Sleep–obesity relation: underlying mechanisms and consequences for treatment

Summary

Short sleep duration has been associated with obesity in numerous epidemiological studies. However, such association studies cannot establish evidence of causality. Clinical intervention studies, on the other hand, can provide information on a causal effect of sleep duration on markers of weight gain: energy intake and energy expenditure. Herein is an overview of the science related to the impact of sleep restriction, in the context of clinical intervention studies, on energy intake, energy expenditure and body weight. Additionally, studies that evaluate the impact of sleep restriction on weight loss and the impact of sleep extension on appetite are discussed. Information to date suggests that weight management is hindered when attempted in the context of sleep restriction, and the public should be made aware of the negative consequences of sleep restriction for weight regulation.

Introduction

The association between short sleep duration (SSD) and obesity is well established. Individuals who report SSD (<7 h/night) have a higher prevalence of obesity and have higher body mass index (BMI) than those who report adequate sleep (7–8 h/night) (1,2). In fact, in the National Health and Nutrition Examination Survey I epidemiologic follow-up surveys, spanning a decade, the average BMI of short sleepers was higher than that of normal sleepers at all assessment time points (3). However, epidemiological findings, despite providing significant information on associations between variables, cannot ascribe causality to the sleep–body weight relation. Nonetheless, several explanations have been put forth to describe this sleep–obesity link: (i) added time awake provides more opportunity to eat; (ii) increased hunger from hormones signalling appetite and reduced satiety from hormones promoting satiation; (iii) altered thermoregulation; and (iv) increased fatigue, implying lower physical activity level (1). A brief overview of the possible causal mechanisms linking SSD and obesity will be discussed herein.

Sleep and food intake regulation

An early study by Spiegel et al. (4) initiated research on the impact of sleep duration on appetite-regulating hormones. In that seminal paper, ghrelin levels were reported to be increased by 28% and leptin reduced by 18% in young normal weight men after a period of sleep restriction (SR; 4 h time in bed [TIB]) relative to 10 h TIB (4). Self-reported feelings of overall hunger and appetite were increased by 24% and 23%, respectively, consistent with the hormone data. The corresponding ratings for calorie-dense and high-carbohydrate foods were increased by 33–45%. A drawback of this study was the lack of food intake measurements. Since then, several studies have been performed to evaluate the effects of SR on hormones involved in food intake regulation along with actual measurements of energy and macronutrient intakes. However, despite additional information about the effects of SR on food intake-regulating hormones, the field is plagued with mixed results relative to the impact of SR on leptin and ghrelin concentrations (5). A recent meta-analysis on this topic concluded to no overall effect of SR on these two hormones, likely as a result of high heterogeneity among studies (6). We had previously proposed that heterogeneity arises from differences because of the sex of the participants and testing conditions (controlled vs. ad libitum feeding, state of energy balance of participants) (5). However, it is important to keep in mind that food intake is regulated at multiple levels, and many more hormones are involved than solely leptin and ghrelin.

Other signalling factors such as glucagon-like peptide 1 and peptide YY have been studied in relation to sleep, albeit to a much lesser extent. We have shown that women have reduced glucagon-like peptide 1 following meals consumed under periods of SR (4 h TIB) relative to normal sleep (9 h TIB) (7). This was not observed in men. However, others have shown that total sleep deprivation delays the glucagon-like peptide 1 peak in response to breakfast consumption in young men, relative to a night of normal sleep (8 h TIB) (8). More recently, Gil-Lozano et al. (9) reported an up-regulation of glucagon-like peptide 1 under total sleep deprivation conditions with light exposure but not under dark relative to sleep (6 h TIB) in young healthy men. No differences in glucagon-like peptide 1 during an oral glucose tolerance test were observed between sleep and total sleep deprivation conditions. Given the preliminary findings suggesting an impact of sleep duration on glucagon-like peptide 1, more studies with larger sample sizes, including men and women, are needed to assess the effects of SR on hormones involved in satiety signalling. This would include other hormones that are released in response to food intake to end consumption.

Hormones, however, are not the only factors that modulate food intake in humans. Individuals eat for a variety of reasons unrelated to hunger. Eating in the absence of hunger has been explained by boredom, stress, fatigue, joy, etc. The appeal of food stimuli to prompt intake following periods of SR and normal sleep has been examined in neuroimaging and behavioural studies. We and others have shown that presentation of food stimuli after a period of SR enhances activity in neuronal centres associated with pleasure and reward compared with viewing the same stimuli after a period of normal sleep (10-12). One study examined resting state functional connectivity to assess salience network activity following a night of total sleep deprivation relative to habitual sleep (13). Compared to baseline and to a non-sleep deprived control group, total sleep deprivation led to an increase in the strength of the connectivity between the dorsal anterior cingulate cortex and bilateral putamen and between the dorsal anterior cingulate cortex and bilateral insula. This increase in functional connectivity between those brain regions was positively correlated with fat intake and negatively correlated with carbohydrate intake measured in the laboratory. The authors concluded that the up-regulation of the salience network co-activation between the dorsal anterior cingulate cortex and the putamen and insula, which predicted increased fat intake and reduced carbohydrate intake, may be in part responsible for guiding poor food choices in sleep deprived individuals and may explain the increased obesity risk associated with poor sleep.

Neuroimaging findings have also been supported by studies examining portion size choice and food purchases in mock conditions. In fact, after a night of total sleep deprivation, participants choose larger food portions both in the fasted and post-meal (sated) states compared to the sleep condition (8 h TIB) (14). Results from this computer task show that SR affects both the homeostatic (fasted state) and hedonic (satiated state) drive to eat. In a separate study, men performed a mock supermarket task in which they were provided $50 USD to purchase food (15). Men purchased more food, expressed both on gramme and caloric bases, after the night of total sleep deprivation relative to normal sleep. Hormone, neuronal and behavioural data suggest enhanced susceptibility to overconsumption under states of inadequate sleep (Fig. 1).

Figure 1.

In the context of sleep restriction, there is an up-regulation of reward, pleasure and salience networks in response to food stimuli. These would predispose one to choose high fat, high carbohydrate foods and have overall increased energy intakes.

At present, several studies have shown that SR leads to increased food intake relative to normal sleep. We, and others, have shown that the degree of overeating is in the order of 300–550 kcal/d. This is reflected by an increased eating frequency and, in some instances, by a shift in food intake to late at night (16). However, a limitation of those studies is their short duration. Nonetheless, a meta-analysis of lifestyle behaviours associated with increased food intake corroborates these conclusions (17). The reader can refer to several reviews related to energy intake in the context of SR (18,19).

Sleep and energy expenditure

The effects of SR on thermoregulation in adults have been assessed in metabolic chamber studies. Three studies have been performed in women alone (20) and in both men and women (21,22). All three studies agree that SR increases total 24-h energy expenditure by 5–7% relative to normal sleep. This elevation in energy expenditure under sedentary conditions largely reflects the added costs of maintaining wakefulness rather than an alteration in metabolic rate as a result of SR. What remains to be determined is free-living energy expenditure. One study reported that outpatient free-living energy expenditure was reduced with SR relative to normal sleep on a single day of assessment, whereas it was not different when participants were kept as inpatients (23). Additional studies in which participants are provided more opportunity to carry on with their daily activities are needed to determine whether physical activity would be reduced in SR, perhaps as a result of increased fatigue resulting from insufficient sleep. Studies that have measured energy expenditure using doubly-labelled water have not found differences between periods of SR and normal sleep on 24-h energy expenditure (24,25). This may be due to the short duration of these studies (≤14 d) and their in-patient nature. Nonetheless, the meta-analysis of data from these randomized clinical studies revealed a trend towards increased energy expenditure as a result of SR (6). The authors commented that the limited number of studies and their small sample sizes restrict conclusions that can be drawn about these outcomes. It is worth noting that the meta-analysis used data from in-patient studies. There is a great need for studies to systematically assess the impact of SR on physical activity energy expenditure.

Net effect of sleep restriction on weight status

On the basis of studies reviewed thus far, it can be hypothesized that SR will lead to positive energy balance: energy intake is increased by at least 300 kcal/d vs. an increase in energy expenditure of approximately 90 kcal/d (derived from in-patient metabolic chamber studies). The net effect, if these consequences of SR persist over time, should be weight gain. Moreover, if one assumes that physical activity energy expenditure is reduced under periods of SR, this degree of positive energy balance is an underestimation.

Very few studies have assessed the change in body weight as a result of SR in the context of free-living energy intake and expenditure. When sleep was maintained at 5.5 or 8.5 h TIB for 14 d, there was no difference in the change in body weight, body fat or fat-free soft tissue when measured by dual energy X-ray absorptiometry (24). Although energy balance was not statistically significantly different between groups, the net difference was approximately 160 kcal/d. It is possible that a longer study duration would have resulted in significant differences in the change in body composition if one assumes that the energy gap between SR and normal sleep is maintained. Interestingly, two 5-d studies noted significant differences in weight change as a result of SR (4 h TIB) relative to normal sleep (9–10 h TIB) (16,22). Body weight gain over the 5-d SR period was similar in those two studies, in the order of 0.8–1 kg. On the other hand, another, slightly longer inpatient study tested the impact of SR relative to normal sleep on energy balance (26). Participants were randomized to either maintain their normal sleep schedule or reduce their sleep duration by delaying bedtimes to achieve two-thirds of their habitual sleep duration. Participants undergoing SR slept an average of 5 h and 12 min for eight nights compared with 6 h and 53 min for the control group. Individuals in the SR group increased their energy intake by 559 kcal/d during the intervention period relative to the lead-in ad libitum sleep period, whereas the control group reduced their intake by 118 kcal/d, for a net difference of 677 kcal/d. Of note is that during the three-night lead-in period, the SR group ate 678 kcal/d less than the control group. As a result, energy intakes during the intervention period were the same in both groups. There was also no difference in accelerometry data between groups. At the end of the intervention, those randomized to SR had an increase in body weight of 0.9 kg vs. 0.6 kg for the control group; group changes were not significantly different. However, information from the prior two studies suggests significant implications of SR for the development of obesity (16,22). Data from the latter are more difficult to interpret because the SR group ate less than the control group during the lead-in period and also in the post-SR, recovery phase (526 kcal/d) (26). In that study, the control group’s energy intake was constant throughout the study whereas the SR group’s varied widely. Nevertheless, these remain very short duration studies, and it is unknown whether positive energy balance would be maintained over a longer period of time, or corrected with time to achieve steady-state weight maintenance.

The longest SR study to date randomized participants to either maintain their normal sleep pattern or restrict their sleep by 1.5 h/night for 3 weeks (27). Men randomized to the SR intervention had an initial, non-significant drop in body weight, which subsequently significantly increased by approximately 1 kg over the final week of the study. The end result was no difference in endpoint body weight or 3-week change in body weight between groups. One must ponder on the meaning and implications of the rise in body weight in the final week. Given that the extent of SR was less than half of that of the other clinical intervention studies (1.5 vs. 3.5–4 h/night), it is possible that a longer intervention duration is required to accumulate a sufficient sleep debt for appetite hormones and neuronal reward networks to become up-regulated to stimulate appetite, food intake and ultimately weight gain. In fact, in the SR group, leptin levels were maintained at baseline levels for the first 2 weeks and then decreased to below baseline by the third week (27). These data support the proposition of a threshold ‘sleep debt’ that must be achieved to result in significant body weight gains from SR. Long-term studies, beyond the 3 weeks performed thus far, are needed to test this hypothesis.

Capers et al. also highlighted this call for longer study durations in this field of research in their meta-analysis (6). Their findings suggested that SR could potentially lead to weight gain, but they could not determine whether meaningful weight gain would occur over a longer study duration given that no study has been performed for more than 3 weeks.

Not only is SSD a risk factor for weight gain due to mostly increases in energy intake, but weight loss may also be hindered in those who restrict their sleep. A small inpatient study of 10 adults examined the impact of energy restriction in the context of SR relative to normal sleep for periods of 2 weeks each (28). In each sleep condition, participants were provided a standard diet designed to meet 90% of their resting metabolic rate determined at the time of screening. The sleep intervention periods consisted of either 5.5 h or 8.5 h TIB. Given the controlled feeding nature of the study, body weight loss was equivalent between phases. Of importance is that the composition of weight lost differed between sleep phases. During SR, weight loss was comprised of more fat-free mass and less fat mass, assessed by dual energy X-ray absorptiometry, than during the normal sleep period. During SR, participants lost less than half as much fat mass as they did during the normal sleep phase. Unfortunately, the authors did not have information on the distribution of fat mass: visceral vs. subcutaneous or central vs. peripheral. Nevertheless, this study had other interesting findings. For example, resting metabolic rate was lower, and fasting and post-meal respiratory quotient were higher, at the end of the weight loss intervention performed in the context of SR relative to normal sleep. Such changes would suggest that individuals with SSD would have more difficulty losing weight and likely have an increased risk of weight regain post-weight loss. It is unfortunate that no other study has been performed to replicate the findings of this initial innovative and informative preliminary study.

Chaput and Tremblay performed a secondary analysis of data from three separate weight loss studies to determine whether sleep duration and quality affected the degree of weight loss achieved by caloric restriction (29). In all three studies, participants were instructed to reduce their energy intakes by 600–700 kcal/d and were seen by a dietitian on a bi-weekly basis. Study duration ranged from 15 to 24 weeks. Sleep duration was positively associated with loss of body fat, after adjusting for age, sex, baseline BMI, study duration and change in energy intakes. For each additional hour of sleep, fat loss increased by 0.77% or 0.72 kg. Sleep quality, assessed using the Pittsburgh Sleep Quality Index, was negatively related to fat mass loss, implying that poor sleep quality is related to lower loss of fat mass. These data, from a larger group of men and women (n = 123), support the findings of the more tightly controlled intervention study described previously (28).

Sleep extension for weight management

Given the negative influence of SSD on energy balance regulation and its potential impact on weight status, questions arise regarding the potential of sleep extension to improve weight management. One study implemented a home-based intervention to extend sleep in overweight individuals sleeping <6.5 h/night (30). Investigators evaluated the effects of their intervention on sleep duration and desire for food. Ten participants received individualized counselling on sleep hygiene and to identify personal, modifiable barriers to sleep extension. The goal was for participants to achieve 8.5 h bedtimes, which would ultimately result in sleep duration of 7–8 h/night. The intervention, which lasted for 2 weeks, was effective in increasing sleep duration in all participants; the average increase was 1.6 h/night. Sleep extension was associated with a 7% increase in daytime activity, assessed by actigraphy. Participants also reported feeling less sleepy and more vigorous. In addition, overall appetite ratings decreased by 14%. This reduction in appetite was largely due to the 62% reduction in appetite ratings for sweet and salty foods; there was no significant change in desire to eat fruits, vegetables or protein-rich foods. This preliminary study demonstrates the feasibility of extending sleep in short sleepers and provides initial evidence of a potential benefit for appetite sensations. Unfortunately, the study lacked a control group or control intervention period, and data following the sleep extension period were compared with pre-intervention baseline data. Food intake was also not assessed, and it is not clear whether energy and macronutrient intakes would be altered as suggested from the self-reported appetite ratings. Nonetheless, this study provides a good starting ground for future research in this area. Furthermore, this clinical intervention study supports findings from a longitudinal study of adults in the Quebec Family Study (31). Participants from that cohort were grouped according to their sleep duration at baseline and 6-year follow-up: either constant short sleepers (≤6 h/night at both time points); short sleepers who increased their sleep to recommended duration (≤6 h/night at baseline and 7–8 h/night at follow-up); constant healthy sleepers (7–8 h/night at both time points). Those who maintained their SSD throughout the study had a greater increase in BMI, body weight and fat mass compared to constant healthy sleepers; those who increased their sleep from baseline to follow-up had similar change in weight status as the constant healthy sleepers. This study shows that sleep duration can change and that increasing one’s sleep, by approximately 1.5 h/night, to reach recommended sleep times, can improve weight management.

Conclusions

There is much evidence supporting a negative impact of SR on food intake regulation. Decision-making relative to food intake is biased by an up-regulation of salience, reward and pleasure networks in response to food stimuli, which would explain greater intakes of high-fat, high carbohydrate energy-dense foods and an overall rise in energy intake relative to normal sleep. Information related to the impact of SR of energy expenditure reflects the multiple ways in which energy expenditure can be measured. There seems to be no fundamental impact of SR on energy metabolism: wake time and sleep time energy expenditure is equivalent regardless of the number of hours spent awake. However, given that wake time energy expenditure is higher than sleeping metabolic rate, the result is greater net energy expenditure when wake time is increased. Not surprisingly then, weight gain occurs when sleep is acutely restricted. On the other hand, milder, longer SR provides contradicting information and suggests a threshold of sleep debt that must be achieved to result in weight gain. This deserves attention. Messages to the general public should caution that weight loss may be hindered in the context of SR and that SSD can have adverse effects for weight management.

Abbreviations:

BMI

body mass index

SR

sleep restriction

SSD

short sleep duration

TIB

time in bed