Childhood nutrient intakes are differentially associated with hepatic and abdominal fats in adolescence: The EPOCH Study

Abstract

Objective:

To examine whether nutrient intakes in childhood are associated with abdominal and hepatic fat depots later in adolescence.

Methods:

Using data from 302 participants in the longitudinal EPOCH study, we constructed energy partition and nutrient density models to examine associations of nutrient intakes in childhood (~10 years), assessed by food frequency questionnaire, with abdominal subcutaneous adipose tissue (SAT), visceral adipose tissue (VAT), and hepatic fat in adolescence (~16 years).

Results:

In energy partition models (energy intake not held constant), total, monounsaturated, and polyunsaturated fat intakes in childhood were associated with higher SAT in adolescence [β(95%) CI: 8.5(0.1,17.1), 25.1 (2.1,48.1), and 59.7(16.1,103.3) mm² per 100 kcal/d], higher starch intake was associated with log-hepatic fat [Back-transformed β(95% CI): 1.07(1.01,1.15) per 100 kcal/d], and, in boys only, higher animal protein intake was associated with VAT [β(95%): 5.3(0.3,10.3) mm² per 100 kcal/d]. Most associations were unchanged when adjusted for energy intake in nutrient density models.

Conclusions:

Childhood nutrient intakes were differentially associated with adolescent body fats; specifically, unsaturated fat intake predicted abdominal SAT, animal protein intake predicted VAT, and starch intake predicted hepatic fat. These nutrient intakes may, therefore, be targets for intervention studies aiming to modify adolescent body fat distribution.

INTRODUCTION:

The high prevalence of childhood obesity, defined by body mass index (BMI)-for-age greater than the 95^th percentile, is an alarming public health issue with associated cardiometabolic risks(1, 2). However, it has also been shown that children with obesity can vary considerably in terms of metabolic dysfunction, despite similar BMIs(3). One factor that may explain this heterogeneity is underlying patterns of body fat partitioning. Specifically, studies have shown that greater abdominal fat deposition, especially visceral fat, and hepatic fat deposition are strong risk factors for insulin resistance and other cardiometabolic risk factors in youth, independent of total adiposity(4, 5, 6). These associations may be particularly relevant during adolescence, a period of development characterized by rapid growth, including changes in body composition(7), and a higher incidence of cardiometabolic diseases, such as type 2 diabetes mellitus (T2DM)(8) and nonalcoholic fatty liver disease (NAFLD)(9), compared to younger children.

Currently, the etiology of body fat partitioning is poorly understood, but lifestyle behaviors, including diet, are likely involved. Of particular interest is the role of diet quantity versus quality. Overfeeding studies have shown that excess energy intake is associated with increases in abdominal and hepatic fat, but often with considerable interindividual differences(10), which may, in part, be explained by qualitative aspects of diet that have also been shown to influence body composition (11, 12, 13). For example, short-term treatment studies (ranging 9 days to 8 weeks) among youth with obesity or clinical NAFLD have shown that isoenergetic modifications to macronutrient composition, particularly reductions in carbohydrate/sugar intake, are associated with lower visceral fat(14) and/or hepatic fat(15). While these findings may be critical in informing treatment strategies among youth with clinical disease, it remains unclear the extent to which macronutrient composition may prevent the accumulation of metabolically adverse body fat depots in settings that are representative of the general pediatric population. Thus, there is a need for additional prospective studies aiming to understand whether nutrient intakes earlier in life, especially among healthy children, are predictive of body fat partitioning patterns later in adolescence, as such findings would be critical in informing primordial prevention strategies.

Our objective was to examine associations of nutrient intakes in childhood (~10 years) with abdominal fat [subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT)] and hepatic fat deposition in adolescence (~16 years), using data from the Exploring Perinatal Outcomes among CHildren (EPOCH) study, a longitudinal cohort study in Colorado. We also tested whether associations were independent of total energy intake (TEI), given potential correlations between energy and nutrient intakes(16), or modified by sex, given established differences in nutrient metabolism(17) and body fat distribution phenotypes (18) for boys versus girls.

METHODS:

Study population

The EPOCH study is a prospective, multiethnic pediatric cohort based in Colorado. Eligible participants were offspring of singleton pregnancies at a single hospital between 1992 and 2002, whose biological mothers were members of the Kaiser Permanente of Colorado Health Plan at the child’s delivery. Eligible participants were invited to two research visits, which were timed approximately 6 years apart (childhood visit = 6–14 yrs old; adolescence visit = 12–19 yrs old)(19, 20). The study was approved by the Colorado Multiple Institutional Review Board. Mothers provided written informed consent and children older than 8 years provided written assent. A flow chart of participant selection for this study is shown in Figure 1. Among the 604 participants enrolled in childhood, 417 returned for a second visit in adolescence. Of those, 18 participants who did not complete the magnetic resonance imaging (MRI) procedure in adolescence (to assess abdominal and hepatic fat depots), two participants with missing dietary data at visit 1, and five participants with extreme TEI at visit 1 (<800 or > 4000 kcal for boys and <500 kcal or >3500 kcal for girls)(21) were excluded, resulting in an eligible sample of 392 participants. We further excluded 90 participants categorized as TEI under-reporters at visit 1 based on the Goldberg method (described next), resulting in an analytical sample of 302 participants. In Table S1, we compared the characteristics of this analytical sample at visit 1 (n=302) to the full EPOCH cohort of children enrolled at visit 1 (n=604).

Figure 1.

Flow chart of the selection of participants from the EPOCH cohort study for this prospective analysis examining associations between childhood nutrient intakes and adolescent abdominal and hepatic fats. ^aAnalytical sample included in “Model 1” regression analyses (unadjusted). ^bSub-sample included in “Model 2” regression analyses (adjusted for potential confounders). ^bSub-sample included in “Model 3” regression analyses (adjusted for abdominal SAT or VAT in childhood).

