Adolescents at high risk of obesity show greater striatal response to increased sugar content in milkshakes

ABSTRACT

Background

Children of overweight or obese parents are at a high risk of developing obesity.

Objective

This study sought to examine the underlying neural factors related to parental obesity risk and the relative impact of sugar and fat when consuming a palatable food, as well as the impact of obesity risk status on brain response to appetizing food images.

Design

With the use of functional MRI, the responses of 108 healthy-weight adolescents [mean ± SD body mass index (kg/m²): 20.9 ± 1.9; n = 53 who were at high risk by virtue of parental obesity status, n = 55 who were low risk] to food stimuli were examined. Stimuli included 4 milkshakes, which systematically varied in sugar and fat content, a calorie-free tasteless solution, and images of appetizing foods and glasses of water.

Results

High-risk compared with low-risk adolescents showed greater blood oxygen–dependent response to milkshakes (all variants collapsed) compared with the tasteless solution in the primary gustatory and oral somatosensory cortices (P-family-wise error rate < 0.05), replicating a previous report. Notably, high-risk adolescents showed greater caudate, gustatory, and oral somatosensory responses to the high-sugar milkshake than to the tasteless solution; however, no effect of risk status was observed in the high-fat milkshake condition. Responses to food images were not related to obesity risk status.

Conclusion

Collectively, the data presented here suggest that parental weight status is associated with greater striatal, gustatory, and somatosensory responses to palatable foods—in particular, high-sugar foods—in their adolescent offspring, which theoretically contributes to an increased risk of future overeating. This trial was registered at www.clinicaltrials.gov as NCT01949636.

INTRODUCTION

Adolescent obesity prevalence has increased from 10.5% between 1988 and 1994 to >20% in 2013–2014 in the United States (1, 2). Adolescence is a critical developmental period for the examination of obesity risk factors given the increased independence over lifestyle and dietary decisions (3). Evidence consistently indicates that elevated palatability of high-sugar and high-fat foods contributes to a positive energy balance (4, 5). However, knowledge of risk factors that predict and maintain hedonically motivated food intake is limited.

fMRI research has begun to uncover the neural correlates of ingestive behaviors in humans, including aberrant eating patterns. Receipt of palatable food elicits an increased BOLD response in regions frequently implicated in hedonically motivated behavior and reward learning (6, 7). Moreover, studies in both humans and animals have reported that eating behaviors and weight status influence the neural response to palatable food (8, 9). Although these studies suggest that neural functioning is correlated with hedonically motivated food intake, these cross-sectional studies cannot differentiate a vulnerability factor from a consequence of overeating behaviors.

One approach to identifying precursors of overeating behaviors is to compare healthy-weight adolescents who are at high or low risk of obesity by virtue of parental overweight or obesity. Parental BMI is a reliable predictor of obesity in their offspring (10, 11), because it reflects both environmental and genetic factors thought to promote weight gain (12). For example, parental obesity has been shown to increase the preference for palatable, high-energy foods in their offspring (13, 14) and a preference for these foods increases obesity risk (15, 16). In support, preclinical experiments found that obesity-prone compared with obesity-resistant animals showed a greater preference for sweet and fatty foods and increased hedonically motivated behaviors before the onset of diet-induced obesity (17, 18). Collectively, these reports suggest that parental overweight or obese status may increase obesity risk in their offspring via increased hedonically motivated food intake.

The utilization of both neuroimaging techniques and parental weight status is, in principle, a strong approach to assess the underpinnings of obesity risk factors during adolescence. Previously, healthy-weight adolescents with overweight or obese parents, compared with those with healthy-weight parents, exhibited a heightened BOLD response to receipt of a palatable milkshake in brain regions associated with gustatory, somatosensory, and reward processes (19). Although that initial study uncovered possible neural contributors underlying obesity risk related to parental weight status, the relative impact of macronutrient content in the tastants consumed and the response to appetizing food images are unknown. Preclinical work suggests that the sugar and fat composition of palatable food imparts differential motivation to eat and weight gain (20). With the use of data from a larger longitudinal cohort, here we used baseline imaging data to examine neural responses to milkshakes that varied parametrically in fat and sugar content and to images of palatable foods in a larger sample of adolescents at high or low risk of obesity. In addition, with the use of 3-y follow-up data, we performed a confirmatory obesity risk analysis to be sure that those in the high-risk group gained more weight than those in the low-risk group.

