Racial/ethnic differences in dietary intake among a diverse sample of adolescents: An experimental study

Abstract

Background:

African-American (AA) and Hispanic/Latina (HL) females have higher obesity prevalence than do non-Hispanic Whites (NHW); this may be due to AA and HL consuming more energy-dense foods in response to stressors.

Objectives:

This study examined racial/ethnic differences in dietary intake under controlled conditions (relaxation and stress) in a diverse sample of adolescent females.

Methods:

Participants included 120 adolescent females (30% AA, 37% HL and 33% NHW) who participated in a laboratory food intake study. Using a randomized cross-over design, ad libitum food consumption was measured following control/relaxation and social-evaluative stress conditions. Food intake was indexed as consumed calories, added sugars and solid fats.

Results:

The effect of laboratory conditions on food intake varied by race/ethnicity, such that AA consumed more energy following relaxation than following stress. For NHW and HL, food intake did not differ between conditions.

Conclusions:

To the best of our knowledge, these findings are the first to directly observe racial/ethnic differences in food intake in response to acute stress, which may contribute to obesity-related health disparities.

1 |. INTRODUCTION

Obesity is among the most prevalent and costly public health problems in the United States.^1,2 Twenty percent of US youth are affected by obesity,¹ setting in motion trajectories of increased risk for cardiovascular and metabolic morbidity and mortality throughout the lifespan.³ However, the health toll of obesity is not equally dispersed, as African-American (AA) and Hispanic/Latina (HL) females experience higher obesity prevalence than do non-Hispanic Whites (NHW).¹ This health disparity is likely determined by an array of proximal and distal neurobiological, psychosocial and behavioural factors, including genetic profiles, epigenetic programming, insulin and glucocorticoid sensitivity, sleep-activity patterns, social and socio-economic inequalities and opportunities to exercise, obtain nutritious food and receive appropriate medical care.^4,5 Racial/ethnic differences in stress-induced dietary intake provides one important but under-studied pathway to obesity disparities. Specifically, social inequalities subject AA and HL youth to greater stress levels,^6,7 which promote the consumption of energy-dense foods.⁸ However, to the best of our knowledge, none of the studies employed experimental designs with direct behavioural observation methods to assess whether AA and HL exhibit greater energy intake following stress exposure than do NHW. The present study addresses this gap in the literature by examining how food intake differs between relaxing and stressful conditions among AA, HL and NHW adolescent females.

Adolescence is a critical period for the development of obesity, marked by changes in food consumption, physical activity, sedentary behaviour and fat distribution.⁹ Simultaneously, adolescence is associated with greater stress exposure⁸ and stronger physiological responses to stress.¹⁰ Racial/ethnic disparities in obesity emerge by adolescence,¹¹ but the negative health effects of obesity have not yet fully developed; therefore, studying obesogenic processes in adolescence is crucial for understanding how racial/ethnic differences in obesity develop and for informing future interventions to reduce obesity-related health disparities.

Greater stress exposure is associated with the development of obesity in childhood and adolescence, and stress-induced eating has been posited as a mechanism accounting for this association.^12,13 AA and HL may be particularly vulnerable to the obesogenic effects of stress-induced eating, due to both the dose and the types of stress to which they are exposed. AA and HL experience more stressful life events than do NHW,^8,9 yielding more frequent opportunities to engage in eating in response to stress. Additionally, the types of stressors to which AA and HL are commonly exposed may be particularly likely to give rise to stress-induced eating. For example, AA and HL adolescents face more stress related to discrimination and institutional racism compared with their NHW peers.¹⁰ Stressors that are uncontrollable and socially threatening, such as discrimination, acutely activate the hypothalamic-pituitary-adrenal (HPA) axis of the stress response system, which regulates appetite and eating behaviour, to promote the consumption of foods high in calories, added sugars and fats.^11,14 Moreover, as such stressors are not typically amenable to direct coping strategies, they may elicit emotion-focused or avoidant coping responses, such as eating highly palatable foods.¹⁵ Repeated exposure to such stressors alters brain function and HPA axis activation to enhance the rewarding nature of energy-rich foods, increase the regulatory effects of pleasurable activities such as eating, and impair inhibitory control of impulsive behaviours.^16,17 Individuals exposed to greater stress levels typically consume more energy-dense foods,^11,18 and females are particularly inclined to eat more in response to stress.^19,20 Therefore, stress exposures may predispose AA and HL young females to consume more palatable, energy-dense foods containing refined sugars and fats, particularly following exposure to stressors.

