Associations of infant feeding practices with abdominal and hepatic fat measures in childhood in the longitudinal Healthy Start Study

Abstract

Background

Infant feeding patterns have been linked with obesity risk in childhood, but associations with precise measures of body fat distribution are unclear.

Objective

We examined associations of infant feeding practices with abdominal fat and hepatic fat trajectories in childhood.

Methods

This study included 356 children in the Healthy Start Study, a prospective prebirth cohort in Colorado. Infant feeding practices were assessed by postnatal interviews and categorized as any human milk <6 mo compared with ≥6 mo; complementary foods introduced ≤4 mo compared with >4 mo; soda introduced ≤18 mo compared with >18 mo. Abdominal subcutaneous (SAT) and visceral adipose tissue (VAT) areas and hepatic fat (%) were assessed by magnetic resonance imaging in early and middle childhood (median 5 and 9 y old, respectively). We examined associations of infant feeding with adiposity trajectories across childhood using linear mixed models.

Results

In the sample of children, 67% consumed human milk ≥6 mo, 75% were introduced to complementary foods at >4 mo, and 81% were introduced to soda at >18 mo. We did not find any associations between duration of any human milk consumption and childhood adiposity trajectories. Early introduction to complementary foods (≤4 mo) was associated with faster rates of change for SAT and VAT during childhood (Slope [95% CI]: 15.1 [10.7,19.4] cm²/y for SAT; 2.5 [1.9,2.9] cm²/y for VAT), compared with introduction at >4 mo (5.5 [3.0,8.0] cm²/y and 1.6 [1.3,1.9] cm²/y, respectively). Similarly, early introduction to soda (≤18 mo) was associated with faster rates of change for all 3 outcomes during childhood (Slope [95% CI]: 20.6 [15.0,26.1] cm²/y for SAT, 2.7 [2.0,3.3] cm²/y for VAT, 0.3 [0.1,0.5] %/year for hepatic fat) compared with delayed introduction (5.4 [2.8,8.0] cm²/y, 1.7 [1.3, 2.0] cm²/y, -0.1 [-0.2,0.0] %/y, respectively).

Conclusions

The timing of introduction and quality of complementary foods in infancy was associated with rates of abdominal and hepatic fat accrual during childhood. Experimental studies are needed to assess underlying mechanisms.

Introduction

Mounting evidence from the field of developmental origins of health and disease research supports that nutritional exposures during vulnerable life stages, particularly the perinatal period, may exert long-lasting effects on offspring health, including a higher susceptibility to obesity and related metabolic disorders [[1], [2], [3]]. Within this paradigm, infant feeding practices are potentially modifiable, nutritional exposures that could be involved in programming later metabolic dysfunction and excess adiposity. Meta-analyses have reported a protective effect of human milk consumption in infancy on later obesity risk [4, 5], especially for durations of ≥6 mo [6]. Practices around the introduction of complementary foods and beverages may also play a role in influencing later obesity risk [7]. In particular, early introduction of complementary foods (≤4 mo) versus typical timing (∼6 mo) has been associated with obesity risk in childhood [[8], [9], [10], [11]]. Beyond timing, another area of public health interest is whether the introduction of certain types of foods early in life, such as sugary beverages [12, 13], is associated with adiposity later in childhood; though, as noted by others, additional high-quality studies are needed to confirm associations [7].

Whereas this prior research supports a potential link between infant feeding practices and adiposity development in childhood, a limitation is that most studies have focused on measures of total adiposity, especially age- and sex-specific BMI, due to its ease of assessment and cost-effectiveness. However, BMI does not capture body fat deposition in certain regions or tissues, such as abdominal adipose tissues or ectopic liver fat, which are metabolically active, insulin-sensitive tissues [14] and may be even stronger risk factors for youth-onset metabolic dysfunction than overall body size or total adiposity [[15], [16], [17], [18]]. Notably, some studies have shown that dietary factors in infancy may be associated with the relative accrual of abdominal subcutaneous and visceral fat early in life [19, 20]; however, whether this tracks later into childhood remains unclear. Filling these gaps in knowledge to better understand the early nutritional origins of childhood adiposity measures beyond BMI will help to inform the design of interventions and public policy aiming to prevent adiposity-related metabolic diseases in youth.

Using data from a prospective, prebirth cohort study based in Colorado (the Healthy Start Study), we aimed to examine whether infant feeding practices are associated with abdominal fat and hepatic fat trajectories across childhood assessed by repeated MRI. The infant feeding variables of interest were determined based on recommendations from the American Academy of Pediatrics [21, 22] and Dietary Guidelines for Americans [23] as follows: any (nonexclusive) human milk consumption for <6 mo compared with ≥6 mo; introduction of complementary foods and beverages at ≤4 mo compared with >4 mo; introduction of soda at ≤18 mo compared with >18 mo. The outcomes were abdominal subcutaneous adipose tissue (SAT) area, visceral adipose tissue (VAT) area, and hepatic-proton density fat fraction (PDFF) MRI [24], all assessed by abdominal MRI. We hypothesized that adiposity trajectories in childhood would differ according to the duration of human milk consumption in infancy and the timing for introducing any complementary foods, particularly soda.

