Emotion Suppression, Coping Strategies, Dietary Patterns, and BMI

Abstract

Objective:

Emotion suppression (ES) is associated with unhealthy coping strategies, such as emotional eating. Physical activity (PA) is a healthy coping strategy that may attenuate the association between emotion suppression and emotional eating (EE). This study evaluated whether: 1) ES is associated with body mass index (BMI) through EE and/or dietary patterns, 2) PA moderates these relationships, and 3) these patterns differ by race/ethnicity and gender.

Methods:

Adult participants (N=1674) of the Family Life, Activity, Sun, Health, and Eating study completed modified versions of the Emotion Regulation, Eating in the Absence of Hunger, and International Physical Activity Questionnaires; a validated dietary assessment; and items on demographics, height, and weight.

Results:

Analyses revealed a serial mediation pathway in the full sample where greater ES was associated with higher BMI through greater EE and lower fruit and vegetable (F&V) intake (B=.0017, CI 95% [.0001, .0042]) after controlling for age, gender, and education. Hedonic snack food (HSF) intake was not a significant mediator of the ES-BMI association. Greater PA attenuated associations of ES and EE with dietary intake and BMI. The serial pathway remained significant for non-Hispanic White women only in subgroup analyses. EE was a significant mediator among women, and PA effects were largely found among Hispanics and men.

Conclusions:

ES was associated with higher BMI through greater EE and lower F&V, but not HSF intake. PA attenuated these associations. Differences in patterns of coping strategies may help to explain disparities in obesity-related health behavior.

1. Introduction

Obesity prevalence is increasing among U.S. adults, and evidence linking obesity to adverse health outcomes is substantial (Hales et al., 2017; Jensen et al., 2014; Lauby-Secretan et al., 2016). Racial/ethnic and gender disparities exist, with Hispanic/Latinx¹ (47%) and non-Hispanic Black (NHB) adults (46.8%) having the highest rates of obesity, followed by non-Hispanic White (NHW) adults (37.9%). Also, women are more likely than men to be obese in NHB and Hispanic populations (Hales et al., 2017).

The rise in obesity is partly driven by nationwide trends in dietary patterns (i.e., eating behavior and food choices) and sedentary behavior (Livingstone & Pourshahidi, 2014; Millen et al., 2016). Despite the importance of energy balance behavior such as diet and exercise, current behavioral interventions have produced inconsistent weight loss results across different contexts and populations (Tate et al., 2016) and mechanistic studies rarely consider multiple health behaviors concurrently with psychological processes to evaluate joint influence on dietary patterns and body mass index (BMI) in diverse samples. Even fewer studies have examined these associations across both race/ethnicity and gender to understand processes that contribute to disparities in obesity and obesity-related conditions.

1.1. Emotion suppression and coping strategies

Emotion suppression (ES) is an emotion regulation strategy in which behavioral expression of an emotion is concealed to reduce subjective emotion experience and potential negative social reactions to emotion expression (Gross & John, 2003). Consequences of ES are: accumulation of unresolved negative emotion, suppression of positive emotion, and continuous cognitive effort to manage emotion responses (Gross & Levenson, 1997; Stepper & Strack, 1993; Strack et al., 1988). ES also contributes to stress-related symptoms, such as subjective negative affect, depression, anxiety, and decreased well-being (Haga et al., 2009; Moore et al., 2008). Accordingly, individuals who suppress emotions often engage in secondary emotion regulation or coping strategies (Folkman & Lazarus, 1984), including both healthy (e.g., physical activity (PA)) and unhealthy (e.g., eating unhealthy foods) behavior. Greater ES has been reported in individuals with obesity compared to those with normal weight (Fernandes et al., 2018). Greater understanding of behavioral coping strategies in response to ES could facilitate a better understanding of the association of ES with obesity.

1.2. Emotional eating and dietary patterns

When ES is not effective in reducing negative emotions, individuals may engage in eating as a means of coping, even if they are not hungry (Haedt-Matt et al., 2014; Macht & Simons, 2000; Macht, 2008). Indeed, ES was associated with emotional eating (EE) among the sample also used in the present study (Ferrer et al., 2017). EE often translates to intake of hedonic foods (high-calorie and energy dense foods with low nutritional value) that produce pleasurable sensations (Heatherton & Baumeister, 1991; Lehman & Rodin, 1989; Macht & Mueller, 2007; Spoor et al., 2007). Findings are mixed as to whether overconsumption associated with EE also includes more or less nutritious, healthy food intake (Ferrer et al., 2017; Larson et al., 2008). EE may be linked to obesity through its influence on hedonic food intake (Laitinen et al., 2002). However, other coping-related health behaviors (i.e., PA) may disrupt the associations of ES and EE with dietary intake and body mass.

1.3. PA and dietary patterns

PA benefits extend beyond reduced morbidity and mortality to include stress reduction, improvements in quality of life, and positive emotion (Dunn et al., 2005; Sarafino, 2006). Regular PA throughout the lifespan is associated with lower incidence and prevalence of obesity (Booth et al., 2000; Myers et al., 2002). Individuals report using PA to cope with stress (American Psychological Association [APA], 2012). In addition to stress and emotion benefits, PA increases positive self-referent thought, reducing reliance on EE (Childs & de Wit, 2014; Emerson & Williams, 2015; Wood et al., 2018). PA may attenuate otherwise positive associations of EE to dietary behavior and BMI (Dohle et al., 2014; Koenders & van Strien, 2011). However, no research has examined whether PA moderates associations between ES and EE to (healthy and unhealthy) dietary behavior and BMI.

1.4. Race/Ethnicity-based differences

Coping strategies may differ among various ethnic groups in the U.S. NHB and Hispanic adults disproportionately experience chronic stressors, such as discrimination and neighborhood factors (e.g., greater access to unhealthy foods, limited access to healthy foods, and safety and availability of PA spaces) (Clark et al., 1999; Grigsby-Toussaint et al., 2010; Mwendwa et al., 2011), which may increase distress and need for emotion regulation strategies. Indeed, racial/ethnic minorities report more frequent use of ES and more frequent attempts to control emotions (Gross & John, 2003; Keltner et al., 2003; Roberts et al., 2008). These differences in ES across race/ethnicity and whether or not they are captured may contribute to inconsistent findings of emotional and binge eating behaviors among NHB and Hispanic women (Kelly, Cotter, et al., 2012; Kelly, Mitchell, et al., 2012; Lydecker & Grilo, 2016; Marques et al., 2011; Striegel-Moore et al., 2000). PA may also be less likely to disrupt the consequences of ES and EE among racial/ethnic minorities, given that these individuals are less likely to meet recommended PA guidelines (Center for Disease Control & Prevention [CDC], 2011; Hawkins et al., 2009). Thus, it is important to identify whether patterns of emotion regulation, behavioral coping and their impact on BMI differ by race/ethnicity.

1.5. Gender-based differences

Associations among ES, EE, PA and BMI may differ by gender. While men may engage in more ES, women are more likely than men to report negative affective outcomes from ES (Flynn et al., 2010), and more likely to engage in EE (Chacko et al., 2015; Tanofsky et al., 1997). Additionally, greater engagement in PA among men (Piercy et al., 2018) may help to explain why ES/EE have not been as strongly associated with BMI among men as a means to cope with negative affect (APA, 2012). Women from racial/ethnic minority groups are most likely to have obesity, report more frequent use of ES, and engage in inadequate PA (Crespo et al., 2001). However, NHW women are more likely to report EE than NHB women (Chacko et al., 2015). These findings underscore the importance of contextualizing gender differences in the context of race/ethnicity instead of examining each variable in isolation.

The current study evaluated whether: 1) ES is associated with BMI through EE and/or dietary patterns, 2) PA moderates these relationships, and 3) these relationships differ by race/ethnicity and/or gender. We hypothesized that greater ES would be associated with higher BMI through greater EE, lower fruit and vegetable (F&V) intake, and greater hedonic snack food (HSF) intake (see Figure 1). We further hypothesized that PA would attenuate these associations and we explored potential differences based on race/ethnicity and/or gender.

Figure 1.

A mediation and moderation model of theorized associations between emotion suppression, emotional eating, dietary patterns, BMI and physical activity

2. Methods 2.1.Participants and procedure

Data were collected as part of the National Cancer Institute’s (NCI) Family Life, Activity, Sun, Health and Eating (FLASHE) study. FLASHE is a cross-sectional internet-based survey of psychosocial, generational, and environmental correlates of diet, PA, and other health behaviors among adolescent-parent dyads, and was approved by the NCI Special Studies IRB. Informed consent was obtained before participants received the surveys. Our analyses focus on the adult sample. Of 1,699 adults who completed both the diet and PA surveys, 1,674 had data on all relevant measures and were included in this exploratory study. Full study methodology and item development are reported elsewhere (Ferrer et al., 2017; Nebeling et al., 2017; Oh et al., 2017) and available at https://cancercontrol.cancer.gov/brp/hbrb/flashe.html.

2.2. Measurements

Adults reported demographic information (gender, educational attainment, race/ethnicity, income, etc.), height, and weight (used to calculate BMI).

2.2.1. ES was assessed with a four-item version of the Emotion Regulation Questionnaire (Gross & John, 2003). These items were: “I keep my emotions to myself,” “When I am feeling POSITIVE emotions, I am careful not to express them,” “ I control my emotions by NOT EXPRESSING THEM,” and “When I am feeling NEGATIVE emotions, I make sure not to express them.” Scores were summed such that higher scores reflected a greater degree of ES, and reliability was acceptable (α=.81) (Ferrer et al., 2017).

2.2.2.EE was assessed using two items from the EE subscale of the Eating in the Absence of Hunger Questionnaire, based on factor loadings in a sample of adolescents (Tanofsky-Kraff et al., 2008). These items were designed to assess eating in response to anxiety and sadness (i.e., negative affect), consistent with the concept of EE (Faith et al., 1997). The items asked “How often do YOU start or continue to eat when YOU’RE not hungry because… (1) You feel sad or depressed? and (2) You feel anxious or nervous?,” rated Never (1) to Always (5)”. These items were summed, and higher scores were indicative of greater EE. Scale reliability was calculated for this sample and was acceptable (α=.83).

2.2.3.Dietary patterns were ascertained through a screener that used items adapted from the National Health and Nutrition Examination Survey, 2009–2010 Dietary Screener (http://epi.grants.cancer.gov/nhanes/dietscreen/) (Smith et al., 2017). The screener asked participants: “DURING THE PAST 7 DAYS, how many times did you eat [food/beverage]?” Daily F&V intake was operationalized as the sum of 5 items reflecting consumption of 100% fruit juice, fruits, green salad, other non-fried vegetables, and cooked beans. Daily HSF intake was the sum of 7 items reporting intake of candy/ chocolate, cookies/cake, potato chips, fried potatoes, frozen desserts, sugary cereal, and processed meat.

2.2.4. PA was assessed using 6 items from the short version of the International Physical Activity Questionnaire (IPAQ) (Craig et al., 2003). Participants provided information on time spent walking and in moderate- and vigorous-intensity activity and were given definitions of these intensities in the item. Variables were calculated from reported daily minutes to yield weekly minutes of moderate, vigorous, and walking, which were combined to derive total weekly PA. Vigorous-level PA was doubled before being summed (Craig et al., 2003), and total PA was converted into hours per week for easier interpretation.

2.3. Data analysis

Descriptive statistics were calculated for all study variables. To correct for extreme values and potential bias in self-reporting, we Winsorized weekly hours of PA (at the 95^th percentile) (Reifman & Keyton, 2010) and BMI scores (1^st and 99^th percentiles). Covariates (e.g., age, gender, highest educational level attained) were selected based on relevant literature and confirmed with bivariate correlational analyses.

Hierarchical linear regression models were estimated to test whether ES was associated with BMI. Covariates were entered in the first step, and ES was entered in the second step. The PROCESS macro (Hayes, 2017) was used to compute moderation and mediation models (see Figure 1). This macro uses bootstrapping methods with 5000 draws to compute 95% confidence intervals to test the null hypothesis. To assess EE (M) as a mediator between ES(X) and BMI (Y), model 4 was estimated. Model 6 was estimated to evaluate whether dietary patterns (F&V and HSF intake) further mediated this relationship in a serial fashion. To examine whether PA moderated the serial pathways between ES, EE, dietary pattern variables, and BMI, model 92 was estimated. Simple interactions using model 1 were estimated to assess whether PA moderates the various relationships between ES, EE, dietary patterns and BMI. The Johnson-Neyman technique was applied to confirm directionality of the interactions. Analyses were conducted on the full sample, then exploratory analyses were conducted separately by race, gender, and subgroups of gender within racial/ethnic groups.

3. Results

3.1. Sample characteristics

Descriptive statistics are presented in Tables 1 and 2. Most participants were 35–59 years of age, and female. This sample was highly educated, with 45.6% of participants reporting completion of a college degree or higher. Approximately equal proportions of participants reported BMIs indicative of normal weight, overweight, and obesity. NHBs reported the highest average BMI, and among NHBs, women had the highest reported average BMI. NHW women reported the lowest average BMI.

Table 1.

Descriptive statistics for participant characteristics and variables of interest

	Body Mass Index (kg/m2)	Total Physical Activity (hrs/wk)	Emotion Suppression	Emotional Eating	Hedonic Snack Food Consumption (daily intake in servings)	Fruit & Vegetable Consumption (daily intake in servings)
	M2(SD)3	M(SD)	M(SD)	M(SD)	M(SD)	M(SD)
Total Sample	28.12(6.78)	15.94(15.71)	10.62(3.70)	4.77(1.98)	4.58(2.19)	3.11(2.17)
Non-Hispanic Whites	27.61(6.60)	15.62(15.24)	10.69(3.67)	4.88(1.95)	4.52(2.10)	3.11(2.06)
Non-Hispanic Blacks	30.67(7.31)	17.51(17.73)	10.36(3.77)	4.40(2.11)	4.77(2.16)	3.17(2.52)
Hispanics	28.07(6.30)	16.63(15.88)	10.78(3.61)	4.77(2.00)	5.00(2.78)	3.15(2.19)
Men	27.91(5.37)	20.04(17.12)	11.46(3.57)	4.31(1.92)	4.94(2.25)	3.11(2.05)
Non-Hispanic Whites	27.84(5.34)	20.39(17.41)	11.60(3.46)	4.33(1.90)	4.97(2.38)	3.22(2.14)
Non-Hispanic Blacks	28.05(5.16)	22.31(17.71)	10.94(3.54)	3.84(1.84)	4.89(1.81)	2.81(2.06)
Hispanics	28.64(5.94)	19.60(17.10)	10.83(4.16)	4.15(1.80)	2.22(1.34)	3.11(1.82)
Women	28.21(7.21)	14.39(14.89)	10.33(3.71)	4.93(1.98)	4.46(2.15)	3.11(2.22)
Non-Hispanic Whites	27.54(7.01)	13.76(13.88)	10.36(3.73)	5.08(1.93)	4.36(1.96)	3.07(2.03)
Non-Hispanic Blacks	31.18(7.57)	16.55(17.62)	10.24(3.82)	4.51(2.14)	4.75(2.23)	3.25(2.61)
Hispanics	27.75(6.56)	14.14(14.67)	10.70(3.25)	5.12(2.02)	4.96(3.11)	3.18(2.39)

²M=mean

³SD= standard deviation

Table 2.

Demographic Characteristics by Race/Ethnicity

	Race/Ethnicity
	Total		Non-Hispanic White		Non-Hispanic Black		Hispanic
	N=1674		N=1229		N=315		N=130
	n	%	n	%	n	%	n	%
Gender
Men	434	26%	330	26.9%	53	16.6%	51	39.5%
Women	1240	74%	899	73.1%	262	83.4%	79	60.5%
Age
18–34	189	11.3%	108	8.8%	62	19.7%	19	14.6%
35–44	729	43.6%	512	41.7%	140	44.6%	77	59.2%
45–59	708	42.3%	570	46.4%	105	33.4%	33	25.4%
60 +	47	2.9%	39	3.2%	7	2.2%	1	.8%
Body Mass Index Categories4
Underweight	22	1.3%	17	1.4%	2	.6%	3	2.3%
Normal Weight	596	36%	489	40.2%	65	20.8%	42	32.6%
Overweight	506	30.5%	357	29.4%	102	32.7%	47	36.4%
Obese	532	32.2%	352	29.0%	143	45.8%	37	28.7%
Physical Activity
>150 min/wk	195	13.2%	136	12.4%	45	17.4%	14	12%
150–300 min/wk	209	14.2%	160	14.6%	32	12.4%	17	14.5%
>300 min/wk	1068	72.5%	801	73%	181	70.2%	86	73.5%
Highest Education Level
≤ High school degree	306	18.4%	232	19%	45	14.4%	29	22.5%
Some college	600	36%	420	34.3%	137	43.8%	43	33.3%
≥ 4-year college degree	760	45.6%	572	16.7%	131	41.9%	57	44.2%

⁴Body Mass Index Categories: underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25–29.9 kg/m²), obese (≥30 kg/m²)

3.2. Association of ES with BMI

Greater ES was significantly associated with higher BMI (R²=.02) (CI 95% [.036, .209]) in the full sample (see Table 3). This association remained significant in subgroup analyses among women (R²=.02) (CI 95% [.034, .247]), NHW (R²=.02) (CI 95% [.018, .221]) and NHB (R²=.07) (CI 95% [.039, .480]) participants. This relationship was also significant among NHW (R²=.03) (CI 95% [.018, .266]) and NHB (R²=.05) (CI 95% [.004, .508]) women, and Hispanic men (R²=.14) (CI 95% [.075, .971]).

Table 3.

Direct and indirect associations between emotion suppression, emotional eating, dietary patterns, and Body Mass Index

	Direct	Indirect			Serial Mediation
	Emotion Suppression	Emotional Eating	Hedonic Snack Food Intake	Fruit and Vegetable Intake	Emotional Eating & Hedonic Snack Food Intake	Emotional Eating & Fruit and Vegetable Intake
	B(p-value)	B(95% CI)	B(95% CI)	B(95% CI)	B(95% CI)	B(95% CI)
Total Sample	.123(.005)	.0810(.0556, .1104)	.0065(−.0037, .0200)	.0095(.0003, .0256)	.0003(−.0027, .0037)	.0017(.0001, .0042)
Non-Hispanic Whites	.119(.02)	.1110(.0752, .1514)	.0016(−.0126, .0206)	.0084(−.0043, .0314)	−.0015(−.0057, .0024)	.0021(−.0003, .0057)
Non-Hispanic Blacks	.259(.02)	.0415(−.0099, 0991)	.0101(−.0275, 0483)	.0090(−.0119, .0465)	.0002(−.0064, .0094)	.0006(−.0038, .0060)
Hispanics	.258(.11)	.0210(−.0703, .1133)	−.0011(−.0182, .0597)	.0374(−.0485, .1606)	.0002(−.0337, .0294)	−.0074(−.0303, .0165)
Men	.053(.45)	.0375(−.0009, .0823)	−.0076(−.0323, .0074)	.0073(−.0099, .0333)	−.0066(−.0177, .0021)	.0000(−.0024, .0032)
Non-Hispanic Whites	.035(.68)	.0458(−.0146, .1096)	−.0289(−.0787, .0045)	.0033(−.1036, .0295)	−.0100(−.0264, −.0001)	−.0008(−.0061, .0037)
Non-Hispanic Blacks	.245(.25)	−.0046(−.1043, .0627)	.0137(−.0829, .1092)	−.0051(−.0822, .0889)	.0021(−.0300, .0403)	.0013(−.0202, .0338)
Hispanics	.523(.02)	.0232(−.1103, .2392)	−.0157(−.1091, .0762)	.1361(−.0533, .4003)	.0065(−.0462, .0364)	−.0055(−.0725, .0550)
Women	.140(.01)	.0889(.0557, .1262)	.0115(−.0003, .0311)	.0079(−.0031, .0270)	.0014(−.0009, .0050)	.0022(.0001, .0055)
Non-Hispanic Whites	.142(.03)	.1233(.0772, .1777)	.0082(−.0030, .0342)	.0080(−.0105, .0413)	.0008(−.0023, .0051)	.0035(.0001, .0091)
Non-Hispanic Blacks	.256(.047)	.0558(−.0050, .1272)	.0156(−.0374, .0647)	.0110(−.0126, .0593)	.0008(−.0064, .0125)	.0011(−.0033, .0075)
Hispanics	.050(.84)	.0374(−.1126, .1833)	.0000(−.0334, .0989)	−.0178(−.0809, .1354)	−.0038(−.0752, .0515)	−.0057(−.0392, .0405)

Outcomes significant at p<.05 are in bold.

3.3. EE and dietary patterns as mediators

EE significantly mediated the relationship between ES and BMI (R²=.06) (see Table 3 for all results). This mediational pathway was significant in subgroup analyses for women (R²=.08) and NHW participants (R²=.08). When considering both race/ethnicity and gender, this mediational pathway was found in NHW women only (R²=.11).

The association between ES and BMI was also mediated by F&V intake in the full sample only (R²=.03). HSF intake was not a significant mediator.

3.4. Serial mediation effects of EE and dietary patterns

There was evidence for serial mediation, such that ES and BMI were linked via ES’s association with EE, the subsequent association of EE with reduced F&V intake, and finally the association of F&V intake with BMI (R²=.08) (see Table 3 and Figure 2). This pathway remained significant in subgroup analyses among women (R²=.09). When considering both race/ethnicity and gender, this pathway remained significant among NHW women (R²=.12). However, among NHW men, an unexpected pathway showed that greater ES was associated with greater EE and HSF intake, which in turn, was associated with lower BMI (R²=.02).

Figure 2. A mediation and moderation model of associations between emotion suppression, emotional eating, fruit and vegetable consumption, BMI and physical activity*.

*Values represent standardized regression coefficients of the direct relationships between variables with p values depicted in parentheses. The values stemming from physical activity represent standardized regression coefficients and p values are depicted in parentheses for interaction effects. The colors of the arrows represent direction of association (green=positive, red= negative). Control variables include participant age, sex, and educational attainment. BMI= body mass index.

3.5. PA as a moderator

PA moderated some direct associations between EE, dietary patterns, and BMI; however, it did not moderate any mediation pathways (see Table 4 and Figure 2). In the full sample, greater amounts of PA attenuated the positive association between EE and BMI (CI 95% [−.0242, −.0037]) and between ES with HSF intake (CI 95% [.0000, .0034]). No other pathways were moderated by PA for the full sample.

Table 4.

Interaction of physical activity on associations between emotional eating, emotion suppression, Body Mass Index and dietary patterns

	Emotion Suppression × Total Physical Activity		Emotional Eating × Total Physical Activity
	B(p-value)	R2	B(p-value)	R2
Total Sample
Body Mass Index	−.00(.73)	.03	−.01(.01)	.08
Hedonic Snack Foods	.00(.04)	.04	.00(.40)	.04
Fruits and Vegetables	−.00(.09)	.05	.00(.22)	.06
Non-Hispanic Whites
Body Mass Index	−.00(.32)	.03	−.01(.01)	.10
Hedonic Snack Foods	.00(.41)	.04	.00(.57)	.05
Fruits and Vegetables	−.00(.14)	.04	.00(.08)	.05
Non- Hispanic Blacks
Body Mass Index	−.00(.95)	.09	−.02(.17)	.13
Hedonic Snack Foods	.00(.24)	.03	−.01(.02)	.06
Fruits and Vegetables	−.00(.96)	.05	−.00(.63)	.09
Hispanics
Body Mass Index	−.00(.85)	.09	.01(.66)	.05
Hedonic Snack Foods	.00(.79)	.06	.03(.00)	.26
Fruits and Vegetables	−.02(.00)	.36	.01(.04)	.28

Outcomes significant at p<.05 are in bold.

PA did not moderate any associations among women. However, PA attenuated the association of ES with HSF intake among men (CI 95% [.0001, .0066]).

PA significantly moderated several relationships in racial/ethnic subsamples (see Table 4). Among Hispanic participants, PA attenuated associations of EE with F&V intake (CI 95% [.0004, .0227]) and HSF intake (CI 95% [.0169, .0448]), and between ES and F&V intake (CI 95% [−.0241, −.0070]). This was also reflected in analyses that considered both race/ethnicity and gender, where associations remained significant for both Hispanic men and women (see supplementary material). The association of EE with HSF intake was attenuated by PA (CI 95% [−.0219, −.0045]) among NHB women.

4. Discussion

Analyses supported several of our hypotheses regarding linkages among ES and EE with dietary behavior and BMI, and the role of PA in attenuating these pathways. As hypothesized, greater ES was associated with higher BMI through greater EE and reduced F&V intake. Interestingly, HSF intake was not a significant mediator in this model. PA attenuated some of these positive associations, including those between EE and BMI, and ES and dietary intake. In subgroup analyses, the serial pathway in which greater ES was associated with higher BMI through greater EE and reduced F&V intake in the full sample remained significant only among NHW women. Moreover, PA generally moderated associations between ES, EE, F&V intake, and BMI among men and Hispanic individuals, but not the mediational pathway.

Results in the full sample are consistent with other studies. The indirect association of ES with HSF and F&V intake via EE in this sample has previously been reported in analyses controlling for BMI (Ferrer et al., 2017). Another study supported a mediated pathway between emotion regulation strategies and BMI via EE (Jones et al., 2019). The present analyses extend these findings by demonstrating a serial pathway between ES and BMI through EE and dietary intake. Moreover, these analyses demonstrate that reduced F&V intake may be more consequential for BMI in the context of ES, compared to increased hedonic eating. However, self-reported dietary screeners, like the one used in this study, may be influenced by social desirability, differences in appropriate portion perception and food choices during mealtimes, or passive overconsumption that are not accurately self-reported and contribute to the development of obesity (Beaulieu et al., 2017; Blundell & King, 1996; Blundell & MacDiarmid, 1997). Thus, more accurate dietary intake measures may yield different results.

Our results are consistent with prior research suggesting that PA may attenuate the association of negative affect with BMI and unhealthy eating behavior (Dohle et al., 2014; Koenders & van Strien, 2011). These findings are important in furthering knowledge of the impact of PA on emotion regulation, EE, food choices, and BMI. The lack of a moderated mediation finding may be due to a reporting bias, as participants tend to overestimate PA levels using the IPAQ (Hagstromer et al., 2010). Alternatively, PA may not impact emotionally influenced eating behavior related to BMI until high levels of PA are achieved, or the associations between PA and the variables outlined in this paper may be complex. For example, although those who experience negative affect are more likely to engage in PA (Liao et al., 2017), PA can cause mood declines among beginners and those who do not exercise regularly or intensely. Future studies using actigraphy data and ecological momentary assessment may also help to clarify temporal relationships, and possibly cyclic patterns, among motives for engaging in PA, intensity and duration of PA, its affective consequences, and other health behavior (Dunton et al., 2020).

These analyses also provide a more comprehensive understanding of the interplay of psychological mechanisms, coping strategies, and BMI across race/ethnicity and gender. The findings among the full sample only remained significant among NHW women. Consistent with previous research that has largely included NHW women, women are more likely than men to report greater emotionality from ES (Flynn et al., 2010), and are more likely to emotionally eat (Chacko et al., 2015; Tanofsky et al., 1997). Additionally, among NHW men, greater ES was related to lower BMI through greater EE and increased HSF intake. This finding may represent potentially greater social and physical environmental supports for healthy lifestyle habits, resilience, and recovery among men (Andrade et al., 2010; Ben-Shlomo & Kuh, 2002). Consistent with previous research, men in this study reported greater amounts of PA. Their PA moderated associations between ES and HSF intake, which may reflect the beneficial impact of PA on the associations among emotions, appetitive urges, and food choices that determine BMI (Dohle et al., 2014; Koenders & van Strien, 2011; Piercy et al., 2018). Although ES was related to BMI for NHWs and NHBs, results suggest that EE may not be a coping strategy that is engaged in or available to all groups. Also, there may be a reporting bias where NHWs are more aware and able to report EE, or that individual factors may be more relevant to health outcomes among NHWs.

Inconsistent findings of ES, EE and PA in other subgroups may reflect unique challenges experienced among different minority groups, and limitations of individual factors as influences on health behavior. Challenges to healthy lifestyle habits exist among racial/ethnic minorities beyond individual-level factors, reflecting the need for ecological frameworks in studies addressing determinants of health behavior (Fleury & Lee, 2006; Mohammed et al., 2016; O’Driscoll et al., 2014). For example, factors such as food availability, instability of work schedules, and acculturation may contribute to obesogenic eating patterns beyond the contribution of emotional influences and PA patterns (Kaplan et al., 2004; O’Driscoll et al., 2014; Hargreaves et al., 2002). Thus, cultural and ecological influences associated with race/ethnicity may be more important to consider when investigating obesity disparities.

The current study was limited by notably smaller sample sizes among NHBs and, especially, among Hispanic individuals. However, given the challenges of recruiting from racial/ethnic minority populations, and the paucity of research that considers multiple racial/ethnic groups, the current findings serve as an exploratory analysis of potential trends in psychological determinants of health behavior underlying racial/ethnic and gender differences. The EE scale used in this study was designed and validated for children and adolescents, which may limit generalizations made to adult populations (Tanofsky-Kraff et al., 2008). Additionally, the self-reported data gathered on PA (Hagstromer et al., 2010), dietary intake (Lechner et al., 1997; Neuhouser et al., 1999), and height and weight (Mozumdar & Liguori, 2016) also present sources of error that yield challenges in being able to reliably extend results, compared to objective measures. However, given the associations that were revealed despite the use of these broad instruments, the use of objective and detailed measurements may help to elaborate on differences between subgroups and how best to address psychological determinants of obesity-related health behavior.

5. Conclusions

ES was associated with higher BMI through greater EE and reduced F&V intake, but not HSF intake. PA attenuated many of these associations. Future studies should consider gender and race/ethnicity concurrently to more precisely identify potential pathways and barriers for weight loss and maintenance interventions through adoption of healthy coping strategies, even in the presence of unhealthy behavior. In addition, ES and EE did not fully explain determinants of dietary behavior in all subsamples, supporting a view that adherence to a healthy lifestyle cannot be achieved by self-regulation alone and is highly influenced by factors outside of the individual (Krieger, 2001; Thorpe Jr et al., 2015; Williams & Jackson, 2005). Future research should also consider other potential contributing factors, such as built environment, epigenetics, education, social and cultural norms.

Highlights:

• Emotion suppression was related to BMI through emotional eating and F&V intake.

• HSF intake was not a mediator between emotion suppression and BMI.

• PA moderated associations among emotion suppression, emotional eating and BMI.

• Coping behavior differences may underlie BMI trends by race/ethnicity and gender.

Abbreviations:

(ES)

Emotion suppression

(EE)

Emotional Eating

(BMI)

Body mass index

(F&V)

fruit and vegetable

(HSF)

hedonic snack food