Association Between Diet Patterns of Human Infants and Growth Differs Across Randomized Formula Groups in the First 18.5 Months

Abstract

Background

The composition of the formula fed to infants and the velocity of weight gain during infancy can impact later weight status. However, whether patterns of food and beverage intake during the complementary feeding period affect this relationship remains unclear.

Objectives

To identify patterns of complementary feeding (primary aim) and to examine how these diet patterns, along with infant formula composition and early weight gain velocity, influence weight status over time (secondary aim).

Methods

Diet records were collected monthly from 0.5 to 12.5 mo and again at 18.5 mo (n = 80) as part of a trial involving healthy infants randomly assigned to receive either cow milk formula (CMF) or an isocaloric extensively hydrolyzed protein formula (EHF). Foods and beverages were assigned to mutually exclusive What We Eat in America food categories, and intake at each age was standardized for body weight (g/kg). Diet patterns were derived using principal component analysis. Generalized estimating equations examined the independent and interactive effects of formula type (fixed effect), rapid weight gain (RWG) category (rapid/nonrapid; fixed effect), and infant diet patterns (time-varying effect) on weight-for-length Z-score (WLZ) over time.

Results

Over the first 18.5 mo, 4 diet patterns were identified: (PC1) “infant formula;” (PC2) “low-sugar cereal, not fried potatoes, grain mixed dishes, fruit, vegetables, meat mixed dishes, 100% fruit juice;” (PC3) “milk, pizza, water, high-sugar cereal and snack/meal bars;” (PC4) “fried poultry, bakery products, fried potatoes, starchy vegetables.” Over time, the effect of RWG category on WLZ was significant for the CMF (χ² = 127.15, P ≤ 0.001) and EHF (χ² = 140.71, P ≤ 0.001) groups. However, the main effect of PC1 (β = 0.189, P = 0.007) and subsequent greater WLZ score over time (χ² = 17.95, P < 0.05) were significant only in the CMF group.

Conclusions

The prolonged feeding of CMF, but not EHF, during complementary feeding was associated with greater WLZ in the first 18.5 mo.

This trial was registered at clinicaltrials.gov as NCT01700205 (https://clinicaltrials.gov/study/NCT01700205).

Introduction

Nutrition plays a vital role in infant development. Infant diet not only affects physical and cognitive growth but also shapes taste preferences and influences long-term health outcomes [[1], [2], [3]]. The composition of the formula fed to infants and the velocity of weight gain during infancy have been shown to impact later weight status. Infants who gain weight rapidly have a greater proportion of their weight gain attributable to fat mass, compared with infants with normal weight gain [4] and they are at a higher risk of having overweight or obesity in childhood and adulthood [[5], [6], [7]].

In prior research, we show that healthy term infants randomly assigned to be fed cow milk formula (CMF), the most commonly consumed type of infant formula [8], experienced accelerated weight gain during the first 4 mo compared with breastfed infants, whereas infants fed an extensive protein hydrolysate formula (EHF) had normative weight gain velocity [9]. The accelerated trajectory of CMF-fed infants resulted from both lower fecal energy loss (early) and higher energy intake during the first several months of feeding [9]. The greatest difference in weight-for-length Z (WLZ) scores occurred between 3 and 4 mo of age, when infant formula was the sole source of nutrition. This early rapid weight gain (RWG), which occurred in a higher percentage of CMF-fed infants (46%) than EHF-fed infants (18%) [9], was independent of maternal feeding style or the number of daily feedings [10]. Indeed, this RWG, compared with normative weight gain, leads to a higher WLZ score in the first year of life for both formula-fed and breastfed infants; however, formula-fed infants who experience RWG have significantly higher odds of being a child with overweight by 12 mo than do breastfed infants with RWG [11]. The difference in WLZ trajectory steadily decreased over time, coinciding with introduction of complementary foods, but remained significantly higher in CMF-fed than in EHF-fed infants as a group through 11.5 mo of age [9]. The timing and pattern of the complementary diet may further contribute to differences in weight gain trajectories, potentially amplifying or mitigating the impact of RWG velocities that occurred during exclusive formula feeding.

The infants in this trial were phenotyped monthly for growth and diet from 0.5 mo to 12 mo, with a follow-up at 18.5 mo. The study spanned from the time the randomized formula was the sole source of their nutrition to when formula was no longer a part of their diet. This contemporary cohort of formula-fed infants provides a valuable opportunity to conduct secondary analyses to identify data-derived diet patterns based on complementary feeding (primary aim) and to examine how these diet patterns and early weight gain velocity influence weight status over time, while accounting for the type of randomized formula fed to the infants during the trial (secondary aim).

Methods

Trial overview

The study design, inclusion and exclusion criteria, and CONSORT table for the original randomized controlled trial (RCT) have been published previously [9]. In brief, advertisements in newspapers, websites, and hospitals were used to enroll females and their infant in the greater Philadelphia region from 2012 to 2015. Inclusion criteria included healthy, full-term infants (gestational age 37–42 wk) with a birthweight between 2500 and 4500 g. The trial was designed to investigate the effect of infant formula composition on growth and measures of energy balance from 0.5 to 12.5 mo, with an 18.5-mo follow-up visit. Term infants whose mothers had decided to formula feed (n = 113) were randomly assigned to be fed either CMF (Enfamil; 67 kcal/100 mL; 48% energy from fat; 44% energy from carbohydrate, 8% energy from protein; n = 59) or EHF (Nutramigen; 67 kcal/100 mL; 48% energy from fat; 41% energy from carbohydrate, 11% energy from protein; n = 54). At each study visit, anthropometric and dietary intake data were measured (see below). Herein, analyses focus on children who completed the 12.5-mo and/or 18.5-mo visit (n = 80; Figure 1).