METHODS

Participants

In total, 133 participants completed the primary longitudinal investigation, of whom 108 had parental weight status for both parents. This analysis included 53 male and 55 female healthy-weight adolescents (mean ± SD age: 15.01 ± 0.86 y). Participants were considered at high risk if they had 2 overweight or obese parents (n = 53) and were considered at low risk if they had ≥1 normal-weight parent (n = 55). We modeled parental overweight status in this manner because it resulted in similar group sizes, putatively increasing sensitivity. Demographic information on the full sample and between risk groups is shown in Table 1. Individuals who reported binge eating or compensatory behavior in the past 3 mo, use of psychotropic medications or illicit drugs, head injury with a loss of consciousness, or Axis I psychiatric disorder in the past year (including anorexia nervosa, bulimia nervosa, or binge eating disorder) were excluded. Scan data were acquired at Oregon Health Sciences University. All parents provided written informed consent, and adolescents provided written assent. The Oregon Research Institute's Institutional Review Board approved all methods (registered at clinicaltrials.gov as NCT01949636).

TABLE 1.

Sample demographic characteristics by obesity risk status¹

	Full sample (n = 108)	High-risk (n = 53)	Low-risk (n = 55)
BMI, kg/m2
Study participant	20.9 ± 1.9	21.1 ± 1.8	20.7 ± 2.0
Maternal	27.0 ± 5.8	31.2 ± 5.1	22.9 ± 2.9
Paternal	28.17 ± 4.4	29.6 ± 4.0	26.8 ± 4.3
Age, y	15.01 ± 0.86	14.91 ± 0.81	15.11 ± 0.9
Sex (M/F), n	53/55	29/24	24/31
Race, %
Non-Hispanic white	75	36	40
Non-Hispanic black	2	1	1
Hispanic	12	11	1
Asian/Pacific Islander	6	2	4
Native American	3	0	3
Multi-race/other	2	1	1
Total calories, kcal/d	2368.8 ± 358.4	2402 ± 440.6	2336.9 ± 255.7
Calories from fat, % of energy/d	35.7 ± 1.1	35.7 ± 0.9	35.7 ± 1.2
Calories from added sugar, % of energy/d	11.6 ± 1.2	11.6 ± 1.3	11.6 ± 1.0

¹Values are means ± SDs unless otherwise indicated.

Experimental design

With the use of data from a larger longitudinal cohort, here we used baseline imaging data as well as height and weight over 3 y of follow-up (for the purpose of validating obesity risk status). The study from which the sample was drawn recruited adolescents in the healthy-weight range to ensure that a history of overeating did not contribute to any differences in BOLD response between groups. Height was measured to the nearest millimeter by using a stadiometer. Weight was assessed to the nearest 0.1 kg by using digital scales, with participants wearing light clothing without shoes at each assessment. BMI (kg/m²) correlates with measures of body fat (r = 0.80–0.90) and with biomarkers of health in young-adult samples (21). BMI was assessed at baseline to confirm that the present effects were not driven by differences in weight and at 1-, 2-, and 3-y follow-ups to confirm that parental history of obesity was related to greater increases in BMI.

