Neurocognitive factors predicting BMI changes from adolescence to young adulthood

Abstract

Objective:

To assess whether inhibitory tasks performance in adolescence could be prospectively related to weight gain in young adulthood. We propose that this association would differ according to the BMI group in adolescence.

Methods:

318 adolescents performed the antisaccade task and 530 completed the Stroop test. Accuracy and reaction time were assessed for each incentive type (neutral, loss and reward) in the antisaccade task and for each trial type (control and incongruent trials) in the Stroop test. Changes in the BMI z-score (ΔBMI z-score) from adolescence to young adulthood were calculated.

Results:

The relationship between the BMI z-score and the antisaccade task accuracy showed an effect on the ΔBMI z-score (β=−0.002, p<0.05). The neutral and loss accuracies were related to ΔBMI z-score in the overweight (all β=−0.004, p=0.05) and obesity (β=−0.006 and β=−0.005, p<0.01) groups. The interaction between adolescents’ BMI z-score with control (β=−0.312, p<0.001) and incongruent (β=−0.384, p<0.001) trials reaction times showed an effect on the ΔBMI z-score. Control (֠β=0.730, p=0.036) and incongruent (β=0.535, p=0.033) trials reaction times were related to the ΔBMI z-score in the OW group.

Conclusion:

Our findings support the hypothesis that cognitive vulnerability could predict the BMI gain from adolescence to young adulthood.

INTRODUCTION

Obesity (OB) is associated with multiple comorbidities, including certain cancer types, diabetes, hypertension, osteoarthritis, metabolic syndrome, atherosclerosis, heart failure, and atrial fibrillation (1,2). The World Health Organization declared OB an epidemic and estimated that 13% of adults worldwide (650 million) were obese and 39% were overweight (OW) in 2016 (3). In Chile, the current prevalence of OW and OB in adulthood is 71% (4), and recent data showed that 48% of adolescents have OW or OB (5).

Behavioral, environmental, and biological factors are determinants of OB. The interactions among these factors are complex and are still far from being fully understood. Many of the OB prevention interventions have focused on behavioral approaches, including healthy dietary practices and physical activity. These programs have shown marginal or no long-term effects on their prevention (6). Specifically in adolescents, adherence to lifestyle intervention programs is low, and those who participate do not achieve or maintain lasting weight loss (7).

Furthermore, OB in adolescence is significantly associated with an increased risk of cardiovascular diseases and type 1 and 2 diabetes in adulthood (8–10). However, data show that even smaller reductions in body mass index (BMI) from childhood to adulthood are related to risk reductions in hypertension, dyslipidemia, and type 2 diabetes in adulthood (11). Recent reports have shown an association between higher BMI during adolescence and increased risk of leukemia, Hodgkin’s disease, and pancreatic cancer in adulthood (12–14).

Nutritional status in adolescence strongly predicts OW and OB in adulthood (15). Studies have shown that 46% of normal-weight (NW) adolescents remain NW in adulthood; 17% of adolescents with OW persist OW as adults, and 78% shift to OB; 14% of adolescents with OB change to OW into adulthood, and 86% of adolescents with OB remain in OB (15,16). Therefore, transition to higher BMI categories is the most common outcome, but it can also be stated that 70-80% of adults with OB were not in the OB group in adolescence (17). Evidence has shown that a decrease in lower weight during adolescence tends to mitigate the progression to OW and OB (15), suggesting that adolescence and the transition to young adulthood are relevant stages for OB intervention and prevention.