Dietary intake assessments:

Dietary intake was assessed at both visits using a modified version of the Block Kids Questionnaire, a semi-quantitative food frequency questionnaire that has been validated in children as young as 8 years old(22, 23), that was developed for the SEARCH for Diabetes in Youth Study(24). Briefly, the questionnaire followed the same general format as the original Block Kids Questionnaire, whereby respondents were asked how many days and the average portion size that an item was consumed over the past week; but, in the SEARCH questionnaire, the number of foods queried was expanded, particularly focusing on additional foods with regional and/or local importance, and the number of nutrients and food groups available for analysis was expanded. In EPOCH, the questionnaires were administered by trained research staff either using a self-administered format or a structured interview format if the staff member determined the participant was having difficulty completing the form. All questionnaires were analyzed using the Nutrition Data System for Research (University of Minnesota, Minneapolis, ME) to estimate TEI in kcal/day and macronutrient intakes in grams/day. Nutrient intakes of interest for this analysis included total carbohydrates, protein, and fat, as well as their sub-types [starch and sugar, animal protein and vegetable protein, and saturated fat (SFA), monounsaturated fat (MUFA), and polyunsaturated fat (PUFA)]. Dietary intake data was also used to calculate Healthy Eating Index-2010 (HEI-2010) total scores as a measure of overall diet quality(25).

Assessment of energy intake under-reporting:

We assessed TEI under-reporting at visit 1 in childhood using the Goldberg method(26, 27). Briefly, we calculated participant-specific ratios of reported TEI (rTEI) to basal metabolic rate (BMR, calculated using Schofield equations) and compared this ratio to a physical activity level (PAL) constant. We chose a PAL constant of 1.55, based on World Health Organization (WHO) recommendations for light activity(26). The cut-off for TEI under-reporting was set as the lower confidence limit for this PAL constant calculated based on the equation described by Black(26). Participants with an rTEI:BMR ratio below this threshold (rTEI:BMR<1.10) were categorized as under-reporters (n=90) and excluded from analyses. Characteristics of the excluded TEI under-reporters compared to acceptable reporters at visit 1 are shown in Table S2. Under-reporters were older and had higher BMI z-scores, abdominal SAT, and abdominal VAT at both visits. Under-reporters also reported lower intakes of energy (as expected), total protein, and animal protein, but higher intakes of starch, vegetable protein, and fiber as a percentage of TEI.

Hepatic and abdominal fat assessments:

MRIs were performed at both visits by trained technicians and research staff at the University of Colorado Anschutz Medical Campus. Abdominal SAT and VAT were assessed at both visits by abdominal MRI using a 3 T HDx Imager (General Electric, Waukesha, WI, USA), as previously described(19). Briefly, participants were placed in a supine position on the scanner and a series of T1-weighted coronal images were taken to locate the L4/L5 plane. Abdominal SAT and VAT areas (mm²) were determined by analyzing one axial T1-weighted image at the umbilicus or L4/L5 vertebrae. All images were analyzed by a single reader who was blinded to each participant’s identity and other measures/assessments. Hepatic fat was also assessed by MRI, but only in adolescence (visit 2), using a breath-hold, 6-point MRI-proton density fat fraction (PDFF) technique(20), whereby hepatic fat fraction was calculated from the mean pixel signal intensity data for each flip angle acquisition using the Osirix, Lipoquant plug-in(28).

Other covariate assessments:

Participant sex, race/ethnicity, and household income were self-reported at the first research visit. Height and weight were measured at both visits and age- and sex-adjusted BMI z-scores were calculated using the WHO growth reference(29). The pubertal stage of participants assessed by self-reported Tanner staging of pubic hair for boys and breast development for girls. When adjusting for pubertal stage in models, we categorized participants as pre-pubertal (Stage I) or pubertal (Stages II-IV) at visit 1, and pubertal or post-pubertal (Stage V) at visit 2. Physical activity was assessed by a validated 3-day physical activity questionnaire(30, 31), which was used to calculate average energy expenditure over three days in metabolic equivalents (METs). Participants were categorized as having exposure to maternal diabetes mellitus (DM) during pregnancy if the mother had a physician diagnosis of gestational DM during pregnancy or type 2 DM before pregnancy, which was ascertained from medical records as previously described(19).

Statistical analysis

Descriptive and univariate analyses

Descriptive statistics were performed to summarize characteristics of the sample using means and standard deviations or medians and interquartile ranges for continuous variables and counts and frequencies for categorical variables. Prior to analyses, we natural log-transformed all hepatic fat values to meet model assumptions of normality in the residuals. Residuals for abdominal SAT and VAT were sufficiently normal and were analyzed without transformation (in mm²).