Methods

Study design

This analysis was based on the Healthy Start Study, an ongoing longitudinal, prebirth cohort study based in Colorado, United States. Initially, 1410 female participants were recruited and enrolled during early pregnancy (<24 wk gestation) from the University of Colorado Hospital obstetric clinics between 2009 and 2014. Other eligibility criteria were: ≥16 y old, expecting a singleton birth, and having no history of serious chronic disease, prior stillbirth, or extremely preterm birth (<25 wk). These females completed prenatal research visits in early (median 17 wk) and mid-pregnancy (median 27 wk), followed by postnatal research visits with their offspring at delivery (median 1 d), 4 to 6 mo (median 5 mo), 18 to 24 mo (median 22 mo), 4 to 8 y (median 5 y, also called the “early childhood” visit), and 8 to 10 y (median 9 y, also called the “middle childhood” visit). The study was approved by the Colorado Multiple Institutional Review Board, and all participants provided written informed consent. Participant selection for this study is summarized in a flow chart in Supplemental Figure 1. Among 1410 initially enrolled mother-child dyads, the eligible sample for this analysis was 356 children who returned for MRI assessments at the early childhood and/or middle childhood visits (n = 220 early childhood only; n = 70 middle childhood only; n = 66 both visits). For the abdominal fat outcomes, 10 of the MRI scans performed at the early childhood visit and 1 of the scans performed at the middle childhood visit were missing the data needed to calculate abdominal SAT and VAT (described in more detail below) and, thus, only hepatic fat was calculated for these scans.

Infant feeding assessments (exposures)

Infant feeding practices were assessed by maternal interviews conducted when offspring were 4 to 6 mo and/or 18 to 24 mo, as previously described [[25], [26], [27], [28]]. At both visits, mothers were queried on current use of human milk, formula, and/or mixed feedings and age at which their child received solid foods or liquids other than human milk or formula on a daily basis (defined as ≥2 consecutive days). At 18 to 24 mo, the questions on complementary feeding practices were more detailed, including the age at which infant cereal was introduced and in what format (i.e., in a bottle or not, liquid used for mixing, etc.), and the age at which a longer list of foods/beverages were introduced (24 items total at 18-24 mo compared with 9 items total at 4–6 mo). We determined the duration of human milk consumption based on data collected at the 4 to 6 mo visit, if sufficient, and then supplemented with data from the 18 to 24 mo visit when necessary, given that self-reported infant feeding practices are valid and reliable within 3 y [29]. For example, for mothers who completed the 4 to 6 mo visit prior to 6 mo but reported current human milk consumption for their infant, we used data from the 18 to 24 mo visit to determine the full duration and exclusivity of consumption. Age at first introduction of any complementary foods was determined based on the earliest reported age when any solid food or liquid other than human milk or formula was introduced daily, aligning with current guidelines [21, 30]. Age at introduction of soda was determined based on two questions at the 18 to 24 mo interview that asked whether the child had received regular sodas (yes/no) and, if yes, at what age they started drinking soda on a daily basis. To avoid bias due to the age at which this visit was conducted, we dichotomized children as having soda introduced at either ≤18 mo (minimum age at the 18 to 24 mo visit) or >18 mo (which also included children who were never introduced to soda).

MRI assessments (outcomes)

MRI was used to assess abdominal adiposity (SAT and VAT) and liver fat content at both the early childhood visit and the middle childhood visit. A Siemens Skyra 3 Tesla system (Siemens Healthineers, Erlangen, Germany) was used for data acquisition. Patients were positioned supine, and a series of eight 5-mm axial T1-weighted images covering the L1/L2 through the L4/L5 disc space were acquired and analyzed to measure SAT and VAT volume in each subject. An image analyst with 12 y of experience used IDL 8.9 (NV5 Geospatial) to segment the SAT and VAT quantities. Additionally, liver imaging was performed in the same scanning session using a 6-echo chemical-shift-encoded water-fat technique, in which fat and water signals were nominally inphase or out-of-phase with respect to each other [31]. Acquired data were passed to a fitting algorithm that modeled the fat signal, estimated fat and water proton densities, and calculated the percentage of fat content in each voxel using a custom toolbox in Horos (www.horosproject.org) developed by the University of California San Diego Liver Imaging Center [[32], [33], [34]]. Hepatic PDFF was extracted from manually drawn regions of interest (ROIs), segmenting all visible regions of the liver in any given slice.

