Introduction

Emerging evidence suggests that poor sleep quantity and quality is associated with increased energy and food intake in both children and adults [1,2,3,4,5,6,7]. However, to date, the individual and environmental mechanisms underlying this relationship are still poorly understood. Factors influencing this relationship may exist at multiple levels [8] including biological (e.g. appetitive hormone disruption), cognitive (e.g. impaired executive functions, increased reward sensitivity), and behavioral processes (e.g. increased impulsivity, impaired decision making). However, one of the mechanisms that has so far received less attention involves emotional pathways, specifically the role of increased negative affect and emotional stress on increased food intake.

Both partial and chronic sleep deprivation have been closely linked to increased emotional reactivity and poorer emotion regulation among healthy individuals [9], and a complex interplay has been suggested to exist between these two dimensions [10]. Sleep deprivation is understood to decrease individuals’ capacity to regulate psychophysiological arousal and reactivity, thus leading to increased emotional lability, and decreases in the capacity for effortful control of emotions [11]. The detrimental impact of sleep deprivation on mood, have also been well documented [9, 12]. Consistent with this, empirical studies have shown that the experimental manipulation of sleep duration can negatively impact both emotional functioning and cognitive performances [13] and several studies have also found that sleep deprivation and sleepiness may impair executive and cognitive functioning such as inhibitory control over food intake, motivation, and mood [14, 15].

Moreover, initial evidence has emerged regarding the role of emotional eating in the relationships between sleep deprivation and increased food intake. Poor sleep quality has been associated with increased emotional eating among women [16]. Experimental inductions of negative affect have led to greater food intake among individuals with high levels of self-reported emotional eating [17]. Similarly, experimentally induced negative affect has been associated with increased food intake among women with high levels of emotional eating reporting sleep deprivation [16]. These findings suggest that emotional regulation may moderate the effects of sleep deprivation on food intake. The findings summarized above may suggest that sleep deprivation increase food intake and trigger overeating potentially through an emotional pathway. One of the limitations of existing studies is that the majority of the experimental investigations have been conducted among healthy individuals, limiting our capacity to understand these processes among individuals with greater difficulties in regulating their food intake including binge eating [18], which has been associate with poor emotion regulation and overeating [18, 19] and chronic sleep deprivation [20]. In addition, recent studies have identified negative affect as one of the main triggers precipitating binge eating [21]. To date, however, the role of poor sleep on food intake among individuals with binge eating symptomatology is poorly understood. The lack of experimental studies investigating the role of acute or partial sleep deprivation on food intake in individuals with binge eating limits our capacity to understand the mechanisms at play in the maintenance of binge eating symptomatology. Thus, the current study aimed to investigate the effect of partial sleep deprivation on food intake in a group of individuals reporting binge-eating symptoms and in a control group of individuals denying any eating disorders symptoms. It was hypothesized that sleep deprivation would increase caloric intake in participants reporting binge eating. In addition, given the previously documented role of emotional eating as a mechanism for increased food intake, we also sought to examine the contribution of emotional eating, expecting that it would increase the effect of sleep deprivation in individuals reporting binge eating.

Materials and methods

Participants

Young adults from the general population were invited through flyers, word of mouth, and electronic advertisements to complete an online self-report questionnaire and provide demographic information (snowball sampling). They were also asked to communicate their e-mail or telephone number if they wanted to participate in a study concerning sleep and eating habits. Individuals reporting symptoms of binge eating or individuals denying any eating pathology based on scores of the self-reported questionnaires described in “Self-report questionnaires”. were eligible for participation. Exclusion criteria were: self-reported food allergies or intolerances, cardio-metabolic illness or particular medical conditions, special diets such as vegetarian or vegan, use of psychiatric medication and underweight status (BMI < 18.5). After the initial online screening, eligible participants for the binge eating group were contacted (N = 40 out of 295). For each participant who agreed to take part in the study, one control participant was contacted and matched for age and gender (35 out of 255). Thirty participants positively answered to this call: 15 reporting symptoms of binge eating (BG), and 15 denying any eating symptomatology (Control Group, CG). All 30 participants signed the informed consent, but 2 out of them (one from each group) were excluded from the analyses due to missing data and drop-out. The final sample thus consisted of 28 participants (age M = 23.75 ± 4.03, 21% male). Each group included 14 participants (11 females and 3 males for each group). All participants received 20 euros (the CG) or an emotion regulation training (the BG) after participation. The study procedure was approved by the Institutional Review Board (Ethical Committee) of the Department of Psychology at Sapienza University of Rome on March 25th 2015.

Procedure

Upon recruitment, participants were invited to an initial lab session when their height and weight were measured. Two appointments were then scheduled: one after a night of habitual sleep (Habitual Night—HN) and one after a night of partial sleep deprivation (Deprivation Night—DN) during which no more than 5 h of sleep were allowed. The order of the sleep manipulation (HN–DN or DN–HN) was counterbalanced across groups and participants in each group were randomly assigned to each sequence. When the DN occurred first, a night of restoration was required before the HN [22]. Upon arrival in the laboratory, on both mornings after HN or DN, participants completed several computer tasks, and then were provided with a buffet breakfast. Food intake was unobtrusively measured by the researchers as described below. Participants were informed of this upon study completion, and provided a second informed consent after the final debriefing for allowing us using these data.

Measures

Self-report questionnaires

All participants completed several self-report questionnaires. The Binge Eating Scale (BES; [23]), in the Italian version by di Bernardo et al. [24] was used to screen for eligibility. The BES is 16-item questionnaire assessing the presence of binge eating behaviours including feeling guilty after eating too much, thoughts about eating, overeating, etc. In this study, Cronbach’s α was 0.93. Criteria for the BG were scores above the cut-off (BES > 17) [25].

Control participants without eating symptomatology using the BES (BES ≤ 17) and the Disordered Eating Questionnaire (DEQ; [26]). The DEQ includes 24 items assessing dysfunctional eating-related behaviour patterns, based on DSM IV-TR criteria [27]. Participants were eligible for the CG if they reported scores below or equal to the clinical cut-off score (DEQ ≤ 30). Cronbach’s α here was 0.94.

Emotional eating after depression (EES-D) was assessed during screening using the Emotional Eating Scale (EES [28]; Italian version by Lombardo and San Martini [29]). This subscale includes 9 items assessing desire to eat after negative emotions (e.g. when they felt depressed, bored, etc.). Cronbach’s α was 0.88.

Participants also completed the Beck Depression Inventory-II (BDI-II; [30];) in the Italian version by Ghisi et al. [31], α = 0.95, and the Insomnia Severity Index (ISI; [32];) in the Italian version by Battagliese and Lombardo [33], α = 0.77, to assess depressive and insomnia symptoms. Menstrual phase was also recorded among female participants.

Height was measured to the nearest 0.1 cm using a standard stadiometer, and body weight was measured to the nearest 0.1 kg using a digital panel indicator scale.

Sleep assessment

Participants were asked to complete sleep diaries upon awakening [34]. They reported at what time they went to sleep, numbers and duration of nocturnal awakenings, at what time they woke up, and subjective sleep onset latency. An electronic portable device called Zeo (Inc., Newton, MA; [35]) was also used to objectively control participants’ compliance to the partial sleep deprivation instruction. Specifically, during the DN, participants were instructed to wear the headband from 10 pm. They were also asked to avoid napping during the day.

Food intake assessment

Participants completed food and beverage diaries. During both laboratory sessions, participants were provided with a buffet breakfast. Food intake was assessed by counting or weighing the food before and after consumption.

Data preparation

Consistent with previous work [36], using the data extracted from sleep diaries, we calculated the following sleep indices: total time in bed (TIB); total sleep time (TST) as TIB minus the minutes required to fall asleep and minus the minutes spent awake after sleep onset; Sleep Efficiency Index (SEI) dividing TST by TIB and multiplying by 100.

Using the website http://sapermangiare.mobi/ [37], sponsored by the Research Centre for Eating and Nutrition (CRA-NUT) and The Ministry of Agricultural, Food and Forestry Policies, foods and beverages consumed in the lab and reported in the food diaries were converted into total caloric (Kcal) and macronutrient (carbohydrate, fat, protein, fibre and alcohol) content.

Data analyses

Data analyses were conducted using the Statistical Package for Social Sciences [38] version 20.0. Since the assumptions of normal distribution, sphericity and homogeneity were met, parametric tests were used. Group differences on demographic and self-reported measures were tested using Independent sample t tests. Several 2 × 2 mixed design factorial ANOVA, Night (HN vs DN) × Group (Binge Eating Group vs Control) were performed on the Total Sleep Time, the Sleep Efficiency Index and the food intake in the lab and during the entire day. Based on the literature, 2 × 2 mixed design factorial ANCOVAs were performed on food intake, controlling for emotional eating. When significant interactions between the covariate and the factors were found, following Pedhazur and Schmelkin’s [39] suggestions, two groups were formed based on a median split: High EES-D and Low EES-D. Then, a series of 2 × 2 mixed design factorial ANOVAs were performed considering Group (High EES-D vs Low EES-D) and Night (HN vs DN) as factors and the total energy and the macronutrients contents eaten at breakfast and throughout the day as outcome variables.

Results

Group characteristics

Table 1 displays the means and standard deviations of the two groups for age, BMI and the relevant self-report measures. A series of t-tests revealed that the groups differed on all the self-report measures as well as on BMI: the BG reported higher levels of binge eating symptoms, disordered eating, emotional eating, depression, and insomnia (all ps < 0.01) and presented a higher BMI (p = 0.014) compared to the CG (Table 1). No significant group differences in menstrual phase were found (χ 2 = 2.533, p = 0.469).

Table 1 Groups’ characteristics and Independent t tests
Full size table

Sleep manipulation

The 2 × 2 mixed design factorial ANOVA Group (BG vs CG) × Night (HN vs DN) on the total sleep time (TST) revealed a significant main effect of the Night, F(1,26) = 141.54, p < 0.001, η 2 = 0.85. During the DN both groups slept less (M = 274.21 ± 31.94 min) compared to the HN (M = 427.43 ± 56.751 min). No main Group effect or interaction was found for the TST. No Group, Night, or interaction effects were found on the Sleep Efficiency Index (SEI).

Food Intake at breakfast

The 2 × 2 mixed design factorial Group (BG vs CG) x Night (HN vs DN) ANOVA on the Total Energy consumed at breakfast in the lab showed no main effect either of the night, F(1,26) = 2.03, p = .166, η 2 = 0.07 or of the group, F(1,26) = 1.93, p = 0.177, η 2 = 0.04 or any significant Group x Night interaction F(1,26) = 0.96, p = 0.34, η 2 = 0.07. When examining fibre consumption, a main effect of the Night, F(1, 26) = 7.71, p = 0.01, η 2 = 0.23 emerged. Thus, after the DN both groups consumed less fibre than after the Habitual Night. The main effect of the Group and the interaction were both non significant. No significant effects were found for carbohydrate, fat, or protein intake during breakfast in the lab (see Table 2).

Table 2 Means and standard deviations of food intake at breakfast, daily food intake and daily snacks intake in binge eating group and control group
Full size table

Based on the existing literature on the putative role of emotional eating [17], we performed a 2 × 2 mixed design factorial ANCOVA, using the EES-D scale as a covariate. Results revealed a significant Night x EES-D interaction, F(1,25) = 6.52, p = 0.017, and a significant main effect of the Night, F(1,25) = 8.69, p = .007, η 2 = 0.21. Given the significant interaction, we created a Low EES-D (EES-D ≤ 15) and a High EES-D (EES-D > 15) group using a median split [39]. The two groups almost perfectly overlapped with the original groups, except for two individuals: one classified into the BG reported low EES-D and one classified into the CG reported high EES-D.

A new 2 × 2 mixed design factorial ANOVA examining Group (Low EES-D vs High EES-D) × night (HN vs DN) effects on Total Energy consumed during breakfast in the lab was performed (see Table 3). Results revealed a significant group × night interaction, F(1,26) = 4.42, p = 0.045, η 2 = 0.15, and a marginal main Group effect, F(1,26) = 3.257, p = 0.083, η 2 = 0.11. Participants reporting high EES-D ate less at breakfast after both nights as compared to the low EES-D group. Analyses of simple effects revealed that individuals reporting low EES-D scores ate less at breakfast when sleep-deprived than after the habitual night of sleep (t = 2.66, p = 0.02) (see Fig. 1).

Table 3 Means and standard deviations of food intake at breakfast, daily food intake and daily snacks intake in High EES-D and Low EES-D
Full size table
Fig. 1
figure 1

Total energy (kcal) consumed at breakfast by Low EES-D and High EES-D

Full size image

Exploratory 2 × 2 mixed design factorial ANCOVAS were also performed on carbohydrates, fat, proteins and fibre intake during the breakfast in the lab. Results for carbohydrate intake revealed a significant night × EES-D interaction, F(1,25) = 7.16, p = 0.013, η 2 = 0.22, and a significant main effect of the night, F(1,25) = 9.53, p = 0.005, η 2 = 0.28. Given this, here too we performed a mixed design factorial 2 × 2 ANOVA considering the Groups defined on the bases of EES-D scores (Low EES-D vs High EES-D) x Night (HN vs DN) on consumed carbohydrates. Results revealed a significant night × group interaction, F(1,26) = 5.96, p = 0.022, η 2 = 0.19, and a main Group effect, F(1,26) = 4.31, p = 0.048, η 2 = 0.14. Simple effects revealed that after the Deprivation Night, the Low EES-D group consumed significantly fewer carbohydrates compared to the Habitual Night (t = 2.77, p = 0.016). Moreover, after the HN, the Low EES-D group consumed more carbohydrates compared to High EES-D group (t = 2.70, p = 0.015). The main group effect indicated that the Low EES-D group consumed overall more carbohydrates compared to the High EES-D group (see Fig. 2).

Fig. 2
figure 2

Carbohydrate content consumed at breakfast by Low EES-D and High EES-D

Full size image

A 2 × 2 mixed design ANCOVA on fats intake revealed a main effect of the Night, F(1, 26) = 5.09, p = 0.033, η 2 = 0.17, and a significant interaction night × EES-D, F(1,26) = 4.28, p = 0.049, η 2 = 0.15. Following the same procedure, a mixed design ANOVA showed no significant results. Only a marginal interaction night × group, F(1,26) = 2.74, p = 0.11, η 2 = 0.10, was found. Simple effects showed that the low EES-D group consumed less fats after the Deprivation Night compared to the habitual night (t = 1.83, p = 0.09). A 2 × 2 mixed design ANCOVA on fibre intake revealed a main effect of the night, F(1,24) = 8.15, p = .009, η 2 = 0.25, and no interaction between Night and the covariate. Both Groups consumed less fibre after the DN compared to after the HN. No significant results were found on proteins intake during the Test Meal in the lab. Since during the DN participants had more time before going to bed (at 1 AM) and therefore, more time to eat, an additional analysis was performed on the number of snacks and caloric content consumed during the evening preceding the HN and the DN. Some of the participants ate a small snack before going to bed (no one during the night) however, no significant night or group effects emerged on the calories consumed.

Daily food intake

The 2 × 2 mixed design factorial ANOVA Group (BG vs CG) × night (HN vs DN) performed on the total daily energy revealed no significant results either for the night, F(1,26) = 0.013, p = 0.91, for the group, F(1,26) = 0.31, p = 0.58, or for the group × night interaction, F(1,26) = 2.18, p = 0.151. Although this interaction was not significant, the means of the two groups differed with the CG eating more after DN compared to the HN (Fig. 3). A marginally significant main effect of the Group, F(1,26) = 4.05, p = 0.055, η 2 = 0.14, was found for daily fibre intake. The BG consumed overall less fibre than the CG. No significant results were found on carbohydrates, fats and proteins content.

Fig. 3
figure 3

Total daily intake (Kcal) of binge eating group and control group

Full size image

The 2 × 2 mixed design factorial ANCOVAs did not reveal any significant main effects or interactions on total energy and macronutrients consumed during the day. Nevertheless, given the potentially low statistically power, we performed Group (High EES-D vs Low EES-D) × night (HN vs DN) ANOVAs. The analysis on Total Daily Energy intake highlighted a significant night × group interaction, F(1,26) = 4.28, p = 0.049, η 2 = 0.14, however, the main effect of the Night and the Group were not significant. Simple effects revealed that the Low EES-D group ate significantly more after the DN compared to the HN (t = − 2.44, p = 0.03, see Fig. 4).

Fig. 4
figure 4

Total daily intake (Kcal) consumed by Low EES-D and High EES-D

Full size image

Results from the exploratory analysis performed on daily carbohydrate intake revealed a marginally significant Night x Group interaction, F(1,26) = 4.21, p = 0.051, η 2 = 0.14. Simple effects revealed no significant differences, although the Low EES-D group consumed more carbohydrates after DN than after HN (t = − 1.66, p = 0.12). The same marginally significant interaction Night x Group was found for protein content F(1,26) = 4.19, p = 0.051, η 2 = 0.14. The analyses of simple effect revealed that after the HN, High EES-D ate more proteins than Low EES-D (t = − 2.09, p = 0.047). A marginal difference (t = − 2.01, p = 0.066) was also found between the proteins content consumed by Low EES-D after the DN, that was higher compared to the HN. Analyses performed on fat and fibre intake did not reveal any differences, however, the group means also consistently displayed a similar pattern with the low EES-D group consuming more after DN than after HN.

Daily snacks

A 2 × 2 mixed design factorial ANOVA Group (BG vs CG)  × night (HN vs DN) on the number of Snacks consumed in the subsequent day revealed a significant effect of the Night, F(1,26) = 7.12, p = 0.013, η 2 = 0.22. After DN, both groups consumed more Snacks compared to after HN. However, findings regarding the Total Energy derived from snack consumption throughout the day did not reveal any significant effects, except for a marginal effect of the Group, F(1,26) = 3.72, p = 0.065, η 2 = 0.13. Overall, the CG ate more compared to BG. A marginal main effect of the Group, F(1,26) = 3.77, p = 0.063, η 2 = 0.13, on carbohydrate intake from snacks was also found. Specifically, the CG consumed more carbohydrates compared to the BG.

The mixed design factorial ANCOVAs revealed no significant effects on Total Energy intake from snacks, or their fat, carbohydrate, and protein content. Results on fibre content controlling for EES-D, showed a significant main effect of the Night, F(1,25) = 4.44, p = 0.045, η 2 = 0.15. Thus, after DN both groups reported more fibre intake from daily snacks than after the HN. No significant results emerged using groups based on EES-D, except for the main effect of the Night F(1,26) = 7.12, p = 0.013, η 2 = 0.22: both groups consumed more Snacks after DN than after HN.

All the analyses were performed separately also controlling for BMI and ISI, which differed significantly between groups. However, results were not affected by these two covariates.

Discussion

This study was the first to evaluate the effect of partial sleep deprivation on food intake in participants reporting symptoms of binge eating vs participants denying any eating disorders symptoms, controlling for the role of emotional eating. We hypothesized that partial sleep deprivation would increase caloric intake observed at breakfast in the lab and self-reported during the day, especially in participants reporting binge eating symptoms. Overall, our findings provided only very partial support for this. We found that partial sleep deprivation may increase snack consumption regardless of binge eating symptoms. In addition, our findings revealed that daily food intake may increase after partial sleep deprivation in individuals who do not report emotional eating.

The sleep manipulation check revealed participants were effectively partially sleep deprived as intended, and that the manipulation affected only sleep duration. In addition, consistent with previous research, the two participant groups differed significantly in terms of BMI, disordered eating, depression, emotional eating and insomnia severity [20, 40, 41]. However, neither the presence of binge eating nor the partial sleep deprivation predicted differences in the caloric intake as hypothesized. The only exceptions were the effect of sleep manipulation on the amount of fibre consumed during breakfast in the lab and the number of snacks consumed during the day after partial sleep deprivation, as well as a near significant effect of the Group on the intake from snacks consumed during the day. These results are partially consistent with those of Nedeltcheva et al. [42], who found that several days of sleep curtailment (5.5 h) was followed by increased intake of calories of snacks.

Accounting for emotional eating [16], partial sleep deprivation mainly affected the group with low emotional eating: individuals reporting low depressive emotional eating consumed significantly fewer Kcal and carbohydrates during breakfast and more Kcal and carbohydrates during the whole day after partial sleep deprivation compared to the day following the habitual night. This is in contrast with our expectation that the sleep manipulation would be associated with increased food intake, particularly in the group reporting binge eating symptoms and high depressive emotional eating. However, this pattern of results is in line with previous results among healthy controls by Markwald et al. [43] who showed that after sleep loss, participants ate smaller breakfasts but ate more carbohydrates, proteins, and fibre over the course of the following day, especially at night after dinner.

In contrast with our expectations, but consistent with previous findings [44] this pattern did not appear among individuals reporting binge eating or emotional eating. Moreover, participants with emotional eating overall consumed fewer Kcal and carbohydrates during breakfast and from snacks during the day, compared to individuals with low emotional eating. A number of factors may account for this. First, our partial sleep deprivation induction (5 h) may have been insufficient given that individuals with binge eating symptoms typically report poor sleep [20]. Second, these individuals may skip breakfast more often than controls [45]. Third, restrained eaters who expect to eat high-calorie foods may be able to activate their dieting goal, thereby limiting their food intake [46]. Given that our participants were asked to maintain their daily habits and behaviors, this may have triggered restrictive goals during the days of the study, resulting in eating overall fewer Kcal and carbohydrates during breakfast and during snacks. In addition, social desirability and shame related to eating in the lab as well as underreporting on food diaries may also have contributed to these findings.

A major limitation of this study is the small sample size that included individuals reporting binge eating symptoms, as opposed to individuals with a diagnosis of Binge Eating Disorder. Another limitation is the use of self-report selection and screening measures, as well as the absence of objective assessment of sleep quantity and quality such as actigraphic measurement. Moreover, the small number of males included in each groups may have limited the generalization of our results. Nevertheless, this preliminary study sheds light on the effect of partial sleep deprivation on food intake and the role of binge eating and emotional eating, which constitute novel findings. Furthermore, the effect size of our significant results showed medium to large effects according to Cohen et al. [47]. In conclusion, our findings suggest that a single night of partial sleep deprivation may increase the daily food intake in people that habitually do not binge and do not eat in response to negative emotions. In contrast, greater sleep deprivation may be necessary to observe the detrimental effects of sleep loss on eating behavior among individuals with binge eating symptoms and emotional eating. Additional studies are needed to extend these preliminary results and to explore the effect of specific sleep stages suppression such as non-rapid-eye-movement (N-REM) sleep on eating behaviour, since experimental studies demonstrated that it may be implicated in the glucose homeostasis and obesity and risk of type 2 diabetes in humans [48].