Smaller subcortical volume relates to greater weight gain in girls with initially healthy weight

Abstract

Objective:

Among 3,614 youth ages 9-to-12-years-old initially without overweight/obesity (12% [n=385] developed overweight/obesity), we examined the natural progression of weight gain and brain structure development during a two-year period at high risk for obesity (e.g., pre- and early adolescence) to determine whether (1) variation in maturational trajectories of the brain regions contributes to weight gain and/or (2) weight gain contributes to altered brain development.

Methods:

Data were gathered from the ABCD Study. Linear mixed-effects regression models controlled for puberty, caregiver education, handedness, and intracranial volume (random effects: MRI scanner, subject). As pubertal development occurs earlier in girls, analyses were stratified by sex.

Results:

For girls, but not boys, independent of puberty, greater increases in BMI were driven by smaller volumes over time in the bilateral accumbens, amygdala, hippocampus, and thalamus, right caudate and ventral diencephalon, and left thalamus (p’s<0.05).

Conclusions:

The results suggest a potential phenotype for identifying obesity risk as underlying differences among regions involved in food intake related to greater weight gain in girls, but not in boys. Importantly, two years of weight gain may not be sufficient to alter brain development, highlighting early puberty as a critical time to prevent negative neurological outcomes.

1. Introduction

Nearly 20% of U.S. children have obesity,¹ and despite increasing efforts, these rates are expected to increase as children transition to adulthood.² Childhood obesity is a multifactorial and preventable disease with devastating preventable medical (e.g., cancer, heart disease, liver disease, dementia) and financial consequences.³ Adolescence is a period in which obesity risk is the highest since infancy,⁴ and extreme weight gain during this time may have greater metabolic consequences.⁵ In part, this is due to pubertal onset where emerging sex-specific hormones trigger rapid changes in body composition.⁴ Pubertal hormones also trigger restructuring of brain regions that regulate appetite and food intake.^6,7 Numerous studies have shown that variations in brain structure and function of regions that control appetite regulation are related to obesity.^8–12 However, it is not clear whether maturational variations in the brain predate or follow weight gain during this period. Understanding the directionality between brain and weight fluctuations during this period may provide insight into critical windows for obesity prevention and intervention efforts.

Substantial brain development occurs postnatally as the brain does not reach its adult size until early childhood.¹³ However, during adolescence, pubertal onset triggers refining (i.e., maturation) of processes for adult-like brain functioning, including changes to regions involved in food intake control that may be necessary to stimulate physical growth for reproductive purposes. Yet, maturation trajectories differ by sex¹⁴ (as girls undergo puberty earlier than boys¹⁵) and by region,⁶ which may have consequences for obesity development. For example, regions involved in motivated behavior and reward processing (e.g., nucleus accumbens) mature slower than those involved in memory, learning, and hormonal control of food intake (e.g., amygdala, hippocampus);⁶ Maturation mismatches translate in the form of excessive risk taking and impulsivity that characterize adolescence,¹⁶ while greater impulsive behavior has also been associated with overeating and weight gain.^17–19 Additionally, neuroimaging studies have noted both structural and functional variation among regions associated with impulsivity and obesity in children and adolescents.^{8–12,20–22} However, the inclusion of youth with obesity in the analyses makes it difficult to disentangle the underlying directional mechanisms involved.

Recently, underlying variations in behavioral assessments of impulsivity have been shown to A) related to weight gain²³ and B) predict unhealthy eating.²⁴ These findings suggest a mechanism promoting obesogenic behavior. However, whether these associations extend to regions of the brain involved in impulsivity and food intake control is unknown. Understanding these associations is important for clinical utility. First, examining whether underlying structural differences predispose youth to weight gain will enhance our knowledge of the causal mechanisms of obesity, which are crucial for effective prevention efforts. Second, excess weight has been associated with changes in brain structure,²² neuroinflammation and cognitive decline.²⁵ While diet, such as the overconsumption of high fat and sugar, is presumed to play a role in changes in the brain via neuroinflammation,²⁶ the mechanisms causing weight gain are multifaceted (e.g., lack of exercise, stress). As such, understanding whether weight gain is related to changes in the brain may be informative from a clinical perspective as weight is a biomarker that can be easily measured in a clinician’s office. Moreover, data from pregnancy models suggest that normative weight gain during pregnancy, but not diet, causes low-grade inflammation.²⁷ Thus, although the mechanisms causing weight gain are important to understand, there is clinical utility in investigating the relationship between weight gain and neurodevelopment. Currently, no studies have evaluated whether weight gain during adolescence is associated with structural changes in the brain.