Covariate assessments

At enrollment, females selfreported their date of birth, race, and ethnicity (which included Hispanic or Latina, American Indian or Alaskan Native, White or Caucasian, Black or African American, Asian or Pacific Islander, or Other), which we view as a social construct that captures the effect of structural social exposures on physiology rather than a biological determinant of health [35]; highest level of education; annual household income; and parity. Maternal age at delivery (y) was calculated from date of delivery and date of birth. Maternal prepregnancy BMI was calculated from most recent prepregnancy weight in the medical record (92%) or selfreported prepregnancy weight and measured height at the early pregnancy visit. Maternal obesity was defined based on a prepregnancy BMI>30 kg/m². During the pregnancy visits and at delivery, the females reported their smoking habits via questionnaires, and females were categorized based on whether they smoked during pregnancy or not. Offspring birth weight and gestational age were abstracted from medical records and birth weight-for-gestational-age z-score was calculated using gestational age at birth and previously published standards [36]. At the in-person visit shortly after delivery, offspring body composition (fat mass and fat-free mass) was also assessed using whole-body air displacement plethysmography (PEAPOD; COSMED, Rome, Italy). The protocol was performed twice, and a third measurement was taken if the first values differed by >2%, as described previously [[37], [38], [39]]. The average of the 2 closest measurements was used to calculate percent fat mass (%FM) based on fat mass (kg) divided by body weight (kg). In childhood, standing height (cm) and weight (kg) were also measured and used to calculate age and sex-adjusted BMI z-scores and percentiles based on the 2000 Centers for Disease Control and Prevention Growth Charts [40]. We categorized children based on their weight status as follows: normal weight (BMI<85^th percentile), overweight (BMI 85^th–<95^th percentile), obesity (BMI ≥95^th percentile).

Statistical analyses

Descriptive statistics were used to summarize the characteristics of the subsample who completed the MRI assessments at early and/or middle childhood visits overall (n = 356) and according to infant feeding categories. To assess selection bias, we examined participant characteristics according to missingness of MRI data at either or both visits in childhood (Supplemental Table 1), which confirmed that the subgroups were relatively similar to each other and the full Healthy Start sample, except for a few differences for maternal traits at enrollment. In particular, the subgroup who completed an MRI at the middle childhood visit only had higher average maternal age, education, and household income at enrollment but lower maternal prepregnancy BMI compared with the other MRI subgroups and the sample overall.

We used linear mixed models, which can account for repeated measurements of the outcome, to estimate associations of each infant feeding variable with abdominal SAT, VAT, and hepatic fat across both visits in childhood. Models included the infant feeding variable, child’s age (y) when the outcome was assessed by MRI, and their interaction term (i.e., infant feeding∗child’s age) as fixed effects. Models also included participant-specific random intercepts, linear age slopes, and an unstructured covariance structure. Degrees of freedom were estimated using the Kenward-Roger method [41]. We used jackknife residuals to test model assumptions and Akaike information criteria to assess model fit [42]. Given potential sex differences in abdominal adipose tissue distribution and hepatic fat in youth [[43], [44], [45]], we assessed for effect modification by sex via a 3-way interaction term in models prior to additional covariate adjustment (child’s sex∗infant feeding∗child’s age). We found no evidence of effect modification based on an interaction alpha level of 0.02 (alpha=0.05/3 infant feeding variables); thus, we decided to only present estimates for all children and adjusted for child’s sex at birth. We also adjusted for Hispanic ethnicity, birth weight-for-gestational age z-score, maternal age at enrollment (y), maternal education (high school or less, some college, college degree), and maternal pre-pregnancy obesity (defined as BMI>30 kg/m²), which were selected based on prior knowledge and bivariate associations with the infant feeding variables. Other covariates that were considered but dropped from the model because they did not appreciably change effect estimates included child’s race, household income, parity, gestational smoking, and gestational age at birth. If the exposure-outcome association was significant in the adjusted models above, we considered an additional model that also included the child’s BMI z-score concurrent to outcome assessment in childhood to assess the extent to which associations of infant feeding with abdominal adiposity and hepatic fat are independent of body size in childhood. Results were reported as the predicted least squares (LS) mean and 95% confidence interval (CI) for each outcome at the median age of the early childhood visit (5 y) and middle childhood visit (9 y) according to infant feeding category. We also calculated the predicted slope and 95% CI for each adiposity outcome across childhood according to infant feeding category.

We conducted a complete case analysis. No data were missing for the covariates listed above, but a small number of participants from the eligible sample were excluded from linear mixed models due to missing data for each infant feeding variable (n = 30 missing data on human milk duration, n = 7 missing age at introduction to complementary foods, n = 49 missing age at soda introduction) (Supplemental Figure 1). In sensitivity analyses, we assessed whether the effects of human milk consumption on childhood adiposity were altered if we defined the exposure based on a shorter duration cut-off (≥4 mo) or as exclusive human milk consumption (no other solids or liquids, including no infant formula) for ≥4 or ≥6 mo. We assessed whether associations of complementary feeding exposures with childhood adiposity were altered if we additionally adjusted for duration of human milk consumption (<6 mo versus ≥6 mo). We also assessed if associations differed if we adjusted for neonatal %FM instead of birth weight-for-age z-score among the subset of children with data for neonatal body composition based on air displacement plethysmography at birth (n = 314). All analyses were performed in SAS Statistical Software (v9.4, Cary, NC, USA).