FIGURE 1.

Flowchart for the diet pattern analysis. CMF, cow milk formula; EHF, extensively hydrolyzed formula; ITT, intent to treat.

The trial was registered online at clinicaltrials.gov prior to start (NCT01700205; 2012–2016), conducted according to the guidelines of the Declaration of Helsinki, and approved by the Office of Regulatory Affairs at the University of Pennsylvania Institutional Review Board. Throughout the trial, the formulas were provided gratis to the participating families.

Dietary intake

Dietary intake was assessed monthly throughout the 12-mo RCT and at the 18.5-mo follow-up visit. Prior to each visit, mothers completed either a 3-d diet record (0.75, 3.5, 12.5, and 18.5 mo) or a 1-d diet record (0.5, 1.5, 2.5, 4.5, 5.5, 6.5, 7.5, 8.5, 9.5, 10.5, and 11.5 mo). At the 0.75, 3.5, and 12.5 mo visits, formula intake was assessed using the weighed bottle method [12]. For all other visits, formula intake was quantified by reported fluid ounces. All other foods and beverages were quantified using household measures (e.g., cups, fluid ounces). Although mothers attended the study visit, trained study personnel reviewed returned records and queried mothers for additional details as needed. Throughout the trial, no instructions were given to mothers on how or when to feed infant formula, solid foods, and/or beverages. Age of weaning (in mo) was operationalized as the first visit in which a food or beverage other than infant formula was reported in the 24-h diet recall or food record. To determine the gram amount and energy content of all foods and beverages consumed, dietary intake data were analyzed using Nutrition Data System for Research software versions 2013–2015, developed by the Nutrition Coordinating Center (University of Minnesota, Minneapolis, MN) [[13], [14], [15]].

To determine data-derived diet patterns, diet data from all time points were compiled into a long format [16]. Foods and beverages consumed were categorized primarily based on the 46 adult food subcategories in the 2017–2018 What We Eat in America (WWEIA) [17]. Baby foods and beverages were assigned the appropriate adult food categories. To better classify the diet, healthier and less healthy versions of the same food were further categorized to distinguish between foods with higher compared with lower levels of added sugars and saturated fats. Consistent with prior research, the subcategory of “French fries and other fried white potatoes” was modified to also include fried sweet potatoes, and the category “white potatoes baked or broiled” was modified to also include sweet potatoes baked or broiled, resulting in the categories “white and sweet potatoes, fried” and “white and sweet potatoes, not fried” [18]. Additionally, ready-to-eat and cooked cereals were combined and categorized based on sugar content (low ≤21.2 g sugar/100 g; high >21.2 g sugar/100 g) [19]. Next, the cumulative frequency of consumption of each food category was calculated (category frequency divided by total frequency of all food and beverages reported, multiplied by 100). Food categories consumed in our infant sample with a cumulative percentage <0.50% were allocated to an alternate food category of similar macronutrient composition or dropped from the analysis prior to diet pattern analyses.

The original RCT had 3 time points (0.75-, 3.5-, and 12.5-mo time points) with 3 d of diet records as did the 18.5-mo follow-up visit. For all other study visits, 1-d diet records were collected. We aimed to include >1 d of diet record per participant for each time point, to better capture intraindividual intake variability. Therefore, with the exception of the baseline visit (0.5 mo), we combined adjacent visits with 1-d records, creating age intervals with 2 d of records: 1.5–2.5 mo (1-d record at 1.5 mo and 1-d record at 2.5 mo combined), 4.5–5.5 mo, 6.5–7.5 mo, 8.5–9.5 mo, and 10.5–11.5 mo (summarized in Table 1).

TABLE 1.

Number and length of usable diet records per age interval (n = 80 infants)¹.

Infant age (mo)	Infants with usable diet records		Infant age interval (mo) for diet pattern analysis	Length of diet record per age interval for diet pattern analysis2
	CMF group	EHF group
0.5	n = 38	n = 37	0.5	1-d record: 100%
0.75	n = 41	n = 35	0.75	1-d record: 3% 2 or 3-d record: 97%
1.5	n = 37	n = 36	1.5–2.5	1-d record: 11% 2-d record: 89%
2.5	n = 39	n = 37
3.5	n = 37	n = 37	3.5	1-d record: 0% 2 or 3-d record: 100%
4.5	n = 35	n = 35	4.5–5.5	1-d record: 15% 2-d record: 85%
5.5	n = 35	n = 36
6.5	n = 35	n = 33	6.5–7.5	1-d record: 32% 2-d record: 68%
7.5	n = 28	n = 27
8.5	n = 33	n = 28	8.5–9.5	1-d record: 30% 2-d record: 70%
9.5	n = 28	n = 28
10.5	n = 33	n = 29	10.5–11.5	1-d record: 41% 2-d record: 59%
11.5	n = 28	n = 29
12.5	n = 36	n = 31	12.5	1-d record: 4% 2 or 3-d record: 96%
18.5	n = 36	n = 37	18.5	1-d record: 16% 2 or 3-d record: 84%

Abbreviations: CMF, cow milk formula; EHF, extensive protein hydrolysate formula.

¹Of the 83 infants who completed the trial, 3 had no diet records at either 12.5 or 18.5 mo and were excluded from analyses, resulting in a final sample size of 80 infants for diet pattern analysis. At each time interval in this study, a small percentage of infants at each visit had missing or unusable data.

²Except for the baseline visit (0.5 mo), for which there was only 1 d of diet record, we aimed to have ≥2 d of diet records at each time interval. Adjacent months of 1-d records were combined; time points in which 2 or more d of diet records were expected (0.75, 3.5, 12.5 and 18.5 mo) were not combined.

Diet patterns

Principal component analysis (PCA) was used to identify data-derived principal components (PCs). The input variable was the gram amount of each food/beverage category consumed divided by the infant’s weight (to account for the infant’s changing body weight over time) [16] and the number of available days of diet data at each time interval (g/kg/d). PCA is commonly used to reduce dimensionality and incorporates data centering, scaling, and orthogonal rotation [20]. Here the PCA created mutually exclusive linear combinations of food and beverage groups (i.e., diet patterns) while maintaining the original variation in the data. The number of PCs retained was determined by visual inspection of the scree plot, which plots the eigenvalue (the proportion of total variance explained by each PC) in decreasing order of magnitude. PCs prior to the final largest break (the point between a relatively large and small eigenvalue) in the scree plot were retained [20].

Infant anthropometry

As described previously [9], each infant’s weight and length were measured by trained personnel at each study visit using a pediatric scale (Scale Tronix) accurate to 0.001 kg and an infantometer (Harpenden Infantometer 702) accurate to 0.1 cm. Each measure was taken 3 times at each monthly study visit, and the average measurement was used. Weight and length measures were converted to Z-scores using WHO growth standards, resulting in WLZ scores. From these measures, we determined the difference in WLZ from 0.5 to 4.5 mo. Those infants who had an increase in WLZ of 0.67 or more comprised the RWG group, and the remaining infants comprised the non-RWG group [9,21,22].

Potential covariates

At baseline, questionnaires were administered to mothers to collect information regarding sociodemographic characteristics. Infant biological sex (male, female) and race (White or Caucasian, Black or African American, other or 2+ races) were reported by mothers. Mothers also self-reported race, education (less than high school, high school, college), income (<$35,000, $35,000–$75,000, >$75,000), and parity (nulliparous, multiparous). Maternal prepregnancy BMI (in kg/m²) was calculated from self-reported prepregnancy body weight and highly correlated (r = 0.93, P ≤ 0.001) with baseline BMI measured at the 0.5-mo postpartum visit.

Statistical analysis

Descriptive statistics (means, frequencies, and percentages) were used to summarize infant and maternal dyad sociodemographic and diet pattern characteristics. Student’s t-tests (continuous variables) or χ²/Fischer’s exact tests (categorical variables) were used to examine baseline differences across sociodemographic characteristics by randomized infant formula group (CMF, EHF).

Generalized estimating equations for repeated measures were used to assess the independent and interactive effects of formula type (fixed effect), RWG category (fixed effect), and infant diet patterns (time-varying effect) on WLZ score over time, from 0.5 to 18.5 mo [23]. The initial equations (Model 1) examined the main and interactive effects of infant formula type and time on WLZ score. Model 2 examined the main and interactive effects of RWG category, formula type, and time on WLZ score. Model 3 determined the effect of the complementary diet, by adding PC scores as a main effect, in addition to the main and interactive effects of RWG category, formula type and time on WLZ score. Next, the main effect of PC scores, in addition to the main and interactive effects of RWG category and time on WLZ score was examined to see if it differed by formula type. To prevent a 4-way interaction and to preserve statistical power, Model 4 was stratified by formula type (Model 4A: CMF; Model 4B: EHF) and further adjusted for potential covariates known to be associated with WLZ score (i.e., infant sex, infant race, age at weaning, household income, parity, prepregnancy maternal BMI) in Models 5A (CMF) and 5B (EHF). All analyses were conducted by using STATA software (version 18.0; STATA Corp).

Results

Participant characteristics

In the original trial [9], 74% of infants (N = 83) completed the 12.5-mo visit (hereafter the 1-y visit) and 69% (N = 78) completed the 18.5-mo follow-up visit (hereafter the 1.5-y visit) [24]. Three of these infants did not have usable dietary intake data for the 1 and/or 1.5-y visit, so our analyses focus on 80 children (n = 42 fed CMF; n = 38 fed EHF) (Figure 1). Despite our efforts to ensure ≥2 d of intake data were available at each time interval in this study, a small percentage of infants (ranging 3% at 0.75 mo to 41% at 10.5–11.5 mo; mean = 17%) had only 1 d of diet record available, because some records were not returned or had missing data (Table 1). At the start of the trial (0.5 mo) the randomized formula groups showed no significant differences in any sociodemographic or anthropometric characteristics (Table 2) [9]. Consistent with the original RCT cohort [9,10,24], the CMF group in these secondary analyses had higher weight gain velocity and a higher proportion of infants with RWG than did the EHF group. Notably, there was no difference in the age at weaning between the CMF and EHF group.

TABLE 2.

Characteristics of infants and their primary caregivers by formula type.

Characteristic	All subjects	CMF group	EHF group	P value
Maternal-infant dyads, n (%)	80	42 (53)	38 (48)
Infants
Male1, n (%)	40 (50)	22 (52)	18 (47)	0.65
Race1, n (%)				0.21
White or Caucasian	19 (24)	12 (29)	7 (18)
Black or African American	46 (58)	25 (60)	21 (55)
Other or 2+ races	15 (19)	5 (12)	10 (26)
Baseline weight-for-length Z-score, mean ± SEM	−0.18 ± 0.91	−0.15 ± 0.17	−0.20 ± 0.12	0.83
Weight gain velocity (g/d) 0.5–4.5 mo, mean ± SEM	27.79 ± 0.71	29.23 ± 0.82	26.21 ± 1.14	0.03
Age of weaning (mo), mean ± SEM	5.22 ± 0.32	5.40 ± 0.30	5.03 ± 0.23	0.32
Mothers
Prepregnancy BMI, mean ± SEM	29.7 ± 1.0	29.58 ± 1.3	29.76 ± 1.5	0.93
Baseline BMI, mean ± SEM	31.13 ± 0.92	31.54 ± 1.22	30.67 ± 1.41	0.64
Multiparous1, n (%)	64 (80)	34 (81)	30 (79)	0.82
WIC enrollment, n (%)	51 (64)	27 (64)	24 (63)	1.00
Household income1, n (%)				0.72
<$35,000	56 (70)	31 (74)	25 (66)
$35,000–$75,000	7 (9)	3 (7)	4 (11)
>$75,000	17 (21)	8 (19)	9 (24)
Education1, n (%)				0.15
Less than high school	7 (9)	6 (14)	1 (3)
High school	53 (66)	25 (60)	28 (74)
College	20 (25)	11 (26)	9 (24)

Abbreviations: CMF, cow milk formula; EHF, extensively hydrolyzed formula; WIC, Special Supplemental Nutrition Program for Women, Infants, and Children.

¹Percentages may sum to >100 because of rounding.

Diet patterns

After food categories with a cumulative frequency below 0.50% were collapsed (e.g., seafood was collapsed into poultry) or dropped (e.g., plant-based protein, condiments), a total of 28 food categories remained for analysis (Table 3) [27]. PCA identified 4 diet patterns over the first 1.5 y (Table 4). Each diet pattern, or PC, was derived based on food categories with strong, positive loadings (≥0.2). For example, PC1 has a high positive loading for infant formula only, and PC2 has high positive loadings for typical infant first foods: low-sugar cereal, not fried potatoes, grain mixed dishes, fruit, vegetables, meat mixed dishes, and 100% fruit juice. Across the 4 PCs, the highest mean factor score through 6.5–7.5 mo was for the infant formula pattern (PC1); from 8.5 to 11.5 mo, the first foods pattern (PC2); and from 12.5 to 18.5, was PC3, with high positive loadings for milk, pizza, water, high-sugar cereal and snack/meal bars (Table 5).

TABLE 3.

Food categories used in the diet pattern analysis.

Food category	Description
100% juice	Citrus juice, apple juice, other fruit juice, vegetable juice with 0 g added sugar
Bakery products	Cakes and pies, cookies and brownies, donuts, sweet rolls, pastries
Cereal, low sugar	Ready-to-eat and cooked (oatmeal, grits, and other cooked cereals) with <21.2 g of sugar/100 g food
Cereal, high sugar	Ready-to-eat and cooked (oatmeal, grits, and other cooked cereals) with >21.2 g of sugar/100 g food, cereal bars, nutrition bars
Cheese	Cheese, cottage cheese, ricotta cheese
Crackers and savory snacks	Crackers, saltines, potato chips, tortilla/corn/other chips, popcorn, pretzels/snack mix
Cured meats and poultry	Cold cuts, cured meats, bacon, frankfurters, sausages
Eggs	Eggs, omelets
Fats and oils	Butter and animal fats, margarine, cream cheese, sour cream, whipped cream, cream, cream substitutes, mayonnaise, salad dressings, vegetable oils
Fruit	Apples, bananas, grapes, peaches and nectarines, strawberries, blueberries and other berries, citrus fruits, melons, dried fruits, other fruits and fruit salads, pears, pineapple, mango and papaya
Grains	White or whole-wheat cooked grains (rice, pasta, noodles, grains), yeast breads, rolls and buns, bagels and English muffins, tortillas, soups and chili
Infant formula	Ready-to-feed, prepared from powder, prepared from concentrate
Meats	Beef, ground beef, pork, lamb, goat, game, liver, organ meats
Milk and cow milk substitutes	Milk (whole, reduced fat, low fat, nonfat), flavored milk (whole, reduced fat, low fat, nonfat), milk substitutes
Mixed dishes – grain and Asian	Rice mixed dishes, pasta mixed dishes, macaroni and cheese, nondessert pot pies and turnovers, fried rice and lo/chow mein, stir-fry and soy-based sauce mixtures, egg rolls, dumplings, sushi
Mixed dishes – meat, poultry, seafood, and Mexican	Meat mixed dishes, poultry mixed dishes, seafood mixed dishes, burritos and tacos, nachos, other Mexican mixed dishes
Mixed dishes – pizza	Pizza
Mixed dishes – sandwiches	Burgers, frankfurter sandwiches, poultry sandwiches, fried poultry sandwiches, egg and breakfast sandwiches, other sandwiches, cheese sandwiches, peanut butter and jelly sandwiches, seafood sandwiches
Potatoes, fried1	Fried white or sweet potatoes
Potatoes, not fried1	White potatoes (baked or broiled), sweet potatoes (baked or broiled), mashed potatoes (white or sweet), potato mixtures (white or sweet)
Poultry and seafood, fried	Chicken patties, nuggets, tenders, fish, shellfish
Poultry and seafood, not fried	Chicken, turkey, duck, fish, shellfish
Quick breads	Biscuits, muffins, quick breads, pancakes, waffles, French toast
Sugars	Sugars and honey, jams, syrups, toppings
Sweetened drinks	Soft drinks, fruit drinks with >0 g added sugar, sport and energy drinks, nutritional beverages
Vegetables	Tomatoes, carrots, other red and orange vegetables, broccoli, spinach, lettuce and lettuce salads, other dark green vegetables, string beans, cabbage, onions, corn, other starchy vegetables, other vegetables and combinations, fried vegetables, coleslaw, nonlettuce salads, vegetables on a sandwich
Water	Tap water, bottled water, flavored or carbonated water, enhanced or fortified water
Yogurt	Yogurt (regular, Greek)

Food categories were established primarily based on the 2017–2018 What We Eat in America (WWEIA) adult food subcategories [17]. Baby foods and beverages were assigned the appropriate adult food categories. Food categories with a cumulative frequency <0.50% that could not be collapsed into a nutritionally similar category were removed from analysis: milk shakes and other dairy drinks, plant-based protein foods, candy, other desserts, diet beverages, coffee and tea, alcoholic beverages, condiments and sauces, sugar substitutes, human milk.

¹These 2 categories were modified from the WWEIA subcategories comprising white potatoes only, to provide increased specificity of food type consumed.

TABLE 4.

Food category loadings for the 4 diet patterns identified in diet record data (n = 80).

Food category	PC1 Infant formula	PC2 Low-sugar cereal, not fried potatoes, grain mixed dishes, fruit, vegetables, meat mixed dishes, 100% fruit juice	PC3 Milk, pizza, water, high-sugar cereal and snack/meal bars	PC4 Fried poultry, bakery products, fried potatoes, starchy vegetables
Infant formula	0.311	−0.09	−0.18	−0.06
Fruit juice (100% juice)	−0.2	0.221	−0.03	−0.25
Bakery products	−0.17	−0.11	−0.12	0.401
Cereal, high sugar, and snack/meal bars	−0.19	−0.12	0.241	−0.03
Cereal, low sugar	−0.08	0.471	−0.18	0.08
Cheese	−0.17	−0.13	0.08	−0.15
Crackers and savory snacks	−0.16	−0.04	−0.1	−0.10
Cured meats and cured poultry	−0.23	−0.12	−0.22	−0.21
Eggs	−0.22	−0.05	−0.16	−0.34
Fats and oils	−0.12	−0.09	−0.24	−0.07
Fried poultry	−0.22	−0.17	−0.02	0.441
Fruit	−0.18	0.281	0.16	0.15
Grains, soups, and chili	−0.2	−0.08	−0.02	−0.03
Meats	−0.15	0.10	0.12	−0.16
Milk and cow milk substitutes	−0.24	−0.05	0.371	−0.02
Mixed dishes – meat, poultry, seafood based	−0.04	0.241	0.16	0.10
Mixed dishes – grain based	−0.17	0.381	−0.06	0.04
Mixed dishes – sandwiches	−0.26	−0.16	−0.08	−0.05
Pizza	−0.17	−0.10	0.321	−0.13
Potatoes, fried	−0.14	−0.11	−0.28	0.311
Potatoes, not fried	−0.09	0.391	−0.27	−0.11
Poultry, seafood	−0.23	−0.01	−0.12	0.10
Quick breads	−0.24	−0.17	−0.22	0.00
Sugars	−0.16	−0.06	−0.18	−0.32
Sweetened drinks	−0.23	−0.08	−0.08	0.16
Vegetables	−0.11	0.261	−0.07	0.211
Water	−0.24	0.00	0.311	0.06
Yogurt	−0.12	0.18	0.19	−0.07

Abbreviation: PC, principal component.

Positive loadings indicate items present in the diet pattern; negative loadings, items less abundant.

¹Denotes principal component analysis loading ≥0.20, indicating the absolute loadings highly correlated within a diet pattern.

TABLE 5.

Mean factor score (±SE) of each diet pattern, by age interval (n = 80).

Infant age interval (mo)	Diet pattern				Percent energy from infant formula
	PC1 Infant formula	PC2 Low-sugar cereal, not fried potatoes, grain mixed dishes, fruit, vegetables, meat mixed dishes, 100% fruit juice	PC3 Milk, pizza, water, high-sugar cereal and snack/meal bars	PC4 Fried poultry, bakery products, fried potatoes, starchy vegetables
0.5	1.7 ± 0.03	−0.6 ± 0.01	−0.1 ± 0.02	−0.1 ± 0.01	100
0.75	1.6 ± 0.02	−0.5 ± 0.01	0.0 ± 0.01	−0.1 ± 0.00	100
1.5–2.5	1.7 ± 0.03	−0.5 ± 0.02	−0.1 ± 0.02	−0.1 ± 0.01	99.3
3.5	1.5 ± 0.02	−0.5 ± 0.03	0.0 ± 0.02	−0.1 ± 0.01	99.0
4.5–5.5	1.4 ± 0.04	−0.1 ± 0.07	−0.0 ± 0.03	0.0 ± 0.03	93.5
6.5–7.5	0.6 ± 0.09	1.0 ± 0.16	−0.2 ± 0.12	0.4 ± 0.10	80.0
8.5–9.5	−0.3 ± 0.19	1.2 ± 0.23	−0.3 ± 0.13	0.5 ± 0.20	66.9
10.5–11.5	−1.5 ± 0.24	0.6 ± 0.32	−0.3 ± 0.22	−0.2 ± 0.21	52.2
12.5	−2.7 ± 0.21	0.1 ± 0.19	0.4 ± 0.22	−0.3 ± 0.19	33.3
18.5	−4.5 ± 0.25	−0.5 ± 0.19	0.6 ± 0.27	0.2 ± 0.25	0.9

Abbreviation: PC, principal component.

Rapid weight gain, diet patterns, and infant weight status

Six models were analyzed to examine the main, and where pertinent, interactive effects of infant formula type [EHF compared with CMF (referent), RWG status, PC scores, time, and potential covariates on WLZ across time] (Table 6). Model 1 examined main and interactive effects of infant formula type and time on WLZ score. There was a decrease of −0.04 [95% confidence interval (CI): −0.46, 0.38] in WLZ scores for infants on EHF compared with CMF, and consistent with prior RCT findings [9,10,24], there was a significant formula type × time interaction. In Model 2, adding RWG category (as a main effect and a 3-way interactive effect of infant formula type × RWG category × time interaction) revealed no significant main effect of formula type, but significant main effects for RWG category and the 3-way interaction. The main effect for RWG showed a decrease of −1.15 (95% CI: −1.75, −0.55) in WLZ score in those with RWG compared with non-RWG.

TABLE 6.

Results from generalized estimated equation models on the influence of formula composition, weight gain velocity, and diet patterns on trajectories of infant weight-for-length z-scores across the first 18.5 mo.

Variables1	Model 1	Model 2	Model 3	Model 4A	Model 4B	Model 5A	Model 5B
				CMF group	EHF group	CMF group	EHF group
Formula group, CMF2	−0.04 (−0.46, 0.38)	0.12 (−0.38, 0.61)	0.13 (−0.36, 0.62)	—	—	—	—
Time3	171.834	327.754	160.204	104.124	110.084	102.654	86.774
Formula group × time3	24.305	29.845	31.395	—	—	—	—
Rapid weight gain group, rapid2	—	−1.15 (−1.75, −0.55)	−1.15 (−1.75, −0.56)	−0.87 (−1.36, −0.38)	−1.15 (−1.77, −0.53)	−0.88 (−1.34, −0.38)	−1.16 (−1.85, −0.47)
Rapid weight gain group × formula group2	—	0.29 (−0.50, 1.09)	0.29 (−0.50, 1.07)	—	—	—	—
Rapid weight gain group × time3	—	295.954	300.293	127.544	172.384	128.654	140.514
Rapid weight gain group × formula group × time3	—	17.636	17.956	—	—	—	—
PC12	—	—	0.15 (0.05, 0.25)	0.20 (0.06, 0.34)	0.10 (−0.03, 0.23)	0.19 (0.05, 0.33)	0.08 (−0.05, 0.21)
PC22	—	—	0.04 (−0.01, 0.09)	—	—	—	—
PC32	—	—	−0.04 (−0.09, 0.01)	—	—	—	—
PC42	—	—	−0.01 (−0.05, 0.04)	—	—	—	—
Infant sex, male2	—	—	—	—	—	−0.03 (−0.39, 0.33)	0.11 (−0.39, 0.61)
Infant race2	—	—	—	—	—
Black or African American						−0.45 (−1.04, 0.14)	0.05 (−1.07, 1.17)
Other or 2+ races						−0.51 (−1.31, 0.29)	0.14 (−0.85, 1.12)
Infant age of weaning2	—	—	—	—	—	0.07 (−0.03, 0.17)	0.04 (−0.16, 0.24)
Multiparous2	—	—	—	—	—	0.28 (−0.20, 0.76)	−0.24 (−0.92, 0.45)
Household income2	—	—	—	—	—
<$35,000						0.07 (−0.60, 0.72)	−0.29 (−1.37, 0.78)
$35,000–75,000						0.23 (−0.53, 1.00)	−0.23 (−1.32, 0.86)
Maternal prepregnancy BMI2	—	—	—	—	—	0.01 (−0.01, 0.03)	0.01 (−0.02, 0.04)

Abbreviations: CMF, cow milk formula; EHF, extensively hydrolyzed formula; PC, principal component.

Symbols: —, excluded from model.

¹The reference group for each categorical variable are as follows: Formula group, reference EHF; Rapid weight gain group, reference nonrapid; Infant sex, reference female; Infant race, reference White; Multiparous, reference primiparous; Household income, reference >$75K.

²Values are beta coefficients (95% CI).

³Values are χ² test statistics.

⁴P ≤ 0.001.

⁵P ≤ 0.01.

⁶P ≤ 0.05.

Next complementary feeding PCs scores were added to the models. Model 3, which added all PC scores to Model 2, showed a main effect for PC1, but not PC2, PC3, or PC4, with an increased in WLZ score of 0.15 (95% CI: 0.05, 0.25) with increasing PCI score. Model 4, which included the main and interactive effects of RWG category and time with PC1 only added as a main effect, was run separately for each randomized formula group. The RWG category × time interaction and the main effect of PC1 were significant for the CMF group (Model 4A). For EHF-fed infants, the RWG category × time interaction was significant but the main effect of PC1 was not (Model 4B). The final model (Model 5), which incorporated all potential covariates (i.e., infant sex, infant race, age at weaning, household income, parity, prepregnancy maternal BMI) into Model 4, yielded findings that were consistent with those of Model 4. That is, the RWG category × time interaction and the main effect of PC1 were significant for the CMF group but only the RWG category × time interaction remained significant for the EHF group. These findings are illustrated in Figure 2. The difference in WLZ score over time, with and without PC1 (the infant formula pattern), in the model was significant for CMF-fed group but not for EHF-fed group.

FIGURE 2.

Visualization of the generalized estimated equation models on the influence of formula composition, weight gain velocity, and diet patterns on trajectories of infant weight-for-length z-scores across the first 18.5 mo (Table 6). Predicted WLZ scores were based on the main and interactive effects of RWG and time, stratified by formula group (A, CMF; B, EHF), without PC in the model (Model 4, hollow symbol, broken line) and with PC1 in the model (Model 5, solid symbol, solid line). The difference in WLZ score trajectory with and without PC 1 in the model was significant for CMF-fed infants (χ² = 17.95; P < 0.05) but not for EHF-fed infants (χ² = 6.51; P = 0.70). The final model (Model 5), which incorporated all potential covariates (i.e., infant sex, infant race, age at weaning, household income, parity, prepregnancy maternal BMI) into Model 4, yielded findings that were consistent with those of Model 4. That is, the RWG category × time interaction and the main effect of PC1 were significant for the CMF group but only the RWG category × time interaction remained significant for the EHF group. CMF, cow milk formula; EHF, extensively hydrolyzed formula; RWG, rapid weight gain; WLZ, weight-for-length z-scores.

Discussion

Among a cohort of formula-fed infants, longitudinally followed from 0.5 to 18.5 mo, we examined infant weight status, assessed as WLZ score, during the transition from an exclusive formula-based diet to a more varied complementary diet. We identified 4 diet patterns: PC1, infant formula; PC2, a first foods pattern; PC3, milk, pizza, water, high-sugar cereal, and snack/meal bars; and PC4, fried poultry, bakery products, fried potatoes, and starchy vegetables. As expected, the mean PC1 (factor) score was highest through 6.5–7.5 mo of age, indicating the predominance of infant formula in the diet that aligns with recommendation to feed exclusively human milk and/or infant formula for the first 6 mo of life. At 8.5–9.5 mo, the mean score was highest for the “first foods” pattern, consisting primarily of low-sugar cereal, nonfried potatoes, grain mixed dishes, fruit, vegetables, meat mixed dishes, and 100% fruit juice, although infant formula still contributed >65% of total energy. By 12.5 mo of age, the highest mean score was for PC3, which included high-sugar cereals and more snacks.

The diet patterns identified in this study of formula-fed infants support findings from previous research reporting both healthy and less healthy food consumption patterns in young children [[25], [26], [27], [28], [29], [30], [31]]. For example, among United States infants and toddlers, healthy patterns were characterized by consumption of fruit and vegetables in one study [26,27] and by fruits, vegetables, whole grains, and seafood in another [25]. Less healthy patterns have been characterized by consumption of sweetened fruit juice, sweet drinks, and other energy-dense foods (e.g., French fries, sweet foods) in several studies [21,26] and by consumption of fried meats and other fried products, snacks, and sweetened milk in another [24]. In studies of Swedish, Norwegian, and Australian infants and toddlers, healthy patterns reported in the literature include “health conscious”[28], “fruits and vegetables”[29]. “family diet pattern”[16,32], “wholesome” [30], “diverse plant and animal foods” [31], “Mediterranean-like diet” [31], and “healthy” [16]. Less healthy foods have also been reported in these same studies to include, such diet patterns as “low variety” [27,33], “high energy density,” “cow’s milk and discretionary combination,” “unhealthy” [25], “Western-like” [28], “palatable energy-dense foods” [31], and “dairy/sugar-sweetened beverages” [25]. Although the names and specific foods within the healthy and less healthy dietary patterns may vary, a consistent finding is that less healthful table food patterns emerge early in life [[25], [26], [27], [28], [29], [30], [31]]. Although infant formula supplies adequate concentrations of iron and zinc, the Dietary Guidelines for Americans (DGA) recommends introducing infants to iron- and zinc-rich foods (e.g., meats seafood) around 6 mo. This prepares them for the transition at 12 mo, when infant formula is no longer recommended. In this cohort, neither meat nor seafood was present in any of the diet patterns identified. This finding aligns with NHANES data showing that by 12 mo, ready-to-eat cereals contribute nearly 20% of iron intake, whereas milk provides nearly 30% of zinc intake [34]. Although fortified ready-to-eat cereals can help toddlers achieve adequate iron and zinc intakes, they also contribute ∼7% of added sugars to the toddler diet [35].

Notably, our analyses revealed that a higher factor score for the infant formula diet pattern (PC1) was significantly associated with higher WLZ score independent of the age of weaning. One possible explanation for this observation is that many first foods or stage 1 baby foods (e.g., carrots, green beans, sweet potatoes, apples) have a lower energy density than infant formula, which could lead to lower energy intake and thus lower WLZ score compared with those with a higher PC1 score (higher energy intake from infant formula). As the infant grows older, the diet shifts toward higher-energy-dense table (noncommercial baby) foods (e.g., pizza, high-sugar cereal, fried poultry, bakery products). Although such foods might typically be expected to contribute to increased infant weight status, our findings in the original cohort revealed no associations between toddler weight status and energy intake from snacks and sweets, sweetened beverages, or fruit [24]. Furthermore, research suggests that energy intake in infants may decrease when meals are more energy dense [36]. Taken together, these findings suggest a means by which infants fed CMF with greater formula intake during the first year may be at increased risk for higher weight status.

An additional mechanism to consider is RWG and its potential programming effects. Infancy is a sensitive period during which early-life nutrition and weight gain can have lasting impacts on metabolism and overall health. Prospective birth cohort studies have consistently shown that RWG during infancy is an independent risk factor for overweight and obesity in both childhood and adulthood [5,21,37,38]. Often, a high WLZ score reached in childhood persists through subsequent years [[39], [40], [41]]. Because CMFs are the most frequently consumed infant formulas in the United States [8], and CMF is recognized as resulting in faster weight gain [9,10], it may be time to consider alternate formula types that support more normative weight gain during early infancy.

Irrespective of any programming effect that an early milk-based diet may have on weight gain trajectories, the diet patterns that emerge remain concerning from a diet quality perspective. Toward the end of the first year, foods that the DGAs [42] recommends in only limited amounts (e.g., high-sugar cereals, bakery products, fried foods) emerge and become predominant, and they persist into the second year of life [35,43]. Because infancy is a sensitive period when flavor preferences form [44,45], early introduction of sweet and salty foods may lead children to prefer such foods later in life [46]. The emergence of less desirable diet patterns before 12 mo of age underscores a need for early recommendations and interventions for healthy diets. Healthcare providers should focus on anticipatory guidance around the time of complementary food introduction and educate parents on continued consumption of nutrient-dense foods and limiting foods high in added sugar, sodium, and saturated fat.

A strength of this study is that it utilizes data from an RCT of exclusively formula-fed infants, with repeated dietary intake measures collected from 0.5 to 18.5 mo of age, allowing capture of dietary transitions. The study population consisted of a diverse and contemporary cohort of formula-fed infants that are representative of the racial/ethnic diversity of Philadelphia (58% Black, 19% other or >1 race), where the RCT was conducted. Another strength is the use of the standardized WWEIA adult food categories to identify diet patterns, which allows others to replicate this research using the same food categories in different populations. A limitation of this study is the use of self-reported diet data, which is subject to error (e.g., under- or overestimating the quantity of food that is actually consumed compared with offered) and proxy reporting bias [47]. Additionally, although PCA is a data-driven approach, several steps of the analytical process are subjective, including the naming of diet patterns; however, as suggested by the general healthy compared with less healthy categories identified in previous studies, we anticipate replication of methods and comparison of findings will show similar groupings across studies among United States infants.

Looking ahead, the prevention of RWG in infancy represents a modifiable target for reducing the risk of overweight and obesity later in life. Future research is warranted to examine the impact of alternative protein sources in infant formula, such as partially hydrolyzed protein, on supporting more normative patterns of weight gain. Additionally, within the context of complementary feeding, there is a need to identify and implement strategies that promote the consumption of healthy, nutrient-dense foods while minimizing the intake of nutrient-poor, highly processed foods during infancy and toddlerhood.

In conclusion, both the composition of infant formula and the rate of weight gain during infancy can influence later weight status. In this cohort of healthy, term infants randomly assigned to receive a standard CMF or EHF, it was not the timing of weaning, but the prolonged feeding of CMF, during the period when complementary foods were introduced, that was associated with higher WLZ in the first 18.5 mo. Moreover, the complementary diet during the first year was characterized by foods high in added sugars and fats, highlighting the importance of establishing healthy eating patterns in the first year.

Author contributions

The authors’ responsibilities were as follows – JCT, JAM: conceived and conducted the original RCT; all authors: contributed to the data analysis plan, data interpretation, and writing of the manuscript; and all authors: reviewed and approved the final manuscript.

Data availability

Data described in the manuscript, code book, and analytic code will be made available upon request pending application and approval.

Funding

This work was supported by NIH grants R01HD072307 and R03HD094908 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development awarded to JAM and JCT. The funding agency was not involved in study design and conduct, data analysis and interpretation, or manuscript preparation.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the author(s) used ChatGPT to improve clarity of select statements. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.

Conflict of interest

JCT and JAM report financial support was provided by National Institutes of Health. The other authors report no conflicts of interest.