In adolescence, higher BMI is associated with multiple comorbidities (8–11), but it is also associated with cognitive functions and alterations in functional networks and brain microstructure (18). A bidirectional relationship between weight status and cognitive functions has also been described (19). Relative to adulthood, the adolescent brain is characterized by greater sensitivity to reward coupled with limitations in inhibitory control (20). The mesolimbic reward system involves brain regions related to the cognitive processing of reward, in which the neurotransmitter dopamine (DA) plays a key role in reward processing, e.g., food products (21). In this process, the brain associates diverse stimuli with an outcome, thereby modifying the individual’s reward-seeking behavior. Reward is linked to incentive salience, which enhances motivation toward reward-related behaviors (22). The mechanism underlying reward sensitivity in adolescence could be related to fluctuations in DA receptor expression and changes in DA projections between the prefrontal cortex and subcortical brain structures (23). The prefrontal cortex is a region that is keenly involved in the modulation of executive functions, including inhibitory control (21). Inhibitory control is the ability to voluntarily suppress dominant, automatic, and preponderant responses in favor of a planned response, which is largely modulated by the DA frontostriatal circuitry, fluctuates in response to reward-driven motivations, and is implicated in preventing impulsive actions to control overeating (21).

Previous work has shown differences in inhibitory control in children and in incentive sensitivity (reward and loss avoidance) in adolescents with OW and OB (24,25), suggesting that in children with weaker inhibitory control, the decision-making process could be affected and could play an important role in overeating behavior (24). On the other hand, adolescents with OW and OB may be particularly sensitive to reward and loss avoidance. Recent evidence supports the relationship between brain functioning and OB with emphasis on reward network, inhibitory control, and emotional regulation-related processes (18).

A significant limitation of studies assessing the association between reward, inhibitory control, and OB in adolescents is the preponderance of their cross-sectional design (25); in fact, longitudinal studies are scarce. A prospective study in adolescents showed that increased reward region responsiveness to palatable food images could predict weight gain (26). In the current longitudinal study, we assessed whether performance in incentivized and inhibitory control tasks in adolescence (16 years) could be related to weight gain in young adulthood (22 years). We hypothesized that adolescents with higher incentive sensitivity and decreased inhibitory control would exhibit increased weight gain (BMI z-score gain) between adolescence and young adulthood. Furthermore, we propose that this association would differ according to the BMI category in adolescents.

METHODS

Participants

Participants were part of a longitudinal cohort studied since infancy as part of an iron deficiency anemia preventive trial. Detailed descriptions of the original study have been described elsewhere (27). In brief, the inclusion criteria for enrollment in infancy were as follows: healthy full-term birth, birth weight ≥3.0 kg, and absence of perinatal complications or chronic diseases. Infants with iron-deficiency anemia at 6, 12, or 18 months were considered for neurofunctional assessments, along with the next “healthy” (nonanemic) control. All participants received oral iron for at least 6 months and had normal hemoglobin levels after treatment. At 16 years of age, a sample of the original infancy cohort was invited to participate in a study on cardiovascular risk and obesity and in a subsequent follow-up at 22 years. Neurocognitive assessments were conducted at the Sleep and Functional Neurobiology Laboratory, INTA, University of Chile, at 16 and 22 years.

Adolescence and young adulthood study protocols were reviewed and approved by the Institutional Review Boards of the University of Michigan, Ann Arbor; University of California, San Diego; and the Institute of Nutrition and Food Technology (INTA), University of Chile, Chile. Parents provided signed informed consent and they gave their written assent in adolescence. In adulthood, the participants provided informed consent.

Procedures

There were 530 adolescents who performed the Stroop test and 318 who completed the incentivized antisaccade task. For this study, the exclusion criteria were: pregnancy, major medical and neurological disorders, psychiatric illness, and drug or alcohol dependency or abuse. All participants completed the neurocognitive task, anthropometric measurements, and questionnaires. From the initial sample, three were excluded from the analysis for movement artifacts during recordings. Excluded participants were not different in demographic and background characteristics.

Incentivized antisaccade task:

The neural circuitry that underlies behavioral performance in this task has been well characterized in animals and humans (28) and has been used to investigate the interaction between cognitive control and the incentive (reward and loss) effect (25). This is an oculomotor test that explores the ability to exert cognitive inhibitory control by employing voluntary suppression of a prepotent saccadic response (fast eye movements) in the presence of “reward”, “loss avoidance”, or “neutral” incentives. Participants had to inhibit eye movement toward a visual stimulus and instead make a planned saccade to its mirror location (antisaccade). Each trial began with 2-or 3-s presentation of one of the three possible incentive types (25):

(a) Reward: An image of a 1,000 Chilean peso bill (US$ 1.5) indicated monetary gain if the participant performed the trial correctly. An error did not result in a “loss of money”.

