Associations of perceived stress and salivary cortisol with the snack and fast-food dietary pattern in women shift workers

Abstract

This study aimed to investigate the association of perceived stress and salivary cortisol levels with the snack and fast-food dietary pattern in a sample of women shift workers. We conducted a cross-sectional study of 539 women aged 18–53 years working in the production line of a poultry processing plant. Stress was assessed with the 10-item Perceived Stress Scale (PSS-10) and by salivary cortisol levels (immediately after waking and upon returning home from work). Dietary patterns were identified by principal component analysis. We used linear and Poisson regression models to assess multivariable-adjusted associations of perceived stress and salivary cortisol levels with the snack and fast-food dietary pattern. After adjustments, women with higher PSS-10 scores had a 28% higher probability (95% confidence interval, 1.04–1.56) of consuming snacks/fast food. Mean (SD) waking cortisol levels were inversely associated with quartiles of the dietary pattern: Q1, 6.63 (0.43) nmol/L; Q2, 6.00 (0.38) nmol/L; Q3, 5.62 (0.40) nmol/L; and Q4, 5.29 (0.35) nmol/L (p = 0.019). Perceived stress was not associated with salivary cortisol levels. The work shift was not associated with perceived stress or cortisol levels. This study demonstrated an association of psychological and physiological measures of stress with a dietary pattern characterized by high intakes of hyper-palatable, energy-dense, ready-to-eat foods among women shift workers.

Keywords:

• Shift work
• salivary cortisol
• perceived stress scale
• dietary pattern
• women
• workers

Introduction

Shift work refers to a work schedule that involves irregular or unusual hours, such as night and rotating shifts, in contrast to a normal daytime work schedule (Morshead, 2002). A recent systematic review revealed that shift workers consume more saturated fats and soft drinks than daytime workers (Souza et al., 2019). The mechanisms linking shift work to poor eating habits are not entirely understood, and some studies have suggested a possible mediating role of stress in this relationship (Nishitani et al., 2009; Nishitani & Sakakibara, 2006). Stress is characterized by coordinated psychological and physiological responses to disruptive or threatening chemical, emotional, physiological, or social stimuli (Stephens et al., 2016). Signals and symptoms of stress can be assessed by tools that measure perceived stress (Lee, 2012) or by cortisol level measurements that indicate the activity of the adrenal cortex as a marker of stress reaction intensity (Hellhammer et al., 2009).

Several studies have demonstrated that shift workers are more susceptible to stress than daytime workers due to the need to reorganize daytime and nighttime activities (Bani-Issa et al., 2020; Hulsegge et al., 2020; Lindholm et al., 2012; Puttonen et al., 2010), especially among women workers (Kakemam et al., 2019). Depending on the intensity of the stress reaction, the adrenal glands might increase cortisol secretion. This glucocorticoid stimulates appetite and increases the intake of highly palatable foods, which suggests a possible mediating role of cortisol levels in the relationship between stress and poor eating habits (Chao et al., 2017).

At the same time, it is well documented that shift workers suffer from circadian misalignment, a type of stress that also involves an altered cortisol secretion (Moreno et al., 2019; Puttonen et al., 2010). Cortisol secretion peaks in the early morning hours and progressively decreases over the course of the day (Kudielka et al., 2004). Previous studies have shown lower waking salivary cortisol levels among rotating shift workers (Kudielka et al., 2004; Wirth et al., 2011). Thus, shift workers are highly burdened by stress, which can affect food intake through interrelated behavioral and physiological mechanisms (Adam & Epel, 2007).

Efforts have been made to investigate the relationship between shift work and eating behavior (Souza et al., 2019). However, few studies have investigated the possible role of stress in this relationship (Hemiö et al., 2020; Nishitani et al., 2009; Nishitani & Sakakibara, 2006), and none has evaluated salivary cortisol levels. Although the literature indicates that higher cortisol levels are linked to increased appetite and intake of highly palatable foods (Chao et al., 2017), it is unclear how lower morning cortisol levels in shift workers can affect these relationships. The present study was therefore designed to investigate the association of stress with a dietary pattern characterized by the consumption of snacks and fast food in a sample of women shift workers. Stress was explored with a perceived stress scale and by salivary cortisol levels. Based on the literature, we hypothesized that women with higher perceived stress would have a higher fast-food consumption.

Methods

Participants and study design

We conducted a cross-sectional study of women aged 18 to 53 years working in the production line of a poultry processing plant in Southern Brazil that operates 24 h a day, 7 days a week. The present study analyzed data from a larger project investigating abdominal obesity and associated risk factors in women workers at a poultry processing plant in Southern Brazil (Garcez et al., 2015). The data were collected in 2011, and participants were interviewed individually in their homes using a pre-tested, standardized questionnaire. We retrieved data on dietary pattern, shift work, and perceived stress for 539 of these workers, and data on salivary cortisol for 480 workers.

The study complied with all ethical requirements for research involving human participants and was approved by the Research Ethics Committee of the University of Vale do Rio dos Sinos.

The snack and fast-food dietary pattern

The dietary pattern analyzed in this study was identified and described in a previous study (Frohlich et al., 2019). Food consumption was determined with a validated qualitative food frequency questionnaire comprising 53 food items. Respondents reported on all food items consumed in the past year, recorded in number of days per week, month, or year. The food frequency data were transformed into an annual consumption rate. Principal component analysis identified 3 dietary patterns, which together explained 24.4% of the total variance. The animal fat and high-calorie food dietary pattern included animal fat, jellies, bread, potatoes and manioc, coffee, eggs, red meat, and sweets. The snack and fast-food dietary pattern included cookies, sausage, crackers, soft drinks, fast food, vegetable fats, and dairy products. The fruit and vegetable dietary pattern included green and yellow vegetables, bananas, citrus fruits, whole grains, and other vegetables and fruits.

To test the hypothesis of the present study, we included only the snack and fast-food dietary pattern in the analysis, as the dependent variable. We explored the score factor as a continuous variable. Subsequently, we ordinally categorized the score factor into quartiles and dichotomized it into highest quartile (fourth quartile [Q4]) vs. lower 3 quartiles (Q1, Q2, and Q3), as suggested in the literature (Newby & Tucker, 2004).

Measurement of perceived stress levels

Stress was assessed with the 10-item Perceived Stress Scale (PSS-10) (Cohen et al., 1983). The PSS-10 consists of 6 negative and 4 positive items about feelings and thoughts related to events and situations that occurred in the past month. Each item is scored on a 5-point Likert scale (never to very often). After reverse scoring the positive items, a total score of 0–40 points is generated. The PSS-10 showed good internal consistency, with a Cronbach’s alpha coefficient of 0.76 and an average inter-item correlation of 0.24.

We explored the PSS-10 score both as a numerical and a categorical (≤ 17 vs. > 17 points) variable. A score > 17 indicates an increased level of perceived stress. This cutoff was based on the study by Aschbacher et al. (2014), in which individuals who scored above the U.S. national mean on the PSS-10 (17 points) were classified as chronically stressed (Cohen & Janicki-Deverts, 2012).

Measurement of salivary cortisol

All women were asked to collect 2 saliva samples during a typical workday: one immediately after waking and the other upon returning home from work. The mean (SD) time between salivary collections was 12.7 (3.1) h. The samples were collected with Salivette® (Sarstedt, Rommelsdorf, Germany). During the interview, all women were instructed on how to use the kit and on how to collect saliva samples at home by rotating the swab in the mouth for at least 2 min and then reinserting it into the tube. They were instructed not to brush their teeth or to eat or drink anything for 30 min before sample collection, to record the exact sampling time, and to keep the samples refrigerated (2–8 °C) until a member of the research team returned to collect them. In the laboratory, the samples were centrifuged at 1,000 × g for 2 min to provide a clear, nonviscous saliva sample, which was aliquoted and stored at −40 °C until assayed. Salivary cortisol levels were measured in ng/mL using a commercial immunoassay kit with chemiluminescence detection (DiaMetra, SRL, Milano, Italy). The highest sensitivity of this assay is 0.33 nmol/L, and measurements showing extreme values were repeated for confirmation. The estimated intra- and inter-assay coefficients of variation were below 14%. Before analysis, the continuous data for salivary cortisol levels were calculated in nmol/L and log-transformed to eliminate the skewness of the distributions. We analyzed 2 parameters of salivary cortisol: cortisol level immediately after waking and cortisol level after work.

Covariates

We investigated the following sociodemographic variables: age (collected as a continuous variable in years and then categorized into 3 age groups of approximately 10 years each: 18–30, 31–40, or > 40 years), self-reported race (white or nonwhite), and marital status (with or without a partner). Socioeconomic status was determined by the number of household items owned and the educational level of the head of the household via the Brazilian Economic Classification Criteria (ABEP); the workers were then categorized into socioeconomic classes. Behavioral variables included smoking (smoker, former smoker, or nonsmoker) and leisure-time physical activity (active or inactive, defined as engaging in regular physical activity for at least 1 year).

Participants reported their bedtime and waking time. Based on these values, we obtained a continuous variable (hours of sleep per day) and dichotomized it for analysis (< 5 and ≥ 5 h/day). Based on the times provided on the day of sample collection, we classified waking time as morning (4:00 AM to 11:59 AM), afternoon (12:00 PM to 7:59 PM), or evening (8:00 PM to 3:59 AM).

Body mass index (BMI) was calculated based on self-reported weight and height and categorized as normal (< 25 kg/m²), overweight (25–29.99 kg/m²), or obese (≥ 30 kg/m²).

Work shifts were categorized as day shift (start time between 6:00 AM and 2:00 PM) or night shift (start time after 6:00 PM, i.e. those who had 90% of their work schedule during the hours of darkness). All participants in the present study worked fixed shifts in the production line. The plant operates 24 h a day, 7 days a week, and employees work 44 h a week with 1 day off, either Saturday or Sunday. We classified the work sector in the production line as more or less fatiguing according to job characteristics.

Statistical analysis

We used descriptive statistics to describe the sample characteristics. We examined the main associations of exposure (perceived stress) and outcome (dietary pattern) with the covariates using one-way analysis of variance (ANOVA) and Pearson’s chi-square test for heterogeneity of proportions (categorical variables) or for linear trend (ordinal variables), as appropriate. We used Poisson regression with robust variance to estimate unadjusted and multivariable-adjusted prevalence ratios (PRs) with 95% confidence intervals (CIs) for the association between perceived stress (numerical and categorical) and the snack and fast-food dietary pattern (Barros & Hirakata, 2003).

We explored the association between perceived stress and both waking and after-work salivary cortisol levels by correlation and linear regression analysis. We compared the mean cortisol levels by one-way ANOVA, using log-transformed data. We performed multiple linear regression analyses to explore associations between the parameters of salivary cortisol levels and the percentile categories of dietary pattern, using log-transformed cortisol data. Variables associated with the dietary pattern at p < 0.2 and waking time were considered confounding factors and included in the multivariable-adjusted model, after which we estimated the adjusted means (SDs).

We analyzed the data in Stata, version 12.0 (StataCorp LP, College Station, TX, USA), and set the level of statistical significance at p < 0.05 for two-tailed tests.

Results

We evaluated 539 women with a mean (SD) age of 33.6 (8.6) years. Most women were 40 years of age or younger (74.6%), white (87.8%), lived with a partner (77.0%), belonged to low socioeconomic class (76.3%), did not smoke (89.0%), were classified as physically inactive (66.1%), had normal BMI (57.9%), slept more than 5 h/day (78.7%), woke up in the morning (54.2%), and worked the night shift (68.3%) in a fatiguing job (71.2%) (Table 1).

Table 1. General characteristics of the sample, and association of demographic, socioeconomic, occupational, and behavioral characteristics with the snack and fast-food dietary pattern and perceived stress in female shift workers, Southern Brazil, 2011. (n = 539).

		Perceived Stress – PSS-10				Snacks and fast-food
Characteristics	n (%)	Mean (SD)	p Value^a	PSS >17n (%)	p Value^b	Mean (SD)	p Value^a	Highest quartile^d n (%)	p Value^b
Age (years)			0.047		0.175		<0.001		0.001
18–30	228 (42.3)	16.27 (5.76)		103 (45.3)		0.18 (1.02)		71 (31.1)
31–40	174 (32.3)	15.09 (5.18)		74 (42.7)		−0.71 (0.91)		42 (24.1)
40–53	137 (25.4)	15.02 (5.76)		53 (38.7)		−0.23 (0.99)		22 (16.1)
Race (skin color)			0.081		0.019		0.279		0.050
White	473 (87.8)	15.41 (5.51)		193 (40.8)		−0.01 (0.09)		112 (23.7)
Black and Brow	66 (12.2)	16.70 (6.12)		28 (56.1)		0.05 (1.08)		23 (34.9)
Marital status			0.213		0.060		0.044		0.487
Without partner	124 (23.0)	16.12 (6.18)		62 (50.0)		0.81 (1.10)		34 (27.4)
With partner	415 (77.0)	15.40 (5.42)		168 (40.5)		−0.03 (0.96)		101 (24.3)
Education (years of schooling)			0.877		0.728		0.002		0.021
<8	183 (33.9)	15.15 (5.75)		77 (42.1)		−0.18 (0.97)		31 (16.9)
8–10	104 (19.3)	15.54 (5.57)		43 (41.3)		0.06 (1.00)		35 (33.7)
>10	252 (46.7)	15.88 (5.53)		110 (43.7)		0.10 (0.99)		69 (27.4)
Economic class (ABEP scale)^c			0.220		0.422		0.260		0.697
B (high-middle)	88 (16.3)	14.99 (5.63)		35 (39.8)		0.19 (0.96)		25 (28.4)
C (middle)	422 (78.3)	15.59 (5.55)		181 (42.9)		−0.64 (0.98)		98 (23.2)
D (low)	29 (5.4)	17.07 (6.09)		14 (48.3)		0.28 (1.19)		12 (41.4)
Smoking status			0.001		<0.001		0.339		0.596
Nonsmoker	480 (89.0)	15.27 (5.46)		190 (39.6)		−0.01 (0.99)		118 (24.6)
Former smoker	42 (7.8)	18.35 (6.18)		29 (69.0)		−0.07 (0.94)		11 (26.2)
Smoker	17 (3.2)	17.11 (6.13)		11 (64.7)		0.25 (1.23)		6 (35.3)
Physical activity			0.141		0.701		0.932		0.310
Active	183 (33.9)	15.49 (6.20)		154 (43.3)		0.05 (0.99)		94 (26.4)
Inactive	356 (66.1)	15.63 (5.28)		76 (41.5)		−0.11 (0.99)		41 (22.4)
Body mass index (BMI kg/m^2)			0.141		0.397		0.010		0.032
<25	312 (57.9)	15.65 (5.45)		138 (44.2)		0.09 (0.97)		94 (30.1)
25–30	157 (29.1)	14.98 (5.48)		60 (38.2)		−0.15 (0.95)		23 (14.7)
≥30	70 (13.0)	16.54 (6.39)		32 (45.7)		−0.08 (1.15)		18 (25.7)
Sleep duration (hours/day)			0.810		0.988		0.658		0.594
>5 h	424 (78.7)	15.54 (5.42)		181 (42.7)		−0.05 (1.00)		104 (24.5)
≤5 h	115 (21.3)	15.68 (6.25)		49 (42.6)		0.15 (0.97)		31 (27.0)
Wake time			0.246		0.068		0.484		0.456
Morning	292 (54.2)	15.20 (5.52)		96 (38.9)		−0.02 (1.0)		67 (22.9)
Afternoon	61 (11.3)	15.82 (5.53)		28 (45.2)		0.02 (1.01)		16 (26.2)
Evening	186 (34.5)	16.06 (5.74)		83 (47.7)		0.02 (0.95)		52 (28.0)
Work shift (n = 521)			0.455		0.346		0.425		0.159
Day	165 (31.7)	15.40 (5.66)		65 (41.1)		−0.05 (0.74)		35 (21.2)
Night	356 (68.3)	15.80 (5.61)		157 (45.6)		0.03 (0.98)		96 (27.0)
Work sector (fatigue) (n = 500)			0.270		0.323		0.817		0.744
Less fatiguing	144 (28.8)	15.97 (6.23)		66 (45.8)		0.05 (1.00)		34 (23.6)
More fatiguing	356 (71.2)	15.36 (5.36)		146 (41.0)		−0.04 (1.01)		89 (25.0)

Wake time: Morning. 4:00AM ≤ hh:hh < 12:00AM; Afternoon. 12:00AM ≤ hh:hh < 8:00PM; Evening. 8:00PM ≤ hh:hh < 4:00AM. Work shift: day work starting between 6:00 AM and 2:00 PM, night shift starting after 6:00 PM.

^a p Value for Student t-test (binary covariates) or ANOVA test (categorical covariates).

^b p Value for Pearson's chi-square test for proportional heterogeneity (categorical covariates) or linear trend test (ordinal covariates).

^c There are no participants in class A.

^d Highest quartile (Q4) vs. lower 3 quartiles (Q1, Q2, and Q3).

Table 1 presents the snack and fast-food dietary pattern and the PSS-10 score as numerical and categorical variables, and their association with demographic, socioeconomic, occupational, and behavioral characteristics. The mean (SD) PSS-10 score was 15.57 (0.24). A higher mean PSS-10 score was observed in women aged 18 to 30 years (mean 16.27, SD 5.76) and former smokers (mean 18.35, SD 6.18). A PSS-10 score > 17 points was more common in nonwhite women (56.1%) and former smokers (69.0%). A higher mean consumption of snacks and fast food was observed in women aged 18 to 30 years (mean 0.18, SD 1.02), without a partner (mean 0.05, SD 1.08), with a higher level of education (mean 0.10, SD 0.99), and with normal BMI (mean 0.01, SD 0.97). The highest dietary-pattern consumption rate (Q4) was more prevalent among women aged 18 to 30 years (31.1%), with 8 to 10 years of education (33.7%), and with normal BMI (30.1%).

There was no statistically significant association between perceived stress and food intake when the PSS-10 score was analyzed as a continuous variable. However, when the PSS-10 score was dichotomized, women with higher scores (> 17 points) had a 30% higher probability (PR = 1.31; 95% CI, 1.07–1.60) of consuming snacks/fast food. After adjustments, the effect measure remained almost unchanged (PR = 1.28; 95% CI, 1.04–1.56), i.e. the effect of the association was independent of the demographic, socioeconomic, occupational, and behavioral characteristics (Table 2).

Table 2. Unadjusted and adjusted prevalence ratios (PRs) for the association of the snack and fast-food dietary pattern and perceived stress among female shift workers in Southern Brazil, 2011. (n = 539).

Perceived stress	Snacks and fast-food			Unadjusted Model		Multivariable Model^a
	Lower 3 quartiles (mean or %)	Highest quartile (mean or %)	p Value*	PR (95% CI)	p Value*	PR (95% CI)	p Value*
PSS-10 score [0–40]			0.162		0.181		0.351
	15.4 ± 5.4	16.2 ± 6.0		1.02 (0.99; 1.05)		1.01 (0.98; 1.05)
PSS-10 score			0.013		0.009		0.017
PSS ≤17	224 (78.9)	65 (21.9)		1.00 (Reference)		1.00 (Reference)
PSS >17	160 (69.6)	70 (30.4)		1.31 (1.07; 1.60)		1.28 (1.04; 1.56)

95% CI: Confidence Interval; PR: Prevalence Ratio;.

Highest quartile (Q4) vs. lower 3 quartiles (Q1, Q2 and Q3).

^dModel adjusted for age, skin color, marital status, education, BMI, wake time and work shift.

*p Values: Wald test for heterogeneity obtained by Poisson regression.

We included 480 women with complete and validated data in the salivary cortisol analyses. Their characteristics did not differ from those of the total sample. Salivary cortisol levels after work were higher in night-shift workers (mean 3.00 nmol/L, SD 3.20) than in day-shift workers (mean 2.42 nmol/L, SD 2.17) (F = 1.78; p = 0.041). However, mean (SD) salivary cortisol levels after work did not show statistically significant differences when stratified by waking time, as follows: morning, 2.72 (2.71) nmol/L for night-shift workers vs. 2.43 (2.23) nmol/L for day-shift workers (F = 0.24; p = 0.180); afternoon, 3.03 (3.03) nmol/L for night-shift workers vs. 3.17 (2.50) nmol/L for day-shift workers (F = 0.23; p = 0.636); and evening, 3.16 (3.56) nmol/L for night-shift workers vs. 1.90 (0.89) nmol/L for day-shift workers (F = 0.49; p = 0.182). Also, there was no statistically significant difference in waking cortisol levels between day-shift workers (mean 6.17 nmol/L, SD 4.87) and night-shift workers (mean 5.64 nmol/L, SD 5.13) (F = 1.12; p = 0.289).

As shown in Figure 1, there was no statistically significant association between perceived stress and waking salivary cortisol levels (PSS-10 ≤ 17: mean 5.63 nmol/L, SD 4.56; PSS-10 > 17: mean 5.97 nmol/L, SD 5.67; F = 0.01, p = 0.905) or after-work salivary cortisol levels (PSS-10 ≤ 17: mean 2.92 nmol/L, SD 3.12; PSS-10 > 17: mean 2.70 nmol/L, SD 3.68; F = 0.07, p = 0.420). The correlation between perceived stress as a numerical variable and salivary cortisol levels was also not statistically significant (waking: r = −0.041, p = 0.368; after work: r = −0.020, p = 0.662).

Figure 1. Comparative bar chart presents the cortisol levels in saliva immediately after waking and upon returning home from work by perceived stress status (PSS-10 ≤ 17 points vs. >17 points). Data are expressed as mean, and error bars represent positive standard deviations. *p Value for Analysis of Variance (ANOVA) test, using log-transformed data (n = 480).

After adjustments, mean (SD) waking salivary cortisol levels were inversely associated with quartiles of the snack and fast-food dietary pattern: Q1, 6.63 (0.43) nmol/L; Q2, 6.00 (0.38) nmol/L; Q3, 5.62 (0.40) nmol/L; and Q4, 5.29 (0.35) nmol/L (p = 0.019). Mean (SD) after-work salivary cortisol levels, however, were not significantly associated with quartiles of the dietary pattern: Q1, 2.86 (0.40) nmol/L; Q2, 2.84 (0.36) nmol/L; Q3, 2.79 (0.37) nmol/L; and Q4, 2.80 (0.38) nmol/L (p = 0.213) (Figure 2).

Figure 2. Comparative bar chart presenting the cortisol levels in saliva immediately after waking and one upon returning home from work by the snack and fast-food dietary pattern categorized in quartiles (Q). Quartiles means: Q1 -1.12 SD 0.34; Q2 -0,39 SD 0.14; Q3 0.16 SD 0.19; Q4 1.34 SD 0.74 Data are expressed as adjusted mean, and error bars represent positive standard deviations. *p Value for linear trend obtained by Linear Regression, using log-transformed data and adjusting for age, skin color, education, BMI, waking time and work shift (n = 480).

Discussion

This study investigated the complex association of psychological and physiological stress with an unhealthy dietary pattern in a population of women shift workers. The results revealed an association between higher levels of perceived stress and the snack and fast-food dietary pattern; however, this association was significant only when PSS-10 scores were dichotomized at 17 points. Lower waking salivary cortisol levels were associated with increased consumption of snacks and fast food, but after-work salivary cortisol levels were not. Perceived stress, as assessed by the PSS-10, was not associated with salivary cortisol levels.

Previous studies have suggested a relationship between higher psychological stress levels and changes in eating behavior toward greater consumption of high-fat, energy-dense foods (Errisuriz et al., 2016; Groesz et al., 2012). Higher cortisol secretion has also been associated with unhealthy dietary patterns (Chao et al., 2017). These findings might be explained by complex interactions between physiological and psychological mechanisms, which influence stress-related eating behaviors. Stress induces the activation of the neuroendocrine hypothalamic-pituitary-adrenal (HPA) axis, thus stimulating the secretion of glucocorticoids (cortisol and corticosterone) from the adrenal glands (Ulrich-Lai et al., 2015). In chronic stress, the chronic activation of the HPA axis alters glucose metabolism and promotes insulin resistance, leading to changes in several appetite-related hormones (e.g. leptin and ghrelin) and feeding-related neuropeptides (e.g. NPY) (Sinha, 2018). Therefore, increased cortisol secretion can lead to chronically stimulated eating behavior, thereby enhancing the propensity to eat high-calorie palatable food (Sominsky & Spencer, 2014).

The results of the present study, however, do not support this hypothesis. After-work cortisol levels were not significantly associated with food intake. Moreover, the lower waking salivary cortisol levels that were found to be associated with greater consumption of snacks and fast food for a long period (1 year) in the present study may be explained by the reward circuitry activation hypothesis. Overconsumption of highly palatable foods is known to reduce reward thresholds and to upregulate the extrahypothalamic corticotropin-releasing factor (CRF) in the amygdala and related limbic-striatal pathways. This may promote food craving and associated greater neural food cue reactivity in these brain regions, thereby increasing the risk of overeating of highly palatable foods. Thus, exposure to a high-fat diet may alter the extrahypothalamic CRF pathways involved in the regulation of stress responses and disrupt brain reward/motivation responses, resulting in increased compulsive food-seeking and stress-induced palatable food-seeking (Sinha, 2018).

Reward circuitry interacts with the HPA axis, meaning that eating “comfort foods” leads to partial suppression of the HPA axis and cortisol secretion. A bidirectional relationship between cortisol levels and behavioral eating would constitute a feedback loop wherein the consumption of these foods would signal a decrease in stress upon reaching satiety (Sominsky & Spencer, 2014). These causal mechanisms are consistent with studies showing that the intake of palatable high-carbohydrate foods is associated with reduced plasma cortisol concentrations in humans (Ulrich-Lai et al., 2015; Wingenfeld et al., 2017). Since we did not evaluate reward circuitry in the present study, further research is warranted to investigate this hypothesis. Also, it is important to note that the average cortisol levels in our study were within normal limits, ranging from 3.6 to 35.1 nmol/L for samples collected at the time of awakening (Hansen et al., 2003).

Our results are consistent with those of previous studies investigating the association between perceived stress and food-related outcomes in shift workers (Heath et al., 2019; Nishitani et al., 2009; Nishitani & Sakakibara, 2006). A recent study exploring the relationship between shift work, mood, and diet reported that higher levels of perceived stress were associated with a higher energy intake and a higher percentage of fat, and that working an afternoon shift was associated with lower levels of stress (Heath et al., 2019). Another recent study with the same objectives found that increased work-related fatigue was associated with higher alcohol intake and that increased night shifts were associated with a higher percentage of energy from fat (Hemiö et al., 2020).

The associations between perceived stress and comfort food intake can be explained as maladaptive coping strategies to alleviate psychological stress. Individuals experiencing stress often perceive themselves as having insufficient resources to handle the demands of their environment. Therefore, the relationship between perceived stress and dietary choices may depend on the perceived ability to effectively manage that stress (Errisuriz et al., 2016). The snack and fast-food dietary pattern includes cookies, sausage, crackers, soft drinks, fast food, and dairy products. These food items are easily accessible and often used as a meal replacement, which suggests that obtaining a quick energy source may be important when experiencing high stress (Errisuriz et al., 2016), especially among women workers, who often bear the responsibility for housework and family care.

It is essential to underscore that the association between PSS-10 scores and the snack and fast-food dietary pattern was significant only when the PSS-10 score was dichotomized. We dichotomized the PSS-10 score based on a cutoff reported in a previous study (Aschbacher et al., 2014), in which individuals who scored above 17 points were classified as chronically stressed. This result might demonstrate a nonlinear relationship between perceived stress and consumption of snacks and fast food in our sample. Only the workers with the highest stress scores (potentially chronically stressed) were more likely to consume snacks and fast food.

In the present study, perceived stress assessed by the PSS-10 was not correlated with salivary cortisol levels. While some studies, investigating different populations, found an association between PSS-10 scores and salivary cortisol levels (Lovell et al., 2011; Murphy et al., 2010; O’Connor et al., 2009), others did not find this association (Putterman & Linden, 2006; Stalder et al., 2011), including a large cohort study of public sector employees that explored this association by means of cross-sectional and longitudinal analyses (Mikkelsen et al., 2017). Regarding shift workers, a study of women health care professionals also found no significant association between cortisol and psychosocial stress measurements (Bani-Issa et al., 2020). These studies claim that the PSS-10 measures the degree to which situations in one’s life are appraised as unpredictable, uncontrollable, and overloading; situations that might lead to a response from the HPA system with increased cortisol secretion. However, the 1-month period for the PSS-10 may be too short to induce longer-lasting changes in the HPA-axis regulation of cortisol secretion (Mikkelsen et al., 2017). This may explain the nonsignificant association found in the present study. Another possible explanation is that, while the PSS-10 data were collected in the prior month, salivary cortisol levels were measured on only 1 day and after the PSS-10 data collection; therefore, the measurements have different time frames. It is also possible that more than one cortisol level measurement might represent the usual parameter. In addition, although the saliva samples for cortisol measurement were collected at the same corresponding time of awakening and upon returning home from work for each group, they were collected on only 1 day. We are aware that some studies collect 2 samples for waking cortisol; however, we were unable to obtain more than 1 sample per person, which may have decreased the accuracy of the measurements. Finally, the hypothesis of the bidirectional cortisol-eating relationship, expressed by lower cortisol levels in women consuming more palatable high-carbohydrate foods, may partially explain this null association and should be evaluated in future research.

Shift workers experience circadian rhythm disruption, and previous studies have shown that waking salivary cortisol area under curve (AUC) values are lower among those working night or afternoon shifts than day shifts (Kudielka et al., 2004; Lindholm et al., 2012; Wirth et al., 2011). In the present study, the work shift was not associated with waking cortisol levels. Still, mean salivary cortisol levels after work were higher in night-shift workers than in day-shift workers in the unadjusted analysis. The highest cortisol levels occur in the early morning hours, progressively decreasing over the course of the day (Kudielka et al., 2004). Therefore, our finding was expected since the night-shift workers collected a saliva sample for cortisol measurement in the morning (end of their work shift). However, it might also suggest that, despite working at night, these workers maintain the circadian rhythm of cortisol secretion.

The work shift was not associated with perceived stress in our analyses. Previous studies have demonstrated an association between shift workers and psychological stress, mainly due to the need to reorganize daytime and nighttime activities (Bani-Issa et al., 2020; Hulsegge et al., 2020; Lindholm et al., 2012; Puttonen et al., 2010). Many of these studies evaluated rotating shift workers who have irregular work schedules (Hulsegge et al., 2020; Lindholm et al., 2012). In the present study, all participants worked fixed shifts. Therefore, working a fixed shift, even though it is the night shift, may help workers organize their daily routine, thus reducing the psychological stress generated by shift work.

Our study has some limitations. First, because information on food intake was collected with a qualitative food frequency questionnaire, there is a risk of information bias. Second, the cross-sectional design of the study limits our ability to infer causality, since we are unable to determine the direction of the relationship between psychological stress and food intake. Third, the usual food intake that characterized the dietary pattern was assessed for the prior year and perceived psychological stress was assessed for the prior month, but cortisol levels were measured on saliva samples collected on only 1 day, which limits our ability to determine the temporality of the associations of cortisol with dietary pattern and perceived stress. Fourth, the lack of assessment of bedtime salivary cortisol levels and cortisol awakening response (CAR) may have limited our findings, given the importance of these measures in the assessment of cortisol circadian rhythm and response to stress exposure (Dahlgren et al., 2009). Finally, we did not evaluate in this study the use of medications that can alter cortisol levels, such as oral contraceptives and steroids.

On the other hand, this study has several strengths. First, as far as we know, this is the first study to evaluate the relationship of perceived and physiological measures of stress with food intake among shift workers. Second, the use of salivary free cortisol levels as a physiological biomarker of psychological stress, which allowed us to standardize and appropriately control the sampling and laboratory analysis of saliva. Salivary cortisol is frequently used as a biological marker of psychological stress (Hellhammer et al., 2009). Also, we used a widely applied validated questionnaire (PSS-10) (Lee, 2012) that has adequate psychometric properties and sufficient reliability to measure perceived psychological stress. Finally, we considered in this study several potential confounders that could have been associated with the exposures and outcomes in the multivariable analysis.

Conclusion

The present study demonstrated an association of psychological and physiological measures of stress with a dietary pattern characterized by high intakes of hyper-palatable, energy-dense, ready-to-eat foods among women shift workers. Higher levels of perceived stress were associated with this dietary pattern. Moreover, lower levels of waking (but not after-work) salivary cortisol were associated with increased consumption of snacks and fast food, highlighting the complex relationship between stress and food intake in shift workers. Further studies examining other types of work and shift schedules are needed to explore the role of behavioral and physiological measures of stress in workers’ eating behavior and food intake in order to better understand the interplay between work shifts, stress, and diet.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was supported by the National Council of Technological and Scientific Development [CNPq, grant numbers 477069/2009-6, 478366/2011-6] and Fundação de Amparo à Pesquisa do Rio Grande do Sul (FAPERGS), grant number FAPERGS/INCT 17/2551-0000519-8. Olinto MT received research productivity grant from the National Council of Technological and Scientific Development [CNPq; grant 307257/2013-4 and 307175/2017-0]. The funders had no role in study design, data collection and analysis, decision to publish, and the preparation or approval of the manuscript.