Multivariable analyses

We constructed two types of linear regression models to examine associations of childhood nutrient intakes with adolescent abdominal SAT, VAT, and hepatic fat, with and without holding energy intake constant. The first type was an energy partition model, which estimates both the energy and non-energy effects of each nutrient intake on the dependent variable. For this model, nutrient intakes were converted from grams/d to kcal/d, and energy intake from all other nutrients was adjusted for as a separate covariate in models. The second type was a multivariate nutrient density model, which estimates the isocaloric effect of an increase in each nutrient intake, offset by a concomitant drop in all other nutrient intakes. For this model, nutrient intakes were converted to nutrient densities, expressed either as a percentage of TEI (%TEI) or g/1000 kcal for fiber, and TEI (kcal/d) was adjusted for as a separate covariate in models. We also adjusted for potential confounders in a stepwise manner to assess whether results were altered: Model 1=adjusted for energy intake from all other nutrients (energy partition models) or TEI (nutrient density models). Model 2=adjusted for age, sex, Hispanic ethnicity, household income (<$50,000, $50,000-$74,999, $>75,000), pubertal stage (pre-pubertal vs. pubertal), physical activity (average METs/d), BMI z-score category (normal/underweight vs. overweight/obesity) at visit 1, maternal DM exposure during pregnancy, and diet quality at visit (HEI-2010 total score). Model 3=adjusted for Model 2 covariates plus abdominal SAT or VAT in childhood (only for models with abdominal SAT or VAT in adolescence as the dependent variable). Model 3 results were not reported for hepatic fat, which was only assessed at visit 2. In unadjusted models, we also tested for effect modification by sex using product terms and reported stratified estimates if p<0.05 for the interaction effect. Results were reported as β-coefficients and 95% confidence intervals (CIs) for associations of a 100 kcal/d increase in each nutrient for energy partition models or a 5% increase in each nutrient for nutrient density models with each outcome in adolescence. For hepatic fat, all estimates were also back-transformed to reflect the ratio of geometric means. To account for multiple testing, we also reported whether p-values were below a Bonferroni-corrected α=0.0167 (α=0.05/3 outcomes). In a sensitivity analysis, we assessed whether results differed if we used the residual method(16) to adjust for energy intake instead of the nutrient density method and found that findings were similar; thus, we only reported results from nutrient density models. All analyses were carried out using SAS statistical software (v9.4, Cary, NC, USA).

RESULTS:

Characteristics of the analytical sample of 302 participants overall and stratified by sex are shown in Table 1. Collectively, 48% (n=145) were boys and 33.8% (n=102) were Hispanic At visit 2 in adolescence, mean levels of abdominal SAT and VAT were 183.5±137.6 mm² and 30.8±19.1 mm², respectively, and the median (IQR) for hepatic fat was 1.8% (1.3–2.5). Mean nutrient intakes in the sample overall and stratified by sex are shown in Table S3.

Table 1:

Characteristics of the analytical sample of youth (n=302) at both visits, overall and stratified by sex.

	Overall (n=302)		Boys (n=145)		Girls (n=157)
Visit 1 Characteristics	Mean or N	SD or %	Mean or N	SD or %	Mean or N	SD or %	p a
Age (yrs), mean (SD)	10.3	1.5	10.3	1.6	10.2	1.5	0.56
Race/ethnicity, n (%)							0.44
Non-Hispanic White	162	53.6%	78	53.8%	84	53.5%
Hispanic	102	33.8%	52	35.9%	50	31.9%
Non-Hispanic Black	22	7.3%	7	4.8%	15	9.6%
Non-Hispanic Other	16	5.3%	8	5.5%	8	5.1%
Household income, n (%)							0.74
<$50,000	75	24.9%	39	26.9%	36	23.1%
$50,000–74,999	48	16.0%	23	15.9%	25	16.0%
>$75,000	178	59.1%	83	57.2%	95	60.9%
Pubertal stageb, n (%)							<0.001
Pre-pubertal (Tanner=I)	139	46.2%	81	56.3%	58	36.9%
Pubertal (Tanner=II to IV)	162	53.8%	63	43.8%	99	63.1%
Physical activityc, mean (SD)	1.9	0.3	1.9	0.3	1.9	0.3	0.20
BMI z-score, mean (SD)	0.04	1.1	0.08	1.2	0.00	1.1	0.55
BMI category, n (%)							0.52
Normal weight	234	77.5%	110	75.9%	124	79.0%
Overweight/obesity	68	22.5%	35	24.1%	33	24.1%
SAT area (mm2)d, mean (SD)	98.3	86.0	89.0	78.0	107.2	92.4	0.07
VAT area (mm2)d, mean (SD)	19.8	13.2	19.2	11.5	20.4	14.7	0.41
VAT/SAT, mean (SD)	0.26	0.12	0.28	0.14	0.23	0.10	<0.001
Visit 2 Characteristics:	Mean or N	SD or %	Mean or N	SD or %	Mean or N	SD or %	p
Age (yrs), mean (SD)	16.6	1.2	16.6	1.2	16.6	1.3	0.67
Pubertal stage, n (%)
Pubertal (Tanner=II to IV)	139	46%	58	40%	81	52%	0.04
Post-pubertal (Tanner=V)	163	54%	87	60%	76	48%
Physical activityc, mean (SD)	1.9	0.4	2.0	0.5	1.9	0.3	0.22
BMI z-score, mean (SD)	0.28	1.1	0.24	1.1	0.31	1.0	0.57
BMI category, n (%)							0.91
Normal weight	222	73.5%	107	73.8%	115	73.3%
Overweight/obesity	80	26.5%	38	26.2%	42	26.8%
SAT area (mm2), mean (SD)	183.5	137.6	141.6	118.0	222.1	143.3	<0.001
VAT area (mm2), mean (SD)	30.8	19.1	29.5	20.8	31.9	17.4	0.27
VAT/SAT, mean (SD)	0.21	0.11	0.26	0.13	0.16	0.06	<0.001
Hepatic fat (%), median (IQR)	1.8	1.3–2.5	1.9	1.3–2.7	1.79	1.3–2.3	0.18

^aP-values calculated by two-tailed Student’s test or Mann-Whitney U Test for continuous variables and Chi-square test for continuous variables. Bolding indicates p<0.05.

^bData on pubertal stage was missing for 1 participant at visit 1.

^cPhysical activity was measured as the average energy expenditure over three days in metabolic equivalents (METs). Data was missing on physical activity for 6 participants at visit 1 and 6 participants at visit 2.

^dData on abdominal SAT and VAT were missing for 12 participants at visit 1.

