|Home | About | Journals | Submit | Contact Us | Français|
Both animals and humans show a tendency toward eating more “comfort food” (high fat, sweet food) after acute stress. Such stress eating may be contributing to the obesity epidemic, and it is important to understand the underlying psychobiological mechanisms. Prior investigations have studied what makes individuals eat more after stress; this study investigates what might make individuals eat less. Leptin has been shown to increase following a laboratory stressor, and is known to affect eating behavior. This study examined whether leptin reactivity accounts for individual differences in stress eating. To test this, we exposed forty women to standardized acute psychological laboratory stress (Trier Social Stress Test) while blood was sampled repeatedly for measurements of plasma leptin. We then measured food intake after the stressor in 29 of these women. Increasing leptin during the stressor predicted lower intake of comfort food. These initial findings suggest that acute changes in leptin may be one of the factors modulating down the consumption of comfort food following stress.
Coincident with increased prevalence and severity of psychological stress in the general population [1, 2], the prevalence of overweight and obesity has reached epidemic levels . Understanding the connections between stress and obesity are critically important. Psychological stress is an increasingly well-established factor implicated in the development of obesity . A key mechanism through which stress likely leads to weight gain and obesity is stress-induced eating. Stress-induced eating of palatable foods is conserved across species. Both humans [5–7] and animals  have been documented to increase their food intake following stress or negative emotion, even if the organism is not hungry [9, 10]. Further, the type of food eaten tends to be high in sugar or fat, or both [11–13], commonly referred to as “comfort food.”
Various psychological and other factors predispose people to stress-induced eating. For example, being female, overweight, or scoring high in dietary restraint (a measure of chronic dieting) puts one at risk for increased food intake in response to a psychological laboratory stressor [5, 14, 15]. As for what physiological factors underlie stress-induced eating, prior research has focused on the stress-responsive hypothalamic-pituitary-adrenocortical (HPA) axis and glucocorticoids. In one study, for example, subjects who responded with the greatest amount of cortisol in response to a laboratory stressor, compared to a session with no stressor, consumed the greatest amount of high fat, sugary food . Furthermore, glucocorticoid administration studies in humans and rodents have documented increased food intake as well as a shift in preference toward sweet and fatty foods [16–18]. Although some studies suggest that glucocorticoids directly increase preference of sweet and high fat food [17, 18], stress affects most bodily systems through multiple pathways [12, 19, 20] and accumulating evidence exists suggesting the role of other systems in modulating consumption during stress towards sweet and fatty foods [8, 9, 12].
Although understanding what triggers increases in eating after stress is important, it is equally important to understand what might protect an individual from stress eating. That is, we must also understand what physiological factors might underlie less food consumption following stress.
One possible modulator of stress eating is leptin, the protein product of the ob gene [21–23]. Leptin circulates in the bloodstream, reflects the amount of fat stores, recent energy balance and dietary macronutrient composition [24–27]. Leptin provides a signal to the hypothalamus of body fat reserves and recent energy intake [21–23, 28, 29], and appears to act as a long-term hormonal signal in the regulation of energy homeostasis [30–33]. Leptin contributes to body weight regulation by affecting both feeding behavior and energy expenditure. In rodents, leptin upregulates thermogenesis in brown adipose tissue. Leptin also exerts effects within the hypothalamus, regulating homeostatic food intake [30–32], and in the ventral tegmental area, reducing dopamine neurotransmission and extinguishing the reward value of food . Paradoxically, obese humans tend to have higher levels of leptin, suggesting a state of leptin resistance [21, 22, 31, 32].
Circulating leptin concentrations increase 4–6 hours after meals in response to nutrient-induced insulin secretion [27, 35]. However, leptin may also be acutely responsive to psychological stress [12, 36]. Brydon et al.  exposed subjects to psychological stress and measured plasma leptin concentrations at baseline, immediately after the stressor, and 45 minutes post-stressor. They found a small but significant increase of leptin, which peaked at 45 minutes following exposure to the stressor and was not correlated with cortisol response to the same task.
These stress-related changes in leptin may also modulate stress-induced eating. One study has examined leptin, stress, and eating together  and reported that higher circulating serum leptin concentrations over one day were related to less overall food consumption following stress compared with days on which the subjects were not exposed to stress. This study suggests that tonic serum concentrations of leptin may be linked to both stress and to decreased consumption.
A remaining question is whether acute changes in leptin under conditions of stress might affect subsequent food intake, or “stress eating.” In the current study, we tested whether leptin modulates stress eating by exposing women to a standardized laboratory stressor, discreetly measuring their food choices and amount eaten following the stressor, and measuring leptin throughout the session. Given the prior research implicating leptin in decreased food consumption, we hypothesized that leptin responses to stress would be negatively associated with comfort food consumption. We also measured cortisol to assess whether leptin effects associated with stress eating were independent of, or associated with, the known effects of cortisol on stress-induced eating behavior.
Sixty-three healthy, non-smoking, post-menopausal women aged 50 to 80 years were recruited throughout the San Francisco Bay area through flyers posted within the University of California, San Francisco (UCSF) campuses and hospitals and the community as well as advertisements in local newspapers and radio stations. The sample was homogeneous in terms of being all women, postmenopausal, and of a limited age range, which reduces variance in these factors which can alter leptin levels [21, 23, 28, 29, 38–41] and eating behavior [15, 42–44]. To capture subjects with a wide range of psychosocial stress, the sample was comprised of caregivers of partners with dementia as well as BMI- and age-matched control subjects. Forty of these subjects took part in the laboratory stress session. Further exclusion criteria included the presence of metabolic or endocrine disease such as diabetes, substance abuse, medications known to affect hormones (e.g. glucocorticoids), current major injuries or illness, and current bipolar or eating disorder.
All procedures were approved by the UCSF Committee on Human Research and all subjects provided written informed consent. Subjects were screened for eligibility over the telephone and in person. To confirm eligibility, fasting laboratory tests ensured normal liver, kidney and thyroid function.
Subjects completed two visits at the UCSF Clinical and Translational Science Institute Clinical Research Center (CTSI-CRC). They were asked to refrain from physical activity and consuming alcohol or caffeine after midnight prior to their first appointment. During the first visit a fasting blood sample was collected shortly after arrival to assess fasting leptin levels. All subjects´ blood was collected between 0730h and 0830h to minimize any potential impact of the diurnal pattern of leptin . Body weight, height, and waist and hip circumferences were also assessed. Subjects were then scheduled to return a week later at noon. Questionnaires assessing sociodemographic variables and individual difference measures (see measures below) were handed out to the subjects to be completed at home and returned during the second visit.
At the second visit, subjects were provided with a standardized lunch at noon to equalize the amount of food intake prior to the experimental session. The caloric content of every subject’s lunch was identical and was prepared by the CTSI-CRC metabolic kitchen. An indwelling forearm catheter was inserted and subjects rested for one hour, followed by a brief questionnaire session during which current negative affect and current hunger were assessed. Immediately after this session, a modified Trier Social Stress Test (TSST)  was used to expose subjects to a standardized psychosocial stressor to assess circulating leptin responses to stress. Three blood samples were collected throughout the stress task and recovery period for measurement of leptin: 0 minutes (baseline), 50 minutes after the onset of the stressor, and 90 minutes after the onset of the stressor. Salivary samples to measure cortisol were taken at 0, 15, 20, 30, 50, and 90 minutes post-stressor onset . All subjects were tested individually between the hours of 1415h and 1715h to limit the diurnal variation in HPA axis activity and leptin levels. Immediately following the laboratory stressor, negative affect and hunger were again assessed. Then subjects were moved to a break room with a snack buffet for a 30-minute period during which food intake was covertly measured.
The modified TSST was a 50-minute session beginning with four 5-minute stressful tasks: (1) introduction to two “trained evaluators” and receiving instructions for the task, (2) a preparation period in which subjects were asked to prepare a 5-minute speech on “their personal strengths and weaknesses,” (3) delivering the speech, and (4) a challenging serial subtraction task. The evaluators were trained confederates who kept neutral facial expressions during the task performance and used a set of standardized comments to increase stress and ensure that each subject experienced the task as demanding. The tasks were followed by 30 minutes of sitting quietly. This test has been shown to reliably result in a short-lived increase of psychological stress with physiological manifestations (i.e., cortisol secretion) and has been used in several previous studies examining the role of stress in eating behavior [9, 12].
During the snacking session (approximately 30 minutes after the stress session and 3 hours since their last meal) subjects were left alone in a room with leisure reading material for 30 minutes and also had access to an array of snacks from which they could choose and eat as much as they wanted. Subjects were not required to but merely invited to eat, and were not aware that their food intake was being measured.
To assess food choice with regard to sweet, salty, low-fat, and high-fat foods, four categories of snack choices were presented on a large platter, without individual packaging (see Table 1). Each serving of different food items was separately weighed to the nearest 0.1 g before and after the subject had left the room to assess how much they consumed. If subjects did not eat the food items in the laboratory but took them home, their data was not analyzed. This study design is a common paradigm to assess food choices in humans after laboratory stressors [10, 12, 15].
Body weight was measured on a digital scale with subjects in light clothing without shoes. Height was measured to the nearest of 0.1 cm using a Harpenden stadiometer. BMI was calculated as weight in kilograms divided by height in meters squared.
Plasma leptin samples during the stressor were obtained from blood samples collected via the inserted indwelling catheter. Blood samples were centrifuged, aliquoted and stored in polypropylene vials at −20°C until analysis. Samples were assayed with a radioimmunoassay kit using a 125I-human leptin tracer and human leptin standards from Linco Research, Inc. (St. Charles, MO, USA) in Dr. Havel’s laboratory. The intra- and interassay variation of the assay were 6.6% and 12.0% respectively. Changes of circulating leptin during the stress session were calculated according to the area-under-the-curve with respect to increase (AUCi) formula according to the recommendations of Pruessner et al.  to calculate time-dependent leptin secretion. In other words, we examined leptin secretion over the test period controlling for baseline leptin values.
Salivary cortisol strongly reflects levels of serum cortisol and indexes the amount of free or biologically viable cortisol . To collect salivary samples, subjects were asked to drool passively through a straw into a tube, which then were kept on ice and then frozen at −20°C degrees. Samples were sent for batch assay to Dr. Kirschbaum’s laboratory (Dresden, Germany). Salivary cortisol was assayed with a chemiluminescence immunoassay using a commercial kit (IBL; Hamburg, Germany). The intra-assay CV was 2.9% for high levels and 7.7% for low levels. The inter-assay CV was 5.7% for high levels and 9.1% for low levels. The sensitivity lower limit was 0.006 µg/dL. As with leptin, we calculated cortisol secretion during the stress session using AUCi.
Information on age and ethnicity was obtained by self-report from the subjects. The number of lifetime diets (resulting in at least a five pound loss of weight) and currently being on a diet (yes/no) were assessed in the questionnaire packet the participants completed at home between the first and second visits. Prior to and immediately after the stressor, ratings of negative affect were obtained as part of a manipulation check using six negative emotions taken from the Affect Balance Scale . The items were answered on a 4-point Likert-type scale with 0 = “ not at all” and 4 = “a great deal.” A total score was obtained as sum of the items (Cronbach’s α= .79). Ratings of current subjective feelings of hunger were assessed on a 4-point Likert-type scale (0 = not hungry at all, 4 = a great deal), prior to and after the stressor. To disguise our interest in subjects’ specific food intake, the item assessing current hunger was interspersed among the mood ratings.
Food intake was quantified as the weight of each food item consumed in grams. To obtain a measure for each food category, the sum of the weight of consumed food items within each category was calculated. Both total consumption and category-specific (i.e. high fat, high sugar) consumption were used as the primary dependent measures.
Statistical analyses were performed using SPPS for Windows 15.0. (SPSS Inc., Chicago, IL, USA). Pearson correlation coefficients were calculated to test the association of outcome variables with potential confounding variables. Paired sample t-tests were computed to perform a manipulation check to confirm that the stressor increased cortisol and ratings of negative affect. The Wilcoxon Ranked Sum Test was used to test whether subjects´ subjective ratings of hunger differed significantly pre- and post-stressor. A one-way Repeated Measure Analysis of Variance was used to investigate whether leptin levels at each timepoint during the stressor differed significantly from each other. Partial correlations were calculated to investigate the relationship between cortisol reactivity and leptin changes controlling for potential confounding variables and to determine whether leptin responses significantly predicted the amount of high fat sweet food consumed. All leptin analyses controlled for BMI.
Forty women completed the laboratory stressor session with complete leptin data. One subject had a fasting leptin concentration greater than four standard deviations above the mean and was excluded from statistical analysis, leaving 39 women. Ten women did not undergo the eating session, leaving 29 women for whom complete data were available. The average age was 62 years (SD = 6.33; range: 51 – 79) and the average BMI was 26.2 and ranged from lean to obese (SD = 5.3; range: 17.7 – 37.5). In terms of BMI categories, 3.3% were “underweight,” 40% were “normal” weight, 33.3% were “overweight,” 16.7% were “obese,” and 6.7% were “morbidly obese.” The majority of subjects self-identified as white (85%), with the remaining 13% Asian/Pacific Islander/Native American, and 3% African American.
The women in this sample had a mean fasting leptin concentration of 18.8 ng/ml (SD = 14.6; range 2.5 — 55.0 ng/ml), which is within the normal physiological range of leptin levels . The majority of the subjects (82%) reported to not be on a diet currently. Of the individuals on a diet, 50% were normal weight, 33.3% were overweight, and 16.7% were obese. The inclusion of caregivers and control subjects resulted in a moderate range of perceived stress scores of 0 to 27 out of a maximum of 30. Being a caregiver had no effect on any of the analyses and thus the two groups were combined.
Q-Q plots indicated leptin and cortisol values at all time points were not normally distributed, and this was corrected via a natural log transformation. As expected, fasting leptin concentrations were positively correlated with BMI (r =.88, p < .001). No significant correlation was found between fasting leptin levels and age (r = .16, p = .32) or current dieting (r = -.05, p = .75).
The stress manipulation was successful – comparison of group means revealed that salivary cortisol levels increased from baseline, reached peak levels 30 minutes after the onset of the stressor and returned to baseline levels by 90 minutes after the onset of the stressor. Further, cortisol levels at peak were significantly higher than cortisol levels at baseline before the onset of the stressor, indicating that the stress manipulation elicited the intended stress reactivity response (t = 8.91, p < .001). In addition, subjects reported a significant increase in negative affect from baseline to the post-stress session (t = 7.30, p < .001) (see Table 4).
Characteristics of the distribution of food intake measures are summarized in Table 2. In general, subjects ate more sweet food than salty food (78% vs. 22%), and more high fat sweet food (34%) than high fat salty food (17%); while few ate low fat salty food (4%; see Table 2 for details). Hunger ratings at baseline (before the stressor) did not correlate with any measures of type or quantity of food intake (all p > .05). This lack of correlation, and that each subject ate a standardized meal before the stressor, suggests that the snacking was non-homeostatic – not related to caloric deprivation. Hunger ratings remained stable during stress (no changes from pre to post-stress task, U = 0.00, p = .99).
Characteristics of the distribution of circulating leptin concentrations during the stressor and during recovery are displayed in Table 3.
We first examined the relationship of leptin during stress to potential confounding variables including fasting plasma leptin concentrations, BMI, age, current dieting, and baseline hunger ratings. As can be seen in Table 4, leptin levels at all three time points during the stressor were significantly correlated with fasting leptin concentrations and BMI (p < .05), while no significant correlation was found between age, current dieting and baseline hunger ratings and leptin levels at all three time points (all p > .05).
Across the entire sample, there were mean increases of leptin concentrations from baseline to 50 minutes post-stressor onset (M = 0.30, SD = 1.64) and from 50 minutes to 90 minutes post-stressor onset (M = 0.40, SD = 1.80). However, neither of these increases was statistically significant (p > .05). Subjects showed great variation in leptin reactivity post-stress, ranging from −5.48 to +4.30 ng/ml. Leptin response was not significantly correlated with cortisol reactivity controlling for the potential confounding variables, BMI and age (all p > .05). Further, leptin response was not significantly correlated with BMI, fasting leptin levels, age, baseline hunger ratings or current dieting (all p > .05).
Next we tested whether changes of leptin with exposure to psychological stress were related to less intake of high fat/high sugar “comfort” foods. The average AUCi of natural-logged leptin was 1.05 (SD = 5.55) and ranged from −13.99 to 12.51. Leptin responses across the test session were negatively related to high fat sweet food intake (r = −.40, p < .05). Leptin responses were not related to intake in any of the other food categories (see Table 5).
Because changes of leptin were related to consumption of high fat/high sugar foods, we conducted analyses to examine factors that might potentially contribute to individual differences in the changes of circulating leptin concentrations. Specifically, we examined the changes of leptin in relation to two variables that, based on prior literature, might contribute to differences in leptin responses: the perceived stressfulness of the task  and the number of episodes of dieting the participants engaged in throughout adulthood [22, 29, 50]. Neither leptin responses during nor after stress were significantly associated with these variables (all p > .05).
Stress-induced eating behavior likely plays an important role in the current obesity epidemic . Although the underlying physiological mechanisms are largely unknown, preliminary initial evidence implicates leptin in the phenomenon of stress-induced eating . The present study investigated whether leptin may be one of the physiological factors that modulates stress-induced eating behavior, which is the characteristic shift in food preference towards high fat and high sugar foods following exposure to stress. We found that leptin increases from baseline in response to acute stress were significantly related to lower consumption of high fat, sweet foods – “comfort” foods. This association between leptin in response to stress and subsequent food intake was found to be specific for high fat sweet food intake and no relationship was found between leptin in response to stress and any other food category. Studies on leptin and salty food consumption is scarce, but this is in line with results reported by Kawai and colleagues  who administered leptin into lean mice and found changes in taste nerve responses to sweet but not sour, salty, or bitter substances‥ More work needs to be done to delineate the effects of leptin on non-sweet substances. However, in studies of rodents  and humans [12, 13], stress appears to predominantly affect sweet food intake (rather than chow or salty foods, respectively), and therefore sweet food intake may be a particularly relevant outcome measure in this specific context of stress eating.
Moreover, in the present study morning fasting leptin levels did not predict food intake following the stressor, underscoring the relevance of leptin reactivity as a novel correlate and potentially important predictor of stress eating, especially since this marker was found to be independent of BMI and fasting leptin concentrations.
Our result diverge from the pattern of findings reported by Appelhans , who found that higher leptin concentrations were related to less total food consumption following stress but not sweet high fat food intake in particular as found in the present study. This may have been a consequence of the different leptin measures the two studies used. Rather than measuring diurnal circulating leptin values, we used leptin measures during stress, again emphasizing leptin reactivity as a potentially more relevant marker implicated in stress eating. As leptin is an energy homeostatic hormone that reduces food intake by acting centrally within the hypothalamus and the ventral tegmental area, extinguishing the reward value of food, less high fat sweet food intake following an acute increase in leptin during stressful periods in the present study might be a consequence of a dampened comforting effect of these foods in stressful situations [30, 31, 51].
At the group level, exposure to the psychological laboratory stressor did not significantly increase mean leptin levels during or after the stressor. This is in contrast with Brydon et al.  who, with a larger sample, reported a small but significant increase in leptin levels following a similar laboratory psychosocial stressor. Our inability to replicate these findings may in part be due to a smaller sample size and a higher average age of the participants in the present study and more importantly the wide range of leptin responses that became evident in our study, with some of the participants showing an increase of leptin during stress while others showed decreasing leptin levels during stress. Our study suggests that a pattern whereby leptin does not increase during stress may lead to greater high fat sweet food intake, and this underscores the importance of further studying the factors that determine whether leptin increases or decreases acutely following stress. In this initial study, we appear to have ruled out the perceived stressfulness of the task and the number of episodes of previous dieting as possible influencing variables predicting acute leptin changes during stress.
There was no correlation between leptin during the stressor and cortisol reactivity, which is in line with findings of by Brydon et al . Cortisol (in combination with insulin) has been implicated in prior studies of stress eating  but our latter finding suggests that any leptin effect on food intake may be an independent process that is not driven by cortisol. This could be even more precisely pinpointed in future studies by examining stressors that do not engage the HPA axis, such as noise exposure , and linking leptin changes to eating without any concomitant increases of cortisol.
A number of limitations of our findings warrant mention. As this study was conducted solely in postmenopausal women and in a laboratory setting, these results cannot readily be generalized to other populations or to other less-controlled situations. Moreover, our sample was small, which reduced both our power and generalizability. We cannot rule out the possibility that subjects may have been aware that their food intake was being monitored. We demonstrate correlation, which does not infer directionality or mechanism. This study did not include an unstressed control group, which leaves the possibility that the reported pattern of leptin change is unrelated to stress and may have occurred under restful conditions over the afternoon in some subjects. Further, our inter-assay variation was larger than our observed effect (12%). Future studies should examine changes of circulating leptin in response to acute stress and compare them with leptin responses during a control session on a separate day to rule out that the observed pattern occurred by chance. We did not have measures of sympathetic nervous system (SNS) activity in this study. The SNS is activated by stress and is a regulator of leptin production  (and vice-versa; leptin is a potent stimulator of the SNS ), and future studies are needed to determine whether the SNS might drive or modulate the effects seen here. Ghrelin, a peptide that has been shown recently to mediate stress eating in mice , similarly was not measured, and might also be a key component driving stress eating, either alone or in concert with leptin, in humans.
This is the first study to document the heterogeneity of leptin responses to acute stress. Increases of circulating leptin concentrations during a psychological laboratory stress task predicted less consumption of high fat, high sugar foods but not other types of foods. This suggests that changes of circulating leptin with exposure to stress may be implicated in reducing stress-induced eating behavior. The relationship between changes of circulating leptin and stress-induced eating behavior appear to be independent of changes of cortisol.
We speculate that leptin may be acting as a modulator of stress eating. When an individual has an adaptable, flexible system that is sensitive enough to change its leptin secretion in response to stress, that individual may not fall prey to the drive to consume comfort food. When the system does not respond, meaning that leptin secretion in response to stress is low, comfort food eating may be triggered. In sum, this study implicates circulating leptin responses in potentially dampening the known shift in food preference to high fat, sweet food following exposure to stress, and points towards its potential as an independent modulator of stress eating. Leptin responses to acute stress show a complex pattern, of which the exact nature, cause and underlying mechanisms remain to be determined.
Humans and animals eat high fat, high sugar foods after stress
Higher leptin responses to a lab stressor were related to less consumption of high fat/sugar food
Leptin in the periphery may be a novel modulator of stress-induced eating
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.