baseline), 3.5, and 12.5 mo indicates that differences in energy expenditure did not contribute to the formula-induced differences in weight gain.

Although the EHF and CMF used in the present study are isocaloric and contain identical fat blends, the total carbohydrate content is slightly lower and the protein content is slightly higher in the EHF than in the CMF group (3). However, the greatest difference is in the form of protein, with the FAA content substantially higher in the EHF than in the CMF group (3). Three hypotheses, not mutually exclusive, may explain how these compositional differences contributed to the differential weight gain of the infants.

First, because EHF has a higher protein concentration than CMF (2.8 compared with 2.1 g/100 kcal), one hypothesized explanation for differences in weight gain may be the amount of protein per se. Koletzko et al. (61) demonstrated greater weight gain during the first 6 mo of life when infants were fed a CMF with a protein content of 2.9 g protein/100 kcal, similar to that of the EHF used in our study, than when fed CMF containing 1.8 g protein/100 kcal. However, in the present study, the protein concentration of CMF was lower than EHF, yet the CMF group had greater weight z scores and were more likely to be categorized as early rapid weight gainers than the EHF group.

Second, the form of the protein may contribute to weight gain differences via an influence, in part, on satiation and thus on energy intake. Compared with CMF, EHF has greater diversity, a 93 times higher concentration of FAA overall, and a 69-times higher concentration of the FAA glutamate (3), a key signal for satiation (62) and, of particular relevance for infant feeding, the most abundant FAA in human breast milk (4). Within-subject experimental studies demonstrate that infants signal satiation sooner and satiate on lower volumes of formula when feeding on CMF with the FAA glutamate added at levels found in EHF (CMF + glutamate) than on CMF alone (19). The differences in satiation were not owing to taste, since there were no perceived differences in the taste of CMF compared with CMF + glutamate and no differences in facial reactivity of the infants during feeding.

Third, the lower carbohydrate content in EHF than in CMF (7.0 compared with 7.4 g/100 kcal) may have played a role in differential weight gain. This hypothesis was not supported by satiation studies in infants, however. Despite the same carbohydrate content, infants consumed more of CMF to satiation compared with CMF + glutamate, while there were no differences in their intake of EHF compared with CMF + glutamate (19). Based on this evidence, it is more likely that differences in the form of the protein, specifically the amount of FAA and small peptides, rather than differences in the amount of protein or carbohydrate or the taste of the formulas, were responsible for the differential weight gain. Because EHF transits the gastrointestinal tract faster than does CMF (57), and because FAAs, particularly glutamate, are known satiation signalers (19, 58) and modulators of gastroduodenal motor functioning (22, 63), we hypothesize that earlier production of neuroendocrine satiation signals (64) and gut nutrient sensing (22) drove the earlier satiation in the short term and the less rapid early weight gain in the longer term. More research is needed to determine the role of the macronutrients (carbohydrate, protein) and other compositional constituents in EHF and breast milk that promote satiation, intake regulation, and healthy weight gain during sensitive periods of development.

In conclusion, randomizing healthy infants to be fed isocaloric formulas that differ in composition altered how rapidly they gained weight, owing to both energy intake and energy loss mechanisms. Almost half of the infants randomly assigned to CMF were early rapid weight gainers, a risk factor for later obesity and other comorbidities (25–31), compared with only one-fifth of those randomly assigned to EHF. Our findings are clinically meaningful because CMF accounts for the vast majority of infant formula sales in America (2) and the infants who are fed breast milk substitutes, like the participants in this RCT, are more likely to be black Americans (32), who are at particular risk of later obesity (33). Research is needed to determine the mechanisms by which diet composition and rapid weight gain during sensitive periods of development (1, 12) program risks for later diseases (28–31).

Notes

All phases of this study were supported by NIH grant R01HD072307 and HD072307-03S1 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development awarded to JAM and JCT, and the National Center for Advancing Translational Sciences of the National Institutes of Health Award UL1TR000003. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The study formulas were provided gratis from the manufacturer, Mead Johnson Nutrition. Neither the funding agency (NIH) nor the donor of the infant formulas (Mead Johnson Nutrition) had any role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation or contents of the manuscript. The Monell Center is a nonprofit research institute that accepts donations or gifts from corporations, one of which is Mead Johnson Nutrition. The Monell Center has a policy to protect the integrity of our research, including that such donations are unrestricted and do not provide financial benefit to individual employees and do not influence the design, conduct, or interpretation of the study findings. Mead Johnson Nutrition donated the study formulas and provided no funds to support personnel or the execution of the trial.

Supplemental Figure 1 and Supplemental Tables 1–3 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.

JAM and JCT contributed equally.

Abbreviations used:

BMIZ

BMI z score

CMF

cow-milk formula

EHF

extensively protein-hydrolyzed formula

EI

energy intake

EL

energy loss

FAA

free amino acid

FDA

Food and Drug Administration

FDR

false discovery rate

FFM

free fat mass

GEE

generalized estimating equation

LAZ

length-for-age z score

RCT

randomized controlled trial

SEE

sleeping energy expenditure

TEE

total energy expenditure

WAZ

weight-for-age z score

WLZ

weight-for-length z score