On the scan day, participants were asked to consume their regular meals but to refrain from eating or drinking (other than water) for 4 h immediately preceding their imaging session for standardization of subjective feelings of hunger, as described in reference 22. Time of scans ranged from 1100 to 1800, with the average start time of 1253; no participants reported fasting longer than the requested 4 h, and the time of scans did not differ between the risk groups (P > 0.05). A block milkshake paradigm was used to assess BOLD response to chocolate-flavored milkshakes with varying fat and sugar content, as follows: high-sugar and high-fat or high-sugar and low-fat (“high-sugar”), low-sugar and high-fat (“high-fat”), and low-sugar and low-fat milkshakes, along with a calorie-free tasteless solution as a control. The chocolate flavor was consistent across milkshakes, but calories and fat and sugar contents of the milkshakes varied as follows: high-sugar and high-fat milkshake (170 kcal, 7.5 g fat, and 23 g sugar/100 mL), high-sugar milkshake (124 kcal, 1.9 g fat, and 23.7 g sugar/100 mL), high-fat milkshake (129 kcal, 9.0 g fat, and 7.3 g sugar/100 mL), and low-sugar and low-fat milkshake (74 kcal, 2.4 g fat, and 8.7 g sugar/100 mL). Details on the block paradigm, scanner acquisition parameters, and the composition of the milkshakes and tasteless solution have been previously published (22). Briefly, the milkshake or tasteless solution was delivered into the scanner via a gustometer. A 1-s cue preceded all milkshake receipt, followed by a fixation cross during the taste administration. Milkshakes were delivered over a 5-s period followed by 3 s to swallow. The administration of the milkshake and tasteless solution occurred in variable length blocks of 4-, 5-, or 7-s milkshake or tasteless administration events. The order of the milkshake or tasteless blocks was randomly assigned. Participants received a rinse of tasteless solution between milkshake tastes, followed by a swallow cue and a variable jittered time span (9–11 s). Two scanning runs were performed, with each run consisting of 3 blocks of each of the 4 milkshake types and tasteless solution in a random order, which created 6 blocks of each of the 5 tastants (22).

A food-picture exposure paradigm assessed BOLD responses to images of appetizing and unappetizing foods and images of glasses of water. To account for the high variability in food preferences, a self-rating approach was used at a session before the scan day. Participants rated 103 food images on a visual analog scale for how appetizing the food looked (0–100). The foods depicted in the pictures varied by macronutrient content and energy density but were matched on size and brightness and had a matched number of pixels. Pictures that were rated as extremely low (0) were eliminated as to not elicit a “disgust” response. During the fMRI scan, the participants were exposed to the 32 pictures they rated most appetizing and to the 32 pictures they rated as the least appetizing, along with 32 images of glasses of water as a neutral stimulus (which were the same for all participants). Although there was between-participant variability in food characteristics depicted in the images, generally the most appetizing foods selected were sweet or savory items (e.g., ice cream, cinnamon rolls, steak, and mashed potatoes), whereas the least appetizing were, for example, a plain bagel, grapes, fish, and plain pasta. Detailed analyses of characteristics of the food and hedonic ratings will be presented in a separate, forthcoming manuscript. All of the pictures were presented for 4.5 s, with a 4- to 11-s (mean: 5.5 s) interstimulus jitter to optimize BOLD recovery. A fixation cross appeared between images to minimize random eye movement.

Scanning was performed using a Siemens Tim Trio 3-T MRI scanner. Functional scans used a T2*-weighted gradient single-shot echo planar imaging sequence (TE = 30 ms, TR = 2000 ms, flip angle = 80°) with an in-plane resolution of 3.0 × 3.0 mm² (64 × 64 matrix; 192- × 192-mm² field of view). Thirty-two 4-mm slices (interleaved acquisition, no skip) were acquired along the anterior commissure - posterior commissure line (AC-PC) transverse oblique plane, as determined by the midsagittal section.

fMRI preprocessing and statistical analysis

Data were preprocessed and analyzed with the use of the FMRIB Software Library Functional MRI of the Brain Analysis Group (FMRIB; Oxford, United Kingdom) Software Library (FSL version 5.0.9) (23). Brains were skull stripped by using FSL's brain extraction tool. Time series were then motion corrected by using MCFLIRT, and the 18 motion regressors created and their derivatives were included in the first-level model. B0 unwrapping was implemented to correct for potential magnetic field inhomogeneities (using a signal loss threshold of 10%), and slice-timing correction was used with a custom slice order file. Functional data were smoothed with a 5.0-mm full-width half-maximum isotropic Gaussian kernel, and a high-pass temporal filter of 90 s was used. Functional images were registered to their T1-weighted structural scan and to the standard Montreal Neurological Institute template brain with the use of FNIRT with a 10-mm warp resolution. Normalization resulted in a voxel size of 2 mm³ for functional images and 1 mm³ for structural images.