Abbreviations: BMI, body mass index; VAT, visceral adipose tissue; SAT, subcutaneous adipose tissue.

Associations of childhood nutrient intakes with adolescent abdominal fats:

Associations of childhood nutrient intakes with abdominal SAT later in adolescence, based on stepwise-adjusted energy partition models and nutrient density models, are shown in Table 2. In energy partition models (energy intake not held constant), higher total fat, MUFA, and PUFA intakes in childhood were positively associated with higher abdominal SAT in adolescence, both in unadjusted models (Model 1) and confounder-adjusted models (Model 2); however, additionally adjusting for childhood SAT (Model 3) attenuated all associations to the null (Table 2). In nutrient density models adjusted for TEI, associations of childhood MUFA and PUFA intakes with adolescent SAT remained significant in unadjusted models (Model 1) and confounder-adjusted models (Model 2), but were again attenuated to the null after adjusting for childhood SAT in Model 3 (Table 2). In comparison, for abdominal VAT in adolescence, in both types of models (energy partition or nutrient density models), there were no associations between nutrient intakes in childhood and abdominal VAT later in adolescence in the full sample (Table 3).

Table 2:

Associations of nutrient intakes in childhood with abdominal subcutaneous adipose tissue (SAT) in adolescence.

	Model 1: Unadjusteda (n=302)		Model 2: + Confoundersb (n=294)		Model 3: + Childhood SATc (n=282)
A) Energy Partition Models:
Nutrient (kcal/d)	β (95% CI) d	p f	β (95% CI) d	p f	β (95% CI) d	p f
Energy	3.3 (−0.2, 6.8)	0.07	1.6 (−1.2, 4.4)	0.26	−0.3 (−2.5, 2.0)	0.81
Carbohydrates	−3.5 (−11.5, 4.6)	0.40	−2.1 (−8.2, 4.0)	0.49	−3.2 (−8.1, 1.7)	0.19
Starch	−1.9 (−22.6, 18.9)	0.86	0.7 (−15.2, 16.7)	0.93	0.2 (−12.5, 12.9)	0.98
Sugar	−5.9 (−16.9, 5.1)	0.29	−4.1 (−12.4, 4.2)	0.33	−4.8 (−11.5, 1.8)	0.16
Protein	−1.9 (−36.3, 32.4)	0.91	1.2 (−25.5, 27.8)	0.93	4.4 (−17.0, 25.8)	0.68
Animal Protein	3.1 (−27.6, 33.8)	0.84	−0.9 (−24.5, 22.8)	0.94	3.5 (−15.5, 22.6)	0.72
Total Fat	15.7 (4.6, 26.9)	0.006 *	8.5 (0.1, 17.1)	0.049	4.4 (−2.5, 11.2)	0.21
SFA	8.4 (−25.9, 42.8)	0.63	−3.6 (−30.2, 23.0)	0.79	1.2 (−20.1, 22.4)	0.91
MUFA	36.3 (6.0, 66.5)	0.019	25.1 (2.1, 48.1)	0.032	14.4 (−4.1, 33.0)	0.13
PUFA	92.6 (37.0, 148.2)	0.001 *	59.7 (16.1, 103.3)	0.008 *	29.3 (−6.3, 64.8)	0.11
B) Nutrient Density Models:
Nutrient (%TEI)	β (95% CI) e	p f	β (95% CI) e	p f	β (95% CI) e	p f
Carbohydrates	−15.4 (−29.2, −1.7)	0.028	−7.5 (−18.0, 2.9)	0.16	−6.5 (−15.0, 1.9)	0.13
Starch	−7.6 (−30.6, 15.4)	0.52	−0.4 (−18.0, 17.3)	0.97	1.4 (−12.6, 15.5)	0.84
Sugar	−14.1 (−28.4, 0.2)	0.05	−8.1 (−18.9, 2.7)	0.14	−7.2 (−15.9, 1.4)	0.10
Protein	4.1 (−35.2, 43.4)	0.84	2.0 (−28.4, 32.4)	0.90	5.7 (−18.8, 30.1)	0.65
Animal Protein	7.9 (−25.1, 41.0)	0.64	−0.2 (−25.6, 25.1)	0.99	4.1 (−16.2, 24.5)	0.69
Total Fat	21.8 (5.1, 38.5)	0.011 *	11.1 (−1.6, 23.9)	0.09	8.6 (−1.6, 18.9)	0.10
SFA	15.8 (−23, 54.5)	0.42	−3.4 (−33.5, 26.8)	0.83	3.9 (−20.2, 28.0)	0.75
MUFA	41.7 (7.1, 76.3)	0.018	26.7 (0.3, 53.0)	0.047	19.2 (−2.0, 40.4)	0.08
PUFA	90.4 (30.7, 150.1)	0.003 *	64.0 (17.0, 110.9)	0.008 *	37.5 (−0.6, 75.7)	0.05
Fiber	16.5 (−29.0, 62.0)	0.48	1.8 (−35.3, 38.9)	0.92	−6.8 (−36.8, 23.3)	0.66

^aModel 1: adjusted for energy intake from all other nutrients (energy partition models) or TEI (nutrient density models).

^bModel 2: adjusted for Model 1 covariates plus sex, age, race/ethnicity, household income, pubertal stage, physical activity, BMI category at visit 1, maternal DM exposure, and HEI-2010 total score at visit 1. 8 participants excluded due to missing data for household income (n=1), pubertal stage (n=1), or physical activity (n=6).

^cModel 3: adjusted for Model 2 covariates plus abdominal VAT in childhood. 12 participants excluded due to missing data for abdominal SAT at visit 1.

^dEstimated per 100 kcal/d increase in each nutrient.

^eEstimated per 5% increase in each nutrient (except fiber, per 5 g/1000 kcal increase).

^fBolding indicates p<0.05. Asterisk (*) indicates if below Bonferroni-adjusted p<0.017 (0.05/3 outcomes).

Abbreviations: SAT, subcutaneous adipose tissue; TEI, total energy intake; SFA, saturated fat; MUFA, monounsaturated fat; PUFA, polyunsaturated fat.

Table 3:

Associations of nutrient intakes in childhood with abdominal visceral adipose tissue (VAT) in adolescence.

	Model 1: Unadjusteda (n=302)		Model 2: + Confoundersb (n=294)		Model 3: + Childhood VATc (n=282)
A) Energy Partition Models:
Nutrient (kcal/d)	β (95% CI) d	p f	β (95% CI) d	p f	β (95% CI) d	p f
Energy	0.4 (−0.1, 0.9)	0.07	0.1 (−0.3, 0.6)	0.56	0.1 (−0.3, 0.5)	0.56
Carbohydrates	−0.2 (−1.3, 0.9)	0.69	−0.2 (−1.2, 0.8)	0.70	−0.2 (−1.0, 0.7)	0.72
Starch	0.5 (−2.4, 3.4)	0.72	0.5 (−2.2, 3.2)	0.73	0.3 (−1.9, 2.6)	0.76
Sugar	−0.5 (−2.0, 1.0)	0.51	−0.4 (−1.8, 1.0)	0.59	−0.3 (−1.5, 0.9)	0.66
Protein	2.6 (−2.2, 7.3)	0.29	2.3 (−2.2, 6.8)	0.32	3.6 (−0.2, 7.4)	0.07
Animal Protein	2.8 (−1.5, 7.0)	0.20	2.0 (−2.0, 6.0)	0.32	2.7 (−0.7, 6.1)	0.12
Total Fat	1.2 (−0.4, 2.7)	0.14	0.4 (−1.1, 1.8)	0.63	0.1 (−1.1, 1.3)	0.89
SFA	0.1 (−4.7, 4.8)	0.98	−1.6 (−6.1, 2.9)	0.50	−1.3 (−5.0, 2.5)	0.52
MUFA	3.0 (−1.3, 7.2)	0.17	1.7 (−2.2, 5.6)	0.40	0.7 (−2.7, 4.0)	0.69
PUFA	5.1 (−2.8, 12.9)	0.20	2.8 (−4.6, 10.3)	0.45	1.1 (−5.2, 7.5)	0.72
B) Nutrient Density Models:
Nutrient (%TEI)	β (95% CI) e	p f	β (95% CI) e	p f	β (95% CI) e	p f
Carbohydrates	−1.5 (−3.5, 0.4)	0.12	−0.7 (−2.5, 1.1)	0.43	−0.6 (−2.1, 0.9)	0.46
Starch	−0.4 (−3.6, 2.8)	0.82	0.1 (−2.8, 3.1)	0.93	0.2 (−2.3, 2.7)	0.89
Sugar	−1.3 (−3.3, 0.7)	0.19	−0.6 (−2.5, 1.2)	0.50	−0.5 (−2.0, 1.1)	0.54
Protein	3.0 (−2.5, 8.4)	0.28	2.3 (−2.8, 7.4)	0.38	3.7 (−0.7, 8.0)	0.10
Animal Protein	3.0 (−1.5, 7.6)	0.19	2.1 (−2.2, 6.4)	0.34	2.7 (−0.9, 6.3)	0.14
Total Fat	1.5 (−0.9, 3.8)	0.22	0.5 (−1.7, 2.6)	0.67	0.1 (−1.8, 1.9)	0.94
SFA	0.9 (−4.5, 6.3)	0.74	−1.1 (−6.2, 4.0)	0.66	−0.9 (−5.2, 3.4)	0.69
MUFA	3.3 (−1.5, 8.2)	0.18	1.8 (−2.7, 6.3)	0.43	0.7 (−3.1, 4.5)	0.71
PUFA	4.6 (−3.8, 13)	0.28	2.9 (−5.1, 10.9)	0.48	1.0 (−5.8, 7.8)	0.78
Fiber	−2.1 (−8.4, 4.2)	0.51	−4.4 (−10.7, 1.8)	0.17	−3.8 (−9.1, 1.6)	0.17

^aModel 1: adjusted for energy intake from all other nutrients (energy partition models) or TEI (nutrient density models).

^bModel 2: adjusted for Model 1 covariates plus sex, age, race/ethnicity, household income, pubertal stage, physical activity, BMI category at visit 1, maternal DM exposure, and HEI-2010 total score at visit 1. 8 participants excluded due to missing data for household income (n=1), pubertal stage (n=1), or physical activity (n=6).

^bModel 3: adjusted for Model 2 covariates plus abdominal VAT in childhood. 12 participants excluded due to missing data for abdominal VAT at visit 1.

^dEstimated per 100 kcal/d increase in each nutrient.

^eEstimated per 5% increase in each nutrient (except fiber, per 5 g/1000 kcal increase).

^fBolding indicates p<0.05. Asterisk (*) indicates if below Bonferroni-adjusted p<0.017 (0.05/3 outcomes).

Abbreviations: VAT, visceral adipose tissue; TEI, total energy intake; SFA, saturated fat; MUFA, monounsaturated fat; PUFA, polyunsaturated fat.

Associations of childhood nutrient intakes with adolescent hepatic fat:

We next examined associations of childhood nutrient intakes with log-transformed hepatic fat in adolescence using the same modeling approach; though, we did not report the results from Model 3 (adjusted for childhood hepatic fat), since hepatic fat was only assessed at visit 2 in adolescence. In energy partition models, higher starch intake in childhood was associated with higher adolescent hepatic fat, but this only reached significance in confounder-adjusted models (Model 2) (Table 4). In nutrient density models adjusted for TEI, the positive association of childhood starch intake with adolescent hepatic fat followed a similar pattern and remained significant in confounder-adjusted models (Table 4).