Results

Study population

The characteristics of the analytical sample for this study overall and according to each infant feeding category are shown in Table 1. About half of the children were boys (51%), 71% selfidentified as white, and 27% as Hispanic ethnicity (Table 1). Also, most children were classified as normal weight (BMI<85^th percentile) at the early childhood visit (86%) and middle childhood visit (77%) (data not shown). Most children consumed human milk for ≥6 mo (67%), were introduced to complementary foods at >4 mo (75%), and were introduced to soda at >18 mo (81%) (Table 1). Duration of human milk consumption was associated with complementary food practices, such that early introduction to complementary foods (≤4 mo) and soda (≤18 mo) were more common among children who consumed human milk <6 mo (46% and 32%, respectively) compared with children who consumed human milk ≥6 mo (14% and 13%, respectively) (Table 1). Hispanic ethnicity, younger maternal age, and lower maternal education were traits consistently associated with lower adherence to all 3 infant feeding guidelines (defined as human milk <6 mo, complementary foods introduced at ≤4 mo or sodas introduced at ≤18 mo) (Table 1). Supplemental Figure 2 shows the observed individual trajectories for each adiposity outcome (SAT, VAT, and hepatic fat) across childhood for the subset of participants in this study who completed MRI assessments at both visits in childhood.

TABLE 1.

Characteristics of the MRI subsample (n = 350) according to categories for the infant feeding variables of interest

	Overall Sample (n = 356)	Any Human Milk <6 mo (n = 109)	Any Human Milk ≥6 mo (n = 217)	Introduction to CF ≤4 mo (n = 88)	Introduction to CF >4 mo (n = 261)	Introduction to Soda ≤18 mo (n = 59)	Introduction to Soda >18 mo (n = 248)
Child traits	Mean (SD)1	Mean (SD)1	Mean (SD)1	Mean (SD)1	Mean (SD)1	Mean (SD)1	Mean (SD)1
Boys, n (%)	181 (51)	45 (41)	119 (5)	42 (48)	134 (51)	29 (49)	125 (50)
Race, n (%)
White	253 (71)	68 (62)	169 (78)	56 (64)	194 (74)	34 (58)	193 (78)
Black	42 (12)	22 (20)	15 (7)	16 (18)	26 (10)	12 (20)	23 (9)
Other	61 (17)	19 (17)	33 (15)	16 (18)	41 (16)	13 (22)	32 (13)
Hispanic Ethnicity, n (%)	97 (27)	45 (41)	40 (18)	30 (34)	63 (24)	31 (53)	47 (19)
Gestational age (wk)	39.3 (1.9)	39.1 (1.6)	39.3 (2.0)	39.5 (1.2)	39.2 (2.1)	39.6 (1.1)	36.2 (2.1)
Birth weight (kg)	3.2 (0.5)	3.1 (0.5)	3.3 (0.5)	3.2 (0.4)	3.3 (0.6)	3.3 (0.5)	3.2 (0.5)
Birth weight-for-age z-score	-0.16 (1.20)	-0.34 (1.13)	-0.05 (1.22)	-0.20 (0.92)	-0.14 (1.28)	-0.06 (0.98)	-0.17 (1.25)
Neonatal %FM	9.5 (3.9)	9.6 (3.9)	9.5 (3.9)	9.0 (3.8)	9.7 (3.9)	9.3 (4.1)	9.6 (3.8)
Infant feeding categories, n (%)
Human Milk <6 mo	109 (33)	-	-	49 (62)	60 (24)	29 (51)	61 (25)
CF Introduction ≤4 mo	88 (25)	49 (45)	30 (14)	-	-	25 (42)	48 (19)
Soda Introduction ≤18 mo	59 (19)	29 (32)	28 (13)	25 (34)	34 (15)	-	-
Maternal traits in pregnancy:
Age (y)	29.1 (5.9)	26.6 (6.2)	30.8 (5.0)	27.6 (5.9)	29.8 (5.7)	26.3 (6.4)	30.2 (5.4)
Education, n (%)
High school degree or less	75 (21)	41 (38)	19 (9)	24 (27)	47 (18)	29 (49)	28 (11)
Some college	85 (24)	35 (32)	43 (20)	31 (35)	53 (20)	18 (31)	52 (21)
College degree	196 (55)	33 (30)	155 (71)	33 (38)	161 (62)	12 (20)	168 (68)
Household Income, n (%)
<$40,000	96 (27)	44 (40)	40 (18)	30 (34)	64 (25)	19 (32)	54 (22)
$40–70,000	64 (18)	16 (15)	45 (21)	16 (18)	47 (18)	11 (19)	48 (19)
>$70,000	146 (41)	21 (19)	119 (55)	28 (32)	118 (45)	7 (12)	129 (52)
Nulliparous, n (%)	169 (47)	49 (45)	106 (49)	43 (49)	121 (46)	25 (42)	124 (50)
Prepregnancy obesity, n (%)	76 (21)	27 (25)	38 (17)	20 (23)	55 (21)	27 (46%)	34 (14)
Gestational diabetes, n (%)	17 (5)	4 (4)	10 (5)	5 (6)	12 (5)	4 (7)	8 (3)
Gestational smoking, n (%)	18 (5)	12 (11)	3 (1)	8 (9)	9 (3)	2 (3)	10 (4)

Abbreviations: CF, complementary foods; %FM, percent fat mass.

