Impaired Brain Satiety Responses After Weight Loss in Children With Obesity.

CONTEXT: Obesity interventions often result in increased motivation to eat.
OBJECTIVE: We investigated relationships between obesity outcomes and changes in brain activation by visual food cues and hormone levels in response to obesity intervention by family-based behavioral treatment (FBT).
METHODS: Neuroimaging and hormone assessments were conducted before and after 24-week FBT intervention in children with obesity (OB, n = 28), or children of healthy weight without intervention (HW, n = 17), all 9- to 11-year-old boys and girls. We evaluated meal-induced changes in neural activation to high- vs low-calorie food cues across appetite-processing brain regions and gut hormones.
RESULTS: Among children with OB who underwent FBT, greater declines of BMI z-score were associated with lesser reductions after the FBT intervention in meal-induced changes in neural activation to high- vs low-calorie food cues across appetite-processing brain regions (P < 0.05), and the slope of relationship was significantly different compared with children of HW. In children with OB, less reduction in brain responses to a meal from before to after FBT was associated with greater meal-induced reduction in ghrelin and increased meal-induced stimulation in peptide YY and glucagon-like peptide-1 (all P < 0.05).
CONCLUSION: In response to FBT, adaptations of central satiety responses and peripheral satiety-regulating hormones were noted. After weight loss, changes of peripheral hormone secretion support weight loss, but there was a weaker central satiety response. The findings suggest that even when peripheral satiety responses by gut hormones are intact, the central regulation of satiety is disturbed in children with OB who significantly improve their weight status during FBT, which could favor future weight regain.

In the United States, the prevalence of children with obesity has more than tripled during the last 4 decades. Similar trends have been recognized in many countries, and it is estimated that 124 million children have obesity worldwide (1). Increasing rates of related comorbidities, such as type 2 diabetes, are equally alarming (2-4). Understanding the underlying mechanisms of dysregulated energy homeostasis is crucial to combat this problem.

Recent behavioral studies demonstrate that uncontrolled eating, such as increased eating in the absence of hunger and poor satiety responsiveness (5, 6), are more frequent among children with obesity (OB), who find food more rewarding and who are also more responsive to food cues than lean children (6). These findings are supported by recent functional magnetic resonance imaging (fMRI) studies, which demonstrated hyperactivation in response to food images in reward regions in obese adults (7, 8) and children (9-11). Furthermore, a growing body of literature demonstrates alterations in appetitive processing within brain regions regulating attentional, reward, salience, and motivational aspects of food consumption in adults (7, 12, 8) and children (10, 9, 13) with obesity. Several gut hormones, such as ghrelin, peptide YY (PYY), and glucagon-like peptide-1 (GLP-1), as well as adipocyte-derived hormones, such as leptin and adiponectin, have been identified as important hormonal regulators of energy intake and homeostasis (14-17). Altered circulating levels of satiety-related hormones in response to food intake are another possible explanation for poor satiety responsiveness among children with obesity (16, 17).

Studies are beginning to assess the interplay of neuroendocrine factors with obesity treatment in children. Family-based behavioral treatment (FBT) with at least 26 contact hours over at least 6 months is the gold standard of childhood obesity intervention (18) and is recommended for children aged 6 or more years old with common obesity (19). Although immediate and long-term improvements in weight status have been demonstrated, the response rates are moderate and variable. It is poorly understood why there are rates up to 50% of nonresponsiveness as we and others have shown (20-22), and why many children regain weight shortly after ending treatment. The goal of our study was to understand changes that occur in response to obesity treatment in young children that potentially contribute to future weight regain, and whether neural responses to food cues within a set of key appetite-processing brain regions may be altered by obesity treatment and weight loss. We and others have previously shown that children with obesity have an attenuated central response to a meal in appetite-processing brain regions, that is, a blunted reduction of brain activation when viewing high-calorie food cues, compared to their healthy weight (HW) peers (9, 13). This indicates a diminished post-meal reduction of brain activation and was present despite intact peripheral gut responses to nutrients in children with obesity, which may predispose them to overconsumption of food and subsequent difficulty with weight loss (13). In addition, children’s brain responses to a meal prior to FBT intervention were related to weight reduction outcomes during FBT (23). Pretreatment failure to reduce activation in appetite-processing regions by high-calorie food cues from before to after eating, was associated with weaker body mass index (BMI) z-score reductions during FBT (23). Previous studies did not address how brain response to food consumption may be altered in response to FBT. Thus, in this current study we tested the hypothesis that changes from before to after FBT in meal-induced brain response and circulating levels of hormones would be both related to weight outcomes and in a direction indicative of further reductions in satiety responsiveness after FBT.

Methods

Participants

This study was performed in 9- to 11-year-old children with obesity (group OB, defined by BMI > 95th percentile for sex and age) with at least 1 overweight parent (BMI ≥ 27 kg/m2) and in children of healthy weight (group HW, defined by BMI between 15th and 85th percentile). Participants were recruited through direct mailings and advertisements. A total number of 369 families were contacted, and eligibility screening was performed in 313 (OB, 258; HW, 55) children of healthy weight and with obesity. Of these, 112 (OB: 102; HW: 10), declined to participate or were lost during follow-up, while 120 (OB: 98; HW: 22), did not meet inclusion criteria or were excluded because of exclusion criteria such as not meeting the BMI criteria, contraindications to MRI (eg, braces and retainers); serious medical conditions (eg, diabetes, cognitive disorders); current use of medications known to alter appetite or body weight (eg, stimulants); and inability to consume study foods (eg, allergies, vegetarianism). Guardians provided consent and children provided assent. Of the 81 children enrolled (OB: 58; HW: 23), 41 children with obesity and 22 children of HW completed both the baseline and 6-month visits. While our previously reported analyses focused on only pretreatment fMRI data (13, 23), the current study included participants with usable MRI data from a pre- and post-meal acquisition at both baseline and 6-month visits. Thirteen children with obesity were excluded from analyses (N = 12 due to unusable MRI in one or more of the 4 required acquisitions; N = 1 due to child baseline BMI < 90th percentile). Five children of HW were excluded from analyses (N = 4 due to unusable MRI in one or more of the 4 required acquisitions; N = 1 endorsing a vegetarian diet (exclusion criteria)). The final sample included N = 28 children with obesity (who all met study-defined intent-to-treat criteria of attending at least 2 FBT sessions) and N = 17 children with HW (Fig. 1). The study was approved by the Seattle Children’s Institutional Review Board.

Figure 1.

Consort diagram.

Study Procedures

Study methods were described previously (13, 23-25). In brief: baseline and 6-month study visits were conducted prior to and at the end of the 24-week FBT intervention for children with obesity, or 6 months after the baseline visit for children of HW. Participants arrived fasting, underwent IV placement and a fasting blood draw, and consumed a standardized breakfast (milk, toast with butter and jam) which was followed by body composition measurements using bioelectrical impedance analysis (BIA; Quantum II. RJL Systems, Detroit, MI) (26, 27). Prior to BIA assessment, participants drank water to ensure adequate hydration status, titrated to body weight (355 mL for OB; 237 mL for HW). Three hours after breakfast, participants underwent the first fMRI, followed by a pre-meal blood draw (time 0). Then they consumed a standardized test meal (macaroni and cheese), underwent the second fMRI, and had blood draws 30 and 60 minutes after the start of the standardized test meal. Finally, they were presented with a 30-minute ad libitum buffet meal. The breakfast and test meal were titrated to represent 10% and 33%, respectively, of each participant’s estimated daily caloric intake at the time of the study visit, based on the Mifflin-St Jeor equation with a standard activity factor of 1.3 (28). With few exceptions, participants consumed the entire breakfast and test meal within the given time of 15 minutes at both study visits. The 30-minute buffet meal offered excess calories (~5000 kcal) and a large variety of different foods familiar to children and of variable nutritional and hedonic properties (eg, pizza, fruit, cookies). Participants were unaware that their food consumption during the buffet meal was monitored until a subsequent debriefing. Macronutrients and kilocalories consumed at the buffet were calculated (ProNutra, VioCare, Inc., Princeton, NJ).

Family-Based Therapy Weight Loss Intervention

The FBT intervention for children with OB was adapted from our prior FBT intervention protocols and more details can be found in our previous publications (13, 23). In summary, children with OB accompanied by at least one parent or caregiver attended 24 weekly in-person sessions. Sessions included a 30-minute meeting with an interventionist and each child/parent pair to individualize treatment, followed by separate child and parent group meetings lasting about 45 minutes. The intervention focused on food and physical activity education, parenting around food, and use of behavioral skills (eg, self-monitoring). Children were encouraged to increase physical activity to 90 minutes/day, and parents up to 60 minutes/day, and to reduce sedentary behavior. Foods and beverages are categorized as Green, Yellow, or Red based on fat and sugar cut-points by food group based on an adaptation of Epstein’s Stoplight Diet (29). Individual goals were set for reducing weekly consumption of Red foods and increasing fruits and vegetables (all categorized as Green) aiming for approximately 0.5 pound per week weight reduction among children and 1.0 pound per week reduction among parents with overweight/obesity (20).

Hunger and Fullness Rating

At the 2 study visit days, visual analog scale (VAS) ratings of hunger and fullness assessed subjective appetite every 30 to 60 minutes (30) using a tablet computer.

fMRI Acquisition, Processing, and Analyses

We used fMRI to assess changes in meal-induced neural responses to food cues before and after FBT intervention using a region of interest (ROI) approach focused on brain areas that are critical to the control of motivation, incentive salience, and reward (31). Details of the fMRI paradigm, acquisition and analyses are published elsewhere (13, 23). We implemented a set of a priori ROIs involved in appetitive processing, including the ventral and dorsal striatum, medial orbitofrontal cortex (OFC), amygdala, substantia nigra/ventral tegmental area (SN/VTA) and insula. These a priori ROIs play a role in appetitive processing and have been established as markers of satiety and predictors of food choice (31, 32). Briefly, responsivity to food cues was tested by presenting child-friendly images, alternating blocks of high-calorie (N = 3) and low-calorie (N = 3) food pictures, and nonfood objects (N = 7), resulting in 13 blocks with 10 images each presented for 2.4 seconds. Attention to the fMRI task on post-scan memory test was similar across all 4 sessions and between groups (on average 84% ± 10% correct). First-level data were modeled (FMRI Expert Analysis Tool, FEAT) and condition effects were estimated from the average response across blocks for each contrast of interest (high-calorie foods vs low-calorie foods, high-calorie foods vs objects, low-calorie foods vs objects). Activation was assessed by parameter estimates (PE). Change in brain activation across a priori ROIs was calculated as follows: Mean activation within each region for contrasts of interest was extracted using functional-anatomic ROI masks as previously described (13, 23). Briefly, the SN/VTA was anatomically defined as published (32), other masks were determined as follows: A functional criteria of a minimum level of responsivity to food cues was applied (high-calorie vs object or low-calorie vs object; before or after meal at both visits; whole brain; P < 0.05, uncorrected), and voxels with no measurable activation by food cues were excluded. The functional activation was restricted to within a priori regions anatomically defined based on the Harvard-Oxford probabilistic atlas (33) using a minimum criterion of 25%. Regional mean PEs were averaged across all a priori regions to obtain for each subject a single measure representing the average activation for each contrast across the 6 a priori brain regions pre- and post-meal.

As a follow-up to our regional analysis, an exploratory voxel-wise analysis was completed outside the a priori ROIs. We sought to identify additional neuroanatomical regions in which changes in meal-induced reduction in brain activation (high-calorie foods vs low-calorie foods) associated with changes in BMI z-score during FBT. First, for each subject, a voxel-wise activation map was generated for activation (post-pre meal) at the baseline and 6-month visit separately and was followed by the generation of an activation map for 6 months (post-pre meal) > baseline (post-pre meal), (fixed effects, uncorrected (P > 0.005)). Then, a group level mixed effects analysis was conducted including all children with OB, to model the association of change in brain activation to a meal to change in BMI z-score (cluster corrected z = 2.3, P < 0.05).

Blood Sampling and Plasma Assays

Plasma glucose was measured by glucose oxidase. Acylated ghrelin levels were assessed by magnetic beads (Luminex xMAP, Millipore, Billerica, CA; Catalog # HMHEMAG-34K, RRID: AB_2910198). Plasma insulin (Catalog # EZHIASF-14K, RRID: AB_2910200), leptin (Catalog # EZHL-80SK, RRID: AB_2894697), total peptide YY (PYY; Catalog # EZHPYYT-66K, RRID: AB_2910201), adiponectin (Catalog # EZHADP-61K, RRID: AB_2801457) and active GLP-1 (Catalog # EZGLPHS-35BK, RRID: AB_2884907) were measured by commercially available immunoassays (Millipore Corp., St. Charles, MO). Soluble leptin receptor was quantified using an immunoassay from R&D Systems (Minneapolis, MN; Catalog # DOBR00, RRID: AB_2910202). Intra-assay coefficients of variation (CV) were < 8% and interassay CVs were < 10%. Unsuccessful IV placement resulted in missing data for all plasma measures in 1 child with OB at baseline and 1 child with HW at both baseline and 6-month visits; other missing plasma measures resulted from failed IV lines at nonfasting draws (ie, pre-meal and post-prandial time 30 and 60 minutes; N is reported for all plasma measures throughout). Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated (34). For each study visit, the change in meal response for insulin, ghrelin, PYY, GLP-1, and glucose concentrations (16, 35) was calculated as the percent change from time 0 (pre-meal) to the mean of the post-meal measures at time 30 and 60 minutes. The ratio for adiponectin (µg/mL) to leptin (ng/mL) was calculated to reflect leptin sensitivity and adipose tissue dysfunction (36). The free leptin index was calculated as the fraction of total leptin (ng/mL) and soluble leptin receptor concentrations (ng/mL) × 100 to assess leptin resistance (37, 38).

Statistical Analyses

For all measures, change from baseline was calculated as: (6-month) – (baseline). For example, a change in BMI z-score that is a negative value reflects a weight reduction and/or height increase from baseline at the 6-month visit. Importantly, we also assessed change in brain response to a meal, which is itself a change measure calculated as (post-meal activation) – (pre-meal activation) within a timepoint. For this “change of a change” measure, a negative value indicates a greater reduction in brain activation by a meal indicating greater satiation at 6 months relative to baseline whereas a positive value indicates a weaker (ie, less reduction in) brain response to a meal at the 6-month visit relative to baseline.

Outcome data are reported as mean ± standard error of the mean and are unadjusted unless otherwise noted. For descriptive variables of group comparisons, analyses by chi-squared test (categorical), linear regression (normally distributed) or by Wilcoxon rank-sum (nonnormally distributed) were performed. Simple and multiple linear regression models tested associations and Pearson’s correlation coefficients were calculated for descriptive purposes. Linear mixed models were conducted with the restricted maximum likelihood estimation to examine group differences in outcomes and included fixed effects (eg, time) and interaction terms. Formal post hoc stratified analyses were performed if a significant group by time interaction was present. Nonnormally distributed variables were transformed for regression analyses. All statistics were completed by STATA (v. 16.1) and graphs were generated in GraphPad Prism (v. 8.4.3).

Results

Characteristics of Study Participants and Changes of Anthropometric and Plasma Measures by Child Weight Groups

Families of children with OB all met study-defined intent-to-treat criteria by attending at least 2 FBT sessions, and on average, attended 19 ± 5 FBT sessions out of 24 (range, 2–24). Groups were balanced for sex and age and did not differ significantly by race or ethnicity (Table 1). On average, children with OB had significant reductions from baseline to 6 months in BMI z-score, BMI % overweight, and BMI % over 95th percentile compared to children of HW whose average BMI z-score did not significantly change between the visits. Children with OB had a reduction in % body fat mass and in increase in % lean body mass, which was significantly different from mean changes in children of HW, who showed an increase in % body fat mass, and decrease in % lean mass over the same time period (Table 1). There were no significant changes in hunger and fullness in response to a meal from baseline to 6 months in either group. However, children with OB consumed more at the ad libitum buffet meal at the second visit compared with their baseline by total calories and by calculated % daily need and consumed more calories at both visits compared with children of HW (Table 1).

Table 1.

Participant characteristics, adiposity measures and satiety measures between children of healthy weight and children with obesity

	Children with healthy weight			Children with obesity
N	17			28
Female, %	47			43			P = 0.78
Age at baseline, y	10.4 ± 0.9			10.5 ± 0.9
Race/ethnicity, % White, non-Hispanic	88			64			P = 0.08
	SV1	SV2	Change	SV1	SV2	Change	P value, mean change by group
BMI, z-score†	-0.06 ± 0.52	-0.09 ± 0.62	-0.02 ± 0.20	2.21 ± 0.36	2.01 ± 0.47*	-0.20 ± 0.22	0.03
BMI, percentile†	47 ± 19	46 ± 22	-1.2 ± 6.3	98 ± 1.2	96 ± 3.0	-1.5 ± 2.2	0.84
BMI, % overweight†	0.03 ± 7.1	0.23 ± 9.7	0.21 ± 3.7	74 ± 44	65 ± 47*	-9.3 ± 11.2	0.006
BMI, % over 95th†	-26 ± 5.1	-26 ± 6.8	-0.4 ± 2.8	29 ± 32	21 ± 34*	-7.7 ± 8.1	0.003
Waist circumference, cm†	64.9 ± 4.9	66.2 ± 6.1	1.2 ± 2.0	97.3 ± 15.6	97.5 ± 17.6	0.20 ± 5.6	0.47
Fat mass, kg†	7.8 ± 3.4	9.2 ± 4.4	1.5 ± 1.6	32.8 ± 15.9	31.8 ± 17.3	-1.0 ± 3.9	0.06
Fat mass, %†	21.9 ± 5.8	24.3 ± 5.8*	2.4 ± 2.8	48.5 ± 7.3	46.1 ± 8.6*	-2.4 ± 3.2	<0.0001
Lean mass, kg†	26.5 ± 3.1	27.5 ± 3.8	1.0 ± 1.2	32.6 ± 5.7	34.1 ± 5.8*	1.5 ± 1.8	0.29
Lean mass, %†	78.1 ± 5.8	75.7 ± 5.8*	-2.4 ± 2.8	51.5 ± 7.3	53.9 ± 8.6*	2.4 ± 3.2	<0.0001
Change in Hunger by a meal, mm	-27 ± 22	-31 ± 27	-3.8 ± 33	-45 ± 34	-29 ± 29	16 ± 37	0.07
Change in Fullness by a meal, mm	25 ± 28	31 ± 23	5.5 ± 32	36 ± 38	24 ± 33	-11 ± 37	0.14
Buffet intake, kcals	777 ± 309	908 ± 415	131 ± 398	1176 ± 421#	1387 ± 484*#	212 ± 254	0.41
Buffet intake, % Daily Need	56 ± 25	62 ± 28	6.0 ± 26	62 ± 20	73 ± 25*	11 ± 13	0.38
Buffet intake, % Protein	11 ± 3.1	11 ± 2.3	0.1 ± 3.1	11 ± 3.6	11 ± 3.2	-0.2 ± 3.3	0.75
Buffet intake, % Fat	36 ± 12	36 ± 13	0.2 ± 9.4	34 ± 9.0	36 ± 8.5	1.8 ± 5.3	0.47
Buffet intake, % Carb	53 ± 13	53 ± 13	-0.3 ± 11	54 ± 11	53 ± 10	-1.1 ± 9.0	0.65

Data are mean ± SD. Groups were compared by chi-squared test for categorical variables and their mean change for continuous variables were compared by t test or Kruskal-Wallis for normally and nonnormally distributed continuous variables (P values in far right column). To assess mean differences within groups or between groups at each study visit, a linear mixed model was applied followed by Bonferroni-corrected posttests. N = 1 OB missing from Hunger and Fullness measures. Data reported for children with obesity is a subset of a larger sample reported in (25).

Abbreviations: BMI, body mass index; HW, healthy weight; OB, obesity; SV, study visit.

*P < 0.01 vs SV1 within same group; †P < 0.001 HW vs OB for all BMI and body composition parameters at SV1 and SV2; #P < 0.01 vs HW at same study visit.

Groups did not differ in mean change of fasting glucose, insulin, HOMA-IR, ghrelin, GLP-1, or PYY between study visits (Table 2). Children with OB had a significant change between study visits in leptin compared with the HW children. There were no group differences in changes in leptin receptor, adiponectin, free leptin index or adiponectin to leptin ratio.

Table 2.

Plasma measures in HW and OB children

	Children of Healthy Weight			Children with Obesity
	SV1	SV2	Change	SV1	SV2	Change	P value, mean change by group
Fasting glucose, mg/dL	94.6 ± 5.5	91.6 ± 4.4	-2.97 ± 5.7	97.1 ± 5.4	95.8 ± 4.2	-1.34 ± 5.8	0.39
Fasting insulin, uU/mL	5.41 ± 2.26	7.67 ± 4.98	2.25 ± 4.41	15.9 ± 15.9#	20.2 ± 13.7#	4.27 ± 14.3	0.57
Fasting ghrelin, pg/mL	69.3 ± 39.3	51.2 ± 45.8	-18.0 ± 44.9	39.6 ± 19.0#	39.0 ± 20.7	-0.58 ± 26.0	0.09
Fasting GLP-1, pM	2.63 ± 1.85	2.15 ± 2.33	-0.48 ± 1.81	3.44 ± 2.91	2.87 ± 4.13	-0.57 ± 4.63	0.55
Fasting PYY, pg/mL	155.8 ± 41.4	160.8 ± 56.4	4.98 ± 44.4	186.2 ± 70.0	192.2 ± 88.4	6.06 ± 71.8	0.96
HOMA-IR	1.3 ± 0.6	1.7 ± 1.1	0.48 ± 1.0	4.0 ± 4.2#	5.0 ± 3.4#	1.0 ± 3.7	0.45
Leptin, ng/mL	4.7 ± 4.3	6.3 ± 9.5	1.6 ± 6.0	36.7 ± 23.0##	29.4 ± 20.6*##	-7.3 ± 16.7	0.02
Leptin receptor, ng/mL	44.8 ± 11.1	43.0 ± 12.8	-1.8 ± 6.4	30.2 ± 6.9##	28.8 ± 6.3##	-1.4 ± 3.7	0.22
Adiponectin, ug/mL	12.1 ± 5.9	11.9 ± 7.3	-0.2 ± 4.3	8.1 ± 2.9	9.8 ± 4.1	1.7 ± 2.8	0.13
Free leptin index	12.9 ± 17.0	20.0 ± 41.2	7.14 ± 25.7	134.1 ± 101.8##	113.5 ± 98.9##	-20.7 ± 65.0	0.10
Ratio adiponectin:leptin	4.0 ± 2.6	3.6 ± 2.3	-0.4 ± 2.9	0.31 ± 0.24##	0.49 ± 0.44##	0.18 ± 0.29	0.92

Data are mean ± SD. Mean change differences between groups were compared by t test or Kruskal-Wallis for normally and nonnormally distributed continuous variables. To assess mean differences within groups or between groups at each study visit, a linear mixed model was applied followed by Bonferroni-corrected posttests. N = 16 for HW, N = 27 for OB. Missing data includes: For HW: N = 1 fasting ghrelin, N = 3 fasting insulin, HOMA-IR, N = 4 adiponectin, ratio adiponectin:leptin; For OB: N = 2 fasting insulin, N = 3 fasting glucose, ghrelin, GLP-1, PYY, HOMA-IR, N = 4 adiponectin and ratio adiponectin:leptin.

Abbreviations: GLP-1, glucagon-like peptide-1; HOMA-IR, homeostatic model assessment for insulin resistance; HW, healthy weight; OB, obesity; PYY, peptide YY; SV, study visit.

*P < 0.05 vs SV1 within same group; #P < 0.05 or ##P < 0.001 vs HW at same study visit.

Relationships Between Changes in Plasma Measures and Changes in BMI z-score and Buffet Intake by Child Weight Groups

In a stratified analysis, % change in fasting leptin and the change in calculated free leptin index was positively associated with change in BMI z-score, such that among children with OB and children with HW greater decreases in fasting leptin and free leptin index were associated with a greater reductions in BMI z-score (children with OB, see Fig. 2A-2B; with HW, β = 0.002%, [95% CI, 0.0001-0.003] P < 0.01; β = 0.006 free leptin index, [95% CI, 0.003-0.009] P < 0.01, adjusted for sex). In children with OB, a greater increase in the calculated ratio of adiponectin (µg/mL) to leptin (ng/mL) was associated with greater decrease in BMI z-score (Fig. 2C). This relationship was not significant in children with HW (β = −0.03, [95% CI, −0.08 to 0.01] P = 0.13 adjusted for sex).

Figure 2.

Changes in circulating leptin in relation to changes in BMI z-score and ad libitum buffet intake among children with obesity undergoing FBT. A greater reduction in BMI z-score in children with obesity after undergoing FBT was associated with a reduction in leptin (A) and the calculated free leptin index (B) as well as an increase in the calculated ratio of adiponectin to leptin (C). Similarly, a decrease in ad libitum caloric intake at a buffet meal after undergoing FBT compared to baseline visit was associated with a reduction in leptin (D) and increase in the adiponectin to leptin ratio (F), but not the calculated free leptin index (E). Free leptin index = (total leptin [ng/mL]/ soluble leptin receptor [ng/mL])*100. Change in all variables calculated as 6-month − baseline data. P values by linear regression. Pearson’s correlation coefficient was calculated for descriptive purposes. N = 27 for A-B and D-E, and N = 23 for C and F. Data, Pearson’s r, P values adjusted for sex.

Among children with OB, but not children with HW, a reduced % change in the meal response of glucose was also associated with greater reductions in BMI z-score (OB: β= 0.01%, [95% CI, 0.00-0.03] P=0.05; HW: β = −0.01% [95% CI, −0.02 to 0.00] P = 0.11), but the test meal-induced change in insulin, ghrelin, GLP-1, and PYY were not related to change in BMI z-score among children with OB or HW (data not shown).

Among children with OB, markers of increased leptin sensitivity at 6 months relative to baseline (ie, decreased % change in fasting leptin and increased calculated ratio of adiponectin to leptin) that were related to changes in BMI z-score (see above and Fig. 2A and 2C) were also associated with greater reductions in ad libitum caloric intake at a buffet meal at the 6 months visit relative to baseline consumption (Fig. 2D and 2F). Change in the free leptin index was not related to change in ad libitum caloric intake (Fig. 2E). No relationships were found among children with HW between the leptin measures and change in ad libitum caloric intake, nor were relationships present among either children with HW or OB for change in the meal response of insulin, ghrelin, GLP-1, or PYY and change in ad libitum calorie intake (data not shown).

Association of Changes in BMI Z-Score Over 6 Months of Observation (HW) or FBT Intervention (OB) With Changes in Brain Activation to a Meal Within A Priori Regions

A significant interaction was found when testing the relationship between change in BMI z-score and 6-month change in meal-induced change in brain activation (Fig. 3A) indicating that the slopes differed between the groups. In children with OB, larger reductions in BMI z-score during FBT were associated with diminished meal-induced suppression of activation by high-calorie vs low-calorie food cues. The interaction between both groups remained significant when adjusting for baseline BMI z-score (P = 0.02 adjusted). Findings were in a similar direction for the high-calorie vs object contrast (Fig. 3B), but not significant. No significant interactions were found to suggest group differences in the association of changes in BMI z-score and changes in brain activation by low-calorie food cues vs object in response to a meal (Fig. 3C), indicating that the significant relationship found for the high-calorie vs low-calorie contrast was not driven by response to low-calorie foods.

Figure 3.

Relationship between change in BMI z-score and change in brain activation by a meal after FBT intervention in children with obesity or 6 months of observation in children of healthy weight. A significant interaction was present between group and change in BMI z-score on the change in brain activation by high-calorie (vs low-calorie) foods in response to a meal (A), but the interaction was not significant for change in brain activation by high-calorie foods (vs object) (B) or low-calorie foods (vs objects) (C). In (A), children with obesity (OB) with a larger reduction of BMI z-score following FBT had less meal-induced suppression of activation by high-calorie food cues (vs low-calorie food cues), with a similar trend found in (B). Change in brain activation across a priori ROIs was calculated as Post-Pre meal activation at the 6-month visit (6M) – Post-Pre meal activation at the baseline visit (BL). P values (P) by linear regression with an interaction term (BMI z-score change*group), followed by simple slope regression by group (P values under plots). N = 45. Data unadjusted.

Association of Changes in Peripheral Hormones With Changes in Brain Activation in A Priori Regions

We next examined how changes in hormones related to changes in brain activation after undergoing FBT among children with OB. Children with OB who had less meal-induced change of brain activation by high-calorie vs low-calorie food cues from before to after the FBT intervention tended (P = 0.08) to have a greater reduction of fasting leptin (Fig. 4A), but no significant associations were found for changes in the free leptin index or the ratio of adiponectin to leptin (Fig. 4B and 4C). Less meal-induced reduction in brain activation by a meal was also associated with greater change in meal-induced reduction in ghrelin after FBT intervention (Fig. 4D) and a positive change in meal-induced stimulation in GLP-1 and PYY (Fig. 4E and 4F). After adjustments for the change in the amount of calorie intake at the test meal, the adjusted P values were 0.08, 0.05, and 0.07, for ghrelin, GLP-1, and PYY respectively. No relationships were found for meal-induced change in glucose (β = −0.03%, [95% CI, −0.14 to 0.08] P = 0.59) or insulin (β = −7.45 %, [95% CI, −39.6 to 24.7] P = 0.64), nor were any relationships detected between these peripheral measures and the change in brain response to a meal among children with HW (data not shown). Importantly, among children with OB, the relationships between increased meal-induced stimulation in GLP-1 and change in meal-induced changes in brain activation remained significant when adjusted for changes in BMI z-score (β = 21.3%, [95% CI, 0.41-42.2] P < 0.05, adjusted).

Figure 4.

Relationship between changes in plasma measures and change in brain activation by a meal after FBT intervention in children with obesity. Measures of increased leptin sensitivity—change in % fasting leptin (A), change in the calculated free leptin index (B) or the ratio of adiponectin to leptin (C) were not significantly related to change in brain activation by high-calorie (vs low-calorie) foods in response to a meal in children with obesity. Changes in the meal response of gut hormones—specifically, reduction of ghrelin (D), and increases in GLP-1 (E) and PYY (F) were significantly related to an increased change (ie, less reduction) in brain activation by high-calorie (vs low-calorie) foods in response to a meal in children with obesity. Change in brain activation (high-calorie > low-calorie) across a priori ROIs calculated as Post-Pre meal activation at the 6-month visit (6M) – Post-Pre meal activation at the baseline visit (BL). Free leptin index = (total leptin [ng/mL]/ soluble leptin receptor [ng/mL])*100 and the meal response of gut hormones was calculated as the percent change from time 0 (pre-meal) to the mean of the post-meal measures at time 30 and 60 minutes. Change in all variables calculated as 6-month − baseline data. P values by linear regression. Pearson’s correlation coefficient was calculated for descriptive purposes. Data, Pearson’s r, P values adjusted for sex (A-C), D-F unadjusted. N = 21-27.

Exploratory Analysis of the Association of Change in BMI Z-Score in Children With OB With Change in Meal-Induced Suppression of Brain Activation Outside of A Priori ROIs

Using a voxel-wise approach, we identified clusters in which increased change in activation was associated with greater BMI z-score reduction (Table 3). Greater reduction in BMI z-score was associated with reductions in change in brain activation to high-calorie vs low-calorie food cues (indicating brain activation was less suppressed by a meal after FBT intervention) in clusters of activation in the occipital pole, parahippocampal gyrus and OFC (Table 3, Fig. 5). No clusters emerged for the association of greater change in activation and an increase in BMI z-score.

Table 3.

Brain regions in which change in meal-induced change in brain activation (high-calorie vs low-calorie food cues) is negatively associated with change in BMI z-score by FBT in children with obesity

Primary gray matter anatomic area of z maxa	b	MNI coordinatesc			Broadman aread	Cluster size (# voxels)	Z maxe	P	Other anatomic areas in cluster by Harvard-Oxford Atlasf
		x	y	z
Occipital pole	L	-27	-100	3	18	1201	3.4	0.004	Cerebral Cortex
Parahippocampal gyrus	L	-26	-31	-16	36	1646	4.58	<0.0001	Hippocampus
Lateral orbitofrontal cortex	R	18	16	-14	11	1971	4.12	<0.0001	Temporal pole, Middle & Superior Temporal Gyrus

Z-statistic images were corrected for multiple comparisons with a cluster-threshold correction (Z = 2.3, P = 0.05). Abbreviations: BMI, body mass index; FBT, family-based behavioral treatment; MNI, Montreal Neurological Institute.

Harvard-Oxford Atlas identified region of local maximum.

Hemisphere of local maximum (L, Left; R, Right).

Montreal Neurological Institute (MNI) coordinates of peak location.

Broadman area of local maximum.

Maximum Z-score.

Other areas identified within the cluster by the Harvard-Oxford Atlas.

Figure 5.

Whole-brain cluster analysis of change in meal-induced change in brain activation outside a priori regions and change in BMI z-score by FBT. Clusters show regions where change in meal-induced change in activation to high-calorie food cues (vs low-calorie) after FBT was negatively associated with BMI z-score change by FBT. Z statistic maps were whole-brain corrected for multiple comparisons with a cluster-threshold correction (z = 2.3, P < 0.05). Color scale represent z-stat range of activation and MRI coordinates are listed for each image. N = 28 children with obesity. Abbreviation: OFC, orbitofrontal cortex. See Table 3 for further details of cluster statistics.

Discussion

The current study focuses on the effect of an obesity intervention on changes in central and peripheral satiety signaling in response to a meal and how such changes are linked to child anthropometric and related hormone changes. In previous studies, we demonstrated evidence for pretreatment disturbed satiety signaling in the brain in children with obesity, specifically a lack of reduction in brain activation by high-calorie food cues after eating, as compared with children of healthy weight (13). We aimed to test if baseline (before starting the intervention) neurobiological factors predict outcomes by FBT and demonstrated that this pattern of a lack of meal-induced suppression of responses in brain regions regulating reward and motivation, when present in children upon entry into FBT, was associated with worse child weight status change outcomes (23). A separate a priori aim was to test if changes in neurobiological factors during FBT (from before to the end of intervention) occur, and if changes in neurobiology relate to intervention outcomes. The findings presented in this paper illuminate how completing an obesity intervention may influence central and peripheral satiety responsiveness, with potential relevance for postintervention behavior and outcomes.

We show that, in children with OB who underwent FBT, greater reductions in BMI z-scores were associated with developing a weaker central satiety response to a meal from before to after FBT. Specifically, children with greater declines in BMI z-score also demonstrated less robust meal-induced reduction of brain activation from baseline to 6 months by high-calorie food cues relative to low-calorie foods, in particular. We interpret this finding as indicating that decreases in adiposity via FBT were associated with a weakening in the satiety responsiveness of key brain regions regulating reward and motivation for food. Clinically, such findings could predispose children to overeating and weight regain after FBT. In fact, children with OB ate more kilocalories and a greater percentage of their estimated daily caloric needs after weight loss when presented an identical ad libitum buffet meal. This was true despite receiving the same standardized meal stimulus at both study visits. These data provide evidence supporting the hypothesis that in the presence of weight loss the brain defends elevated body weight (39), even in young children and implicates a mechanism that involves suppression of satiety responsiveness and persistent attention and motivation for high-calorie foods after eating.

Prior studies in adults also suggest that weight loss through behavioral intervention, but not bariatric surgery, heightens central reward responses to food cues with negative implications for weight loss maintenance. In adults with obesity (N = 25) imaged after 12 weeks of behavioral weight loss, activation by high-calorie food images within the SN/VTA, putamen, and fusiform cortex were positively associated with weight regain over the subsequent 9-month period (40). Moreover, also in adults undergoing weight loss via caloric restriction, the amount of weight loss achieved correlated with the degree of increase in activation by appetizing food cues (vs scenery) from baseline to 1 month of weight loss in regions including the dorsolateral prefrontal cortex, inferior frontal gyrus, and dorsal anterior cingulate cortex, among others (41). Increases in activation in cognitive control regions promoted further weight loss despite evidence of heightened appetite stimulation via peripheral hormones (eg, ghrelin) (41). In contrast, studies of bariatric surgery consistently show postsurgical reductions in activation by visual food cues or related tasks in regions governing reward and motivation (42-45), suggesting that satiety responsiveness is enhanced by the procedure in a manner that supports long-term weight loss maintenance.

Another important observation in our study was that the above-described changes in central satiety were related to changes in peripheral appetite-regulating signals such as gut hormones and peptides of the leptin-melanocortin system. We also queried the relationship of changes in satiety-related brain responses to changes in peripheral satiety signals, such as gut hormones evoked by FBT. A weaker, that is, lesser suppression of brain activation by high-calorie vs low-calorie food cues in response to a meal, from before to after the FBT intervention, was seen in children with OB after significant weight loss at the end of FBT intervention, despite an improvement of gut hormone responses (ie, increase in meal-induced reduction of ghrelin, and increases in meal-induced stimulation of GLP-1 and PYY). What this tells us is that FBT intervention and weight loss lead to a meaningful change of peripheral satiety-regulating hormones (46, 47). However, at the same time, a weaker central response to a meal was observed. This could mean that even when peripheral satiety responses by gut hormones are intact among children with OB who are successful in FBT, their central regulation of satiety is disturbed.

In a previous fMRI study in fasted lean adults, De Silva et al demonstrated that the combined administration of PYY and GLP-1 attenuated brain areas that are involved in appetitive behavior, including amygdala, caudate, insula, nucleus accumbens, OFC, and putamen, thereby supporting that pharmacologic provision of these hormones (48, 49), affects activation within brain areas included in our study. In another fMRI study following a test meal, ghrelin levels were associated with activation in the insula, amygdala, and OFC, while GLP-1 was associated with reduced activation in insula and OFC (50), demonstrating a potential relation between meal-induced changes of gut hormones and changes in brain areas involved in appetitive behavior.

As in our prior research papers (13, 23) we assessed central satiety responses using change in BOLD signal from before to after a test meal across key satiety-processing brain regions, including the dorsal striatum, insula, amygdala, and dopaminergic reward regions (ventral striatum, SN/VTA, medial OFC) in response to viewing pictures of high-calorie food cues. In addition, we performed an exploratory voxel-wise analysis in areas located outside the a priori ROIs to assess the association of change in brain activation to a meal to change in BMI z-score in brain regions excluded by our ROI approach. We identified 3 brain regions in which less meal-induced change in brain activation by a meal (high-calorie vs low-calorie food cues) after FBT was associated with greater reduction in BMI z-score in children with obesity: the occipital pole, the parahippocampal gyrus, and the OFC. While our ROIs include medial portions of the OFC, the exploratory cluster analysis identified associations consistent with lateral OFC. The lateral OFC has distinct neural connections differentiating its function from the medial OFC (51). The lateral OFC responds to unpleasant taste stimuli but is also important in modifying behavioral responses to rewarding food choices, thereby halting consummatory behavior (52). Enhanced lateral OFC responses after FBT among children who demonstrated the greatest lowering of both BMI z-scores and ad libitum buffet meal intake relative to other children in the program may indicate that they are engaging self-regulatory cognitive mechanisms successfully while viewing high fat and high sugar foods (“Red foods”). In addition, a cluster emerged in the parahippocampus, a region which has previously been linked to a lack of satiation and food anticipation (53) and in the occipital cortex which may indicate increased visual attention to high-calorie food cues. In adults, prior studies demonstrate that weight loss success occurs despite enhanced hormonal and neural compensatory responses and is related to engagement of inhibitory control (41). The current findings suggest similar mechanisms promote success in children who reduce adiposity during FBT. Specifically, they engage cortical regions that modify behavioral responses to rewarding stimuli and counteract the reductions in satiety responsiveness observed within reward and appetite-regulating brain regions.

Elevated circulating leptin levels are found in obesity and the failure of these signals to suppress adipose tissue accumulation sufficient to maintain normal weight has been attributed to leptin resistance. It had been suggested that this is related to increased SOCS3 activity (54), endoplasmic reticulum stress, and hypothalamic inflammation (55, 56). In children with OB vs those with HW, we found significantly higher fasting levels for insulin and leptin, and it could well be that central resistance to these hormones plays an important role in the central dysregulation of satiety at baseline. After the intervention, children with obesity reduced circulating leptin in conjunction with decreases in adiposity, as expected. However, those children in whom the intervention increased leptin and/or decreased adiponectin/leptin ratio had higher calorie intake at a buffet meal. In other words, when circulating leptin levels remained high despite the intervention, children increased food intake over their baseline, while children who reduced BMI z-score accompanied by declines in circulating leptin did not eat more, potentially because they had engaged inhibitory control centers. Similar to our study, Reinehr et al found lower circulating soluble leptin receptor in OB vs HW children. They also demonstrated an increase of circulating soluble leptin receptor in OB children with weight loss (57), which we did not replicate.

We used the free leptin index which is a biomarker of leptin resistance and the status of leptin action (37, 38), and the adiponectin to leptin ratio, where a low value indicates adipose tissue dysfunction (36, 37). We found that a greater decrease in fasting leptin and the calculated free leptin index, and an increase in the calculated ratio of adiponectin to leptin were associated with greater reduction in BMI z-score. However, when testing the relationships of changes of these hormones and changes in food-related brain activation, we found only a trend demonstrating that the stronger the leptin was reduced after FBT, the weaker was the brain satiety response, aligning with our results associating BMI z-score reductions with reduced central satiety responsiveness.

While these results need to be interpreted with caution, they are consistent with prior studies showing that a significant reduction in leptin, which is well described after weight loss, can lead to a compensatory caloric overconsumption and weight regain, which can potentially be reversed by exogenous leptin supplementation (58, 59). However, most recent research in rodents demonstrated that a partial reduction in plasma leptin levels can actually lead to improved leptin sensitivity (60). Hence, although the exact mechanisms need further exploration, it seems that the primary defect is not overconsumption in these children, but the fact that homeostatic brain responses actively combat return to a healthier level of adiposity.

Although the present study included a relatively large sample size for a neuroimaging study in young children with minimal loss of scans due to excessive motion, there were some limitations that need to be mentioned. First, only subjects with 4 completed fMRI scans (one before and after a test meal, at baseline, and at the 6-month time point) could be included, which resulted in some loss of data. Second, the order of pre-meal/post-meal scans could not be counterbalanced, but this approach allowed us to simultaneously measure fasting and post-meal stimulated hormone levels in OB and HW children. Third, while we conducted screening on 369 children and their families, only 58 children with obesity and 23 children of healthy weight enrolled, as many declined to participate or did not meet inclusion criteria for the child or parent. While we are not aware of any differences between study participants and excluded families, we cannot exclude that selection bias might have influenced our sample or results. Finally, while it is unlikely that participants had an underlying monogenic obesity syndrome, we did not include genetic screening in our study.

In conclusion, we demonstrated a disconnect between the peripheral hormonal signaling and central satiety regulation in children with OB, implicating a possible impairment of central action. Furthermore, we demonstrated evidence that after effective weight loss intervention in young children with obesity, brain satiety responses are disturbed. This finding underscores the need for supporting cognitive control of reduced caloric intake and long-term obesity intervention protocols for children to prevent rapid regain after weight loss interventions driven by, potentially, homeostatic compensatory mechanisms that ultimately overwhelm inhibitory control. Therefore, it is possible, that even in children after 6 months of obesity intervention, these obesity-related set-points defending a higher body weight are quite robust and require long-term lifestyle and behavioral interventions or pharmacological approaches that intervene upon the related central mechanisms.