To test the BOLD response to the receipt of the tastant, we created a general linear model with regressor and temporal derivatives for receipt of each milkshake type (4 regressors) and the tasteless solution as regressors of interest. The milkshake cue, the tasteless cue, the receipt the rinse, and their derivatives were included as confounders. A similar approach was used for the picture paradigm. All of the regressors were convolved using the double-γ hemodynamic response function. The model was prewhitened with FSL's FILM tool (24). All of the milkshakes combined were compared with the tasteless solution (all milkshakes > tasteless). In addition, each type of milkshake was compared with the tasteless solution individually (high-sugar and high-fat milkshake > tasteless, high-sugar and low-fat milkshake > tasteless, high-fat and low-sugar milkshake > tasteless, low-sugar and low-fat milkshake > tasteless). To assess brain regions associated with viewing appetizing foods, the appetizing food pictures were compared with pictures of water (appetizing food pictures > pictures of glasses of water). This contrast was chosen for the picture paradigm because it most closely resembles the milkshake contrasts used (milkshake receipt > tasteless solution receipt) in which a palatable or appetizing beverage or food is contrasted against a water-type control. The block nature of the milkshake receipt paradigm was not designed to adequately test for cue-elicited anticipatory response (i.e., the predicative milkshake cue). As such, there were not enough events or exposures to evaluate a contrast such as, for example, the predicative milkshake cue compared with the tasteless cue, given that the latter cue had only 6 events total. Post hoc analyses tested each milkshake compared with each other milkshake (e.g., high-sugar and high-fat > high-sugar and low fat, high-sugar and high-fat > low-sugar and high-fat).

Individual-level data (fixed-effects) were averaged and taken to group-level (mixed-effects) analysis to assess the difference between the high and low risk of obesity groups. Between-group t tests were used to test differences in the previously mentioned contrasts between the high- and low-obesity-risk groups with the use of FSL's FLAME 1. To best correct for multiple comparisons, nonparametric permutation testing was performed using FSL's randomise tool with the use of the threshold free cluster enhancement (TFCE) algorithm with 5000 permutations (25). Clusters after randomization were considered significant if a cluster had a family-wise error rate P value (P-FWE) <0.05 and a cluster k > 10 (26, 27). All stereostatic coordinates are presented in Montreal Neurological Institute space. Analyses of all nonimaging data were performed in R (version 3.3.1).

Confirmatory analyses of risk status

The purpose of this report was to examine obesity risk status. All of the analyses were performed in R (version 3.3.1). To confirm obesity risk status, 2 tests were used; hierarchal linear modeling and a t test at the last point of BMI data collection. Individual change in BMI was modeled by using a mixed-effects model with visit (baseline, year 1, year 2, and year 3) and participant as a random slope and intercept, respectively (lme4, version 1.1-12). These models offer a flexible and powerful technique for modeling change in continuous variables and use maximum likelihood estimation to accommodate missing data (28). To build these models we performed the following: 1) examined plots, 2) fit a conditional means model, 3) fit a conditional linear slope model, and 4) fit polynomial interpolation models (i.e., second- and third-degree spline models). A likelihood ratio test of the linear slope relative to spline models indicated that the splines did not fit the data significantly better than a linear slope (2-spline P = 0.07, 3-spline P = 0.36). We compared the linear and spline models for model fit quality with the use of a likelihood ratio test. Post hoc 1-factor ANOVAs were used to assess risk-dependent differences in BMI at the year 3 follow-up, with the use of Tukey’s honestly significant difference test to correct for multiple comparisons (stats, version 3.4.2). BMI trajectory analyses of risk status showed that, over the 3-y follow-up, high-risk adolescents’ linear slope or polynomial interpolation of BMI did not differ from that of low-risk adolescents (P = 0.21), As such, tests at individual time points are presented (Figure 1).

FIGURE 1.

No difference in BMI was seen between the high- and low-risk groups (n = 53 and 55, respectively) at baseline (P = 0.28) (A). There was a significant difference in BMI at the 3-y follow-up visit as a function of obesity risk status: high- compared with low-risk participants (n = 42 and 46, respectively) had a significantly higher BMI (*P = 0.002) (B). Values are means ± SEs. Post hoc 1-factor ANOVA was used to assess risk-dependent differences in BMI at the year 3 follow-up; Tukey’s honestly significant difference test was used to correct for multiple comparisons.

Assessing the palatability of milkshakes

The purpose of this analysis was to test differences in perceived palatability of all of the milkshakes between the high- and low-risk groups. Hierarchal linear modeling assessed differences in palatability ratings for each milkshake between the high- and low-risk groups, with risk by milkshake type as a random slope (lme4, version 1.1–12). To build this model we performed the following: 1) normalized palatability ratings, 2) examined plots, 3) fit a conditional linear slope model, and 4) performed a post hoc Bonferonni correction for multiple comparisons (lsmeans, version 2.27-2). BOLD activation in response to the individual milkshakes has been previously published in a subset of this sample (22).

RESULTS

Relation of risk status to neural response to palatable food receipt

Brain regions in which the high-risk adolescents showed a significantly greater BOLD response during receipt of the various milkshake and tasteless solution comparisons relative to their low-risk counterparts are summarized in Table 2. High- compared with low-risk adolescents exhibited an increased BOLD response during receipt when assessing the combination of all 4 milkshakes compared with the tasteless solution in the anterior insula (Figure 2A; r = 0.29, P-FWE = 0.04), extending into lateral orbital frontal cortex (Figure 2B; r = 0.39, P-FWE = 0.03), postcentral gyrus (r = 0.42, P-FWE = 0.02), and the ventral precentral gyrus (r = 0.39, P-FWE = 0.04). When using the high-sugar and low-fat milkshake compared with the tasteless solution contrast, high-risk compared with low-risk adolescents showed greater BOLD response in the caudate (Figure 2C; r = 0.35, P-FWE = 0.03), central operculum (Figure 2D; r = 0.37, P-FWE = 0.04), superior temporal gyrus (r = 0.42, P-FWE = 0.03), juxtapositional lobule (r = 0.28, P-FWE = 0.03), and thalamus (r = 0.30, P-FWE = 0.03). No significant effect of obesity risk status was observed when testing the high-sugar and high-fat milkshake, the high-fat and low-sugar milkshake, or the low-sugar and low-fat milkshake compared with the tasteless solution contrast. In post hoc analyses comparing each milkshake with each other milkshake (e.g., high-sugar and high-fat > high-sugar and low-fat, high-sugar and high-fat > low-sugar and high-fat), no significant effect of obesity risk status was seen.

TABLE 2.

Brain regions with a significantly greater BOLD response to milkshake stimuli in adolescents at high risk of obesity compared with those at low risk of obesity¹

Contrast and regions	k	1-p	z Score	r	x	y	z
All milkshakes2 compared with the  tasteless solution
Postcentral gyrus	2003	0.98	4.34	0.42	68	−8	24
Central operculum	—	—	4.09	0.39	50	−18	20
Parietal operculum	—	—	4.04	0.38	50	−30	20
Orbital frontal cortex	257	0.97	4.07	0.39	40	22	−8
Precentral gyrus	63	0.96	4.20	0.39	48	0	40
Insula	30	0.96	2.99	0.29	46	−4	−2
High-sugar milkshake compared  with the tasteless solution
Superior temporal gyrus	3133	0.97	4.37	0.42	70	−20	6
Planum temporal gyrus	—	—	4.12	0.39	50	−34	16
Parietal operculum	—	—	4.00	0.28	46	−26	22
Central operculum	40	0.96	3.86	0.37	44	8	6
Juxtapositional lobule	24	0.96	2.91	0.28	6	−2	64
Thalamus	13	0.97	3.16	0.30	12	−10	6
Caudate	11	0.97	3.68	0.35	10	10	14

¹ n = 53 and n = 55 for adolescents at high and low risk of obesity, respectively. P values and clusters were generated by using threshold free cluster estimation through FSL's randomize (version 5.0.9, Functional Magnetic Resonance Imaging of the Brain group) with the use of a 2-sample t test. Local maxima were calculated with FSL's cluster tool in randomize. Local maxima are listed under the cluster within which they fall. Coordinates are presented in Montreal Neurological Institute space. k = cluster size at 2 × 2 × 2; r = effect size: Z/√n.

²The “All milkshakes” category included high-fat, low-fat, high-sugar, and low-sugar milkshake tastants.

FIGURE 2.