Table 4:

Associations of nutrient intakes in childhood with log-transformed hepatic fat in adolescence.

	Model 1: Unadjusteda (n=302)		Model 2: + Confoundersb (n=294)
A) Energy Partition Models:
Nutrient (kcal/d)	β (95% CI) c	p e	β (95% CI) c	p e
Energy	1.01 (0.99, 1.02)	0.14	1.00 (0.99, 1.02)	0.53
Carbohydrates	1.00 (0.98, 1.03)	0.76	1.01 (0.98, 1.03)	0.71
Starch	1.07 (0.99, 1.14)	0.05	1.07 (1.01, 1.15)	0.039
Sugar	0.98 (0.95, 1.02)	0.33	0.98 (0.95, 1.02)	0.39
Protein	0.96 (0.86, 1.08)	0.53	0.91 (0.82, 1.02)	0.12
Animal Protein	0.99 (0.89, 1.09)	0.80	0.94 (0.85, 1.04)	0.26
Total Fat	1.02 (0.98, 1.06)	0.27	1.01 (0.98, 1.05)	0.51
SFA	0.98 (0.87, 1.09)	0.68	0.96 (0.86, 1.08)	0.48
MUFA	1.06 (0.96, 1.18)	0.22	1.05 (0.95, 1.16)	0.31
PUFA	1.09 (0.91, 1.31)	0.36	1.10 (0.91, 1.33)	0.32
B) Nutrient Density Models:
Nutrient (%TEI)	β (95% CI) d	p e	β (95% CI) d	p e
Carbohydrates	1.00 (0.95, 1.04)	0.85	1.01 (0.96, 1.06)	0.71
Starch	1.07 (0.99, 1.15)	0.08	1.09 (1.01, 1.17)	0.025
Sugar	0.97 (0.92, 1.01)	0.17	0.98 (0.93, 1.02)	0.31
Protein	0.96 (0.85, 1.09)	0.53	0.90 (0.79, 1.02)	0.10
Animal Protein	0.98 (0.88, 1.09)	0.75	0.93 (0.84, 1.04)	0.22
Total Fat	1.01 (0.96, 1.07)	0.71	1.00 (0.95, 1.06)	0.89
SFA	0.95 (0.84, 1.08)	0.43	0.93 (0.82, 1.06)	0.28
MUFA	1.05 (0.94, 1.17)	0.42	1.04 (0.93, 1.16)	0.51
PUFA	1.07 (0.88, 1.30)	0.50	1.09 (0.89, 1.34)	0.42
Fiber	1.00 (0.86, 1.16)	0.99	0.96 (0.82, 1.13)	0.65

^aModel 1: adjusted for energy intake from all other nutrients (energy partition models) or TEI (nutrient density models).

^bModel 2: adjusted for Model 1 covariates plus sex, age, race/ethnicity, household income, pubertal stage, physical activity, BMI category at visit 1, maternal DM exposure, and HEI-2010 total score at visit 1. 8 participants excluded due to missing data for household income (n=1), pubertal stage (n=1), or physical activity (n=6).

^cEstimated per 100 kcal/d increase in each nutrient. Beta-coefficients have been back-transformed and reflect the ratio of geometric means for hepatic fat.

^dEstimated per 5% increase in each nutrient (except fiber, per 5 g/1000 kcal increase). Beta-coefficients have been back-transformed and reflect the ratio of geometric means for hepatic fat.

^eBolding indicates p<0.05. Asterisk (*) indicates if below Bonferroni-adjusted p<0.017 (0.05/3 outcomes).

Abbreviations: TEI, total energy intake; SFA, saturated fat; MUFA, monounsaturated fat; PUFA, polyunsaturated fat.

Sex-specific findings:

We found evidence of effect modification by sex on associations of total and animal protein intake in childhood with abdominal VAT in adolescence, both in energy partition models (p=0.024 for sex*total protein interaction and p=0.024 for sex*animal protein interaction) and nutrient density models (p=0.013 for sex*total protein interaction and p=0.020 for sex*animal protein interaction). Sex-stratified estimates are shown in Table 5. In all models, intakes of total and animal protein in childhood were associated with higher VAT in adolescence in boys, but not girls (Table 5).

Table 5:

Sex-specific associations of total protein and animal protein intakes in childhood with abdominal visceral adipose tissue (VAT) in adolescence in the analytical sample (n=302).

		Model 1: Unadjusteda (n=302)		Model 2: + Confoundersb (n=294)		Model 3: + Childhood VATc (n=282)
A) Energy Partition Models:
Nutrient (kcal/d)	Sex	β (95% CI) d	p f	β (95% CI) d	p f	β (95% CI) d	p f
Total Protein	Girls	−0.9 (−6.8, 4.9)	0.75	−1.3 (−6.9, 4.2)	0.64	−0.6 (−5.3, 4.0)	0.79
	Boys	5.8 (0.5, 11.2)	0.033	4.8 (−0.2, 9.8)	0.06	6.4 (2.2, 10.6)	0.003 *
Animal Protein	Girls	−0.9 (−6.5, 4.6)	0.74	−2.0 (−7.4, 3.4)	0.46	−2.3 (−6.8, 2.3)	0.33
	Boys	6.9 (1.6, 12.2)	0.011 *	5.3 (0.3, 10.3)	0.036	6.7 (2.6, 10.8)	0.002 *
B) Nutrient Density Models:
Nutrient (%TEI)	Sex	β (95% CI) e	p f	β (95% CI) e	p f	β (95% CI) e	p f
Total Protein	Girls	−3.6 (−11.2, 4.1)	0.36	−4.1 (−11.2, 3.0)	0.26	−1.9 (−8.0, 4.1)	0.53
	Boys	10.2 (2.5, 17.8)	0.009 *	8.8 (1.6, 15.9)	0.016 *	9.2 (3.2, 15.1)	0.003 *
Animal Protein	Girls	−1.5 (−7.6, 4.6)	0.64	−2.5 (−8.2, 3.3)	0.40	−2.0 (−6.9, 2.8)	0.41
	Boys	9.4 (2.6, 16.3)	0.007 *	7.7 (1.3, 14.0)	0.018	8.3 (3.0, 13.6)	0.002 *

^aModel 1: adjusted for energy intake from all other nutrients (energy partition models) or TEI (nutrient density models).