(b) Loss avoidance: A torn bill image of the same amount indicated a monetary loss if the saccade was incorrect. The correct response did not result in a “gain of money”.

(c) Neutral: A green rectangle indicates no incentive, i.e., no money was “gained” or “lost”, and regardless of performance, the amount of money remained the same.

Following one of these incentive images, a peripheral target (a small yellow dot) appeared for 1.0 s at one location (to the left or right of the screen center) to indicate that an antisaccade must be completed (response phase). Finally, a central stimulus appeared for 1.0 s to center the subject’s gaze before the next trial. They were encouraged to perform the task as well and as quickly as possible, regardless of the incentive type. During the task, the participants did not receive feedback about their performance. Twenty reward, 20 loss avoidance, and 20 neutral trials were presented in random order.

Saccades were recorded using an eye-tracking system (Eye-Trac 6; Applied Science Laboratories, Bedford, MA) that uses a corneal reflection method with bright pupil technology. The point-of-gaze was determined by the corneal reflection of an infrared beam, which was projected to the center of the illuminated pupil that rotates with each eye movement. Visual stimuli were displayed on a computer monitor using the E-Prime software (Psychology Software Tools, Pittsburgh, PA). In a darkened room facing the stimulus monitor, the participants remained comfortably seated at 60 cm from the center of the monitor. Before the eye-tracking session, a 9-point calibration was performed. Standardized instructions were carefully provided by trained personnel, and the recording began after the participants demonstrated their understanding.

Saccades were scored off-line using ILAB software (Northwestern University Medical School and V.A. Healthcare System, Chicago, IL) and MATLAB (MathWorks, Natick, MA) to calculate the latency and accuracy of correct saccades. In each trial, the first saccade was chosen as the response. A correct response was defined as a saccade with velocity ≥ 30°/s made toward the mirror location of the peripheral target and extended beyond a 2.5°/visual angle from the central stimulus. An incorrect response occurred when the saccade was directed toward the peripheral target and exceeded the 2.5°/visual angle from the central stimulus. Trials that were excluded from the analysis were those in which the eye movement latencies were < 70 ms or there was no fixation on the central stimulus at the onset of the trial.

Sensitivity for each incentive type was assessed using the following behavioral variables: percentage of correct responses (accuracy), percentage of correction, and latency (ms) of correct and incorrect responses.

Stroop task:

The Stroop test has been extensively used to evaluate cognitive inhibition (29). Stimuli were presented on a computer monitor. Participants responded using a computer keyboard. The targets were the words ‘red,’ ‘blue’ and ‘green’ or a string of ‘X’s’ displayed in red, blue, or green. Color words were stimuli in the incongruent trials (for example, the word ‘red’ displayed in blue), and strings of ‘X’s’ were used in the control trials. Half of the items consisted of incongruent trials and the other half were control trials. Each item was displayed in one of the three colors an equal number of times. Stimuli were displayed on a computer monitor using the E-Prime software (Psychology Software Tools, Pittsburgh, PA). Three keys from the computer keyboard were covered with a color patch, and participants were instructed to press the key that corresponded to the color of the target as soon as it appeared on the screen. Participants had a practice block of 12 trials, followed by an experimental block of 60 trials. Each trial began with a fixation point (‘+’ sign) presented for 500 ms in the center of the screen, after which the stimulus appeared on the monitor until the subject’s response (24).

Reaction times (RT) and percentages of correct responses (accuracy) for the control and incongruent trials were calculated. In addition, the effect of the preceding trial’s type on cognitive inhibitory processes was assessed by comparing RTs for incongruent trials that followed two successive control trials (control–control–incongruent: estimate of conflict adaptation) and RTs for control trials that followed two successive incongruent trials (incongruent–incongruent–control: estimate of conflict inhibition) (24).