Greater response in the left anterior insula (MNI coordinates 46, −4, −2; z score = 2.99, P-FWE = 0.04, k = 30) (A) and the left orbital frontal cortex (MNI coordinates 40, 22, 8; z score = 4.07, P-FWE = 0.03, k = 257) (B) in the high-risk group (n = 53) compared with the low-risk group (n = 55) during receipt of all 4 milkshakes compared with a tasteless solution. Greater activation in the left caudate (MNI coordinates 10, 10, 14; z score = 3.68, P-FWE = 0.03, k = 11) (C) and the left central operculum (MNI coordinates 44, 8, 6; z score = 3.86; P-FWE = 0.04, k = 40) (D) in the high-risk group (n = 53) compared with the low-risk group (n = 55) during receipt of the high-sugar and low-sugar milkshake compared with a tasteless solution. Bar graphs represent mean PEs extracted from the local peak response denoted by the cross-hairs; error bars denote SDs. P values and clusters were generated by using threshold free cluster estimation through FSL's randomize (version 5.0.9, Functional Magnetic Resonance Imaging of the Brain group) with the use of a 2-sample t test. Local maxima were calculated with FSL's cluster tool. Coordinates are presented in MNI space. FWE, family-wise error rate; MNI, Montreal Neurological Institute; PE, parameter estimate.

Relation of risk status to neural response to appetizing food images

No significant differences between risk groups were observed when viewing pictures of self-selected appetizing foods compared with pictures of glasses of water.

Confirmatory analyses of risk status

Confirmatory analyses to ensure that fMRI results were not driven by current weight status or other confounders (race, sex) showed that baseline BMI did not differ as a function of risk status (P > 0.05). Demographic characteristics also did not vary by risk status (Figure 1A; Table 1). In support of the validity of the obesity risk status definition herein, a post hoc test showed that high-obesity-risk participants were significantly heavier than were low-risk participants at the 3-y follow-up visit (Figure 1B, P-corrected = 0.002), although risk groups did not differ on BMI at the 1- and 2-y follow-ups.

Assessing the palatability of milkshakes

Analysis of palatability showed no significant differences in the perceived palatability of the milkshakes between risk groups. Differences in palatability between milkshake types in the whole sample are shown in Supplemental Table 1 and Supplemental Figure 1.

DISCUSSION

Adolescents at high risk of developing obesity showed an elevated BOLD response to receipt of a palatable milkshake in brain regions associated with encoding gustatory and reward processing when compared with those at lower obesity risk, independent of current BMI. Furthermore, higher sugar content in the milkshake consumed elicited a more robust response in regions that are thought to encode gustatory processing, salience of stimuli, attention, hedonically motivated behaviors, and reward learning. The finding that food with higher sugar content elicits a greater striatal response dovetails with a previous report from a subsample of these participants that shows that when calories were consistent across milkshakes, higher sugar density elicited a stronger BOLD response than did higher fat content (22). Preclinical work further supports these results, with sugar but not fat, increasing and maintaining hedonically motivated eating behavior (e.g., compulsive eating) (20, 29). This potent capacity of sugar to elicit a strong response in hedonically motivated behaviors has led some research to draw parallels to neural response to substances of abuse (20, 29). Given the above findings, sugar's robust ability to elicit response from reward circuitry, and the knowledge that fat provides more energy density (kilocalories per gram), it is reasonable to suggest that sugar is more associated with creating and maintaining habitual overeating and that the fat typically paired with sugary foods more readily contributes to excess calories (20, 29). Moreover, adolescence is a critical time point in the development of dopaminergic systems, resulting in overperformance of hedonically motivated behaviors (30), in risk taking, impulsivity, and reward seeking (31, 32). Thus, theoretically, this is a critical period for developing and establishing habits around repeated consumption of high-sugar, highly palatable foods, which might be escalated in adolescents with overweight or obese parents.

A viable alternative interpretation of the relative contribution of sugar compared with fat focuses on the impact of higher fat content. Results presented here could indicate that a high fat content threshold acts to dampen, for example, the caudate response. Although fat and sugar both provide calories and are associated with high palatability, they are fundamentally different in terms of hedonic sensation and mouth-feel (33). For example, fat is more associated with texture and mouth-feel, whereas the ingestion of sugar is more readily associated with hedonic aspects of taste. As such, it is possible that repeated exposure to high-fat milkshakes over the course of the experiment decreased the caudate response due to the creamy and thick texture of fat imparting sensory-specific satiety (34, 35). This dampening or sensory-specific satiety effect could be more robust than the differential brain response (by risk status) seen when consuming the higher-sugar variants. This notion is supported by the observed effect of risk status when grouping all milkshakes was observed, in which the aggregate tastant contained proportionally a high amount of sugar, although the increased number of events in this contrast also provides greater statistical power, increasing the ability to observe smaller effects.