^bModel 2: adjusted for Model 1 covariates plus sex, age, race/ethnicity, household income, pubertal stage, physical activity, BMI category at visit 1, maternal DM exposure, and HEI-2010 total score at visit 1. 8 participants excluded due to missing data for household income (n=1), pubertal stage (n=1), or physical activity (n=6).

^cModel 3: adjusted for Model 2 covariates plus abdominal VAT in childhood. 12 participants excluded due to missing data for abdominal VAT at visit 1.

^dEstimated per 100 kcal/d increase in each nutrient.

^eEstimated per 5% increase in each nutrient (except fiber, per 5 g/1000 kcal increase).

^fBolding indicates p<0.05. Asterisk (*) indicates if below Bonferroni-adjusted p<0.017 (0.05/3 outcomes).

Abbreviations: VAT, visceral adipose tissue; TEI, total energy intake.

Post-hoc analyses:

Because the associations between total fat, MUFA, and PUFA intakes in childhood and abdominal SAT in adolescence were attenuated after adjusting for childhood SAT, we performed a post-hoc analysis to examine associations between nutrient intakes in childhood with change in SAT from childhood to adolescence to assess whether the associations reflected an increase in SAT from childhood or reflected an association already present in childhood. As shown in Table S3, higher PUFA intake in childhood was associated with an increase in SAT from childhood to adolescence, but only in nutrient density models [β (95% CI): 39.4 (2.8, 76.0) mm² per 5% TEI]. There were no other associations between total fat intake or MUFA intake with change in abdominal SAT from childhood to adolescence (Table S3), suggesting most associations between these nutrient intakes and abdominal SAT were already present earlier in childhood.

DISCUSSION:

The etiology of body fat partitioning in youth is complex and multifactorial, but lifestyle factors – including diet – are likely important contributors to inter-individual differences. In this study, we examined the influence of dietary intakes in childhood on future body fat deposition in adolescence, particularly in terms of abdominal and ectopic liver fats measured by MRI. Our analyses revealed that certain nutrient intakes in childhood exhibit differential associations with abdominal and hepatic fat deposition later in adolescence. Specifically, we found that childhood MUFA and PUFA intakes were associated with higher abdominal SAT in adolescence and childhood starch intake was associated with higher hepatic fat in adolescence. In boys only, we also found that childhood total protein intake, particularly as animal protein, was associated with higher abdominal VAT in adolescence. Importantly, most findings were similar when adjusted for potential confounders, including TEI in nutrient density models, supporting independent pathways linking these nutrient intakes to specific body fat depots. However, associations of childhood MUFA and PUFA intakes with adolescent abdominal SAT were markedly attenuated after adjusting for childhood abdominal SAT, which suggests that associations were partially already present in childhood and that interventions aiming to reduce abdominal SAT may need to target these intakes even earlier in childhood. It should also be noted that the above findings were based on estimates and 95% CI, and only a few associations survived multiple hypothesis testing corrections. Thus, most associations we found between childhood nutrient intakes and adolescent abdominal or hepatic fat should be considered modest and will need to be confirmed in other prospective studies.

Although many studies have examined the nutritional determinants of childhood obesity in general, far fewer have examined the determinants of body fat partitioning in youth. This is particularly true for large, prospective cohort studies, due to the cost and time requirements of the imaging techniques (such as MRI) needed to accurately measure specific body fat depots. Regarding abdominal fat deposition, we found a novel association between childhood unsaturated fat intakes and adolescent abdominal SAT in this study. This effect was particularly strong for PUFA intake, which predicted an increase in abdominal SAT from childhood to adolescence, and may relate to the pro-inflammatory and adipogenic potential of n-6 PUFAs(32), especially in the context of a high dietary ratio of n-6 to n-3 PUFAs, which is characteristic of a Western Diet. Future studies will be needed to fully elucidate the underlying mechanisms at play, as well as to determine whether the effect of PUFAs on adolescent SAT depends on the nutrient being substituted (i.e., carbohydrates versus fat).

We also found that higher childhood protein intake, especially from animal sources, was associated with higher adolescent VAT, but only in boys. Although the literature on animal protein intake and VAT is limited, other studies in children(33, 34) and adults(35, 36) have also observed an association between higher intakes of animal protein and/or certain amino acid-derived metabolites with abdominal adiposity measured by anthropometrics (i.e. waist circumference or waist-to-height ratio). Our findings, therefore, add to this body of literature by showing that animal protein intake in childhood may be specifically associated with MRI-measured abdominal VAT in boys. One proposed mechanism for these associations is the ability of animal protein to upregulate insulin and insulin-like growth factor-1(37), which stimulates adipocyte proliferation and differentiation(38), and interacts with growth hormone to regulate energy metabolism in both the liver and adipose tissue(39). Our finding that this association was only in boys may reflect the established sexual dimorphism of visceral adiposity, and suggests that animal protein intake may interact with mechanisms that predispose boys to more VAT compared to girls, including differential levels of reproductive and growth hormones; differential distribution of estrogen receptors in abdominal versus peripheral fat; and/or differential expression of lipolytic (β1–2) and antilipolytic (α2) adrenergic receptors in VAT(18).