Therefore, the following study fills gaps in the literature by examining the bidirectional relationship between developmental trajectories of subcortical volume and weight gain during a two-year period among initially healthy weight youth (aged 9-to-12-years-old) stratified by sex. We focus on volumetric changes to the subcortex as many of these regions are directly involved in appetite regulation and impulsivity. We examined: 1) whether underlying differences in subcortical volume growth relates to increases in body mass index (BMI, where a positive increase is considered a proxy for increased adiposity); and 2) whether weight gain over two-years relates to variations in subcortical volume maturation. We hypothesize that variation in subcortical volume predates weight gain, across both sexes, however, that weight gain effects on subcortical volume would be more robust in girls given the combination of the chronological ages of the youth studied and earlier pubertal maturation in girls.^4,15 Understanding these associations during adolescence is of critical importance, as rodent models suggest that neuroinflammation during this period of development may have irreversible effects on the brain and cognition.²⁸ The data were obtained from the Adolescent Brain Cognitive Development Study (ABCD Study^® www.ABCDStudy.org; n=2,794, ages 9/10-to-11/12-years-old). Notably, by the two-year period, 12% of youth transitioned to have overweight or obesity, allowing for insights into the natural progression of weight gain by sex during early puberty.

2. Methods

2.1. Study design and participants.

The 10-year, 21-site ABCD Study^® started between 2016 and 2018 when the youth were 9/10-years-old. Recruitment matched the American Continuing Survey demographics of the United States; multiple studies have described the study design, assessments, and objectives (see www.ABCDStudy.org for a list of publications and protocol design). The ABCD Study is responsible for obtaining caregiver consent and youth assent, and protocols were approved by a centralized Institutional Review Board at the University of California San Diego. The current manuscript utilizes the 4.0 data release (https://doi.org/10.15154/1503209) and focuses on data collected at the baseline and two-year follow-up visits. The ABCD Study had minimal exclusion criteria but to obtain a sample optimal for the current manuscript’s analyses, additional exclusion criteria were applied (e.g., had overweight/obesity at baseline, met diagnostic criteria for an eating disorder; see Supplemental Materials).

2.2. Physical assessments.

Height and weight were assessed by a trained researcher and converted into BMI (kg/m²). Our analyses included raw BMI (i.e., not standardized scores) as the variable of interest as standardized scores should not be utilized in longitudinal research questions.²⁹ Puberty (i.e., Tanner staging) was assessed by caregiver and youth reports. Additional details are reported in the Supplemental Materials.

2.3. Demographics.

Caregivers reported on the youth’s race and ethnicity, sex at birth, date of birth, family structure, and the caregiver’s highest education.

2.4. Handedness.

Handedness (left, right, or ambidextrous) was assessed via the Youth Edinburgh Handedness Short Form.

2.5. Image acquisition and preprocessing.

MRI data were collected with 29, 3T MRI scanners. Structural MRI data were collected with a T₁-weighted acquisition and processed with FreeSurfer to obtain estimates of subcortical volume (mm³). The ABCD Study Data Analytics Core was responsible for preprocessing, conducting basic quality control and providing exclusion recommendations. As such, data were excluded if the T₁ image was not usable, FreeSurfer failed, or if derived results did not exist (quality control variable labeled: imgincl_t1w_include). The present study examined data from 8 bilateral subcortical volume regions of interest (ROIs; e.g., the accumbens area, amygdala, caudate, hippocampus, thalamus, pallidum, putamen, and ventral diencephalon [ventral DC]).

3. Statistical Analyses

Analyses were conducted using Python (version 3.10.11). The demographic characteristics were analyzed using chi-square tests and t tests where appropriate (Table 1, Figure 1 shows the distribution of BMI by sex and year). The data were checked for implausible values, skewness, kurtosis, and multicollinearity using the variance inflation factor (statsmodels, version 0.13.5). Outliers were removed if the values were 3 standard deviations below or above the mean. Linear mixed-effects models were conducted using the pymer4 (version 0.8.0) (https://github.com/ejolly/pymer4)³⁰ to examine: Model 1) whether variations in BMI (i.e., higher weight gain) were explained by differences in subcortical volume for each ROI over Time (e.g., subcortical development trajectories; ROI*Time); and Model 2) whether variation in subcortical volume trajectories was associated with changes in BMI over time (BMI*Time). Both models included fixed effects for puberty, highest caregiver education, handedness, and intracranial volume, with crossed random intercepts accounting for MRI scanner and subject ID random effects. A crossed random effect structure was chosen over nested to account for sites that had multiple scanners and that youth may have been scanned on more than one scanner at each site. Covariates were selected based on the literature on potential confounders for our outcomes. Continuous independent variables (e.g., puberty, intracranial volume, ROI [Model 1 only], BMI [Model 2 only] were transformed into standardized scores using scikit’s standard scaler package (version 1.2.1). Fixed effects were effects coded. Multiple comparison correction was conducted with the Benjamini-Hochberg method to control for false discovery rate.

Table 1.

Participant characteristics in the whole and subsample, stratified by sex.

	Whole sample (n=11,878)	Subsample (n=3,614)		Boys (n=1,834)	Girls (n=1,780)
Variable			p			p
Sex, n (%)
Male	6193 (52.1)	1834 (50.7)	0.210	1834
Female	5684 (47.8)	1780 (49.3)			1780
Missing	3 (0)
Race, n (%)
White	7525 (64.3)	2580 (72)		1311 (72.2)	1269 (71.9)	0.864
Black	1870 (16)	382 (10.7)	<0.001	201 (11.1)	181 (10.2)
Asian	275 (2.3)	87 (2.4)		40 (2.2)	47 (2.7)
AIAN/NHPI	78 (0.7)	22 (0.6)		12 (0.7)	10 (0.6)
Other	525 (4.5)	127 (3.5)		62 (3.4)	65 (3.7)
Multi-race	1434 (12.2)	384 (10.7)		190 (10.5)	194 (11.0)
Ethnicity, n (%)
Latino/a/x	2411 (20.6)	568 (15.9)	<0.001	267 (14.7)	301 (17.1)	0.057
Non-Latino/a/x	9314 (79.4)	3012 (84.1)		1550 (85.3)	1462 (82.9)
Caregiver report of education, n (%)
<HS	568 (4.8)	99 (2.7)	<0.001	49 (2.7)	50 (2.8)	0.353
HS/GED	1080 (9.1)	217 (6)		121 (6.6)	96 (5.4)
Some College	2978 (25.1)	754 (20.9)		397 (21.6)	357 (20.1)
BA degree	2969 (25)	1032 (28.5)		519 (28.3)	513 (28.8)
Postgraduate degree	3988 (33.6)	1512 (41.9)		748 (40.8)	764 (42.9)
Missing	295 (2.5)			1311 (72.2)	1269 (71.9)
Baseline Weight Class, n (%)
Underweight	468(3.9)			100 (100)	100 (100)
Healthy Weight	7602(64)	3614 (100)	<0.001
Overweight	1802(15.2)
Obese	1992(16.8)
Missing	16(0.1)
Y2 Weight Class, n (%)
Underweight	468 (3.9)
Healthy Weight	7602 (64)	2811 (77.8)	<0.001	1407 (76.7)	1404 (78.9)	0.255
Overweight	1802 (15.2)	354 (9.8)		181 (9.9)	173 (9.7)
Obese	1992 (16.8)	31 (0.9)		19 (1.0)	12 (0.7)
Extreme Weight Gain, n (%)	686 (5.7)	224 (22.8)		99 (20.8)	124 (24.8)	0.164
Handedness, n (%)				227 (12.4)	191 (10.7)
Right	9081(79.6)	2907 (80.4)
Left	817(7.2)	248 (6.9)	<0.001
Both	1506(13.2)	459 (12.7)
Age in months, mean (SD)
Baseline	119 (7.5)	119 (7.4)	0.836	118.9 (7.4)	119.1 (7.4)	0.435
Y2	143.5 (7.8)	143.2 (7.7)	0.044	143.1 (7.7)	143.3 (7.7)	0.435
Puberty, mean (SD)
Baseline	2 (0.8)	1.9 (0.8)	<0.001	1.6 (0.5)	2.2 (0.9)	<0.001
Y2	2.7 (1)	2.7 (1)	<0.001	2.0 (0.7)	3.4 (0.7)	<0.001
BMI, mean (SD)
Baseline	18.8 (4.2)	16.9 (1.4)	<0.001	16.9 (1.3)	17.0 (1.5)	0.006
Y2	20.6 (4.9)	18.7 (2.2)	<0.001	18.4 (2.1)	18.9 (2.3)	<0.001

Note. AIAN/NHPI = American Indian, Alaska Native/Native Hawaiian, Pacific Islander; HS = high school; GED = Generalized Education Degree; BA = bachelor’s degree; Y2 = year 2 follow-up (ages 11/12-years-old; BMI = body mass index; Descriptive statistics are displayed by self-reported race only for interpretation of sample diversity. P-values reflect chi-squared and t-tests where appropriate. Data are mean (SD) or frequency and percentage (%). p<0.05 (bold font) for chi-squared and t-tests were appropriate. Bold font indicates significant differences between groups.

Figure 1.

A) Distribution of body mass index (BMI) at baseline (9/10-years-old) for boys (blue) and girls (pink). B) Distribution of BMI at year 2 for boys and girls. C) Changes in BMI over time for boys and girls. D) Distribution of BMI by baseline and year 2 follow-up (Y2; ages 11/12-years-old) for both boys and girls.

Model 1 (Variation in subcortical volume precedes weight gain): BMI ~ ROI + Time + ROI*Time + Puberty + Caregiver’s highest education + handedness + intracranial volume + (1 | MRI Scanner ID) + (1 | Subject)

Model 2 (Weight gain causes changes in subcortical volume): ROI ~ BMI + Time + BMI*Time + Puberty + Caregiver’s highest education + handedness + intracranial volume + (1 | MRI Scanner ID) + (1 | Subject)

4. Results

4.1. Sample characteristics.

There were 3,614 youth who had usable data. At baseline, all youth were selected because they were of a healthy weight, but by the two-year follow-up, 354 youth (9.8%) had transitioned to have overweight (≥85^th BMI percentile) and 31 (0.9%) had obesity (≥95^th BMI percentile). The sample was 50.7% male (n=1,834), 80.4% right-handed; 15.9% of caregivers-reported Latino/a/x, and 29% had caregivers with an education of less than a bachelor’s degree. On average, youth were 9.9 years-old at baseline with an average BMI of 16.9 (kg/m²). At year 2, the mean age was 11.9 years, and the average BMI was 18.7 (kg/m²). All the demographic data are reported in Table 1. There were no differences between boys and girls in our analyses on any sociodemographic variables (e.g., race, ethnicity, education). As expected, girls had higher scores for puberty and BMI. Developmental trajectories for each ROI are plotted separately by sex in Figures S1 and S2.

4.2. Does underlying variation in subcortical volume relate to weight gain?

4.2.1. Boys.

There was no evidence of associations between any ROI*Time and BMI.

4.2.2. Girls.

Mixed models, controlling for puberty, highest caregiver education, intracranial volume, and handedness, revealed significant ROI*Time interactions with respect to BMI for the bilateral accumbens, amygdala, hippocampus, and thalamus (all p’s<0.05). There was also a significant ROI*Time interaction for the right caudate and ventral DC and left pallidum (p’s<0.05; Table 2 details the list of effects). These patterns of results suggest that in comparison to baseline (ages 9/10-years-old), greater change in BMI, was driven by smaller subcortical volume change over time (see Figure 2; estimated regression coefficients are presented in Figure 3). No significant interactions or main effects were observed for the left caudate, right ventral DC, or bilateral putamen.

Table 2:

The relationship between subcortical volume and body mass index (BMI) over time.

BMI ~ ROI*Time + puberty + education + intracranial volume + handedness + (1|MRI scanner / subject ID)
		Boys				Girls
		ROI (ME)		ROI*Time		ROI (ME)		ROI*Time
ROI	H	β (95% CI)	p	β (95% CI)	p	β (95% CI)	p	β (95% CI)	p
Accumbens	L	−0.06 (−0.12, 0.01)	0.12	−0.07 (−0.16, 0.01)	0.10	−0.04 (−0.12, 0.03)	0.27	−0.17 (−0.27, −0.08)	<0.001*
	R	−0.04 (−0.11, 0.03)	0.27	−0.04 (−0.12, 0.04)	0.33	−0.08 (−0.15, 0.0)	0.06	−0.18 (−0.27, −0.09)	<0.001*
Amygdala	L	−0.04 (−0.12, 0.03)	0.37	−0.07 (−0.16, 0.01)	0.11	−0.06 (−0.15, 0.02)	0.16	−0.19 (−0.28, −0.09)	<0.001*
	R	−0.01 (−0.09, 0.07)	0.87	−0.03 (−0.12, 0.05)	0.48	−0.01 (−0.1, 0.08)	0.81	−0.17 (−0.27, −0.07)	0.001*
Caudate	L	−0.02 (−0.1, 0.05)	0.55	−0.06 (−0.14, 0.02)	0.17	0.02 (−0.07, 0.11)	0.62	−0.06 (−0.16, 0.03)	0.17
	R	−0.0 (−0.08, 0.07)	0.87	−0.04 (−0.12, 0.04)	0.38	−0.01 (−0.1, 0.08)	0.82	−0.1 (−0.19, −0.01)	0.03*
Hippocampus	L	−0.08 (−0.16, 0.01).	0.07	−0.07 (−0.15, 0.02)	0.13	0.05 (−0.05, 0.15)	0.32	−0.2 (−0.29, −0.1)	<0.001*
	R	0.01 (−0.08, 0.09)	0.92	−0.02 (−0.1, 0.06)	0.66	0.03 (−0.07, 0.13)	0.55	−0.21 (−0.3, −0.11)	<0.001*
Pallidum	L	−0.05 (−0.12, 0.02)	0.14	−0.09 (−0.17, −0.0)	0.04	−0.01 (−0.09, 0.08)	0.85	−0.17 (−0.27, −0.07)	0.001*
	R	−0.05 (−0.12, 0.03)	0.24	−0.1 (−0.18, −0.01)	0.03	0.08 (−0.0, 0.16)	0.05	−0.08 (−0.18, 0.01)	0.10
Putamen	L	−0.0 (−0.08, 0.08)	0.93	−0.01 (−0.09, 0.07)	0.93	0.04 (−0.05, 0.12)	0.41	−0.04 (−0.13, 0.06)	0.45
	R	−0.01 (−0.09, 0.06)	0.74	−0.02 (−0.1, 0.07)	0.72	−0.0 (−0.09, 0.09)	0.98	−0.1 (−0.2, −0.01)	0.04
Thalamus	L	0.02 (−0.07, 0.1)	0.68	−0.05 (−0.14, 0.03)	0.21	0.01 (−0.09, 0.11)	0.80	−0.21 (−0.3, −0.11)	<0.001*
	R	0.02 (−0.07, 0.1)	0.76	−0.06 (−0.14, 0.02)	0.15	0.09 (−0.02, 0.19)	0.10	−0.16 (−0.25, −0.06)	0.001*
Ventral DC	L	0.1 (0.01, 0.19)	0.02	−0.06 (−0.15, 0.02)	0.13	0.09 (−0.01, 0.19)	0.09	−0.09 (−0.19, 0.01)	0.07
	R	0.1 (0.0, 0.18)	0.04	−0.07 (−0.15, 0.02)	0.14	0.04 (−0.06, 0.14)	0.42	−0.15 (−0.25, −0.05)	<0.001*

Note. Estimated regression coefficients along with 95% confidence intervals (CI) from the linear mixed-effects models examining the impact of variation in subcortical structure on the rate of change in body mass index (BMI). Models were controlled for puberty, caregiver education, handedness, and intracranial volume. Random intercepts were modeled for MRI scanner and subject ID. For all significant interactions, there was a main effect of time (not reported). ROI=Region of interest; ME = main effect; H=hemisphere; L=left; R=right;

^*= survived Benjamini Hochberg correction.

Figure 2.

Visualization of findings for girls from Model 1, examining whether variation in brain structure was related to changes in BMI (dependent variable), including two-way interaction effects for each region of interest (ROI) by Time (e.g., baseline [9/10-years-old], two-year follow-up [11/12-years-old]) on body mass index (BMI). A-F represent observed data for each ROI at levels of −1 standard deviation (SD; low, solid black line with circle end points), the mean (e.g., med, blue, large, dashed line with x end points), and +1 SD (e.g., high, red mini dashed line with square end points). The mixed-effects model was adjusted for sex, puberty, and caregiver education. Fixed effects were effects coded and crossed random effects were modeled for MRI ID (i.e., serial number) and subject ID. Low corresponds to lower performance.

Figure 3.

Forest plot of results, for boys (left) and girls(right), from Model 1 examining the impact of variation in brain structure on changes in BMI over time. Regression coefficients for the main effects of each region of interest (ROI, orange boxes) on body mass index (BMI) and the interaction between Time (e.g., baseline [9/10-years-old], two-year follow-up [11/12-years-old]; purple boxes) and ROI on BMI are presented. Mixed-effects models controlled for puberty, highest caregiver education, handedness, and intracranial volume. Fixed effects were effects coded and crossed random effects were considered for MRI scanner ID and subject ID. LH = left hemisphere; RH = right hemisphere. Purple lines and asterisks correspond to effects that survived multiple comparisons testing.

4.3. Does weight gain over time cause changes in subcortical volume?

4.3.1. Boys.

The results revealed significant main effects of BMI on bilateral ventral DC volume (Left: β=16.07, 95% CI=[5.57,26.57], p=0.003; Right: β=16.39, 95% CI=[5.95,26.83], p=0.002], while controlling for puberty, highest caregiver education, intracranial volume, handedness, and the interaction between BMI*Time. This finding suggested that a greater BMI (regardless of time) was related to a greater subcortical volume in the ventral DC, but that this pattern did not change over time. No other significant main effects or interactions were observed (see Supplemental Table S2).

4.3.2. Girls.

There were no associations between BMI*Time and any ROIs two years later (see Supplemental Table S2).

5. Discussion

Here, we showed that in a sample of youth who were initially healthy weight (12% overweight/obese two-years later), by the two-year follow-up smaller subcortical volumes were related to greater increases in BMI, a proxy for weight gain. The subcortex is largely responsible for food intake control,³¹ while variations in its structure and function have been proposed to play a role in overeating.³² However, no studies have evaluated these associations, particularly amongst youth who do not have obesity. Thus, by following a sample of youth who were of a healthy weight at baseline, our study sheds light on a potential neural phenotype that may predispose some youth to gain excess weight. Our findings provide some support for the idea that underlying variation in the brain especially among regions associated with controlling food intake, may provide a platform that promotes overeating and weight gain. Yet, additional studies are needed to explore these mechanisms further. Importantly, our findings showed sex-specificity. For girls, smaller subcortical volumes indicated a relatively stronger relationship with weight gain over the two-year period (i.e., 9/10-to-11/12-years-old), which may be driven by variation in subcortical maturation. Interestingly, these associations were not observed in boys. Additionally, there was no evidence that two-years of typical developmental weight gain had significantly influenced subcortical growth trajectories in either boys or girls. This is important because in rodents, weight gain has been shown to have irreversible consequences for brain structure and function that facilitate a vicious cycle of overeating.²⁸ As such, early puberty may be a critical time for intervention and prevention efforts, especially for individuals who may be predisposed to overeat due to underlying differences in brain regions involved in food intake.

Overall, adolescence is associated with age- and sex-specific changes in the subcortex (e.g., growth of the pallidum, amygdala, hippocampus; reduction of thalamus, caudate, putamen),^14,33 and this change is a sign of maturation (i.e., a refinement of processes). However, smaller volumes³⁴ and attenuated growth and reduction^35,36 in regions such as the amygdala, hippocampus and nucleus accumbens have been associated with various adverse outcomes (e.g., depression, anxiety, stress regulation). In our study, we observed similar patterns; that is, smaller volumes of the accumbens, amygdala, and hippocampus by ages 11/12-years-old were associated with greater gains in BMI over two-years (i.e., 9/10-to-11/12-years-old). These associations were likely driven by attenuated growth of areas that increase in size during puberty (e.g., amygdala, hippocampus), and accelerated reductions of those that decrease in volume during puberty (e.g., caudate, putamen).³³ However, additional studies are needed to confirm this. These findings are consistent with literature suggesting that altered brain structure and maturation are associated with adverse outcomes, but this is the first study to observe these relationships and weight gain among a healthy weight early adolescent sample.

Several theories exist to explain the relationship between weight gain and altered subcortical structure via the proposed mechanisms of overeating. For example, some studies have suggested that variation in reward response may promote overeating and subsequent weight gain.^{17,20,37–41} Indeed, this forms the foundation of the Dual Processes Model of Overeating,³² which states that overeating occurs when the brain’s response to a rewarding stimulus (i.e., appetitive food) is greater than its ability to control the impulse to achieve the reward (i.e., eat the food). In fact, among a similar and overlapping sample of healthy weight youth, it has been shown that variation in behavioral impulsivity is related to weight gain,²³ thus providing support for the Dual Processes Model of Overeating. Our current findings provide additional support for this model at the level of the brain regions associated with impulsive behaviors (e.g., caudate, putamen, pallidum, ventral DC). However, alternative explanations exist to explain these relationships. For example, the amygdala and accumbens are thought to play a role in impulsivity through their involvement in fear and stress processing, ^42,43 while stress can physiologically increase food intake.⁴⁴ Therefore, perhaps an underlying variation in these regions may be related to altered stress responses, and as such, increased food intake and subsequent weight gain. Additionally, in the rodent literature, there is extensive evidence suggesting that variation in the hippocampus occurs in response to a diet high in fat and sugar,^45,46 and in humans, diets high in these components are associated with greater weight gain.⁴⁷ However, examining associations between brain structure and emotional/stress processing was beyond the scope of this paper, and future research is needed in this area. This is the first to show that smaller subcortical volumes among regions involved in memory, learning, and hormonal food intake control are associated with weight gain in human adolescents.

Importantly, we observed robust associations between smaller subcortical volumes and weight gain. This is not surprising, given that much of the subcortex is differentially involved in food intake through various mechanisms, including emotion regulation (e.g., stress response; e.g., amygdala, hippocampus, accumbens), memory and reinforcement learning (e.g., hippocampus), reward sensitivity (e.g., caudate, putamen, thalamus, pallidum, ventral DC), taste-aversion learning (e.g., amygdala), and the hormonal control of food intake (e.g., hippocampus, ventral DC).^43,45,48 Although our study cannot provide insight into the causal mechanisms driving the associations between underlying variations in the subcortex (and its development) and weight gain, these findings can aid in our understanding of how the brain may guide decisions that increase the susceptibility to weight gain. Future studies are needed to understand whether underlying variation in the subcortex leads to poor food intake decisions and subsequent weight gain or whether other factors are at play. Notably, this was the first study to examine the relationship between weight gain and subcortical development across the entire subcortex, particularly among youth who do not yet have obesity.

Interestingly, the association between the relationship between the subcortex and weight gain was observed only in girls, perhaps due to differences in pubertal onset by sex (typically ~1 year earlier in girls).⁷ At baseline, girls were in the early stages of puberty (i.e., Tanner stage 2) and heavier than their male counterparts, who were more likely to be in pre-puberty (i.e., Tanner stage 1); by year 2, there was still a difference in the Tanner stage between girls and boys (where girls were more likely to have advanced to stage 3). As the ABCD Study did not collect data prior to 9/10-years-old, this limits the insight into understanding of these associations prior to early puberty in girls. However, our findings were independent of puberty, so although smaller subcortical regions in the subcortex are associated with weight gain, these associations may only emerge after pubertal onset. Future studies will be needed to further disentangle the role of puberty from the associations between the brain and weight gain. Importantly, as the boys in this study age, this may reveal important information about how pre- vs. early-puberty may be important for understanding the relationship between brain structure and weight gain.

It is important to note that this study focused on weight gain. Regardless of weight status (e.g., healthy weight, obese), unhealthy weight gain can have metabolic and neurologic consequences⁴⁹ that may affect maturation trajectories of the brain.²² Understanding the neurobiological mechanisms of weight gain itself, does in part give rise to understanding obesity. This approach also allows for insight into an under researched community at high risk for metabolic consequences, namely, adolescents with unhealthy weight gain at any weight class category (e.g., healthy weight, obese).

5.1. Strengths and limitations

The strength of this study lies in the ability to assess the relationship between the natural progression of weight gain and subcortical brain development during a critical period for weight gain risk. This study has some limitations. First, although the ABCD Study aimed to include a diverse sample, by excluding youth who had overweight or obesity at baseline, we may have limited the generalizability of our findings. Although this may have led to bias in our sample, it was important to restrict the sample to include only those with a healthy weight, as weight gain affects the brain. Second, the ABCD Study did not collect markers of inflammation or more precise measures of adiposity, which could have helped us better understand the biological mechanisms underlying weight gain. Third, since the inclusion criteria for the ABCD Study started at 9-years-old, we lacked the ability to draw inferences prior to the age of 9, thus, it could be that girls had already started to undergo puberty prior to enrollment in this study. However, because our findings showed that weight gain occurred after the baseline visit, these findings still have merit. Fourth, the caregiver and youth report estimate of puberty may not accurately reflect pubertal development as a trained clinician did not perform an exam. Tanner staging has been correlated well with clinical exams,⁵⁰ but we acknowledge the limitations of this measure that may lead to misrepresentation of pubertal development. Although we note this limitation, pubertal development was considered a covariate in these analyses, and therefore, not an effect of interest. However, to better understand the mechanisms driving these associations, future studies will be needed to assess pubertal development more precisely via clinical examination or hormone data. Last, the causes of obesity are multifactorial and as such, there could be other unmeasured factors that might also explain the relationships between brain development and weight gain.

5.2. Conclusions.

For the first time, we demonstrated that smaller subcortical volumes were associated with greater increases in weight gain. This finding suggested that variations in brain structure of food intake regions may predate weight gain, while altered maturation of these regions may be a risk factor for overeating and subsequent weight gain. However, this was observed only in girls, possibly due to differences in pubertal onset (boys were largely prepubertal). As such, early puberty may be a critical time for prevention efforts to decrease continued weight gain for those who may be predisposed to overeat due to variation in brain regions associated with food intake control. Notably, weight gain over two-years was not substantial enough to cause changes in the brain, which is important given that weight-induced brain changes are thought to facilitate a vicious cycle of overeating. Importantly, this was the first study to evaluate the natural progression of weight gain, prior to obesity and the maturation of brain regions involved in food intake, allowing for insight into the cause vs. consequence mechanisms of weight gain. As such, our findings are extremely relevant as we begin to map out the mechanisms that are related to weight gain during a time that is most at risk: adolescence.

Key Points.

What is already known about this subject?

Variations in brain structure among regions involved in food intake are thought to play a role in obesity development. However, former studies have included youth with obesity in the analyses, so the cause-specific vs. consequence-specific relationships, which may be sex-specific, are unknown.

What are the new findings in this manuscript?

Our results highlight a neural phenotype at risk for obesity that may emerge at puberty, but two years of weight gain was not associated with altered subcortical development. Importantly, these analyses were conducted among youth with initially healthy weight, while 12% transitioned to having overweight/obesity. As such, this study design captures the natural progression of weight gain and brain development to shed light on the cause vs. consequence mechanisms of obesity and the brain.

How might these results change the direction of research of the focus of clinical practices:

Animal models have shown that adolescent weight gain causes irreversible effects on neurocognitive development, but here, weight gain was not yet associated with negative brain outcomes. Thus, early puberty may be a critical time in development for interventions to target to avoid negative consequences on the brain. Equally important is that because underlying brain differences may predispose some youth to weight gain, early puberty may be a critical time for prevention efforts to mitigate weight gain risk.