Chronic Sleep Restriction Differentially Affects Implicit Biases Toward Food Among Men and Women: Preliminary Evidence

Summary

Chronic sleep restriction and obesity are two major public health concerns. This study investigated how chronic sleep restriction changes implicit attitudes towards low-and high-calorie foods. In a randomized, counterbalanced crossover design, seventeen participants (8 females, 9 males) underwent two laboratory testing sessions where they were either sleep restricted for 3 weeks (i.e., underwent 3 weekly cycles of 5 nights of 4 hours of sleep followed by 2 nights of 8 hours of sleep opportunity) or received 3 weeks of control sleep (i.e., 8 hours of sleep opportunity per night for 3 weeks). There was evidence for a significant sex × sleep condition interaction (F(1, 20)=4.60, p=.04). After chronic sleep restriction, men showed a trend towards significant decrease in their implicit attitudes favoring low-calorie foods (p =.08), whereas women did not show a significant change (p = .16). Men may be at increased risk of weight gain when sleep deprived due to a reduced bias toward low-calorie foods.

The high prevalence of obesity and chronic sleep restriction are serious public health concerns in the United States (Altevogt and Colten, 2006, Flegal et al., 2010) . While the National Sleep Foundation recommends between 7–9 hours of sleep per night for healthy adults (Hirshkowitz et al., 2015), 44% of individuals sleep less than 7 hours a night during the weekday and “catch up” on sleep during the weekends (National Sleep Foundation, 2012). Additionally, 68% of U.S. adults are either overweight or obese (i.e., have a body mass index (BMI) of 25 or higher) (Flegal et al., 2010). Importantly, sleep restriction is a risk factor for future weight gain, especially for men (Wu et al., 2014, Spaeth et al., 2013, Spaeth et al., 2014, Cedernaes et al., 2015, Copinschi et al., 2014). While the mechanisms underlying the effects of sleep loss on weight gain are likely multifaceted, one insufficiently explored possibility is that sleep restriction may disinhibit normal cognitive control mechanisms, allowing implicit negative attitudes toward healthier food options to dominate and potentially affect food choices. Recently, we demonstrated that several weeks of chronic sleep restriction increased negative implicit biases toward Arab Muslim men using the Implicit Association Test (Alkozei et al., 2017) (IAT), a well-validated technique that capitalizes on the implicit organization of conceptual associations within cognition. Here, we utilize a food IAT to test the hypothesis that three weeks of chronic sleep restriction would similarly unmask implicit negative attitudes towards healthier low-calorie foods (versus less healthy high-calorie foods) compared to three weeks of normal sleep.

Method

Participants

Seventeen healthy normal sleeping adults (8 females; 9 males; mean habitual sleep duration = 8.2 hours, SD = 32.4 min), ranging in age between 19–31 years (M = 24.53 years, SD = 4.20) were enrolled. Seven participants self-identified as Caucasian (41.2%), eight as African American (47.1%), one participant identified as Asian (5.9%) and one as Native American (5.9%). All women reported a regular menstrual cycle and none were using oral contraceptives. Participants ranged in BMI between 20.1–29.75 (M =24.71, SD = 2.91). The full list of inclusion and exclusion criteria are presented in Simpson et al. (2016). This study was approved by the Institutional Review Board for the Protection of Human Subjects at Beth Israel Deaconess Medical Center. All participants provided written informed consent.

Measures

Food Implicit Association Test (IAT).

The Food IAT was used as a measure of implicit negative attitudes towards low/high calorie foods. The IAT was composed of five blocks. The first two blocks and the fourth block were practice blocks during which participants were presented with one type of category (i.e., high/low calorie food words [e.g., high: doughnut, potato chips, cheeseburger; low: celery, apple, shrimp] or positive/negative valence category [e.g., positive: wonderful, peace, freedom; negative: war, hatred, pain]). After the first two practice blocks, participants were presented with the first critical trial where they were presented with the categories “Low calorie and Positive” and “High Calorie and Negative” and were asked to categorize words into these two categories. Participants then completed another practice run of categorizing negative and positive words. The final critical trial asked participants to sort words into the categories “Low calorie and Negative” and “High calorie and Positive”. The two critical blocks included 40 trials each. The order in which the critical blocks were presented was counterbalanced across participants, and across time points.

Scores on the IAT were analyzed in line with Greenwald et al.’s (2003) IAT scoring algorithm in order to obtain a “d” score. A d-score of zero indicates no valence bias favoring either low or high calorie food, positive scores indicate a slight (~.15), moderate (~.35), or large (~.65) bias associating healthier low calorie foods with positive valence (Greenwald et al., 2003). Negative scores indicate a bias against low calorie foods (i.e., linking low calorie foods with negative valence and unhealthy high calorie foods with positive valence).

Food habit questionnaire.

Participants were asked how much they craved certain types of food when generally hungry before they entered the study on a scale from 1 (never) – 10 (not at all). Food types included carbohydrates, protein, fats, fruit/vegetables and sweet and salty foods.

Sleep restriction paradigm.

In a counterbalanced cross-over design, participants underwent two 25-day in-hospital stays (restricted sleep condition and sleep control condition) each separated by at least 2 months. The study paradigm is described in detail in Simpson et al. (2016). During the rested sleep control condition, participants were provided 8 hours of enforced sleep opportunity per night (2300h-0700h) for 3 weeks. During the sleep-restricted condition, participants completed three weekly cycles during which they were allowed to sleep for 4 hours (0300–0700h) for 5 consecutive nights followed by two 8-hour nights of recovery sleep each week. This cycle was repeated three times over the 25-day period. Polysomnographic data were collected on 5 nights during the sleep control condition, and on 2 sleep-restricted and 3 recovery nights for the sleep-restricted condition. For the sleep control condition, average total sleep time was 6.9 hours (SD = 31.31 min). For the sleep-restricted nights, average total sleep time was 3.75 hours (SD = 17.55 min); for the recovery nights, average total sleep time was 6.96 hours (SD = 37.12 min). Prior to the in-hospital stays, participants were given a menu plan to mark their meal preferences. Meals were prepared by the Nutrition Core of the Research Center and controlled for macronutrients (15% proteins, 30% fats, 55% carbohydrates) and micronutrients (3gm NA+, 3 gm K+ adjusted for calories). Participants’ caloric needs were established using the Harris-Benedict Equation with an average activity factor of 1.4, but participants’ caloric needs were adjusted throughout the study to prevent weight gain or loss. Caffeine and simple sugars were excluded from the menu plan and no unscheduled snacks were served. Meals were served at standardized hours (breakfast at 0800h, lunch at 1200h, dinner at 1800h, and a light snack at 2200h) throughout each in-hospital stay. The IAT was completed at ~1500h on Day 21 of each 3-week in-hospital lab stay (i.e., 4^th sleep restriction or control sleep day of the third week of the protocol).

Results

Hypothesis Testing

Change in IAT d-scores between the sleep control and sleep restricted conditions was analyzed using a linear mixed model, including the effects of sleep condition, age, sex, BMI, as well as phase order (i.e., whether participants first were allocated to the sleep control or sleep restriction condition) and their interactions as fixed effects. Age, sex and BMI were centered before entering them as fixed effects. Two participants had missing data for the sleep control and two participants had missing data for the sleep restricted condition. There was no main effect of condition (F(1, 20)=.61 p=.44), sex (F(1, 20)=.59, p = .45), BMI (F(1, 20)=.19, p=.67), or age (F(1, 20)=3.15, p=.09). However, there was a significant condition × sex interaction (F(1, 20)=4.60, p=.04), but no significant interaction with BMI (F(1, 20)=.87, p=.36), age (F(1, 20)=.02, p=.89), or phase order (F(1, 20) = .84, p = .37). Post-hoc analyses showed that there was a significant difference in food preference for men and women when fully rested (t(19.74) = −2.14, p = .04), suggesting that men have a greater bias favoring low (versus high) calorie foods than women when well rested. This was not evident when sleep restricted (t(19.86) = 1.00, p = .32). Figure 1 shows that compared to being fully rested, women showed an increase in positive implicit bias favoring low (versus high) calorie food when sleep restricted (albeit not significantly, t(10.89) = −1.48, p = .16), whereas men showed a decrease in implicit bias for low (versus high) calorie foods (approaching significance, t(9.63) 1.89, p = .08). In order to rule out any effects of differences in menstrual cycle phase, we re-ran the analysis in only the female participants and added menstrual cycle phase (i.e., follicular vs. luteal) as a covariate of interest. There was no significant effect of menstrual cycle phase (F(1, 5) = .30, p = .61) nor a significant sleep condition × menstrual cycle phase interaction (F(1, 5) = .00, p = .99). In addition, there were no significant differences in general food preferences (i.e., for carbohydrates, fats, protein, vegetables/fruit, salty or sweet foods) between men and women (i.e., for all t’s, −1.30 < t < .14; all p’s > .20).

Figure 1.

There was a significant sex × sleep condition interaction for IAT d-scores (F(1, 20)=4.60, p=.04).

Discussion

Our results did not fully support our initial hypothesis that sleep restriction would increase negative implicit biases towards low calorie foods. On the whole, participants tended to show a bias associating low calorie foods with positive traits compared to high calorie foods. However, we also found preliminary evidence of a sex difference in response to sleep restriction, suggesting that implicit biases toward low calorie foods are affected differently for men and women. We also found support for a difference between men and women in food preference when fully rested. While men showed stronger implicit preference for low-calorie food than women when fully rested, this difference was no longer visible when participants were sleep-restricted. It is possible that women show less of an implicit preference for low-calorie versus high-calorie food than men, because of greater conflicting experiences with such foods. As women are more likely to diet than men (Wardle et al., 2004), they might implicitly associate healthy foods with unpleasant or disagreeable experiences, such as food restriction and its associated sensations. However, it should be pointed out that, both men and women associated healthier low-calorie food more strongly with positive evaluations than high-calorie foods. This finding suggests that overall, high-calorie foods are associated with greater negative implicit attitudes, possibly due to frequent exposure to public health campaigns about the negative health consequences of such foods (Wakefield et al., 2010).

However, whereas no significant change in implicit biases for high- and low-calorie food was observed for women, men showed a trend toward decreased implicit preferences for low-calorie foods when sleep restricted; a change that approached significance (p = .08). Thus, men may be at greater risk of increased high-calorie food consumption when sleep restricted due to altered expression of implicit biases, which could potentially lead to greater weight gain. In fact, studies have shown that men gain more weight and consume more calories than women when sleep restricted (Spaeth et al., 2013, Spaeth et al., 2014). In the current study, meals were controlled in each participant (i.e., for levels of macronutrients, micronutrients, and caloric intake) in order to maintain body weight/composition throughout the study and prevent excess consumption during sleep loss. Given that implicit preferences for certain foods predict actual consumer behavior (Richetin et al., 2007), it seems reasonable that men’s decreased preference for healthy, low-calorie foods when sleep restricted might have been reflected in increased consumption of higher calorie foods had they been given the opportunity to consume more. As women also gain weight after sleep deprivation, albeit not as much as men (Spaeth et al., 2014), future work will be necessary to investigate how sleep deprivation not only changes implicit preferences for low-calorie foods, but also how this is related to actual food choices in men and women separately. It is also unclear how these findings would generalize to other times of the day and/or to different levels of hunger/satiety, which may be important avenues for future research to explore.

It should be noted that our current sample size was relatively small and these findings remain preliminary until replicated. However, this limitation was minimized via a highly controlled, counterbalanced, cross-over study design that permitted within-subject comparisons, and which was ecologically valid and highly reflective of the type of sleep pattern that many individuals regularly adopt. In addition, while habitual sleep duration of the sample was slightly longer than 8 hours, participants’ total sleep time during the sleep control condition, and on recovery nights during the sleep-restricted condition, was close to 7 hours. This implies that some participants may have been slightly sleep restricted during the control sleep condition. However, sleep data were only obtained on 5 out of 21 nights and it is therefore difficult to draw firm conclusions. With due consideration to these limitations, these preliminary findings suggest that men may be at greater risk than women for experiencing subtle declines in cognitive bias toward less healthy food options when chronically restricted of sleep. Future work will need to explore methods for potentially counteracting such biases when optimal sleep is not possible.