Regarding adolescent hepatic fat, we found a positive association with childhood starch intake, which is also difficult to interpret given many different foods contain starch. Since we did not find an association between childhood fiber intake and adolescent hepatic fat, we hypothesize that this association was more likely driven by high-starch foods that are low in fiber, such as refined grains, but this will need to be tested in the future. Unexpectedly, we found no associations between childhood total sugar intake and adolescent abdominal VAT or hepatic fat, which conflicts with experimental studies in children showing that dietary sugar restriction is associated with reductions in these body fat depots (14, 15), as noted in the introduction. We also found no associations with childhood fiber intake, despite dietary fiber often being associated with a more optimal body fat distribution in adolescence (40, 41). These discrepant findings may be due to differences in sample characteristics, since most other studies in this area have focused on youth with obesity, compared to the generally healthy sample of youth in EPOCH. It may also be due to the prospective nature of this study, with approximately 6 years of follow-up between exposure and outcome assessments, suggesting that intakes of these nutrients (i.e., sugar, fiber) more proximal to adolescence may be more relevant to body fat partitioning patterns than intakes earlier in childhood.

Another potential explanation for some null findings is that, even after excluding TEI under-reporters, there was still some degree of under-reporting in the sample. A limitation of this study is, therefore, our reliance on self-reported dietary intake data, which can be prone to social desirability bias, particularly in individuals with obesity(42), and may contribute to dietary under-reporting, resulting in attenuated associations. Thus, our findings may be conservative estimates of true associations. There may also be measurement error specifically associated with data derived from FFQs due to incomplete food lists or inaccuracies in frequency or portion size estimations(43). At the same time, FFQs are the most common and feasible approach for large epidemiological studies, such as EPOCH, due to their low respondent burden and ability to rank individuals according to longer-term, habitual intake, which is most relevant to the development of chronic diseases, such as NAFLD. In addition, the FFQ used in EPOCH was modified for and validated in children (22, 23) and we took several additional steps to limit error, such as excluding under-reporters and adjusting for energy intake in nutrient density models(44).

Other limitations include the observational nature of this study, which limits causal inference. We did not measure hepatic fat at visit 1 in childhood; therefore, we were unable to adjust for childhood levels of hepatic fat, similar to the modeling approach used for abdominal SAT and VAT. We relied on other self-reported variables, as well, such as for pubertal stage and physical activity, which may be prone to measurement error. The sample was from one geographic region in the United States (Colorado) and participants were selected based on exposure to maternal DM during pregnancy; while we adjusted for the latter exposure in models, this may reduce the generalizability. In addition, we mainly interpreted our findings based on raw estimates and 95% CIs to avoid type 1 error(45), but it should be noted that several associations did not survive Bonferroni-correction and should be interpreted with caution.

Strengths include the use of complementary modeling strategies to examine associations between nutrient intakes and body fat deposition before and after adjusting for energy intake. Specifically, using energy partition models, we were able to evaluate the effect of an absolute increase in each nutrient, and using multivariate nutrient density models, we were able to evaluate the isocaloric effect of each nutrient holding total energy intake constant. We used state-of-the-art MRI technology to accurately assess abdominal SAT and VAT depots and hepatic fat in adolescence, the primary outcomes of interest, ensuring the reliability and reproducibility of findings. The longitudinal nature of the cohort enabled us to better establish temporality in evaluating associations between childhood nutrient intakes and future adiposity outcomes in adolescence, which not only limited potential reverse causality but also provided insights that may be used to inform future prevention efforts. Lastly, the EPOCH cohort was well-characterized in terms of anthropometric, lifestyle, behavioral, and biological variables, multi-ethnic, and included both lean individuals and individuals with overweight or obesity. Thus, as a general risk population, our findings represent associations that are present before the development of severe body fat distribution phenotypes and may be used to inform dietary guidelines.

In conclusion, the results from this prospective analysis suggest that childhood nutrient intakes exhibit differential associations with adolescent abdominal and hepatic fats; whereby PUFA and MUFA intakes were associated with adolescent abdominal SAT, total and animal protein intakes were associated with adolescent abdominal VAT in boys, and starch intake was associated with adolescent hepatic fat. These findings may be used to inform dietary interventions aiming to promote a healthier body fat distribution in youth. Because some associations, especially between childhood fat intakes and abdominal SAT, were attenuated after adjusting for childhood adiposity, this suggests that such interventions may be particularly effective if implemented earlier in childhood, before the progression of body composition phenotypes into adolescence.

STUDY IMPORTANCE:

What is already known on this subject?

• Regional body fat distribution, especially as abdominal and/or ectopic fat, appears to be a strong risk factor for cardiometabolic dysfunction in youth.

• Diet and nutrition in childhood may be predictive of adverse body fat distribution later in life, but few studies have assessed this using a prospective study design and more sophisticated measures of body composition.

What are the new findings in your manuscript?

• Using data from a longitudinal cohort in Colorado, we showed that nutrient intakes in childhood (~10 years) are differentially associated with different types of abdominal and ectopic fat deposition later in adolescence (~16 years).

• Specifically, unsaturated fat intake in childhood predicted higher abdominal subcutaneous fat, higher animal protein intake in childhood predicted higher abdominal visceral fat, and higher starch intake in childhood predicted higher hepatic fat in adolescence.

How might your results change the direction of research or the focus of clinical practice?

• This study provides insight into the potential influence of nutrient intakes earlier in childhood on future abdominal and hepatic fat deposition in adolescence and may therefore be used to inform interventions aiming to modulate body fat distribution patterns in youth.