Anthropometry

Trained personnel recorded weight and height measurements using standardized techniques (without shoes, wearing underwear, and in the Frankfurt position) on the same calibrated machine every day. Weight to the closest 0.1 kg and height to the closest 0.1 cm were measured using a SECA scale (model 700, Seca, Hamburg, Germany). BMI was calculated for each participant in adolescence and young adulthood as the ratio of weight in kilograms divided by the square of height in meters. Sex-and age-specific BMI z-scores were calculated and categorized according to the cut-off points recommended by the World Health Organization as NW (BMI z-score ≥ −2 to < 1 SD), OW (BMI z-score ≥ 1 SD), and OB (BMI z-score ≥ 2 SD). WHO BMI-for-age at 19 years was used for the BMI at 22 years. Changes in the BMI z-score from adolescence to young adulthood were calculated (adulthood-adolescence).

Questionnaires

Self-report instruments filled out in adolescence were as follows: (a) Epworth sleepiness scale for daily sleepiness assessment, participants were asked to rate how likely they would be - in recent times - to fall asleep in eight situations; higher scores indicate increased sleepiness (30); (b) dietary intake questionnaire to evaluate the quality of the food intake, adolescents assessed their usual food intake by meal frequency, an increased score is related to better quality of dietary intake (31); and (c) physical activity score to measure usual physical activity level, five items were assessed (recumbent, seated, walking, playing outdoor, sports), and higher scores suggest greater physical activity (32).

Data analysis

We conducted ANOVA and repeated measures to examine the main effects and interaction of performance on neurocognitive tasks (Incentivized antisaccade task and Stroop test) and group (NW, OW and OB). The within-participants factor was incentive type (reward, loss avoidance, and neutral), and the between-participants factor was group (NW, OW and OB). Generalized linear models were calculated to explore the relationship between BMI and performance in neurocognitive tasks in adolescents with changes in the BMI z-score from adolescence to young adulthood (ΔBMI z-score). The response variable was the ΔBMI z-score, adolescent neurocognitive performance was the moderator variable, and the predictor variable was BMI in adolescence, first the BMI z-score and then BMI category. All analyses were adjusted for sex, age at the neurocognitive test, physical activity (32) and diet quality assessment (31) scores in adolescence, and socioeconomic status. Socioeconomic status was assessed using a modified Graffar index based on ten items that included family size and structure, father presence, educational level of the head of the household, and home ownership (33). The Bonferroni correction for the multiple comparison method was applied. A p-value < 0.05 (corrected for multiple comparisons) was considered statistically significant. Statistical analyses were conducted using SPSS software version 19.0 (SPSS Inc., Chicago, IL, USA).

RESULTS

From adolescence to young adulthood, participants exhibited an increase in the prevalence of OW and OB (Table 1). Of NW adolescents, 67% remained NW and 28% moved to OW in young adulthood. Among OW adolescents, 57% were OW and 37% shifted to OB in young adulthood. Regarding OB adolescents, 90% remained in the OB group in adulthood. From young adulthood, 48% of young adults in the OB group were not OB in adolescence; of them, 36% were in the OW group in adolescence.

Table 1.

Background and descriptive characteristics of the participants in adolescence and young adulthood

	Adolescence n=530	Young adulthood n=530
Female n (%)	264 (50.20)
Birht weight (Kg)	3.56 ± 0.02
Iron-deficiency anemia in infancy n (%)	132 (24.90)
Socioeconomic status a	33.80 ± 0.30
Age at test	16.74 ± 0.02	22.74 ± 0.03
BMI (kg/m2)	24.05 ± 0.20	26.93 ± 0.24*
BMI-z score	0.73 ± 0.05	1.21 ± 0.05*
Delta BMI-z score		0.48 ± 0.03
Normal weight n (%)	310 (58.50)	216 (40.70)
Overweight n (%)	138 (26.00)	174 (32.80)
Obesity n (%)	82 (15.50)	140 (26.40)
Epworth sleepiness scale	7.90 ± 0.20	9.00 ± 0.21*
Formal education (y)		11.50 ± 0.10
High school graduation n (%)		464 (88.00)

^*Paired sample t-test, p<0.001

^aGraffar score at 10 years

In adolescence (Table 2), the BMI groups were similar in background characteristics and by design, differed in BMI measurements. The OB group had a significantly lower physical activity score than the NW (p<0.01) and OW (p=0.03) groups, and a smaller change in BMI z-score (Δ) (p=0.06) than the NW and OW groups.

Table 2.

Background and descriptive characteristics of the participants by BMI group in adolescence

	Normal weight n=310	Overweight n=138	Obesity n=82	p-value
Female (%)	146 (47.10)	76 (55.07)	45 (54.88)	0.227
Birth weight (Kg)	3.53 ± 0.02	3.58 ± 0.03	3.63 ± 0.04	0.064
Iron-deficiency anemia in infancy n (%)	81 (26.13)	32 (23.19)	21 (25.61)	0.791
Socioeconomic status a	34.14 ± 0.44	33.14 ± 0.59	33.24 ± 0.67	0.322
Age at test	16.75 ± 0.02	16.78 ± 0.02	16.72 ± 0.04	0.495
BMI (kg/m2)	21.04 ± 0.11	26.10 ± 0.11	32.03 ± 0.35	0.000b
BMI-z score	−0.05 ± 0.04	1.43 ± 0.02	2.52 ± 0.05	0.000b
Change BMI-z score (Δ)	0.54 ± 0.04	0.45 ± 0.05	0.35 ± 0.07	0.061
Epworth sleepiness scale	7.87 ± 0.20	8.27 ± 0.28	7.13 ± 0.39	0.058
Physical activity score	4.20 ± 0.10	4.17 ± 0.15	3.55 ± 0.16	0.009c
Food intake questionnaire (score)	5.23 ± 0.07	5.09 ± 0.13	5.50 ± 0.16	0.085

One-way analysis of variance (ANOVA) and p-values adjusted using the Bonferroni multiple-comparison test.

^aGraffar score at 10 years.

^bAll groups were significantly different among them (Bonferroni test p<0.05).

^cObesity group vs. Healthy-weight and Overweight groups (Bonferroni test p<0.05).

Behavioral performance in neurocognitive tasks in adolescents

In the Incentivized antisaccade task, there was no significant effect of the BMI group (Table 3). There was a significant effect of the BMI group (F=3.85, p<0.05) on the accuracy of incongruent trials on the Stroop task, in which the OB group showed decreased accuracy compared with the NW participants (p<0.05) (Table 3).

Table 3.

Behavioral performance of adolescents in neurocognitive tasks by BMI group.

	Total n=530	Normal weight n=310	Overweight n=138	Obesity n=82	p-value
Stroop task
RT control trials (ms)	697.82 ± 7.17	691.32 ± 8.84	708.24 ± 14.06	704.83 ± 21.74	0.556
RT incongruent trials (ms)	739.16 ± 9.07	732.68 ± 11.05	743.33 ± 17.71	756.62 ± 28.46	0.629
RT conflict adaptation (ms)	198.60 ± 5.80	202.66 ± 7.71	182.20 ± 9.60	210.37 ± 16.94	0.225
RT conflict inhibition (ms)	169.46 ± 5.31	165.81 ± 6.94	179.42 ± 10.60	166.79 ± 13.10	0.543
Correct control trials (%)	96.33 ± 0.19	96.52 ± 0.24	96.49 ± 0.41	95.40 ± 0.56	0.121
Correct incongruent trials (%)	95.28 ± 0.46	96.23 ± 0.47	94.63 ± 1.12	92.81 ± 1.39	0.022
Incentivized antisaccade task a
Neutral incentive
Correct saccade (%)	51.18 ± 1.47	53.67 ± 1.61	52.09 ± 2.65	47.78 ± 3.12	0.245
Corrected saccade (%)	42.72 ± 1.43	41.40 ± 1.57	42.08 ± 2.58	44.68 ± 3.05	0.633
Latency correct saccade (ms)	400.71 ± 5.24	401.60 ± 5.72	409.44 ± 9.59	391.08 ± 11.05	0.456
Latency incorrect saccade (ms)	304.35 ± 3.26	301.11 ± 3.55	310.08 ± 5.86	301.86 ± 6.97	0.415
Loss incentive
Correct saccade (%)	49.99 ± 1.64	51.90 ± 1.79	48.35 ± 2.96	49.71 ± 3.49	0.561
Corrected saccade (%)	43.63 ± 1.56	42.61 ± 1.71	46.40 ± 2.82	41.86 ± 3.33	0.460
Latency correct saccade (ms)	404.25 ± 5.06	402.87 ± 5.52	403.27 ± 9.27	406.62 ± 10.67	0.952
Latency incorrect saccade (ms)	307.13 ± 3.37	302.35 ± 3.68	311.77 ± 6.06	307.26 ± 7.21	0.396
Reward incentive
Correct saccade (%)	52.15 ± 1.60	54.41 ± 1.76	52.81 ± 2.90	49.37 ±3.42	0.419
Corrected saccade (%)	41.40 ± 1.48	39.95 ± 1.62	41.78 ± 2.67	42.64 ± 3.15	0.709
Latency correct saccade (ms)	404.63 ± 4.55	402.63 ± 5.72	404.68 ± 8.34	406.58 ± 9.60	0.928
Latency incorrect saccade (ms)	309.10 ± 3.05	306.45 ± 3.32	316.80 ± 5.48	304.06 ± 6.52	0.210

Mean ± SE

One-way analysis of variance (ANOVA) and p-values adjusted using Bonferroni multiple-comparison test.

^aANOVA with repeated measures, normal weight n=193, overweight n=71, and obesity n=54.

Behavioral performance in neurocognitive tasks in adolescence and changes in the BMI z-score to young adulthood

Incentivized antisaccade task

The interactions between the BMI z-score and accuracy for neutral, loss, and reward incentives in adolescence showed significant effects on the ΔBMI z-score (all β=−0.002, p<0.05). These effects were also significant for the BMI group and accuracy for neutral (W[3] = 13.36, p=0.004) and loss incentives (W[3] = 12.29, p=0.003). A negative relationship was established between ΔBMI z-score and accuracy for neutral and loss incentives in the OW (β=−0.004, p=0.014 and β=−0.004, p=0.024, respectively) and OB (β=−0.006, p=0.007 and β=−0.005, p=0.004, respectively) groups. Given the coefficient values and according to the adjusted model, examples of predictive means of the ΔBMI z-scores were calculated for loss incentive accuracy and the BMI z-score model (Figure 1) and then for the BMI group model (Figure 2).

Figure 1. Predicted values of the ΔBMI z-score according to the adjusted model for loss incentive accuracy (%) of Incentivized antisaccade task and the BMI z-score in adolescents.

Predictive means of the ΔBMI z-score with 95% confidence intervals for the BMI z-score and loss incentive accuracy in adolescence (10% accuracy: dot line, 40% accuracy: short dash line, 70% accuracy: long dash line, and 100% accuracy: solid black line).

Figure 2. Predicted values of the ΔBMI z-score according to the adjusted model for loss incentive accuracy (%) of Incentivized antisaccade task by BMI group in adolescence.

Predictive means of the ΔBMI z-score with 95% confidence intervals for loss incentive accuracy by BMI group in adolescence (NW: circle and solid light gray line, OW: square and dark gray line and OB: triangles and solid black line).

Stroop task

The relationship between the adolescent’s BMI z-score with the RT for control (β= −0.312, p<0.001) and incongruent (β= −0.384, p<0.001) (Figure 3) trials had a significant effect on the ΔBMI z-score. Also, by BMI group, the RT of control (W[2] = 8.69, p=0.011) and incongruent trials (W[2] = 8.62, p=0.013) were related to the ΔBMI z-score. There was a positive relationship between the RT of control trials and the ΔBMI z-score in the OW (֠β=0.730, p=0.036) and OB (֠β=0.629, p=0.015) groups (Figure 4). In addition, a positive relationship was found between the RT of incongruent trials and the ΔBMI z-score in the OW group (֠β=0.535, p=0.033) (Figure 5).

Figure 3. Predicted values of the ΔBMI z-score according to the adjusted model for reaction time (ms) of control and incongruent trials of the Stroop task and the BMI z-score in adolescents.

Predictive means of the ΔBMI z-score with 95% confidence intervals for the BMI z-score and reaction time of (A) control [400 ms: dot line, 700 ms: short dash line, 1000 ms: long dash line, and 1300 ms: solid black line] and (B) incongruent [500 ms: dot line, 800 ms: short dash line, 1100 ms: long dash line, and 1400 ms: solid black line] trials in adolescence.

Figure 4. Relationship between reaction time (ms) of control trials on the Stroop task in adolescents and changes in the BMI z-score by BMI group in adolescence.

(A) There was no significant association between reaction time (RT) of control trials and ΔBMI z-score in the NW group (֠β=−0.378, p=0.195). (B) and (C) Positive association between RT of control trials and ΔBMI z-score in the OW and OB groups, respectively. Lines represent the fit values and 95% confidence intervals.

Figure 5. Relationship between reaction time (ms) of incongruent trials on the Stroop task in adolescents and changes in the BMI z-score by BMI category in adolescence.

(A) There was no significant associations between reaction time (RT) of incongruent trials and ΔBMI z-score in the NW (֠β=−0.339, p=0.130) and (C) OB (֠β=0.401, p=0.113) groups. (B) Positive association between RT of incongruent trials and ΔBMI z-score in the OW group was significant. Lines represent the fit values and 95% confidence intervals.

DISCUSSION

This study highlights the prospective relationship between inhibitory control and incentive sensitivity in adolescence with changes in the BMI z-score between adolescence and young adulthood and contributes to the understanding of the longitudinal interaction between cognitive vulnerability and BMI gain. Our results showed that this relationship differed according to the adolescents’ BMI. Considering that almost half of the OB young adults were not OB adolescents, our results deserve attention: the OW and OB groups in adolescence shared some neurocognitive features, but the OW group showed a specific relationship between decreased inhibitory control and greater increase in BMI z-score.

Regarding the results of incentive sensitivity, the coefficient values were low; however, they have practical significance when considered within the scenario of cognitive vulnerability in adolescents. Taking this into consideration, our data indicated that in the Incentivized antisaccade task, adolescents with a greater BMI z-score and higher accuracy will show lower increases in BMI z-score into adulthood compared with adolescents with lower accuracy. This correlation was observed in the OW and OB groups but was slightly higher in the OB group. These results are consistent with the model that suggests that repeated intake of energy-rich foods in individuals could trigger or exacerbate deficits in striatal reward DA circuits, thus increasing the risk of weight gain (21,34).

On the other hand, adolescents showed a relation between higher BMI z-score values, decreased cognitive inhibitory control, and greater increases in the BMI z-score in young adulthood. Specifically, the OW group showed how inhibitory control was interwoven with increases in the BMI z-score. This may be related to differences in DA receptors activation in the prefrontal cortex, which modify top-down inhibition, a neural process necessary for self-regulation (35). In environments with easy access to high-caloric food, the ability to inhibit food seeking is key to modulating risky behaviors and attenuating the temptation of overeating (21,36).

Additionally, adolescence is a period of maturation of the DA reward and cognitive control pathways, resulting in increased sensitivity to reward, less inhibitory control, and a tendency toward impulsive behaviors (37,38). Response inhibition and incentive sensitivity, which are components of the neurocognitive tasks used in this study, are processes engaged during more complex decision making in adolescents (24,25). These characteristics could trigger the potentiation of the adverse neural effects of obesogenic environments in the brain and influence neurofunctional development in adolescents (39).

This sensitive period of cognitive development and maturation is aligned with the fact that even though the transition to a higher BMI is the common outcome, adolescence is still considered a key moment to transition to a lower BMI and avoid obesity (15,17). As we have already described, our results of neurocognitive tasks in adolescence and changes in the BMI z-score into young adulthood were significant in the OW and OB groups. In this regard, evidence suggests that in early adolescence: (a) differences in the performance of executive functions may precede weight gain, (b) increases in BMI are related to decreased inhibitory control (40), and (c) higher BMI and impulsive traits are related to greater intake of fat and added sugar, leading to weight gain (41). In line with these findings, at an early age our participants showed differences in inhibitory control (24). This could be related to inappropriate food-based decisions and greater weight gain during adolescence. Therefore, we suggest that mainly in adolescents with higher BMI values (OW and OB groups), having a slightly better cognitive control performance at 16 years could have a greater impact on BMI changes in adulthood.

In this study, we proposed that chronic and repeated exposure to highly palatable food and the downregulation of DA receptors in the striatal– prefrontal pathway, could generate a reduced capacity for self-regulation of dietary intake (21,39). Although our participants were exposed to the same food environment, those with a higher BMI in adolescence might have a different onset and duration of OW or OB (42). This exposure to excess weight, and the probable prolonged consumption of an unhealthy diet and the accumulation of adipose tissue (43), occurred before 16 years of age and could generate a greater neural vulnerability in these groups.

Studies in preschoolers have shown that executive functions and inhibitory control ability can predict BMI at school age (44,45). This result received substantial support in a recent meta-analysis showing that inhibition could predict weight loss, mainly at an early age (46). An imaging study in adolescents showed limited evidence regarding reward region responsivity to palatable food and its potential value in predicting weight gain (26). To the best of our knowledge, this is the first study to show that differences in inhibitory control and sensitivity to incentives in adolescence could predict weight changes in young adulthood.

However, limitations must also be acknowledged. We relied on BMI to establish OW and OB status and changes in weight gain, but measures such as body composition or biochemical indices might classify individuals more accurately. However, almost all OB studies in adolescence and disease risk in adulthood have used BMI. Thus, in order to compare BMI changes with respect to adolescence, we used the WHO BMI-for-age at 19 years reference for z-score calculation in adulthood; other studies have also used this reference for young adults (26,42). Self-report of impulsivity traits, dietary intake and physical activity, factors related to OW and OB, could be assessed by different methods (47). However, all these variables were included in our data analysis using validated questionnaires or behavioral neurocognitive tasks. On the other hand, to understand brain functioning in adolescents with OB, imaging methods could be included in future studies (48). These limitations do not allow us to extrapolate our results to the entire population of Chilean adolescents. Despite these limitations, the strengths of the study include its prospective design, objective measures of inhibitory control and incentive sensitivity, height and weight taken by trained personnel, and adjustment of analysis for potential confounders.

CONCLUSION

To conclude, our findings support the hypothesis that cognitive vulnerability predicts BMI gain from adolescence to young adulthood. The transition to higher BMI categories from adolescence to adulthood is the most expected outcome (15). However, with an important proportion of adults with OB who were not in the OB group in adolescence, interventions to promote lower weight gain during adolescence could mitigate the tendency of transitioning into the OB category in early adulthood. In conclusion, our results support the reinforcement of adequate development of cognitive functions, strengthening of executive functions, and inclusion of the neurocognitive dimension in interventions focused on OB prevention.

Study Importance Questions.

What is already known about this subject?

• In adolescence, a higher BMI is associated with cognitive functions and alterations in functional networks and brain microstructure.

• Recent evidence supports the relationship between brain functioning and OB, with an emphasis on reward networks, inhibitory control, and emotional regulation-related processes. However, a significant limitation of studies evaluating this association in adolescents is the preponderance of their cross-sectional design.

What are the new findings in your manuscript?

• This study highlights the prospective relationship between inhibitory control and incentive sensitivity in adolescents with changes in the BMI z-score between adolescence and young adulthood.

• Our data contributes to the understanding of the longitudinal interaction between cognitive vulnerability and BMI gain.

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

• We propose the reinforcement of adequate development of cognitive functions in adolescence and the inclusion of the neurocognitive dimension in interventions focused on obesity prevention.