Evidence is mixed as to whether AA and HL youth consume more energy-rich diets than do NHW, with some studies indicating that AA adolescent females consume more calories than do NHW and/or HL,^21,22 but other studies failing to find differences in energy consumption between AA and NHW children.^23,24 However, less is known regarding racial/ethnic differences in stress-induced eating specifically. Rather, stress has been linked to energy-rich food consumption separately in NHW, AA and HL adult samples, with few studies comparing the magnitude of effects across groups.²⁵ Among children and adolescents, stress is associated with less healthy eating for AA girls²⁶ and more emotional eating for HL;²⁷ additionally, sugar intake moderates the relationship between stress (measured via HPA axis activation) and visceral adipose tissue among AA and HL adolescents.²⁸ Moreover, the existing literature has relied largely on non-systematic self-report to quantify food intake in response to stress, which introduces methodological limitations such as under-reporting, dependence on participant memory and common method bias. To the best of our knowledge, only one study has examined the effects of experimentally manipulated stress exposure on adolescents’ observed food consumption, finding that adolescents with obesity (45% AA and 55% NHW), showed a reduction in caloric consumption following a stressor, though this study did not assess for racial/ethnic differences in this effect.²⁹

The present study examines differences in food consumption among AA, HL and NHW adolescent females. Food consumption is assessed under laboratory conditions, which include both a relaxation control condition and the administration of a standardized psychosocial stressor. We predicted that racial/ethnic groups would differ in food consumption, such that AA and HL would consume more calories, added sugars and solid fats compared to NHW under both relaxing and stressful laboratory conditions. Additionally, we anticipated that the effect of in-lab condition (control vs. stress) on food consumption would vary by racial/ethnic group, such that AA and HL would show greater increases in food intake under the stress condition.

2 |. METHODS

2.1 |. Participants

A sample of 130 adolescent females was recruited from paediatric primary care clinics and affiliated community agencies in an urban area between December 2017 and December 2019, for a study on the influence of race/ethnicity on dietary patterns. Eligibility criteria included self-identification as AA, HL or NHW, ages 13–17 years, Tanner Stage ≥3 and body mass index (BMI) in the normal-to-obese range. Exclusionary criteria included low BMI (<5th percentile), pregnancy, medical problems (ie, endocrine disorders or unstable cardiac, pulmonary or renal conditions), use of medication that influences appetite or HPA axis, scoring below 50% on a visual analogue scale (VAS) to assess perceived pleasantness of foods or food preference provided during the study, or having allergies to any food presented on the VAS. Adolescents provided assent and their parents provided consent; all procedures were approved by the Institutional Review Board (Protocol #2017–3441). Of these 130 participants, two withdrew or were disqualified before completing any food consumption measure (one due to parent withdrawing, one due to child being underweight). One parent chose not to provide socioeconomic status (SES) information. Of the remaining 127 participants, 120 completed at least one laboratory condition; 117 completed the relaxation condition, 115 completed the stress condition and 112 completed both conditions.

2.2 |. Study design

Participants who met inclusion criteria were scheduled for initial assessments, which included demographic information, height, weight, and medical and psychiatric histories. Subsequently, participants completed two separate food consumption lab visits (relaxation and stress conditions, presented in a randomized sequence), scheduled 530 days apart and supervised by a registered dietitian. Participants were asked to fast overnight (from 10 PM) until their laboratory visits (at 8–9 AM). During each visit, a standardized 210 calorie breakfast was provided, consisting of a cereal bar and juice, followed by a 2.75-hour relaxation period during which participants watched movies with neutral content. In the control/relaxation condition, participants then watched a 15-minute nature video; in the stress condition, participants instead were administered the Trier Social Stress Test (TSST), a 15-minute acute social-evaluative stressor.³⁰ Following the stress/control manipulation, a buffet lunch was provided (30 min after the assigned manipulation, timed to detect effects of stress on eating behaviour),³¹ which consisted of a wide assortment of foods that youth commonly eat, selected to differ greatly in macronutrient composition and caloric value, and are based upon food choices provided to youth in similar studies.³² The 32 foods offered included low-calorie healthy (eg, green leafy salad), high-calorie healthy (eg, raw almonds), low-calorie unhealthy (eg, diet Coke) and high-calorie unhealthy (eg, donuts) options. Participants were instructed to eat whatever they wanted and given 40 min to eat ad libitum in a private room. After lunch, participants who had completed the stress condition were debriefed.

2.3 |. Measures

2.3.1 |. Anthropometry

Height and weight were measured using the National Health and Nutrition Examination Survey (NHANES) III methodology,³³ which takes the average of three recordings. Height was measured on a free-standing stadiometer and recorded to the nearest 0.1 cm. Weight was obtained with subjects in light clothing without shoes, on a digital platform scale and recorded to the nearest 0.1 kg. BMI percentiles and z-scores were calculated from height, weight and age using the Centers for Disease Control and Prevention (CDC) growth charts.³⁴

2.3.2 |. Food consumption

The amount of food consumed was calculated in grams by weighing each food item before and after the meal and computing the difference. Calories, added sugar and fat content of consumed foods was then calculated using the Nutrition Data System for Research (NDSR),³⁵ a computer-based program, which uses the University of Minnesota Nutrition Coordinating Center (NCC) Food and Nutrient Database of food energy and nutrient content.

2.3.3 |. Psychosocial stressor

The TSST,³³ a standardized social-evaluative stress protocol for adolescents, was administered in the stress condition. The TSST involves relaying an improvised 5-minute story following a 5-minute preparation period and completing a 5-minute mental arithmetic task aloud in front of a panel of evaluators.

2.3.4 |. Participant demographics

Participants reported their self-identified race/ethnicity. Parents reported participant age and information used to compute SES; SES was calculated based on the occupation and educational level of up to two parental figures.³⁶

2.4 |. Statistical analysis

Prior to conducting analyses, the distribution of data was reviewed. Complete cases were used for each analysis. Values greater than three standard deviations beyond the mean for each outcome variable (1 for each of calories, sugars and fats consumed in the relaxation condition; 2 for each of calories and sugars consumed in the stressful condition) were reduced to M + 3 SD. Primary analyses were conducted using analysis of covariance (ancova) to examine racial/ethnic differences in food consumption under controlled conditions in the lab, adjusting for SES, BMI and age. The interaction between racial/ethnic group and lab condition was tested using a mixed ancova, with group (AA, HL and NHW) as a between-subjects variable and condition (control, stress) as a within-subjects variable. Effect sizes were calculated as partial-eta squared $\left({\eta }_{\text{p}}^2\right)$. Significant effects were probed using Bonferroni-corrected post-hoc tests and presented as mean difference and confidence interval; when significant ancovas suggested group differences that failed to reach significance in post-hoc tests, Cohen’s d effect sizes were calculated to determine whether larger samples are likely to detect significant differences between groups. Post-hoc power analyses reveal that the collected sample is adequate to detect moderate effects (Cohen’s f = 0.29–0.30) with 80% power (α = 0.05).

3 |. RESULTS

3.1 |. Participants and descriptive statistics

Participant demographics are displayed in Table 1. Of the 120 participants (M_age = 15.13 years) who completed food consumption procedures, 30% identified as AA, 37% as HL and 33% as NHW. Participants were diverse with respect to BMI and SES. Groups did not differ significantly in terms of age or BMI. Racial-ethnic groups differed in SES, F(2,117) = 3.54, p = 0.03; however, post-hoc analyses of this effect failed to reach significance (HL < AA, −5.98, CI = −12.05, 0.09, p = 0.055; HL < NHW, −5.20, CI = −11.10, 0.71, p = 0.10). Food consumption did not differ based on randomization sequence for calories (control: t(115) = −1.18, p = 0.24; stress: t (113) = −0.12, p = 0.96), added sugars (control: t(115) = −0.43, p = 0.67; stress: t(113) = −0.82, p = 0.41), or solid fats (control: t(115) = −1.10, p = 0.27; stress: t(113) = −0.43, p = 0.67) consumed.

TABLE 1.

Characteristics of participants at study initiation

Characteristics	Overall (N = 120)	AA (N = 36)	HL (N = 44)	NHW (N = 40)	F	PHC
Age in years mean (SD)	15.13 (1.45)	15.31 (1.41)	15.00 (1.41)	15.10 (1.55)	0.44	-
BMI percentiles mean (SD)	72.43 (22.64)	75.42 (23.41)	72.36 (24.79)	69.80 (19.46)	0.58	-
Socioeconomic status mean (SD)	43.99 (11.36)	46.44 (10.70)	40.47 (12.81)	45.66 (9.36)	3.54*	HL < AA†
Weight status no. (%)
Normal weight: BMI <85th percentile	74 (62%)	20 (56%)	23 (52%)	31 (78%)	-	-
Overweight: BMI 85th to 94.9th percentile	26 (22%)	10 (28%)	13 (30%)	3 (7%)	-	-
Obese: BMI >95th percentile	20 (17%)	6 (17%)	8 (18%)	6 (15%)	-	-

Abbreviation: AA, African-American; HL, Hispanic/Latina; NHW, non-Hispanic White, PHC, post-hoc comparison.

^*p < 0.05.

^†p < 0.10.

3.2 |. Food intake

All models adjusted for BMI, SES and age. In the control (relaxation) condition, racial/ethnic groups differed in consumption of calories, F(2, 111) =3.95, p = 0.022, ${\eta }_{\text{p}}^2=0.07$ and solid fats, F(2, 111) =3.95, p = 0.022, ${\eta }_{\text{p}}^2=0.07$, such that AA had a higher energy intake than NHW (+246.44 kcal, CI = 27.80, 465.09, p = 0.02) and higher fat consumption than HL (+7.38 grams, CI 0.54, 14.21, p = 0.03; Table 2). Groups did not significantly differ in consumption of added sugars, F(2,111) = 2.50, p = 0.09, ${\eta }_{\text{p}}^2=0.04$.

TABLE 2.

Food intake across relaxation control and acute stress conditions

Condition	Overall (N = 120)a	AA (N = 36)	HL (N = 44)	NHW (N = 40)	F	PHC
Control
Calories Mean (SD)	1025.90 (398.95)	1177.42 (452.61)	1000.56 (356.90)	917.86 (357.67)	3.95*	AA > NHW*
Added sugars Mean (SD)	27.48 (16.72)	32.95 (19.03)	25.75 (14.97)	24.48 (15.55)	2.50†	-
Solid fat Mean (SD)	22.38 (12.35)	26.90 (14.20)	20.43 (9.47)	20.48 (12.60)	3.95*	AA > HL* AA > NHW†
Stress
Calories Mean (SD)	1036.09 (389.36)	1088.24 (445.18)	1071.44 (350.08)	954.68 (367.86)	1.25	-
Added sugars Mean (SD)	26.80 (15.70)	29.88 (17.60)	27.31 (15.73)	23.52 (13.45)	1.59	-
Solid fat Mean (SD)	23.25 (11.50)	25.34 (12.19)	22.09 (9.41)	22.50 (12.68)	0.84	-

Note: All analyses included age, socioeconomic status and body mass index as covariates. Calories are measured in kilocalories; added sugars and solid fats are measured in grams.

Abbreviations: AA, African American; HL, Hispanic/Latina; NHW, non-Hispanic White; PHC, post-hoc comparison.

^aN in each analysis varies; see the “Participants” section 2.1 for a description.

^*p < 0.05.

^†p < 0.10.

Groups did not differ in food consumption following the TSST with respect to kcal, F(2, 109) = 1.25, p = 0.29, ${\eta }_{\text{p}}^2=0.02$, sugars, F(2, 109) = 1.59, p = 0.21, ${\eta }_{\text{p}}^2=0.03$ or fats, F(2, 109) = 0.84, p = 0.44 ${\eta }_{\text{p}}^2=0.02$, (Table 2). However, there was a racial/ethnic group by condition interaction, such that the effect of in-lab condition (relaxation vs. stress) on calorie consumption varied by group, F(2, 106) = 3.47, p = 0.04, ${\eta }_{\text{p}}^2=0.06$ (Figure 1A). AA consumed fewer calories under the stress condition compared to the control condition (control > stress: +105.82 kcal, CI = 8.05, 203.58, p = 0.03, Cohen’s d = 0.36), whereas NHW (control < stress: −30.44 kcal, CI = −122.24, 61.36, p = 0.51, Cohen’s d = 0.11) and HL (control < stress: −67.48 kcal, CI = −161.48, 26.53, p = 0.16, Cohen’s d = 0.23) did not differ in calories consumed by condition. Group and condition did not significantly interact to predict consumption of added sugars, F(2, 106) = 0.89, p = 0.41, ${\eta }_{\text{p}}^2=0.02$ (Figure 1B), or solid fats, F(2, 106) = 2.50, p = 0.09, ${\eta }_{\text{p}}^2=0.05$ (Figure 1C).

FIGURE 1.

Group × condition interaction, displaying the effect of relaxation control versus stress condition on food consumption for African-American (AA), Hispanic/Latina (HL) and non-Hispanic White (NHW) adolescent females

4 |. DISCUSSION

The present study assessed differences in food consumption among AA, HL and NHW adolescent females, under controlled (relaxation and stress) conditions. Compared to HL and NHW, AA consumed more calories than NHW and more solid fats than HL following relaxation. The effect of condition varied by race/ethnicity, such that AA consumed fewer calories under stressful, as compared to relaxing, condition. No effect of stress on food consumption was observed for HL or NHW adolescents nor was significant racial/ethnic group by condition interaction observed for sugars or fats consumed.

Direct observation of ad libitum eating supports that AA may consume more calories and fats than NHW and HL, respectively, under controlled conditions (small-to-medium effect sizes). This finding partially mirrors data from national surveys conducted in the 1990s, which found that AA consumed more calories and a greater percentage of calories from fat than did HL and NHW in the naturalistic environment.²⁴ These racial/ethnic differences in eating behaviour may create energy imbalances that contribute to the development of obesity among AA. For example, AA consumed on average 196.56 more calories than NHW; if similar racial/ethnic differences in food consumption were repeated each day in free-living conditions, this difference would account for nearly 0.5 lbs. of additional weight per week for an AA adolescent compared to NHW adolescent with equivalent energy expenditure. Identifying risky patterns of energy intake may be particularly important for understanding racial/ethnic disparities in obesity, as, unlike NHW, greater energy expenditure did not protect AA girls from developing obesity in a longitudinal naturalistic observed behaviour study.²⁵

Findings support that stress shapes caloric consumption among AA adolescents; however, effects were in the opposite direction as predicted: AA adolescents consumed fewer calories after an acute social-evaluative stressor than they did after relaxation. Although reducing energy intake is also a common response to stress,³⁷ a prior investigation of AA female adolescents, using self-reported measures of stress and eating, found the opposite effect: that stress increases emotional eating.²⁹ This discrepancy highlights the need for observed, rather than self-reported, measures of food consumption.³⁸

Several factors may account for the appetite-suppressing effect of stress observed among AA adolescents. First, AA youth may be less inclined to binge eat as an emotional coping strategy,³⁹ such that obesity disparities are produced by eating under affectively neutral or positive conditions, rather than by eating under stress. Second, AA may exhibit different patterns of HPA axis reactivity to the stressor, which influence their food consumption. Previous studies have shown that high cortisol reactors eat less following acute stress.^32,34 As AA youth have greater exposure to the early life and chronic stressors, such as discrimination, that shape HPA axis reactivity to acute stress, future studies should examine whether HPA activity accounts for racial-ethnic differences in stress-induced eating. Altered physiological stress reactivity may also affect the time course of stress influencing appetite. AA adolescents may experience appetite suppression in the short term but subsequent appetite increases; therefore, future research should examine racial/ethnic differences in food consumption over the course of the day following stress. Finally, AA may show increased food consumption only in response to those forms of social stress that more closely mirror stressors to which they are more frequently exposed, such as discrimination. Although the TSST involves interpersonal scrutiny from experimenters who may be perceived as occupying positions of authority, it may present a more general form of social stress. Future studies should utilize stress paradigms that involve experimentally manipulated social status and/or social exclusion by in-group versus out-group members,e.g., Reference 40 to examine whether racial-ethnic groups differ in food consumption following those forms of stress that disproportionately impact AA and HL.

While AA and HL females both experience high obesity prevalence, they did not display similar patterns of energy consumption. AA exhibited more obesogenic dietary intake than HL under relaxing conditions, while HL and NHW did not significantly differ from each other. Additionally, caloric consumption varied by stress condition for AA but not for HL. However, HL showed a non-significant mean increase in calories consumed between the control and stress conditions. This small effect was not statistically significant in the present study, which was powered to only detect moderate or larger effects; future studies should employ larger samples to determine whether HL adolescents display modest stress-induced appetite increases. Alternatively, both HL and NHW may only show stress-induced increased energy consumption when other moderators, such as emotional eating, are also present.⁴¹

Although the present study provides preliminary evidence for racial-ethnic differences in eating in response to social-evaluative stress, more work is needed to explicate mechanisms that may under-pin these differences. For instance, racial/ethnic differences in the overall history of chronic stress exposure⁹ may sensitize AA to alter eating behaviours in response to a variety of stressors.^17–20 Alternatively, racial/ethnic differences in particular types of stress exposure, such as discrimination and institutional racism, may sensitize AA to obesogenic eating patterns only selectively, in response to similar stressors. Future research should examine how stress history, including both exposures to chronic stressors of all types and exposure to discrimination-related stress specifically, might account for racial/ethnic differences in stress-induced eating, as differences in stress exposure early in development may shape one’s propensity for stress-induced eating and obesity life-long.⁵ Additionally, future studies should examine this mechanism in conjunction with the myriad genetic, neurobiological, psychosocial and behavioural factors that simultaneously contribute to racial/ethnic obesity disparities.^4,6 Finally, the present study did not assess cultural variables or the degree of acculturation among participants; future studies should also examine the degree to which cultural practices or values, such as adaptability to changing food environments or socialization into racial/ethnic food traditions,^42,43 may account for group differences in eating behaviour under controlled conditions.

This study has several limitations, which include not assessing for mechanisms that might account for racial/ethnic differences in energy consumption, such as chronic stress exposure, physiological reactivity and genetic factors. Additionally, we were unable to measure participants’ eating behaviours after leaving the lab, precluding assessment of stress effects on appetite throughout the day. It is also not clear whether post-stressor eating behaviour in the controlled environment reflects stress-induced eating in the natural environment, where adolescents can eat covertly or across a longer period. Similarly, these data do not measure the degree to which adolescents eat in response to naturally occurring stressors; although the TSST provides an ecologically valid social-evaluative stressor for adolescents, the controlled nature of stress exposure limits this study’s generalizability to other forms of stress in other settings. Finally, our sample size allowed only for the detection of moderate or larger effects. Despite these limitations, the present study offers initial data of observed behaviour under experimental conditions supporting that adolescent females of different racial/ethnic groups diverge in their energy consumption in ways that may differentially contribute to the development of obesity. Given that AA and HL are at elevated risk for both obesity and elevated stress exposure, understanding how stress differently shapes dietary patterns for young women of colour may help explain the equifinality and multifinality in obesity health disparities.

In summary, in a study examining food consumption under controlled relaxing and stressful conditions, we found racial/ethnic differences in adolescent females’ food consumption, such that AA adolescents consumed more calories than NHW and more solid fats than HL following relaxation. Additionally, only AA showed differences in food consumption between relaxing and stressful conditions, such that they had lower energy consumption following an acute social-evaluative stressor. These findings support the presence of racial/ethnic differences in dietary intake during adolescence that may contribute to disparities in obesity and related health outcomes. The findings highlight that racial/ethnic groups may be differentially susceptible to effects of stress on eating behaviour, which may partially account for group differences in obesity prevalence. Future observational and experimental studies of racial/ethnic differences in stress-related food consumption conducted over longer periods and using a wider variety of innovative food consumption and stress measurement methodologies are needed to further explicate this possible risk mechanism.