¹Descriptive statistics are presented as means and standard deviations (SDs) unless otherwise indicated.

Duration of human milk consumption and childhood adiposity

The duration of human milk consumption in infancy was not associated with differences in the trajectory for any adiposity outcome in childhood (Table 2). Supplemental Figure 3 shows the predicted trajectory for each outcome in childhood according to the duration of human milk consumption from linear mixed models.

TABLE 2.

Least squares means and mean differences for abdominal adipose tissues and hepatic fat from early to middle childhood according to duration of any human milk consumption in infancy

Adiposity Outcome:	Any Human Milk <6 mo	Any Human Milk ≥6 mo	Mean Difference, <6 versus ≥6 mo
Abdominal SAT (cm2), n = 318	Estimate (95% CI)1	Estimate (95% CI)1	Estimate (95% CI)1
LS-Mean at 5 y2	51.3 (46.4, 56.3)	51.1 (46.8, 55.3)	0.3 (-5.8, 6.4)
LS-Mean at 9 y3	89.2 (70.3, 108.1)	76.5 (64.1, 88.9)	12.7 (-9.7, 35.1)
Slope across childhood4	9.5 (5.3, 13.7)	6.4 (3.7, 9.0)	3.1 (-1.9, 8.1)
Abdominal VAT (cm2), n = 318
LS-Mean at 5 y2	12.0 (11.1, 12.8)	12.3 (11.6, 13.1)	-0.4 (-1.4, 0.7)
LS-Mean at 9 y3	19.2 (16.6, 21.7)	19.3 (17.7, 21.0)	-0.2 (-3.2, 2.8)
Slope across childhood4	1.8 (1.2, 2.4)	1.7 (1.4, 2.1)	0.0 (-0.6, 0.7)
Hepatic Fat (%), n = 326
LS-Mean at 5 y2	2.0 (1.7, 2.3)	2.2 (1.9, 2.5)	-0.2 (-0.6, 0.2)
LS-Mean at 9 y3	1.5 (0.8, 2.2)	2.2 (1.8, 2.7)	-0.7 (-1.5, 0.1)
Slope across childhood4	-0.1 (-0.3, 0.0)	0.0 (-0.1, 0.1)	-0.1 (-0.3, 0.1)

Abbreviations: SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue.

¹Estimates are from linear mixed effects models adjusted for child’s age when adiposity was assessed by MRI (y) and child’s sex, Hispanic ethnicity, birth weight-for-age z-score, maternal age, maternal education, and maternal prepregnancy obesity. Models also include an interaction term between child’s age at outcome assessment and each infant feeding variable.

²Median age at the early childhood visit (5 y old).

³Median age at the middle childhood visit (9 y old).

⁴Slope estimates are the predicted average rate of change per year in each adiposity outcome from early to middle childhood according to infant feeding category.

Timing of complementary foods introduction and childhood adiposity

The predicted average trajectories for each outcome in childhood according to timing of introduction to complementary foods during infancy from linear mixed models are shown in Figure 1. At 5 y old (median age at the early childhood visit), there were no significant differences in abdominal SAT, VAT, and hepatic fat according to timing of complementary foods introduction (Table 3). However, trajectories for SAT and VAT differed sharply in slope across childhood, such that those who were introduced to complementary foods earlier (≤4 mo) had significantly greater rates of change per year in SAT and VAT (predicted slope [95% CI]: 15.1 [10.7, 19.4] cm²/y for SAT and 2.5 [1.9, 2.9] cm²/y for VAT), compared with those who were introduced to complementary foods at >4 mo (5.5 [3.0,8.0] cm²/y for SAT and 1.6 [1.3,1.9] cm²/y for VAT) (Table 3). Due to more rapid increases in SAT and VAT across childhood visits, children who were introduced to complementary foods earlier in infancy (≤4 compared with >4 mo) had significantly higher abdominal SAT and VAT by 9 y old (median age at the middle childhood visit) (Adjusted Mean Difference [95% CI]: 37.3 [14.6, 60.1] cm² for SAT and 3.1 [0.2, 6.0] cm² for VAT) (Table 3). For hepatic fat, a similar pattern of findings was observed, but confidence intervals crossed the null for all effect estimates (Table 3). In models additionally adjusted for child’s BMI z-score concurrent to outcome assessment by MRI, differences in abdominal SAT and VAT trajectories across childhood according to age at introduction of any complementary foods remained significant, but the magnitude of most effect estimates was attenuated (Supplemental Table 2).

FIGURE 1.

Trajectories for abdominal SAT (A; n = 341), abdominal VAT (B; n = 341), and hepatic fat (C; n = 349) from early to middle childhood according to age at introduction of complementary food and beverages during infancy (“early” ≤4 mo or “appropriate” >4 mo). Trajectories are predicted from linear mixed models adjusted for child’s age when adiposity was assessed by MRI (y) and child’s sex, Hispanic ethnicity, birth weight-for-age z-score, maternal age, maternal education, and maternal prepregnancy obesity. Models also include an interaction term between child’s age at adiposity assessment and the infant feeding variable. The predicted average rate of change (slope) for adiposity trajectories across childhood according to age at introduction of complementary foods in infancy are shown in Table 3. Abbreviations: CF, complementary foods SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue.