The above results replicate previous findings that show the relation of parental overweight status on greater caudate region response to tastes of a palatable food from an independent sample, examined on a different MRI and using a different paradigm and a different analytic approach, meeting the replication criteria that have been suggested for full generalizability and conceptual replicability (36). We did not observe an effect of parental weight status when responding to self-selected food images. There are multiple interpretations of why risk status affected (higher sugar) food receipt but not response to food images. First, on a theoretical level, these results support tenets of the dynamic vulnerability model of obesity, suggesting that parental obesity may underlie the initial obesity risk factors (37, 38). In the dynamic vulnerability model, the initial stage is associated with greater striatal (caudate, putamen, nucleus accumbens, and olfactory tubercle) response to taste that potentiates overeating, and therefore it is theorized that initially there is not a heightened response to food cues, in line with the present results and results from our earlier study in which healthy-weight youth at high risk of obesity only showed greater reward region response to receipt, but not a picture of, a chocolate milkshake, which signaled impending receipt of milkshake tastes (37, 38). This logic is in line with multiple obesity and overeating theories and reports that support the notion that increased response to cues is developed over the time of overeating and weight gain (7, 9, 39). An alternate explanation lies in the manner in which the stimuli were selected. Specifically, the images were selected according to individual preference, not as a function of food characteristics (e.g., fat content or sugar content), in which image categories based on objective food characteristics would be more analogous to the tastants presented. Given the nature of the food image paradigm design, attempting to analyze the data by macronutrient content is not viable because participants would have a varied number of exposures. Last, as mentioned in Methods, the milkshake receipt paradigm was not designed to adequately test for cue-elicited anticipatory response (i.e., the predicative milkshake cue). There were not enough events or exposures of cues and thus anticipatory responses were not tested.

We did not observe an effect of risk when directly contrasting the high-sugar and low-fat with the high-fat and low-sugar milkshakes, which would be the most direct test of our abovementioned hypotheses on sugar and fat. These 2 milkshakes had the same flavor and were eucaloric, and thus only varied on macronutrient composition. The lack of effect here may be a function the highly conservative nature of the milkshake to milkshake comparison. This low-sensitivity issue when comparing very similar stimuli is prevalent in the literature and particularly challenging to overcome when appropriately correcting for multiple comparisons. The lower sensitivity could have also contributed to the null finding when testing against the low-sugar and low-fat milkshake. Another possible cause for null results when comparing the conservative contrasts, such as the high-sugar and low-fat milkshakes with the low-sugar and high-fat milkshakes and vice versa, could be due to insufficient sample sizes in the high- and low-risk groups (40, 41). As such, presenting and comparing tastants that only contain sugar with a tastant that only contains fat with a larger sample size may be an ideal strategy in future work. In addition, it should be noted that previous exposure to palatable foods has been found to modify the BOLD response (42). Although we did not see a difference between the high- and low-risk groups in intakes of added sugar, fat, or total calories (Table 1), this could be a potential confounder. Last, it should be noted that the time of day the scans were performed occurred over a 7-h window. There were no between-risk-group differences in time of day the scan was performed, although there remains a small chance that an individual-level time of day by risk group by individual-level milkshake (compared with tasteless receipt) response interaction influenced the presented results. Because time of day has been suggested to affect striatal functioning (43), this is a potential confounder that may have affected the present results.

In conclusion, our results indicate that parental weight status was associated with a greater BOLD response to the consumption of milkshakes. In particular, high-sugar content was more robustly implicated in these effects in regions thought to encode hedonically motivated behavior via dopaminergic functioning. These data indicate that this elevated responsivity of regions implicated in reward and gustatory processing may underpin heritable obesity risk factors that theoretically contribute to overeating.

Notes

Supported by NIH grants DK-080760 and DK-112317.

Supplemental Table 1 and Supplemental Figure 1 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn.