Specific Infant Feeding Practices Do Not Consistently Explain Variation in Anthropometry at Age 1 Year in Urban United States, Mexico, and China Cohorts^,

Abstract

Infant feeding practices generally influence infant growth, but it is unclear how introduction of specific foods affects growth across global populations. We studied 3 urban populations in the Global Exploration of Human Milk study to determine the association between infant feeding and anthropometry at 1 y of age. Three hundred sixty-five breastfeeding mother-infant pairs (120 US, 120 China, and 125 Mexico) were recruited soon after the infant’s birth. Enrollment required agreement to breastfeed ≥75% for at least 3 mo. Weekly, 24-h, food frequency data were conducted on infants for 1 y and exclusive breastfeeding (EBF) duration and timing of specific complementary food introduction were calculated. Weight and length were measured at age 1 y and anthropometry Z-scores calculated using WHO standards. Cohorts in the 3 urban populations (Shanghai, China; Cincinnati, USA; and Mexico City, Mexico) differed by median EBF duration (5, 14, and 7 wk, respectively; P < 0.001), timing of introduction of meat/eggs/legumes (4.8, 9.3, and 7.0 mo, respectively; P < 0.0001), and other feeding practices. By age 1 y, infants in Shanghai were heavier and longer than Cincinnati and Mexico City infants (P < 0.001). Adjusting for nonfeeding covariates, the only feeding variable associated with anthropometry was EBF duration, which was modestly inversely associated with weight-for-age but not length-for-age or BMI Z-scores at 1 y. Although feeding variables differed by cohort, their impact on anthropometry differences was not consistent among cohorts. Overall, across these urban, international, breast-fed cohorts, differences in specific feeding practices did not explain the significant variation in anthropometry.

Introduction

Although many developing countries continue to struggle with infant malnutrition and growth faltering, childhood obesity is an increasing concern worldwide. In developed nations in North America and Western Europe, childhood obesity is increasingly prevalent (1). Countries undergoing rapid dietary transitions from traditional to Western diets, including Mexico and China, are simultaneously experiencing increases in early-onset obesity (2–4) coupled with the persistent issues of growth faltering (5, 6) and malnutrition (7). The role of infant feeding in this context is important to understand potential ways to prevent obesity while maintaining sufficient growth in infancy.

Several studies suggest that the early introduction of any solid food (typically prior to 4 mo of age) is associated with later obesity in childhood (8) and throughout life (9), but some studies refute this (10, 11) and a recent systematic review suggested the overall relationship is weak (12). In addition, a longer duration of breastfeeding is associated with later introduction of solid food, potentially confounding this effect. The WHO recommends infants be exclusively breastfed for 6 mo with introduction of complementary foods after that point and continued breastfeeding until age 2 y (13, 14); breastfeeding itself has been associated with slower weight gain during the first year of life compared with formula feeding (11, 15). Attempting to disentangle these effects, previous studies have noted that among breast-fed infants, timing of introduction of any solid food may not affect obesity (16) in contrast to formula-fed infants (8) or those breastfed for a short duration (17). This may be related to total energy intake, because the early introduction of solid food tends to replace human milk in the diet (18) while it adds to intake among formula-fed infants (19).

The timing of the introduction of specific food groups may, however, affect growth among breast-fed infants. Human milk contains low amounts of protein and a lower protein infant formula appears to mimic the growth effects of breastfeeding (20), suggesting relationships between protein intake and growth in infancy. Indeed, there is increasing consensus around the health and growth benefits of meat as a first complementary food (21). The progression of food introduction to infants differs across global populations and may affect growth patterns. The WHO Multicentre Growth Reference Study (MGRS)⁸ reported wide variation in complementary food introduction patterns among worldwide populations (22). However, this study did not evaluate the role of complementary food introduction patterns on infant growth.

The Global Exploration of Human Milk (GEHM) study was conducted in the United States (Cincinnati, OH), Mexico (Mexico City), and China (Shanghai) to examine growth and health outcomes in diverse, urban cohorts of predominantly breastfeeding mothers and infants from birth to 2 y of age in relation to human milk composition and complementary feeding. In this first report, we examine the differences in maternal and infant characteristics and specific infant feeding practices across the sites as determinants of infant anthropometry by age 1 y within the context of breastfeeding. We hypothesized that differences in the timing of solid food introduction and the progression of specific foods introduced would significantly affect infant anthropometry by age 1 y.

Methods

The GEHM Study is a multi-country, prospective cohort study of 365 breastfeeding mothers and infants recruited between January 2007 and December 2008, following the same study protocol, unless otherwise noted. Recruitment at all sites focused on a single, large birth hospital, with additional community-based recruitment in Cincinnati and Mexico City. All cohorts represent urban populations, with mothers in Mexico recruited from a lower income area than mothers in Cincinnati and Shanghai to achieve socioeconomic and lifestyle diversity across sites while representing local populations. None of the mothers in Mexico City participated in the Nutrisano supplemental food program and no mothers in Shanghai participated if any supplemental food program, while 31 mothers in Cincinnati were enrolled in the Women’s, Infants’ and Children’s supplemental food program after the infant’s birth. The Women’s, Infants’ and Children’s supplemental food program provides support for appropriate infant nutrition but does not offer specialized foods or supplements not otherwise available. Eligibility was determined in 2 phases. The initial eligibility criteria, evaluated when the infant was ∼2 wk old, included mothers aged 18–49 y intending to breastfeed at least 75% for at least 3 mo and without medical issues that would interfere with breastfeeding. Mothers in Cincinnati also had to reside within 25 miles of Cincinnati Children’s Hospital. Infants were included if they were singletons born at ≥37 wk of gestation, with a birth weight ≥2500 g and without medical issues that would interfere with breastfeeding. Final eligibility was determined at the 4-wk study visit, when mothers not achieving the 75% breastfeeding goal by 4 wk postpartum were excluded and replaced in the cohorts to achieve recruitment goals. All mothers provided written informed consent and this study was approved by the Institutional Review Boards of Cincinnati Children’s Hospital Medical Center, the National Institute of Medical Sciences and Nutrition in Mexico City, and Shanghai Children’s Hospital of Fudan University.

Study procedures.

The 2-y prospective study included eligibility and baseline questionnaires, 5 in-person visits, and weekly telephone surveillance in y 1. The data for this analysis focused on infant feeding data from the weekly interviews and anthropometric assessments completed at the 1-y (52 wk) study visit. Pregnancy characteristics, including prepregnancy weight, gestational diabetes, and gestational weight gain, were assessed by maternal recall at the baseline visit.

Infant’s weight, length, head circumference (HC), and mid-upper arm circumference (AC) were assessed at every in-person visit. All sites used the same equipment and were trained by one of the coauthors (B.S.D.), with direct observation of methods. Infant weight was measured in duplicate ± 0.1 kg (Baby Checker Scale, Medela) with the infant unclothed; a third measurement was taken if measurements differed by ≥0.3 kg. Supine length, head, and arm circumferences were measured in duplicate ± 0.1 cm; if measurements differed by ≥0.3 cm, a third measurement was taken. Supine length was measured using an infant length board with a fixed headboard and a moving footboard (Folding Lightweight Measuring Board, Hopkins Medical Products). HC was measured using a nonstretching flexible tape, taking the largest occipitofrontal diameter of the infant’s head. AC was taken at the midpoint between the infant’s right shoulder and elbow using a nonstretching flexible tape. Weight and length measurements at age 1 y were used in this analysis. BMI was calculated as weight (kg)/length (m)². Age- and sex-specific Z-scores for BMI (BMIZ), weight-for-age (WAZ), length-for-age (LAZ), weight-for-length, HC-for-age, and AC-for-age Z-score were calculated using the 2005 WHO infant growth standards (23).

Assessment of infant feeding.

Between 2 and 52 wk of age, weekly phone surveillance assessed current breastfeeding status and problems, a 24-h food frequency recall of infant feeding, maternal and infant health updates, and maternal work and infant daycare arrangements. All mothers had working telephones and provided an optimal day and time for the weekly call. Up to 3 attempts were made to contact mothers each week before that weekly interview was considered missing. The 24-h food frequency recall recorded the number of times the infant had been fed food or liquids classified into 21 specific categories [e.g., breast milk, infant formula, whole milk, other milk (specify), water, tea, rice water, cereal or porridge (atole in the Mexico questionnaire), soft drink, juice, baby foods, vegetable, fruit, egg, chicken or turkey, red meat, bread/cookie/cracker/tortilla, oral rehydration solution, yogurt or probiotic, fermented milk, and other (specify)] but did not assess portion size or nutritional value of intake. The Mexico questionnaire did not include questions on baby food or other milks and included separate questions for fruit and vegetable juice. All foods listed in the “other” category were translated into English (as necessary) and classified into one of the existing categories (e.g., fish and shrimp in China were classified with chicken or turkey). Mixed foods (e.g., pizza) were classified using their dominant component (e.g., bread). Baby food versions of single foods (e.g., squash) were classified into the corresponding food category (e.g., vegetable). Breastfeeding intensity was determined during the first year of follow-up as the number of times the infant had been fed human milk divided by the total number of items fed (multiplied by 100), with 100% indicating exclusive breastfeeding (EBF). Provision of oral rehydration solution was extremely rare (<1% of all weekly reports) and did not disqualify any infants under our definition of EBF.

EBF duration was defined as the infant’s age at the last report of EBF (24). Any breastfeeding (ABF) duration was defined as the infant’s age when the mother reported breastfeeding cessation, which was cross-validated with the last 24-h recall reporting human milk feeding. Mothers continuing to breastfeed into the second year of life were identified and asked to provide a weaning date at the time of their 2-y visit or contacted specifically to obtain this information. Mothers still breastfeeding at the 2-y visit had an unknown end date for ABF and data up to the last study visit was included.

Introduction of solid food was defined at the first reported intake of any solid or semi-solid food. The 20 food and liquid categories assessed besides breast milk were combined into 9 groups: infant formula, nonhuman milk (e.g., cow or soy milk), nonmilk dairy foods (e.g., yogurt, fermented milk, cheese), non-nutritive drinks (e.g., water, oral rehydration solution, tea), fruit/vegetable juice, soft drinks, fruits or vegetables, high-protein foods (e.g., egg, meat, fish, legumes), cereals (e.g., bread, tortillas, rice, cereal, pasta), and sweet or salty snacks (e.g., cake, cookies, chips). Food and liquid introduction (other than breast milk) was considered in the first year only, because weekly surveillance was discontinued after the 1-y visit (95% of 1-y visits completed by 56 wk of age). Feeding was reported in weeks of age or converted to months of age as 4.3 wk/mo.

To assess breastfeeding intention, mothers were asked eligibility questions (how many months they intended to breastfeed their child, whether they intended to feed any formula before 3 mo of age, and the percent they expected to feed them formula daily by 1 mo of age). In addition, the baseline questionnaire, administered at 2 wk of age, asked them to indicate the mix of human milk and formula that they expected to feed their infants at 3 and 6 mo (5 choices ranging from all breast milk to all formula).

Statistical analysis.

Data were analyzed using SAS v.9.3 (SAS Institute). Descriptive data were compared among sites using omnibus ANOVA or χ² tests for continuous or discrete variables, respectively, followed by post hoc pairwise comparisons. The distribution of maternal intended mix of milk feedings at 3 and 6 mo among sites was assessed using χ² tests with 4 (3 sites by 3 responses used) and 8 (3 sites by 5 responses used) df, respectively.

The duration of breastfeeding and timing of the introduction of any solid food were analyzed using log-rank survival analysis, followed by post hoc pairwise comparisons. The median age at introduction of specific food components was compared using an omnibus Kruskal-Wallis test due to non-normality of the timing of introduction of foods, followed by post hoc pairwise comparisons. Mothers not introducing specific foods by their 1-y study visit or loss to follow-up were not included in the calculation of median food introduction or percent of mothers introducing a food category. Sensitivity analyses limited to participants with complete data to age 1 y demonstrated similar results. Mean anthropometry Z-scores at 52 wk were compared using omnibus ANOVA tests, followed by post hoc pairwise comparisons. For all analyses, omnibus P ≤ 0.05 was considered significant, with a Bonferroni-adjusted P ≤ 0.017 significant for pairwise comparisons among cohorts.

Analyses of infant feeding characteristics with infant anthropometry Z-scores at 1 y were conducted using linear regression applying a staged modeling approach. First, infant anthropometry outcomes were regressed on infant feeding practices. Feeding variables of interest included timing of introduction of each of the food groups except groups not frequently introduced (e.g., formula use, nonmilk dairy, nonhuman milk, soft drinks, sweet/salty snacks), which were modeled as ever vs. never introduced within the first year. Feeding variables were individually tested, adjusting only for cohort site, and included in a multivariable model if P ≤ 0.20. The best final set of feeding variables, adjusting only for cohort site differences, was determined using lowest model Akaike Information Criterion.

Next, nonfeeding covariates were separately modeled to determine a reduced set of covariates to add as a group to the feeding model. The variables considered for model inclusion were: maternal age at delivery, maternal education level, maternal prepregnancy BMI, gestational weight gain, gestational diabetes, type of delivery, infant birth weight, and infant sex, with cohort site automatically included. The multivariable covariate set was selected using the lowest model Akaike Information Criterion. Finally, both feeding and nonfeeding models were combined and nonfeeding covariates trimmed using backward selection to eliminate nonsignificant (P > 0.05) terms.

Further analyses were conducted to determine whether divergent feeding practices among sites could account for divergent 1-y anthropometry. For these analyses, models were constructed for each pairwise difference among cohorts, using the variables from the above combined model. Pairwise cohort differences for each anthropometry outcome were first estimated without other variables. The cohort differences were then evaluated after the addition of either feeding variables alone, nonfeeding variables alone, or all variables from the final models. The β estimate for cohort differences was examined for significant (>10%) change in magnitude, indicating partial or complete mediation by the set of variables in the model.

Results

A total of 365 infants were recruited into the study (Cincinnati and Shanghai, n = 120; Mexico, n = 125). By 1 y of age, the cohorts included 92 (77%) mother-infant pairs in Cincinnati, 104 (87%) in Shanghai, and 89 (74%) in Mexico City. Those remaining in the study did not differ from those dropping out, with the exception that mothers dropping out of the Cincinnati site were less likely to be exclusively breastfeeding at 4 wk (P < 0.001) and were less likely to have intended to feed “all breast milk” at 3 mo (P = 0.04) and 6 mo (P = 0.02) postpartum. Despite consistent eligibility and recruitment methods, the cohorts significantly differed on most baseline characteristics (Table 1), as anticipated. With regard to anthropometry, Cincinnati mothers had a higher obesity prevalence (28%) than either Mexico City (4.5%) or Shanghai (0.8%; omnibus P < 0.0001) mothers. However, Shanghai mothers reported a greater gestational weight gain compared with the other cohorts (P < 0.001). Infant birth weight was higher in Cincinnati and lower in Mexico compared with Shanghai infants (omnibus P < 0.001), but infant gender distribution was similar across the sites.

TABLE 1.

Demographic and socioeconomic description of cohorts of infants and their mothers in Shanghai, Cincinnati, and Mexico City at baseline¹

	Shanghai	Cincinnati	Mexico City	Omnibus P value2
n	120	120	118
Household characteristics
Married or common union	118 (98.3)a	100 (83.3)b	91 (77.1)b	<0.001
Maternal education				<0.001
≤High school graduate	22 (18.3)	6 (5.0)	110 (93.2)
Attended or graduated from college	87 (72.5)	94 (78.3)	7 (5.9)
Post-college education	11 (9.2)	20 (16.7)	1 (0.8)
Smoked during pregnancy	1 (0.8)b	3 (2.5)ab	11 (9.3)a	0.003
Household member smoked during pregnancy	45 (37.5)a	14 (11.7)b	58 (49.2)a	<0.001
Employed at time of pregnancy	85 (71.4)a	97 (80.8)a	48 (40.7)b	<0.001
Infant characteristics
Sex (male)	64 (53.3)	55 (45.8)	64 (51.2)	0.49
Birth weight, kg	3.43 ± 0.47b	3.56 ± 0.46a	3.11 ± 0.37c	<0.001
Maternal delivery characteristics
Age at delivery, y	29.3 ± 3.7b	31.5 ± 5.2a	24.4 ± 5.6c	<0.001
Prepregnancy BMI, kg/m2	20.6 ± 2.3c	27.5 ± 6.6a	23.9 ± 3.2b	<0.001
Prepregnancy BMI category				<0.001
Underweight (BMI <18)	11 (9.2)	0 (0.0)	1 (0.9)
Normal (BMI 18–24.9)	103 (86.6)	54 (45.8)	76 (67.9)
Overweight (BMI 25–29.9)	4 (3.4)	31 (26.3)	30 (26.8)
Obese (BMI ≥30)	1 (0.8)	33 (28.0)	5 (4.5)
Gestational weight gain, kg	16.2 ± 5.3a	13.8 ± 5.1b	11.2 ± 5.3c	<0.001
Maternal medical conditions at delivery, % yes
Any conditions indicated	51 (42.5)ab	62 (51.7)a	36 (30.5)b	0.004
Gestational diabetes	4 (3.3)ab	13 (10.8)a	3 (2.5)b	0.009
Asthma	2 (1.7)b	14 (11.7)a	2 (1.7)b	<0.001
Allergies	42 (35.0)a	37 (30.8)a	9 (7.6)b	<0.001
Other diagnosis	8 (6.7)b	24 (20.0)a	25 (21.2)a	0.003
Delivery type, % vaginal	36 (30)c	92 (76.7)a	72 (61.0)b	<0.001
Time infant in hospital after birth, d	4.5 ± 1.6a	2.1 ± 1.0b	1.8 ± 1.2b	<0.001

¹Values are mean ± SD for continuous variables or n (%) for categorical variables, n = 358. Baseline data were available for only 118 mothers in Mexico City. Means or percentages in a row without a common letter differ, P < 0.017.

²Omnibus P values for comparisons across cohorts presented from ANOVA or χ² tests.

Infant feeding.

At the baseline visit, >90% of mothers in each cohort intended to breastfeed “all” or “mostly” breast milk at 3 mo and 80–90% intended to feed all or mostly breast milk at 6 mo of age. Despite these similar intentions (omnibus P > 0.4), the actual breastfeeding durations significantly differed by site (Fig. 1A). The median EBF durations were 4.9 wk in Shanghai (IQR: 1.6, 11.9 wk), 13.7 wk in Cincinnati (IQR: 4.6, 19.3 wk), and 7.1 wk in Mexico City (IQR: 0, 18.0 wk) (omnibus P < 0.001, pairwise comparisons of Cincinnati vs. Shanghai or vs. Mexico City, P < 0.017). The median ABF duration (Fig. 1B) was also shorter in Shanghai (37.1 wk) than in Mexico City (52.0 wk) and Cincinnati (49.5 wk; omnibus P < 0.001; pairwise comparisons of Shanghai vs. Mexico City or vs. Cincinnati, P < 0.017). The median introduction of any solid food (Fig. 1C) was >1 mo earlier in Shanghai (18 wk) than in Cincinnati (23 wk) or Mexico (25 wk; omnibus P < 0.001; pairwise comparisons of Shanghai vs. Mexico City or vs. Cincinnati, P < 0.017).

FIGURE 1.

Duration of EBF (A), ABF (B) and timing of introduction of any solid food (C) for infants in the Cincinnati, Mexico City, and Shanghai cohorts. Noncensored observations are presented as survival curves by Shanghai (n = 120), Cincinnati (n = 120), and Mexico City (n = 125) cohorts. The horizontal reference line at 0.5 represents the median. ABF, any breastfeeding; EBF, exclusive breastfeeding.

The introduction of specific solid foods and beverages markedly differed by site in timing and progression of introduction (Fig. 2). By 1 y of age, most mothers introduced infant formula (99, 73, and 67%, in Shanghai, Cincinnati, and Mexico City, respectively) and fruit or vegetable juice (91, 73, and 100%, respectively). The introduction of soft drinks by age 1 y was not reported in Shanghai and rare in Cincinnati (2%) but was more common in Mexico (40%). The introduction of non-nutritive drinks and juice in Shanghai and Mexico City were significantly earlier than in Cincinnati; introduction of formula was also earlier in Shanghai than in Cincinnati. The median introduction of nonhuman milk occurred at ∼1 y of age in the Cincinnati and Shanghai cohorts but was significantly earlier in Mexico. However, few Shanghai mothers reported introducing nonhuman milk at all in the first year (12%).

FIGURE 2.

Median and IQR of infant age in months of introduction of food and liquid categories to infants in Shanghai, Cincinnati, and Mexico City, and the number and proportion of mothers in each cohort introducing each food or liquid category within the first 56 wk of life. The number in each cohort included in each analysis varies and is indicated as the denominator in the right-most column. *P ≤ 0.05 by omnibus tests; labeled medians within a category without a common letter differ, P < 0.017. ORS, oral rehydration solution.

Within the first year, nearly all infants were introduced to cereals/grains, fruit/vegetables, and meat/eggs/legumes in all cohorts. In all cohorts, the first introduction of cereals/breads and fruits/vegetables generally occurred between 4.3 and 6.7 mo of age, but both were earlier in Shanghai than in Cincinnati or Mexico (omnibus P < 0.0001 for both categories). The introduction of high-protein foods was earlier in Shanghai (4.8 mo) than in Mexico City (7.0 mo) or Cincinnati (9.3 mo; P < 0.0001). As with nonhuman milk, few Chinese mothers reported introducing nonmilk dairy foods (16%) in the first year. Sweet and salty snack foods in the first year were rare in Shanghai (3%) but more common in Cincinnati (34%) and extensive in Mexico (78%). More generally, in Shanghai, nearly all food categories were introduced in quick succession between 4 and 6 mo of age, whereas introductions of foods in the Mexico City and Cincinnati cohorts were widely distributed across the first year of life.

Infant anthropometry at 1 y.

Of the 365 infants recruited, 285 had 1-y anthropometry assessments. Anthropometric Z-scores differed across the 3 cohorts by 1 y of age (all omnibus P ≤ 0.004) (Table 2). Significant differences among all cohorts were evident for WAZ, LAZ, and weight-for-length Z-score, with infants in Shanghai significantly higher than the other cohorts in pairwise comparisons and infants in Mexico significantly lower than other cohorts. The BMIZ and head circumference-for-age Z-score were also significantly higher among the Shanghai infants compared with the other 2 cohorts and AC-for-age Z-scores were significantly larger in Shanghai compared with Cincinnati infants. The prevalence of stunting was rare in both Shanghai (0%) and Cincinnati (4.4%) but occurred in nearly 16% of Mexico City infants (omnibus P < 0.001). Wasting was rare in all cohorts, but overweight was present in 9.6% of infants in China, 2.2% in Cincinnati, and 3.4% in Mexico (omnibus P = 0.05) by age 1 y.

TABLE 2.

Anthropometric characteristics relative to WHO growth charts at age 1 y among infants born in Shanghai, Cincinnati, and Mexico City¹

	Shanghai	Cincinnati	Mexico City	Omnibus P value2
n	104	92	89
WAZ	0.87 ± 0.09a	0.16 ± 0.10b	−0.51 ± 0.10c	<0.001
LAZ	0.54 ± 0.10a	−0.21 ± 0.10b	−1.04 ± 0.11c	<0.001
Prevalence of stunting3	0 (0.0)b	4 (4.4.)b	14 (15.7)a	<0.001
WFLZ	0.85 ± 0.10a	0.35 ± 0.09b	−0.01 ± 0.10c	<0.001
Prevalence of wasting3	0 (0.0)	0 (0.0)	2 (2.3)	0.10
BMIZ	0.78 ± 0.09a	0.38 ± 0.09b	0.13 ± 0.10b	<0.001
Prevalence of overweight3	10 (9.6)	2 (2.2)	3 (3.4)	0.05
HCZ	0.61 ± 0.09a	0.06 ± 0.10b	0.17 ± 0.11b	<0.001
ACZ	0.51 ± 0.09a	0.04 ± 0.12b	0.19 ± 0.10ab	0.004

¹Values are mean ± SEM for continuous variables or n (%) for categorical variables, n = 285. ACZ, arm circumference-for-age Z-score; BMIZ, BMI-for-age Z-score; HCZ, head circumference-for-age Z-score; LAZ, length-for-age Z-score; WAZ, weight-for-age Z-score; WFLZ, weight-for-length Z-score. Means or percentages in a row without a common letter differ, P < 0.017.

²P values from omnibus comparisons across cohort, from ANOVA or χ² tests with 2 df.

³Definition of categorical variables: stunting, LAZ <−2; wasting, WFLZ <−2; overweight, BMIZ >+2.

Associations of infant feeding with 1-y anthropometry.

Feeding variables were individually assessed (adjusting only for site mean differences) for association with WAZ, LAZ, and BMIZ at 1 y of age, with univariate P < 0.20 considered for multivariable analysis. Of all feeding variables tested, lower WAZ was significantly or marginally associated only with longer duration of EBF (P = 0.007), later introduction of high-protein foods (P = 0.06), and later introduction of fruits/vegetables (P = 0.10) (Table 3). Similarly, shorter LAZ was marginally associated with longer duration of EBF (P = 0.08), with later introduction of high-protein foods (P = 0.19), and later introduction of fruits/vegetables (P = 0.12) also meeting the univariate threshold. Lower BMIZ was associated with longer duration of EBF (P = 0.02) and later introduction of high-protein foods (P = 0.15) also meeting the univariate threshold, after adjusting only for mean differences among sites. Sensitivity analyses excluding the China site revealed the same patterns (data not shown).

TABLE 3.

Univariate, partially adjusted, and fully adjusted associations between infant feeding practices and WAZ, LAZ, and BMIZ scores in infants from Shanghai, Cincinnati, and Mexico City Cohorts, combined, at age 1 y¹²

	WAZ			LAZ			BMIZ
Univariate	Feeding model	Full model	Univariate	Feeding model	Full model	Univariate	Feeding model	Full model
n in model	285	285	282	285	285	285	285	285	277
Adjusted R2		0.28	0.36		0.29	0.34		0.08	0.11
	β ± SE
Longer duration: EBF, mo	−0.08 ± 0.03**	−0.07 ± 0.03*	−0.07 ± 0.03*	−0.05 ± 0.03†	−0.05 ± 0.03	−0.05 ± 0.03	−0.07 ± 0.03*	−0.07 ± 0.03*	−0.04 ± 0.03
Later introduction: protein, mo	−0.05 ± 0.03†	−0.04 ± 0.03	−0.04 ± 0.02	−0.04 ± 0.03†			−0.04 ± 0.03†
Later introduction: fruits and vegetables, mo	−0.07 ± 0.04†			−0.07 ± 0.05†			−0.04 ± 0.04
Cohort: Shanghai		1.23 ± 0.15***	1.11 ± 0.16***		1.53 ± 0.15***	1.30 ± 0.15***		0.59 ± 0.14***	0.34 ± 0.15*
Cohort: Cincinnati		0.85 ± 0.16***	0.69 ± 0.17***		0.89 ± 0.15***	0.56 ± 0.16***		0.32 ± 0.14*	0.03 ± 0.16
Mother’s age, y			−0.02 ± 0.01*
Gestational weight gain, kg									0.02 ± 0.01*
Infant sex (male)			−0.36 ± 0.11***						−0.31 ± 0.11**
Birth weight, kg			0.69 ± 0.13***			0.65 ± 0.14***			0.28 ± 0.13*

¹BMIZ, BMI-for-age Z-score; EBF, exclusive breastfeeding; LAZ, length-for-age Z-score; WAZ, weight-for-age Z-score.

^2†P > 0.05 and < 0.20 for univariate models, *P > 0.01 and P ≤ 0.05, **P > 0.005 and P ≤ 0.01, ***P < 0.005.

When univariate feeding associations (P < 0.20) were included in a combined feeding model (all feeding variables plus cohort mean differences), most associations became nonsignificant, with the exception of longer EBF associated with lower WAZ (P < 0.05) and lower BMIZ (P < 0.05). Further adjustment for mother’s age, maternal gestational weight gain, infant sex, birth weight, and mode of delivery in the adjusted feeding models resulted in EBF remaining associated only with WAZ. Cohort-specific analyses did not reveal factors specific to individual sites (data not shown).

Accounting for anthropometry differences among cohorts.

Figure 3 presents changes in pairwise anthropometric differences between cohorts when adjusting for either feeding or nonfeeding sets of variables. Significant, unadjusted, pairwise differences exist between all cohorts for WAZ (Fig. 3A) and LAZ (Fig. 3B) and between the Shanghai site and the other 2 sites for BMIZ (Fig. 3C) at 1 y. When adjusting for only feeding variables, anthropometry differences between Cincinnati and Mexico City infants either increased (WAZ and BMIZ) or did not change (LAZ). When adjusting for nonfeeding covariates, the differences between Cincinnati and Mexico City infants were substantially reduced (by 20, 38, and 122% for WAZ, LAZ, and BMIZ, respectively). By contrast, adjustment for only feeding variables reduced the estimated anthropometric differences between Shanghai and Cincinnati infants by 54, 18, and 36% for WAZ, LAZ, and BMIZ, respectively, whereas adjustment for only nonfeeding variables had little impact on cohort differences. Patterns of change between Shanghai and Mexico City infants are not as striking but typically are reduced more by nonfeeding variables in the models.

FIGURE 3.

Pairwise differences (β estimates ± SE) in WAZ (A), LAZ (B), and BMIZ (C) between Shanghai and Cincinnati (n = 193), Cincinnati and Mexico City (n = 181), and Shanghai and Mexico City (n = 196) cohorts in 4 modeling scenarios: unadjusted, adjusting for feeding variables, adjusting for nonfeeding variables, and in fully adjusted models. Feeding models included the following factors: WAZ: duration of EBF (mo) and timing of introduction of meat/eggs/legumes (mo); LAZ: duration of EBF (mo); BMIZ: duration of EBF (mo). Nonfeeding models included the following factors: WAZ: maternal delivery age (y), infant sex, and infant birthweight (kg); LAZ: infant birthweight (kg); BMIZ: maternal gestational weight gain (kg), infant sex, and infant birthweight (kg). Full models included all factors in both the feeding and nonfeeding models. For all comparisons, the cohort with the lower mean Z score (e.g., Mexico City or Cincinnati) was the reference category. * indicates a β estimate difference of ≥ +10% or ≤ −10% compared with the unadjusted model. BMIZ, BMI Z-score; LAZ, length-for-age Z-score; WAZ, weight-for-age Z-score.

Discussion

Dramatic differences in infant feeding practices were found across the breastfeeding mother-infant dyads enrolled in the international study cohorts. Significant differences in anthropometry among the sites were also noted by 1 y of age, with Shanghai infants longer and heavier and Mexico City infants shorter and lighter than infants in Cincinnati. Only the duration of EBF was significantly associated with WAZ and BMIZ at 1 y of age and remained significant only for WAZ when adjusting for demographic and other birth covariates. Nevertheless, although specific feeding practices measured in this study were only modestly associated with anthropometry at age 1 y across the cohorts, accounting for these practices notably reduced the estimated differences between Cincinnati and Shanghai babies’ anthropometry, suggesting partial mediation of these differences by infant feeding; however, this mediation was not consistent between cohorts.

Infant feeding practices across sites.

We noted dramatic differences in the timing and progression of complementary foods introduced at each site. In the Shanghai cohort, a rapid decline in EBF between 4 and 6 mo was matched by a sharp rise in the introduction of solid food and nutritive beverages. Furthermore, nearly all food groups were introduced in quick succession in Shanghai, suggesting that these infants experienced a nearly complete change in their food intake during this brief period of development. In both Mexico City and Shanghai, high-protein foods were concurrently introduced with cereals, whereas in Cincinnati, high-protein foods were introduced nearly 4 mo after cereals.

The MGRS is a longitudinal, 6-country (US, Oman, Ghana, Norway, Brazil, India) study of breast-fed infants that formed the basis for the WHO growth standards (23). The MGRS limited enrollment to high socio-economic women and required adherence to strict behavioral and health criteria in order to create an international standard for optimal growth (25). However, the MGRS data on complementary feeding, reported only for those who complied with all dietary guidelines, also demonstrate notable variety in the introduction of specific food categories worldwide (22). More than 50% of infants were introduced to grains by 6 mo of age (similar to the present study), fruits/vegetables by 9 mo (later than the present study), and flesh foods by 12 mo of age (later than the present study), and the introduction of legumes/nuts, milk/dairy foods, and eggs was highly variable across sites.

The present study provides critical new data about urban Chinese and Mexican infants who were not included in the MGRS study. The introduction of foods in Shanghai differed distinctly from any of the MGRS sites with a much earlier introduction of meat/legumes/eggs and a later and more infrequent introduction of dairy milk and nonmilk dairy foods. A survey of >20,000 rural, Chinese, mother-infant pairs similarly reported high breastfeeding rates, early introduction of complementary foods, and early and frequent consumption of protein foods such as meat and eggs (26), suggesting that such practices are typical in China. In Mexico, the 1999 National Nutrition survey collected cross-sectional data on breastfeeding and specific complementary food introduction in nearly 3200 infants. In that study, the breastfeeding rates were significantly lower and timing of introduction of specific complementary food categories was earlier than in our cohort, in which enrollment was based on intention of at least predominant breastfeeding (27). Nevertheless, the progression of food introduction is similar to the present Mexico cohort (e.g., fruit/vegetables first, followed by cereals, then meats, at the median). The 2008 Feeding Infants and Toddlers Study is a cross-sectional study of 3200 infants and toddlers in the United States (28). Similar to the present study, the Feeding Infants and Toddlers Study found that U.S. infants were introduced first to cereals, then fruits/vegetables, then meats/protein (at the median), and that the timing of introduction was similar to the Cincinnati cohort (29).

The reasons for the variation in feeding practices among sites may include differences in cultural beliefs or physician recommendations about feeding. Notably, the introduction of nonhuman milk (e.g., cow milk) was delayed until age 1 y in the Cincinnati cohort, consistent with American Academy of Pediatrics recommendations (30). The rare introduction of nonhuman milk and dairy products in Shanghai may reflect perceived predispositions to lactose intolerance (31) or negative perceptions of dairy products (32) among Asian populations. Likewise, U.S. mothers may delay the introduction of foods such as eggs, fish, and nuts despite the recent consensus (33) that this is not necessary for allergy prevention.

Infant feeding and growth.

Chinese infants were significantly heavier, longer, and larger than infants in our Mexico and U.S. cohorts by 1 y of age. Given the prevalence of early-onset obesity in the US, this finding was unanticipated. However, studies of child growth suggest that early adiposity is increasingly common in China (34).

The present study does not provide support for the hypotheses that timing of solid food introduction, timing of specific food introduction, or temporal progression of food introduction has a significant impact on infant growth to age 1 y within the context of breastfeeding. This is consistent with previous studies suggesting that the timing of introduction to any solid food may affect weight only among formula-fed infants (8), potentially through increased, rather than offset, energy intake (19). It contrasts, however, with a small case-control study in China that found early introduction of solid food (<4 mo) was associated with a 10-fold odds of obesity in preschool (35); that study, however, did not simultaneously account for breastfeeding duration. It is important to note, however, that the present study’s endpoint was at 1 y of age, which may have been too early to detect differential growth effects from food groups introduced only weeks earlier (e.g., high-protein foods in Cincinnati or dairy products in Shanghai).

Despite the modest relationships between specific feeding practices and anthropometry adjusting for cohort membership, this study also revealed that both feeding and nonfeeding variables did account for a notable proportion of the observed differences between cohorts. Our finding contrasts with a Dutch study that did not find that differential feeding practices accounted for differential anthropometry among specific ethnic groups (36). The Dutch study, however, examined several ethnic groups located in the same country, which likely had more environmental and cultural similarities than the present study’s cohorts. Interestingly, differences between Shanghai and Cincinnati infants’ anthropometry were largely accounted for by differences in feeding practices, whereas differences of Mexico City from the other cohorts may have been driven more by demographic or socioeconomic differences. These results suggest that careful examination of how countries or societies differ may provide clues to why their children grow differently. Importantly, within the context of cohorts selected for similarity in both socioeconomic profile and feeding recommendations, the MGRS found striking similarity in linear (length) growth among its international cohorts (37) despite noted differences in feeding practices (22), although the effect of feeding practices on weight or length was not specifically reported. Nonetheless, the present study and the MGRS findings suggest that factors that affect variability in growth rates within a given population may differ from factors that account for between-population differences in growth but that differences between populations are potentially overcome by encouraging optimal feeding and environmental conditions for growth. However, caution may be warranted when extrapolating the findings from one culture to another with respect to which infant feeding practices will optimally affect growth.

The introduction of meat/eggs/legumes in Shanghai at 4–6 mo occurred much earlier than at other sites, so the effect of this behavior could not be modeled across sites. However, it is possible that introduction of higher protein foods could still have potentially influenced the larger anthropometry observed in Chinese infants. Indeed, a recent study notes that meat consumption is associated with lower likelihood of stunting in toddlers in low-income societies (21) and high-protein diets in infancy are associated with greater weight gain (20) and later development of overweight (38). In addition, early introduction of high-protein foods may also affect Chinese infants’ micronutrient status, particularly with regard to zinc and iron (39), which are notably low in human milk. Evaluation of differences in zinc and iron status is therefore of interest for future study, even in the absence of growth effects.

The GEHM cohorts benefit from being concurrently conducted using standard enrollment criteria, protocols, and questionnaires and detailed prospective follow-up through the first 2 y of life, enabling direct comparison of international sites. In addition, detailed, weekly, 24-h food recall regarding infant diet permits the prospective evaluation of both breastfeeding duration and exclusivity as well as the precise timing of feeding specific complementary foods in the first year of life. Despite the many strengths of the GEHM study, limitations should be considered. First, the intensity of follow-up limited the sample size and led to lower than expected retention to 1 y of age, potentially limiting statistical power. However, similarity between those completing and not completing the 1-y assessment with regard to infants’ birth weight, sex, and race and maternal education, age at delivery, marital status, prepregnancy BMI, and gestational weight gain, as well as breastfeeding intentions in the China and Mexico cohorts, argues against retention bias. Data are not currently available to assess the impact of infectious disease on growth. The food recall of infants’ diets did not assess either the quantity or dietary composition of the infants’ diets, so it was not possible to control for the infants’ caloric intake or dietary adequacy with regard to macro- or micro-nutrients. Finally, all infants in the cohorts were breastfed at least 75% for at least 4 wk, so these findings may not be generalizable to nonbreastfed or minimally breast-fed cohorts.

The present study suggests several directions for future research. In particular, this study was not able to evaluate infant feeding practices with respect to the broader set of international guidelines regarding introduction of complementary foods (13), nor did it determine whether specific foods have a window of timing for introduction that optimizes infant growth or development. The impact of timing of introduction of specific complementary foods should be evaluated relative to infant nutritional status (e.g., circulating concentrations of iron, zinc, and other vitamins and minerals) through international trials. In addition, it is possible that the aggregate infant feeding pattern may have an impact on anthropometry in ways that could not be identified here by analysis of individual foods but may benefit from a systems-based approach to analyzing infant nutrition patterns.

In conclusion, this study in 3 urban, international, breast-fed cohorts demonstrates that infant feeding practices vary considerably both in the timing and progression of specific food categories introduced. However, although the timing of solid food introduction or other complementary feeding practices do not, by and large, significantly influence weight or length at 1 y of life, these practices may help explain the growth differences between cohorts in infancy.