TABLE 3.

Least squares means and mean differences for abdominal adipose tissues and hepatic fat from early to middle childhood according to age at introduction to complementary foods or beverages during infancy

Adiposity Outcome:	Introduction to Any Complementary Foods ≤4 mo	Introduction to Any Complementary Foods >4 mo	Mean Difference, ≤4 versus >4 mo
Abdominal SAT (cm2), n = 341	Estimate (95% CI)1	Estimate (95% CI)1	Estimate (95% CI)1
LS-Mean at 5 y2	52.3 (46.8, 57.7)	53.0 (49.4, 56.7)	-0.7 (-6.7, 5.3)
LS-Mean at 9 y3	112.5 (92.6, 132.4)	75.1 (63.6, 86.6)	37.3 (14.6, 60.1)
Slope across childhood4	15.1 (10.7, 19.4)	5.5 (3.0, 8.0)	9.5 (4.5, 14.5)
Abdominal VAT (cm2), n = 341
LS-Mean at 5 y2	12.1 (11.2, 13.0)	12.6 (11.9, 13.2)	-0.5 (-1.5, 0.5)
LS-Mean at 9 y3	22.0 (19.4, 24.5)	18.9 (17.5, 20.4)	3.1 (0.2, 6.0)
Slope across childhood4	2.5 (1.9, 3.0)	1.6 (1.3, 1.9)	0.9 (0.2, 1.5)
Hepatic Fat (%), n = 349
LS-Mean at 5 y2	2.2 (1.9, 2.5)	2.0 (1.8, 2.2)	0.2 (-0.2, 0.6)
LS-Mean at 9 y3	2.5 (1.8, 3.2)	1.9 (1.5, 2.3)	0.6 (-0.2, 1.4)
Slope across childhood4	0.1 (-0.1, 0.2)	0.0 (-0.1, 0.1)	0.1 (-0.1, 0.3)

Abbreviations: SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue.

¹Estimates are from linear mixed effects models adjusted for child’s age when adiposity was assessed by MRI (y) and child’s sex, Hispanic ethnicity, birth weight-for-age z-score, maternal age, maternal education, and maternal prepregnancy obesity. Models also include an interaction term between child’s age at outcome assessment and each infant feeding variable.

²Median age at the early childhood visit (5 y old).

³Median age at the middle childhood visit (9 y old).

⁴Slope estimates are the predicted average rate of change per year in each adiposity outcome from early to middle childhood according to infant feeding category.

Timing of soda introduction and childhood adiposity

The predicted trajectory for abdominal SAT, VAT, and hepatic fat according to timing of soda introduction during infancy/toddlerhood are shown in Figure 2. Early soda introduction (≤18 mo) was associated with significantly greater rates of change across childhood for SAT, VAT, and hepatic fat (predicted slope [95% CI]: 20.6 [15.0, 26.1] cm²/y for SAT, 2.7 [2.0,3.3] cm²/y for VAT, 0.3 [0.1,0.5] %/y for hepatic fat) compared with delayed introduction (>18 mo) of soda (5.4 [2.8, 8.0] cm²/y for SAT, 1.7 [1.3, 2.0] cm²/y for VAT, -0.1 [-0.2, 0.0] %/y for hepatic fat) (Table 4). These associations between early soda introduction and trajectories for abdominal SAT, VAT, and hepatic fat in childhood also remained significant in models additionally adjusted for child’s BMI z-score concurrent to outcome assessment; however, again, the magnitude of several effect estimates was attenuated (Supplemental Table 3).

FIGURE 2.

Trajectories for abdominal SAT (A; n = 300), abdominal VAT (B; n = 300), and hepatic fat (C; n = 307) from early to middle childhood according to age at introduction of soda during infancy/toddlerhood (“early” ≤18 mo or “delayed” >18 mo). Trajectories are predicted from linear mixed effects models adjusted for child’s age when adiposity was assessed by MRI (y) and child’s sex, Hispanic ethnicity, birth weight-for-age z-score, maternal age, maternal education, and maternal prepregnancy obesity. Models also include an interaction term between child’s age at adiposity assessment and the infant feeding variable. The predicted average rate of change (slope) for adiposity trajectories across childhood according to age at introduction of regular sodas in infancy/toddlerhood are shown in Table 4. Abbreviations: SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue.

TABLE 4.

Least squares means and mean differences for abdominal adipose tissues and hepatic fat from early childhood to middle childhood according to age at introduction to sodas during infancy/toddlerhood

Adiposity Outcome:	Introduction to Soda ≤18 mo	Introduction to Soda >18 mo	Mean Difference, ≤18 versus >18 mo
Abdominal SAT (cm2), n = 300	Estimate (95% CI)1	Estimate (95% CI)1	Estimate (95% CI)1
LS-Mean at 5 y2	56.7 (50.3, 63.0)	51.0 (46.6, 55.4)	5.7 (-2.1, 13.5)
LS-Mean at 9 y3	139.0 (114.3, 163.7)	72.6 (60.7, 84.4)	66.4 (39.1, 93.7)
Slope across childhood4	20.6 (15.0, 26.1)	5.4 (2.8, 8.0)	15.2 (9.1, 21.3)
Abdominal VAT (cm2), n = 300
LS-Mean at 5 y2	12.5 (11.4, 13.6)	12.3 (11.6, 13.1)	0.1 (-1.2, 1.5)
LS-Mean at 9 y3	23.1 (19.9, 26.3)	19.0 (17.4, 20.5)	4.1 (0.6, 7.6)
Slope across childhood4	2.7 (2.0, 3.3)	1.7 (1.3, 2.0)	1.0 (0.2, 1.8)
Hepatic Fat (%), n = 307
LS-Mean at 5 y2	2.1 (1.7, 2.4)	2.2 (1.9, 2.5)	-0.1 (-0.6, 0.3)
LS-Mean at 9 y3	3.1 (2.3, 4.0)	1.9 (1.5, 2.3)	1.2 (0.3, 2.2)
Slope across childhood4	0.3 (0.1, 0.5)	-0.1 (-0.2, 0.0)	0.3 (0.1, 0.6)

Abbreviations: SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue.

¹Estimates are from linear mixed effects models adjusted for child’s age when adiposity was assessed by MRI (y) and child’s sex, Hispanic ethnicity, birth weight-for-age z-score, maternal age, maternal education, and maternal prepregnancy obesity. Models also include an interaction term between child’s age at outcome assessment and each infant feeding variable.

²Median age at the early childhood visit (5 y old).

³Median age at the middle childhood visit (9 y old).

⁴Slope estimates are the predicted average rate of change per year in each adiposity outcome from early to middle childhood according to infant feeding category.

Sensitivity analyses

In sensitivity analyses, most results were unchanged if we categorized participants based on a shorter duration of consumption (≥4 mo) or based on duration of exclusive human milk consumption. One exception was that the magnitude of the association between human milk consumption in infancy and abdominal SAT trajectories in childhood became stronger if we categorized participants based on exclusive human milk consumption. For example, exclusive human milk <4 mo, compared with ≥4 mo, was associated with faster accrual of abdominal SAT across childhood (Slope Difference [95% CI]: +4.4 cm²/y [-0.1, 8.8]) (Supplemental Table 4). For complementary feeding practices, adjusting for duration of human milk consumption did not appreciably change any effect estimates (data not shown). Results were also similar if we adjusted for neonatal %FM instead of birth weight-for-age z-score (data not shown).

Discussion

Using data from a prospective, prebirth cohort study, we showed that the timing and quality of complementary foods introduced during infancy and toddlerhood were associated with differential trajectories for abdominal SAT and VAT and hepatic fat deposition in childhood. More specifically, early introduction to complementary foods by 4 mo was associated with faster rates of change in abdominal SAT and VAT from early to middle childhood. In turn, this corresponded to higher predicted levels of SAT and VAT by middle childhood (∼9 y). The relationship between early soda introduction and trajectories for SAT, VAT, and hepatic fat in childhood followed a similar pattern, whereby children who were introduced to soda by 18 mo had faster rates of change in all 3 outcomes across childhood, resulting in higher SAT, VAT, and hepatic fat by middle childhood. Whereas, we did not observe strong associations between a longer duration of human milk consumption in infancy (≥6 mo) and abdominal or hepatic fat trajectories in childhood.

The potential influence of infant feeding on child overweight and obesity risk has been an area of great research interest [4]. This study contributes to the literature by examining precise measures of adiposity assessed by a gold standard method (MRI) that reflect where body fat is deposited, particularly as abdominal fat or ectopic organ fat (i.e., hepatic fat). Contrary to our hypothesis, we did not find a strong association between human milk consumption in infancy and abdominal adiposity or hepatic fat in childhood. The exception was one result from our sensitivity analyses showing that exclusive human milk consumption for at least 4 mo was associated with faster accrual of abdominal SAT across childhood. Still, our findings align with another study based on the Generation R cohort that examined associations of infant feeding practices with MRI-based adiposity assessments in childhood (at 10 y) [46]. This study initially found positive associations of shorter or nonexclusive human milk consumption with childhood visceral and hepatic fat, but these associations were attenuated when adjusted for family-based confounders, especially maternal education [46]. Regarding hepatic fat in particular, we can also compare our findings to studies on the association between infant nutrition and risk of a clinical metabolic dysfunction-associated steatotic liver disease (MASLD), previously called nonalcoholic fatty liver disease. For example, one study examined associations between any or exclusive human milk consumption with MASLD risk in young adults (24 y), assessed based on transient elastography and the controlled attenuation parameter score, and again found only weak associations that crossed the null in adjusted models [47]. This pattern of findings should not minimize other reasons to promote lactation and human milk feeding in infancy [21, 22], but in terms of abdominal and hepatic fat susceptibility, findings from this study and others suggest that this may be just one factor among a larger cluster of correlated maternal and child characteristics and perinatal nutritional exposures at play.

We did, however, find associations between the complementary feeding practices assessed, particularly early introduction of any complementary foods at or before 4 mo and early introduction of soda at or before 18 mo, and faster accrual of abdominal SAT and VAT (for either exposure), and hepatic fat (for early soda) across childhood. Though it is not clear why adiposity differences according to complementary feeding exposures became stronger over time and were only significant in middle childhood (∼9 y) but not early childhood (∼5 y), one possible explanation is that this pattern may reflect an accumulation of risk or chain of risk pathway, whereby early-life exposures are correlated with and exacerbated by other environmental insults occurring across the life course [48]. For example, it is plausible that children who were exposed to soda by 18 mo continued to drink soda throughout childhood, aligning with studies showing a link between complementary feeding practices and diet quality later in life [49] and that the cumulative exposure to a high-sugar diet throughout infancy and childhood contributed to faster accumulation of abdominal and hepatic fat. Our finding that associations were attenuated when adjusted for concurrent BMI z-score aligns with this hypothesis and supports that excess body weight after infancy, which may be attributed to continued poor diet and other lifestyle factors, accounts for some variation in these relationships. More research will be needed to assess the joint, mediating, and modifying effects of dietary intakes across life stages on childhood adiposity, including dietary intake more proximal to the timing of adiposity assessment in childhood.

We also found that, for the abdominal adipose tissues, associations tended to be stronger for SAT than VAT. This conflicts with findings from another longitudinal cohort showing that higher juice intake in infancy (another dietary source of free sugar) was more strongly associated with VAT than SAT in middle childhood and early adolescence (∼8 to 13 y). This discrepancy could be due to the younger age range when adiposity was assessed in our study (∼4 to 10 y), which may be a developmental period when fat deposition is more likely to occur as abdominal SAT, whereas VAT (and hepatic fat) accumulation may be more likely to occur later in adolescence and puberty, coinciding with the saturation of abdominal SAT, as shown by others [18].

A limitation of this study is that only a subgroup of children from the larger cohort underwent the abdominal MRI assessments in childhood, as this required an additional in-person study visit following the primary study visit. Though we compared the participants who completed the MRI at either visit to each other and the overall sample and found that most characteristics were similar across groups, the modeling assumption for the linear mixed models that data were missing completely at random cannot be tested. The Healthy Start Study is comprised of mothers and children at general risk levels, with relatively low rates of overweight/obesity compared with the overall US population [50, 51]. This may impact the generalizability of our findings to higher-risk subgroups. We relied on self-reported data to assess infant feeding practices, which may be subject to recall bias or social desirability bias, especially for parents of infants with overweight or obesity. Such bias may have attenuated our effect estimates, making our findings conservative. We also did not have detailed information on the method of human milk consumption (i.e., breastfeeding or chestfeeding compared with bottle-feeding of expressed milk), nor the frequency or dosage of soda and other complementary foods introduced during infancy, which could further modify the strength of associations. This study has several strengths. It was based on a relatively large, prospective study of mother-child dyads who have been completing extensive assessments since pregnancy. This allowed us to adjust for key confounding variables in our analysis, though residual confounding is still possible. We used MRI to assess the outcomes of interest (abdominal SAT, VAT, and hepatic fat) with high accuracy. Another strength was the inclusion of MRI-based outcome data from two study visits in early childhood (4 to 8 y) and middle childhood (8 to 10 y), which enabled us to assess the effects of infant feeding on adiposity trajectories across a range of ages in childhood.

Conclusions

This study showed that early introduction of complementary foods by 4 mo and early introduction of sodas by 18 mo were associated with faster accrual of abdominal and ectopic fat deposition in childhood. These observational findings should be tested in high-quality intervention studies. Mechanistic studies aiming to understand the underlying causal pathway are also warranted. The clinical and public health implications of these findings include a greater emphasis on research and education on the timing and quality of complementary foods and beverages introduced during infancy and toddlerhood. This study also provides additional support for recommendations to delay the introduction of soda, an energy-dense food with no nutritional value, during this vulnerable life stage.

Author contributions

The authors' responsibilities were as follows—CCC and DD conceived and designed the research. CCC analyzed the data. KKH and DHG assisted with the modeling strategy and statistical considerations. CCC wrote the first draft of the manuscript. HH assisted with interpretation of the imaging data. WP and KS assisted with results interpretation and revising the manuscript. CCC and DD had primary responsibility for final content. All authors have read and approved the final manuscript.

Funding

The Healthy Start Study is supported by National Institutes of Health (NIH) grant no. R01-DK076648, UH3-OD023248, and R01-DK133235 to Dr. Dabelea. Dr. Cohen was supported by NIH grant no. F32-DK131757. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data availability

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

Conflict of interest

The authors have no conflicts of interest to disclose.

Appendix A. Supplementary data

The following is the Supplementary data to this article: