Perceived stress and eating behaviors by gender, obesity-status, and stress vulnerability: findings from the Vitamins and Lifestyle (VITAL) study

Abstract

Stress has been associated with eating patterns in human studies with differences due to the type and duration of stressor, type of food, as well as individual susceptibility factors. Laboratory and smaller epidemiological studies have reported stress-associated preferences for foods high in sugar and fat; associations have been found more consistently among women and the obese. Larger studies are needed to sufficiently test these relationships. The aim of this study was to evaluate associations between self-reported amount of stress and dietary nutrient intakes (% calories from fat, carbohydrates, added sugar) and dietary behaviors (number of eating occasions and servings of fruits and vegetables, high-fat snacks, fast food items, and sweetened drinks) by gender, obesity status, and stress vulnerability. Linear regression was used to estimate associations of perceived stress with eating patterns among 65,235 older adults while adjusting for demographic factors, body mass index, physical activity, alcohol intake, number of comorbidities, and other relevant covariates. Higher perceived stress was associated with greater intake of calories from fat, high-fat snacks, and fast food items as well as lower intake of calories from carbohydrates (all p for trend <=0.002). Among those with high perceived stress vulnerability, perceived stress was associated with fewer eating occasions (p for interaction <0.0001). Although associations were small, significant relationships were found for perceived stress arising from everyday experiences among an older mostly white population. These findings have public health implications and suggest that stress may be important to consider in programs promoting healthy eating.

INTRODUCTION

Obesity continues to be a major public health problem that may be attributable, in part, to modern lifestyles characterized by sedentary behaviors and the overconsumption of high fat and high sugar foods^1,2. Understanding what drives these obesogenic behaviors is essential to prevention efforts. Stress may be a bio-behavioral mechanism through which modern lifestyles promote obesogenic eating behaviors and, ultimately, obesity risk. In a U.S. representative sample, multiple dimensions of stress were associated with 10-year weight gain among men and women, especially among those who were already overweight or obese³.

Yet, explaining how stress relates to eating is complex; most people eat more in response to stress while some eat less^4,5. Stress can be conceptualized as comprising multiple bio-behavioral cascades. Stress “begins” when environmental demands overwhelm the resources of an individual and perceptions of stress arise which may result in physiological, cognitive, and behavioral processes designed to maintain allostasis in the short-term⁶. Stress as a construct has been measured, therefore, as counts of potential sources of stress (i.e. environmental demands), perceived stress, or behavioral or physiologic responses to stress. Chronic activation of the stress response can lead to dysregulation that has been associated with increased appetite^7,8, preference for foods high in sugar and fat⁴, visceral fat accumulation and deposition^4,8,9, and obesity^6,8. The type and severity of stressor may be important to associations with eating⁸. Also, repeated exposure to stressors that threaten one’s social self (e.g. stressors associated with social position) are thought to especially contribute to this dysregulation ⁴.

Greeno and Wing suggest that individual differences in learning history, attitudes, and biology are integral to stress-eating models as such differences may impact the susceptibility of an individual to stress-related eating¹⁰. Many laboratory studies have evaluated the relationship of stress and eating with respect to gender, obesity-status, and disinhibited eating (e.g. eating in response to external cues or emotional states)⁴. Obese individuals may be more susceptible to hunger while the prevalence of disinhibited eating is higher both among women and the obese^4,11. Overall, laboratory studies suggest that women, the obese, and those who display disinhibited eating are more likely to engage in stress-related eating⁴. It is conceivable that disinhibited eating may occur among those who are more vulnerable to stress as a means of coping and allostasis^4,12. There is some evidence that preference of foods high in fat and sugar may influence opioid releases in the brain as a coping “reward” after a stressful situation which produces behavioral reinforcement^4,13.

Ultimately, curbing the obesity epidemic depends on successful intervention strategies aimed at long-term maintenance after weight-loss¹⁴. Further understanding how stress relates to eating in everyday life could inform broad-level obesity prevention strategies. While smaller observational studies suggest relationships between stress and obesogenic eating behaviors^15–17, additional evaluation in large population subgroups is needed. To address this gap, associations between perceived stress and dietary factors associated with high energy intake were evaluated in a large well-characterized epidemiologic cohort. We hypothesized that greater perceived stress would be related to obesogenic eating behaviors and that these relationships would vary by gender, obesity-status, and perceived vulnerability to stress.

MATERIALS AND METHODS

Participants

The Vitamins and Lifestyle (VITAL) is a prospective study that was established to investigate the association of vitamin and mineral supplement use and other lifestyle factors with cancer risk. Men and women were eligible to join if they were between the ages of 50 and 76 years and lived within western Washington State. Names of eligible individuals were acquired through purchased mailing lists and baseline questionnaires querying supplement use, health history, cancer risk factors, and diet were mailed to 364,418 individuals. Of those, 77,718 men and women passed questionnaire quality control checks and were enrolled into the study between 2000 and 2002. Further details of study design are reported elsewhere¹⁸. Individuals were excluded from the present analyses if they reported intestinal malabsorption disorders (n=45) which may have a substantial effect on eating patterns (e.g. gastric bypass surgery). Individuals were also excluded if they had missing data on self-reported stress (n=1,620), stress vulnerability (n=303), body mass index (BMI) (n=3,569), covariates (listed below) (n=1,401), or if they failed quality control checks on the FFQ (described below) (n=5,545) resulting in a sample of 65,235. Mean perceived stress was similar between those dropped (3.43) and retained (3.43). This project was reviewed and approved by the Fred Hutchinson Cancer Research Center Institutional Review Board.

Perceived stress (independent variable)

The amount of perceived stress was measured by single item: “In the past year, how would you rate the amount of stress in your life (at home and at work)?” with possible responses ranging from 1 (“no stress”) to 6 (“extreme stress”). To capture more chronic stress as well as correspond to the time period over which dietary data were collected, we asked the question in reference to the past year¹⁹. This item was validated within a subsample of the cohort using validated instruments¹⁹: 1) a modified 53-item questionnaire based on the Hassles and Uplifts Scale²⁰ (r=0.50); 2) the 4-item Perceived Stress Scale (PSS-4)²¹ (r=0.36); and 3) a 10-item questionnaire on major life events based on the Women’s Health Initiative Life Events questionnaire²² (r=0.38). The 3-month test-retest reliability for this perceived stress item was moderate (weighted kappa=0.66)¹⁹.

Nutrient intake and dietary behaviors (outcomes)

Nutrient intake and dietary behaviors were assessed via a semi-quantitative food frequency questionnaire (FFQ) based on the FFQ developed for the Women’s Health Initiative and other studies²³ which asks about the frequency of consumption and portion sizes of 120 foods or food groups over the last year. The FFQ analytic program is based on nutrient values from the Minnesota Nutrition Data System for Research²⁴. Participants were excluded from dietary calculations if they did not complete at least 5 items per page of the FFQ or if their energy intake was lower than 800 kcal/day or higher than 5,000 kcal/day for men or lower than 600 kcal/day or higher than 4000 kcal/day for women. Nutrient outcomes included percent energy from fat, from carbohydrates, and from added sugar. Added sugars included dietary fructose, galactose, glucose, lactose, maltose, and sucrose not from whole foods.

Dietary behaviors that were outcomes in this study included the frequency of: 1) eating occasions; 2) servings of fruits and vegetables; 3) servings of high-fat snacks; 4) servings of fast food items; and 5) servings of sweetened drinks. Total eating occasions was assessed via a single item, “On average, how many times a day did you eat (meals and snacks)?” Participants were instructed to not count drinking beverages alone as snacks except beverages with milk. Total fruit and vegetable intake was calculated by summing all servings except fruit juice and potatoes, with adjustment for serving sizes and self-reported total fruit and vegetable consumption. High-fat snack items included servings of: chips, muffins, croissants, scones, biscuits, and chocolate, candy bars, and other candy. Consumption of fast food items was defined as the number of servings per week of foods typically served in fast food restaurants and included servings of: regular fat hamburger, regular fat hotdogs, fried chicken, fried fish, pizza, and French fries²⁵. Sweetened drink servings included both regular and diet varieties of soda as well as fruit-flavored drinks (not juice)²⁶.

Potential effect modifiers

Gender, BMI, and perceived vulnerability to stress were evaluated as potential effect modifiers of associations between perceived stress and dietary outcomes. BMI was based on self-report and calculated as current weight at baseline in kilograms divided by maximum height in meters squared. Obese individuals were defined as having a BMI >= 30.0 kg/m² which translated to being at or above the top 25%ile for this variable distribution. The ability to handle stress, referred to as ‘perceived vulnerability to stress’ in this paper, was assessed via a single item, “On a scale of 1 to 6, how would you rate your ability to handle stress?” (from 1=“I can shake off stress” to 6=“stress eats away at me”). This item has demonstrated good reliability within a subsample of this cohort (weighted kappa=0.71)¹⁹. High perceived vulnerability to stress was defined as being at or above the top 25%ile for this variable distribution (score: 5–6).

Covariates

Additional covariates were selected a priori based on probable or established associations with perceived stress and included: age, gender, race, education, marital status, laughter²⁷, psychological distress (defined as self-assessed or taking medications for depression/anxiety)^28,29, number of comorbidities (of 9 major health conditions)^32,33, physical activity^15,28, alcohol intake²⁹, and smoking status³⁰. Nutrient intake variables were also adjusted for energy intake.

Statistical analysis

A multiple linear regression model was generated to estimate the mean perceived stress and 95% confidence intervals associated with levels of selected potential covariates. For each dietary outcome, adjusted mean differences (i.e. linear regression coefficients) and 95% confidence intervals associated with each of 4 categories of perceived stress (1–2, 3, 4,5–6) were estimated using linear regression models (with the lowest category as referent group) adjusted for variables listed above and in the table footnotes. Tests for a trend of the dietary outcome by levels of perceived stress used a continuous variable with the original 6 levels of perceived stress. Wald tests were used to generate P-values. Tests for interactions by gender, obesity-status, or perceived vulnerability to stress in models of perceived stress and dietary outcomes were evaluated using an interaction term of perceived stress (as 6-level continuous variable) multiplied by a dichotomous potential effect modifier variable (described above). To compensate for multiple comparisons, a Bonferroni correction for 32 comparisons (8 main effects and 3 interactions for each main effect) was used to assess statistical significance such that only P-values at the 0.002 level were considered significant. All analyses were conducted using Stata SE version 12.1 released 2011 (StataCorp, College Station, TX).

RESULTS AND DISCUSSION

Table 1 presents the characteristics of the study population and the association of those characteristics with perceived stress. All characteristics (each adjusted for other characteristics in the table) were significantly associated with the amount of perceived stress reported by individuals. Specifically, individuals who were younger, female, Native American or Alaska Native, more highly educated, widowed or divorced, perceived themselves as vulnerable to stress, laughed less, were psychologically distressed, had comorbidities, obese, performed low levels of physical activity, drank less alcohol, or were current smokers reported higher levels of perceived stress.

Table 1.

Adjusted mean amount of perceived stress^a by demographic, psychosocial, and health factors among 65,235 adults in Vitamins and Lifestyle (VITAL) Study

Factors	N	(%)	Adjusted Meanb	95% CI	P-value
Age (years)					<0.0001
50–59	31,039	47.6	3.79	3.78, 3.81
60–69	22,482	34.5	3.21	3.20, 3.23
70+	11,714	18.0	2.93	2.91, 2.95
Gender					<0.0001
Male	32,355	49.6	3.35	3.33, 3.36
Female	32,880	50.4	3.52	3.52, 3.54
Race/Ethnicity					<0.0001
White	61,261	93.9	3.44	3.43, 3.45
Hispanic	538	0.8	3.39	3.28, 3.50
Black or African-American	669	1.0	3.43	3.34, 3.53
American Indian or Alaska Native	964	1.5	3.62	3.54, 3.70
Asian or Pacific Islander	1,427	2.2	3.32	3.26, 3.39
Other	376	0.6	3.52	3.39, 3.65
Educational attainment					<0.0001
High School or less	12,044	18.5	3.33	3.31, 3.36
Some college	24,766	38.0	3.43	3.41, 3.44
College graduate	16,681	25.5	3.44	3.42, 3.46
Advanced degree (e.g. Master’s or Doctoral degree)	11,744	18.0	3.57	3.55, 3.59
Marital status					<0.0001
Married or living with partner	51,337	78.7	3.42	3.41, 3.43
Never married	2,106	3.2	3.26	3.21, 3.31
Separated or divorced	7,481	11.5	3.51	3.48, 3.54
Widowed	4,311	6.6	3.60	3.56, 3.64
Perceived stress vulnerabilityc					<0.0001
1 to 2	25,977	39.8	3.14	3.12, 3.15
3	18,870	28.9	3.41	3.39, 3.43
4	13,219	20.3	3.73	3.71, 3.75
5 to 6	7,169	11.0	4.07	4.04, 4.10
Laughter (times/day)					<0.0001
1 to 2	17,726	27.2	3.49	3.47, 3.51
3	22,608	34.7	3.41	3.40, 3.43
4	12,388	19.0	3.42	3.40, 3.44
5	12,513	19.2	3.43	3.41, 3.45
Psychological distressd					<0.0001
No	54,485	83.5	3.34	3.33, 3.35
Yes	10,750	16.5	3.94	3.92, 3.96
Comorbiditiese					<0.0001
No	44,483	68.2	3.42	3.40, 3.43
Yes	20,752	31.8	3.49	3.47, 3.51
BMI (kg/m2)					<0.0001
Normal weight (<25)	22,282	34.2	3.37	3.36, 3.39
Overweight (25 to <30)	28,847	41.1	3.45	3.43, 3.46
Obese (30+)	16,106	24.7	3.52	3.50, 3.54
Physical activity (MET hours/week)					<0.0001
<1.3	16,471	25.3	3.50	3.48, 3.51
1.3–6.0	16,742	25.7	3.46	3.44, 3.48
6.1–15.4	15,714	24.1	3.41	3.40, 3.43
15.5+	16,308	25.0	3.39	3.37, 3.41
Alcohol intake					<0.0001
1	16,309	25.0	3.46	3.44, 3.48
2	16,309	25.0	3.47	3.45, 3.49
3	16,309	25.0	3.44	3.42, 3.46
4	16,308	25.0	3.39	3.37, 3.41
Smoking status					<0.0001
Never	31,008	47.5	3.45	3.44, 3.47
Former	28,926	44.3	3.41	3.39, 3.56
Current	5,301	8.1	3.52	3.49, 3.56

BMI: body mass index; MET: metabolic equivalent of task

^aAmount of perceived stress measured on a 6 point scale ranging from 1= “No stress” to 6= “Extreme stress”

^bMeans for each factor adjusted for all factors in the table in addition to alcohol intake (continuous grams/day), and smoking status (never, former, current).

^cPerceived stress vulnerability measured on a 6 point scale ranging from 1= “I can shake off stress” to 6= “Stress eats away at me”

^dPsychological distress defined as taking drugs for anxiety or depression or reporting feeling stress or anxious during half of the previous year

^eComorbidities defined as the number of reported chronic conditions at baseline (e.g. coronary artery disease, stroke, diabetes, rheumatoid arthritis, etc.). Variable presented as dichotomous in table only.

For main effects models including covariates only, R square values ranged between 0.0228 and 0.1932. When adding perceived stress to these models, R square values ranged between 0.0229 and 0.1939. Associations between the perceived stress and dietary variables are presented in Table 2. Higher levels of perceived stress were associated with higher fat intake as a percent of calories consumed, greater intake of high-fat snacks, more fast food consumption as well as lower carbohydrate intake as a percent of calories consumed and fewer eating occasions (p <0.002 for all). Intakes of added sugars, fruits and vegetables, and sweetened drinks were not significantly associated with amount of perceived stress.

Table 2.

Associations between perceived stress and dietary intake among 65,235 adults in VITAL.

Dietary Intake	Categories of Perceived Stress Ratings							P trendb
	1 (Score:1–2) N=19,023	2 (Score:3) N=15,303		3 (Score:4) N=15,303		4 (Score:5–6) N=15,606
	Ref	βa	95% CI	βa	95% CI	βa	95% CI
Nutrients
Fat (% energy)c	--	0.13	−0.03, 0.29	0.25	0.08, 0.42	0.54	0.36, 0.72	<0.0001
Carbohydrates (% energy)c	--	−0.16	−0.40, 0.30	−0.36	−0.60, −0.20	−0.75	−1.10, −0.50	<0.0001
Added sugars (% energy)c	--	0.01	−0.10, 0.12	0.01	−0.10, 0.12	−0.04	−0.15, 0.08	0.31
Behaviors
Eating Occasions (per week)	--	−0.05	−0.24, 0.13	0.11	−0.08, 0.30	−0.43	−0.63, −0.23	0.001
Fruits and Vegetables (servings/week)	--	−0.01	−0.06, 0.04	−0.01	−0.07, 0.04	−0.04	−0.09, 0.01	0.36
High Fat Snacks (servings/week)	--	0.04	−0.02, 0.10	0.09	0.03, 0.16	0.10	0.03, 0.17	0.002
Fast Food Items (servings/week)	--	0.07	0.03, 0.10	0.13	0.09, 0.17	0.22	0.17, 0.26	<0.0001
Sweetened Drinks (servings/week)	--	−0.04	−0.17, 0.10	0.01	−0.14, 0.15	0.21	0.05, 0.37	0.008

^aBeta for each category of perceived stress (modeled as indicator variables) adjusted for: age (continuous), gender, race (White, Hispanic, Black or African-American, American Indian or Alaska Native, Asian or Pacific Islander, other), education (high school or less, some college, college graduate, advanced degree), marital status (married or living with partner, never married, separated or divorced, widowed), perceived stress vulnerability (ordinal with low=1 and high=6), laughter (ordinal with 1= “less than once a day” to 5= “10 or more times per day”), psychological distress (yes/no), comorbidities (continuous; range:0–9), BMI (continuous), physical activity (continuous metabolic equivalent of task (MET) score), alcohol intake (continuous grams/day), and smoking status (never, former, current).

^bP trend across original 6 levels of stress (ordinal with low=1 and high=6)

^cModel also adjusted for energy (kcal/day)

To determine whether specific subgroups were more susceptible in relationships between perceived stress and dietary intake, we evaluated the interaction effects of perceived stress with gender, BMI, and perceived vulnerability to stress. There was no evidence that gender (p for interaction >0.11 for all dietary variables) or BMI (p for interaction >0.05 for all dietary variables) modified associations between perceived stress and diet. There was only one significant interaction of perceived vulnerability to stress in the relationship between perceived stress and eating occasions (p for interaction <0.0001); perceived stress was associated with fewer eating occasions only among those with higher perceptions of vulnerability to stress (Table 3).

Table 3.

Test of interaction of perceived stress vulnerability in association between perceived stress and dietary intake among 65,235 adults in VITAL.

Dietary Intake	Categories of Perceived Stress Ratings								Interaction c
	1 (Score:1–2) N=19,023	2 (Score: 3) N=15,303		3 (Score: 4) N=15,303		4 (Score: 5–6) N=15,606		P trendb
	Ref	βa	95% CI	βa	95% CI	βa	95% CI
Nutrients
Dietary Fat (% energy)d									0.19
Not stress vulnerable	--	0.16	−0.0, 0.35	0.22	0.02, 0.42	0.61	0.40, 0.83	0.014
Stress vulnerable	--	0.06	−0.31, 0.42	0.20	−0.15, 0.54	0.31	−0.03, 0.66	0.04
Carbohydrates (% energy)d									0.03
Not stress vulnerable	--	−0.24	−0.46, −0.02	−0.39	−0.63, −0.15	−0.85	−1.11, −0.59	<0.0001
Stress vulnerable	--	0.12	−0.31, 0.56	−0.08	−0.49, 0.33	−0.35	−0.76, 0.07	0.003
Added sugars (% energy)d									0.03
Not stress vulnerable	--	0.00	−0.12, 0.12	−0.06	−0.19, 0.07	−0.13	−0.27, 0.01	0.04
Stress vulnerable	--	0.16	−0.08, 0.40	0.32	0.09, 0.55	0.27	0.03, 0.50	0.13
Behaviors
Eating Occasions (per week)									<0.0001
Not stress vulnerable	--	−0.13	−0.33, 0.07	0.10	−0.12, 0.33	−0.22	−0.46, 0.02	0.511
Stress vulnerable	--	−0.09	−0.51, 0.32	−0.06	−0.45, 0.33	−0.75	−1.15, −0.35	<0.0001
Fruit and Vegetables (servings/week)									0.04
Not stress vulnerable	--	−0.01	−0.06, 0.04	0.00	−0.06, 0.06	0.00	−0.06, 0.06	0.66
Stress vulnerable	--	−0.02	−0.13, 0.09	−0.06	−0.16, 0.05	−0.12	−0.22, −0.01	0.05
High Fat Snacks (servings/week)									0.24
Not stress vulnerable	--	0.04	−0.03, 0.12	0.08	0.01, 0.15	0.06	−0.02, 0.14	0.03
Stress vulnerable	--	0.11	−0.02, 0.25	0.19	0.06, 0.32	0.21	0.07, 0.34	0.02
Fast Foods (servings/week)									0.32
Not stress vulnerable	--	0.07	0.03, 0.12	0.15	0.11, 0.20	0.23	0.18, 0.28	<0.0001
Stress vulnerable	--	0.07	−0.01, 0.16	0.11	0.03, 0.19	0.18	0.10, 0.26	<0.0001
Sweetened Drinks (servings/week)									0.11
Not stress vulnerable	--	0.00	−0.15, 0.15	0.02	−0.16, 0.19	0.19	0.00, 0.38	0.06
Stress vulnerable	--	−0.01	−0.33, 0.31	0.11	−0.19, 0.42	0.27	−0.05, 0.60	0.04

^aBeta for each category of perceived stress (modeled as dummy variables) adjusted for: age (continuous), gender, race (White, Hispanic, Black or African-American, American Indian or Alaska Native, Asian or Pacific Islander, other), education (high school or less, some college, college graduate, advanced degree), marital status (married or living with partner, never married, separated or divorced, widowed), perceived stress vulnerability (ordinal with low=1 and high=6), laughter (ordinal with 1= “less than once a day” to 5= “10 or more times per day”), psychological distress (yes/no), comorbidities (continuous; range:0–9), BMI (continuous), physical activity (continuous metabolic equivalent of task (MET) score), alcohol intake (continuous grams/day), and smoking status (never, former, current).

^bP trend across original 6 levels of stress (ordinal with low=1 and high=6)

^cInteraction term modeled as perceived stress (ordinal with low=1 and high=6) x perceived stress vulnerability (ordinal with low=1 and high=6) for test of interaction effect; categorization of stress vulnerability based on using cutpoint at the 75%ile of variable distribution.

^dModel also adjusted for energy (kcal/day)

The aim of this study was to evaluate whether perceived stress was associated with obesogenic dietary intake and whether these relationships varied by gender, obesity-status, and perceived stress vulnerability among a large sample of free-living older adults. These findings suggest that everyday perceptions of stress may influence dietary patterns that are linked to obesity. This is consistent with laboratory and smaller observational studies that have found relationships between measures of stress and intake of dietary fat^4,17, high fat snacks³¹, and fast food^15,25. However, perceived stress was associated with decreased carbohydrate consumption, while one study found a positive relationship between perceived stress and carbohydrate intake among Puerto Rican adults living in Boston who were of similar age to our cohort ³². Perceived stress was not associated with decreased fruit and vegetable consumption in this study as other studies have suggested^15,31. Although an association between perceived stress and % calories from added sugar (i.e. sugar from non-whole foods) was not evidenced, the data did suggest an association between perceived stress and consuming more sweetened beverages which is also consistent with other studies^17,25,31; however, the association in this study did not meet the more stringent criteria for statistical significance.

Perceived stress was also associated with fewer eating occasions including meals and snacks, although only among those who perceived themselves as vulnerable to stress. Despite higher perceived stress being associated with higher frequency of snacking^31,33, it has also been associated with eating fewer main meals in prior studies ³¹. Increases in both snacking and main meals, however, have been associated with higher energy intake³⁴. As the measure of eating frequency combined all types of eating occasions, it is difficult to draw conclusions with respect to obesity. Yet, the data may also suggest that higher perceived stress was still related to obesogenic dietary behaviors (i.e. greater % calories from fat, fewer fruits and vegetables, and higher intake of fast food items, high fat snacks, and sweetened drinks) among these same individuals with higher perceived vulnerability to stress. In other words, although individuals with higher perceived vulnerability to stress ate less frequently with higher perceived stress, they may still have been more likely to consume more calorically-dense foods with perceived stress.

A dependency in the relationship between perceived stress and diet by gender or obesity-status was not found, contrary to other studies^4,15,25,31; there was some evidence, however, that the diet-perceived stress association depended on perceived stress vulnerability. Of the three previously published non-laboratory studies, O’Connor and colleagues did report that relationships between perceived stress and higher intake of high-fat snacks as well as fewer main meals and vegetables were not restricted to subgroups of individuals³¹ and Groesz and colleagues found correlations between perceived stress and fast food items including soda among all study participants²⁵. Perceived stress may therefore be an important correlate of obesogenic dietary behaviors for many people.

This study had several limitations. These analyses are cross-sectional in nature and the directionality of associations cannot be established. Given the hedonic effects on the brain associated with fat and sugar as well as experimental evidence that stress leads to certain adverse eating behaviors⁴, however, it may be more likely that people eat to relieve feelings of stress. Secondly, this sample consisted of older and primarily white and well-educated individuals which may not only limit generalizability to other ethnic or socio-economic groups, but potentially impact the strength of detected associations. Relationships between perceived stress and eating may be greater given that perceived stress tends to be lower among older, non-minority populations³⁵. Thirdly, measurement error and residual confounding could influence these findings. The single-item measures of perceived stress and perceived stress vulnerability used in this study may not accurately reflect these constructs. Perceived stress was measured over a year period to provide a long-term average to best estimate chronic everyday stress. It is possible, however, that underpinning sources of stress were acute and/or chronic in nature. Given the greater correlation of the perceived stress item with the Daily Hassles and Uplifts Scale among a subsample of VITAL participants, however, it may be more likely that the perceived stress measure does capture more chronic stress exposure. Our measure of perceived stress vulnerability may also measure the intensity of the stressor if those experiencing more acute stressors report higher perceived stress vulnerability. However, post-hoc analyses showed that the mean number of major life events did not differ by perceived stress vulnerability. Our findings were also consistent with another study which used a 10-item measure of perceived stress¹⁵. Our dietary measures may also be inaccurate given that intake tends to be underestimated using the FFQ, especially among the obese³⁶. This may be a contributing factor to not detecting differences in perceived stress-related eating by obesity status. Strengths of this study are that analyses were performed in a very large cohort of over 65,000 individuals while accounting for numerous potential confounders.

CONCLUSION

These findings provide information on relationships between everyday stress perceptions and dietary behaviors among free-living adults. This may have significant public health implications given that more chronic levels of perceived stress associated with everyday living may contribute to greater intake of dietary fat. Addressing stress in obesity prevention efforts, therefore, may be important for the adoption of healthier eating behaviors.