Obesity and short sleep: unlikely bedfellows?


J Horne, Sleep Research Centre, Loughborough University, Loughborough, Leicestershire LE11 3TU, UK. E-mail: j.a.horne@lboro.ac.uk


The link between habitual short sleep and obesity is critically examined from a sleep perspective. Sleep estimates are confounded by ‘time in bed’, naps; the normal distribution of sleep duration. Wide categorizations of ‘short sleep’, with claims that <7 h sleep is associated with obesity and morbidity, stem from generalizations from 5 h sleepers (<8% of adults) and acute restriction studies involving unendurable sleepiness. Statistically significant epidemiological findings are of questionable clinical concern, even for 5 h sleepers, as any weight gains accumulate slowly over years; easily redressed by e.g. short exercise exposures, contrasting with huge accumulations of ‘lost’ sleep. Little evidence supports ‘more sleep’, alone, as an effective treatment for obesity. Impaired sleep quality and quantity are surrogates for many physical and psychological disorders, as can be obesity. Advocating more sleep, in these respects, could invoke unwarranted use of sleep aids including hypnotics. Inadequate sleep in obese children is usually symptomatic of problems not overcome by increasing sleep alone. Interestingly, neuropeptides regulating interactions between sleep, locomotion and energy balance in normal weight individuals, are an avenue for investigation in some obese short sleepers. The real danger of inadequate sleep lies with excessive daytime sleepiness, not obesity.


An increasing number of epidemiological studies are reporting small but significant links between short sleep and obesity, often with the implication is that this ‘lack of sleep’ is another reason for obesity, although none has been able to establish sleep as an underlying cause. Given the concerns by the public and media towards the ‘obesity epidemic’, this link has the potential to focus attention on sleep rather than to exercise and diet as behaviours to promote weight loss and a healthy body weight. In taking a sleep-oriented perspective, this review has an underlying theme that as sleep quality and quantity are largely a litmus test for a variety of physical and psychological disorders and problems, any association between short sleep and obesity is probably by virtue of common underlying factors, and is less likely to be causal. Thus, e.g. before more sleep is advocated as a putative method for stabilizing or reducing weight, one might consider whether this extra period of intended sleep would better be spent in undertaking some more effective method for weight reduction, such as comparatively brief periods of exercise

The recent, comprehensive and critical review in this journal, by Nielsen et al. (1), of short sleep as a possible cause of obesity, concluded that there was a consistent relationship, here, for children and young adults, but not so evident for older adults. Various factors affecting potential causality were appraised, especially those concerning: time lags and bidirectional effects between the two variables, confounders vs. mediators, the need to consider sleep quality as well as quantity. The present review complements and expands upon some of these key topics, and draws attention to other relevant aspects of sleep also having a major bearing on this still controversial area and its clinical ramifications.

As will be seen, any potential sleep-related real risk of an adult becoming obese or overweight is not apparent until habitual sleep is down to around 5 h per 24 h and if, say, 7 h sleep per day is advocated, this difference could accumulate to >700 h extra sleep annually. In contrast, the ‘worst case’ for weight gain that could be attributed to short sleep over a year is <2.0 kg, which could be worked off in very much shorter periods of brisk walking. Other ramifications include the possibility that hypnotics and other sleep aids might be seen as a feasible method for obese short sleepers in facilitating this extra sleep and putative weight loss. Moreover, by advocating the need to avoid short sleep, to prevent weight gain and other health consequences, this may well add to the worries of those people with primary insomnia, having undue concerns about inadequate sleep (2), further increasing their demands for sleep aids.

Vgontzas et al. (3) pointed out that obese people free of obstructive sleep apnoea, but having short sleep, often complain of insomnia accompanied by chronic emotional stress. However, no difference was found by these authors in self-reported sleep duration between those obese and non-obese individuals who appeared free of chronic stress and subjective sleep disturbance. Moreover, chronic emotional stress had a stronger relationship with reported sleep duration than did BMI, with the conclusion that self-reported short sleep in obese individuals may be a surrogate marker of emotional problems. Indeed, depression and obesity are often clearly inter-linked (4) with, e.g. (5) obesity associated with a 25% increase in mood and anxiety disorders. As short sleep can also be a diagnostic criterion for depression, then a disproportionate number of short sleepers would be expected to be depressed, irrespective of obesity (6). Other behaviours linking short sleep with obesity include stress-induced lassitude and ‘comfort eating’ (cf. current reviews –(7,8)).

Recent surveys of normal adult sleep point to habitual mean daily sleep durations of around 7–7.5 h, as found, e.g. in: a UK survey of 2000 UK adults (9) reporting a mean of 7.04 h (SD: 1.5 h); a larger US survey (10), of around 110 000 adults, having a similar values (Fig. 1). Implications that the current ‘obesity epidemic’ is associated with adults sleeping fewer hours nowadays, are without solid foundation (11), as this 7 h average has changed little, and not over the last 50 years in the UK (12–14). Those habitual sleep reductions that have been reported, mostly from Scandinavia, have been very small, e.g. 5.5 min less daily sleep per 10 years (15) since 1972 (no change in the more extreme sleep durations), and by only 15 min in Swedish women over the last 36 years (16). In the USA, the proportion of adults sleeping <6 h per day since 1975 has increased by only 1.7% (7.6% to 9.3% in 2006, (17)). Further claims from the USA (e.g. (18)) of today's ‘sleep debt’, are often based on historic records, not of sleep duration but of people's responses as to whether they felt ‘rested’ and ‘energetic’ when awake, i.e. terms only implicating adequate sleep. Current assertions (cf. (18)) that we slept for around 9 h per night, 100 years ago, stem from a commonly cited, but misquoted study not on the sleep of adults, but of 2692 children and teenagers (19), where the mean duration was 9 h (Fig. 2) – see later. There is no solid evidence that adults, then, slept 9 h, especially as many were impoverished, worked much longer hours and slept under conditions that few of us would tolerate today.

Figure 1.

Distribution of sleep durations in 110 000 US adults – adapted from Krueger & Freedman (2009) –(10).

Figure 2.

Mean sleep durations by age for 2692 schoolchildren – from Terman & Hocking (1913) –(19).

Other claims of chronically insufficient sleep in today's adults have just relied on surveys involving leading questions (cf. (20)) asking people if they would ‘like to have more daily sleep’, which encourages a positive answer, but fail to probe the intent by asking, e.g. ‘what waking activities would you forfeit in order to sleep more?’ We (20) found that such positive answers are usually not accompanied by elevated daytime sleepiness (a cardinal sign of insufficient sleep) or an increased willingness to forego waking pleasures and needs.

Habitual daily sleep duration is approximately normally distributed, of course (e.g. Fig. 1). Thus, there are genotypes who may be longer or shorter sleepers, and others who are phenotypically this way, whereby some reduction to sleep can safely occur without increasing daytime sleepiness or jeopardizing health when there are competing waking pressures. Clearly, there are others who are indeed chronically sleep-deprived, having failed to adapt to this inadequate sleep and suffering as a consequence. Although, as will be seen, links between sleep and obesity only become more apparent with those habitually sleeping <5 h, only around 8% of adults sleep this amount (Fig. 1), with an unknown portion being below phenotypic adaptation, who endure excessive sleepiness, maybe an increased likelihood of obesity, morbidity and even falling asleep at the wheel.

Brain mechanisms (see below) common to the regulation of feeding, foraging, sleep and locomotion, allow for phenotypic adaptation of sleep, depending on environmental pressures, but without impacting on the levels of waking vigilance. Safe and well-fed mammals in zoos, and as pets, sleep longer than those in the wild where, e.g. more time is required in foraging and hunting for food. But as their wild counterparts have to remain vigilant, they are unlikely to be endangered by chronic sleepiness, despite sleeping fewer hours; even less likely to become obese, as happens in captivity.

Epidemiological evidence linking obesity to short sleep is often accompanied by claims that this is not necessarily because of sleepiness-induced indolence, but to more specific changes in leptin and ghrelin (cf. (21)), largely emanating from laboratory findings (22) with healthy young adults acutely restricted to 4 h sleep per day for 6 days. However, significant effects, including impaired glucose tolerance and signs of metabolic syndrome were only evident when compared with subsequent recovery sleep, not with normal sleep days (22). Restriction was stressful as corticosteroid output was elevated (there was no stress controls). Such an inadequate amount of sleep causes excessive sleepiness and can not be tolerated for more than a few days. As 4 h sleep is rarely found habitually (Fig. 1), the extent to which these findings can be generalized even to 5 h sleepers is unclear.

Methodological issues

As noted by Nielsen et al. (1), and reflected by all the papers covered in a recent meta-analysis (23), sleep duration is generally based on subjective responses to a single question, with a failure to clarify what is meant by ‘sleep’; a term easily confused with ‘time in bed’. Such data should also require ‘sleep diary’ confirmation with respect to times of: going to bed, sleep onset, morning awakening and getting up, which are seldom obtained. It could be argued that subjective estimates contain random errors that average out, but this has yet to be demonstrated. Lauderdale et al. (24) compared self-reports of sleep duration vs. those obtained by wrist actimetry (a reliable and valid measure of sleep duration and quality –(25)) in 699 adults, and found that people subjectively overestimated their sleep by up to an hour, with R2 between these two indices being only 0.22. People estimating their sleep at 5 h usually only slept about 4 h, and those estimating 7 h only slept about 6.6 h. The discrepancies largely depended on health and socio-demographic factors.

Studies (e.g. (26)) still combine all those sleeping below a certain sleep duration (usually 7 h), and in finding significantly more obesity compared with those sleeping 7–8 h, it is assumed that this includes those sleeping, e.g. 6–7 h. Whereas the <5 h sleepers may be at greater risk, this can not be generalized to 6–7 h sleepers without actual findings to this effect. From the more limited data available, there is little or no significant evidence demonstrating that alert, 6 h sleepers are any more likely to become obese than 7 h sleepers (27). Although other studies have banded sleep durations ostensibly in 1-h intervals, typically as: ‘5 h or less, 6 h, 7 h, 8 h and 9 h or more’ (e.g. (28,29)), there is an apparent anomaly, as e.g. a 5.1-h sleeper (who may be at a greater risk than a 6-h sleeper) will fall within the 6-h category, indicating that 6 h sleep is riskier than is the case.

Statistical vs. clinical significance

This topic covers many aspects of medicine (cf. (30)), where epidemiological studies comprising many thousands of participants, find small differences that are highly statistically significant, leading to the assumption that this is synonymous with a more worrying ‘clinical significance’. In the case of short sleep, such statistical findings can further buttress the hazards of ‘sleep debt’. Another concern involving statistical interpretations, here, is that many epidemiological studies describe as odds ratios (ORs) the likelihood of short sleepers becoming obese, rather than as relative risks (RR), usually with 7 h–8 h sleep as the norm. RRs are more logical and interpretable (cf. (31)), but can not always be calculated in some designs such as case-controls (e.g. selecting participants from the outcome measurement rather than the exposure). As outcomes can differ depending on the method (31), it is often prudent to examine the actual data if presented, as will be done here, when possible.

Any significant weight differences between short and normal sleepers accrue only after many years of sleeping in this manner. For example, a BMI difference of 2 kg m−2 between short (5 h) and normal (7–8 h) sleepers (32,33), among the most substantial reported, developed over 10 years, following some hundreds (probably thousands) of hours of accumulated (presumably) lost daily sleep over this period, when compared with the norm. This is even more apparent with the small, highly statistically significant 1 kg m−2 BMI difference found (34) with 1469, 32–62 years old women habitually sleeping either more or less than 6 h per day, over several years.

Patel et al.'s (35,36) prospective study of 68 183 female nurses, over 16 years, found for each year measured, that ≤5 h sleepers consistently weighed more than those sleeping longer. Specifically, in the first year (Table 1 in ref (36)), ≤5 h, 6 h, 7 h and 8 h sleepers weighed, respectively: 69.7 kg, 68.4 kg, 67.1 kg and 67.8 kg. Easily overlooked is that 16 years later, despite remaining heavier, the overall weight gain for the ≤5 h sleepers was only 2.8 kg more than that of the 7 h sleepers, and 3.2 kg more than the 8 h sleepers (increases being 7.6 kg, 5.9 kg, 4.8 kg and 4.4 kg for ≤5 h, 6 h, 7 h and 8 h sleepers, respectively), despite there being some (potentially) 10 000 h accumulated difference in sleep duration between <5 and 7 h sleepers over the 16 years.

The recent meta-analysis by Cappuccio et al. (37) concluded that a habitual sleep reduction of 1 h per day below 7 h is associated with a 0.35 kg m−2 increase in BMI, over an unknown period of years. Similarly, Patel et al. (38) in an actigraphically monitored sleep of 3055, 67–96 years old people, found that compared with 7–8 h sleepers, a habitual sleep duration as low as <5 h was only associated with 2.5 kg m−2 and 1.8 kg m−2 higher BMIs in men and women, respectively, which is small considering that this had probably accumulated over tens of years. In comparing BMIs of 7 h habitual sleepers with <5 h 6 h, 8 h and >9 h durations, Stranges et al. (39) found that compared with 7 h sleepers, those sleeping <5 h had a statistically significantly higher, but small BMI increase averaging 1.1 kg m−2 for both men and women, and a greater waist circumference averaging <2.3 cm. However, the authors conceded that the effect was small (as can be seen in Fig. 3, shown here), with probably no temporal relationship between short sleep and BMI changes. Interestingly, 25% of these men and 33% of women <5 h sleepers were depressed, compared with around 8–14% for men and women in other hourly sleep categories. Other, very recent case–control studies (40,41) again cast doubt on any causal association between short sleep and obesity (obstructive sleep apnoea excluded), as does the recent review by Marshall et al. (42). A study of twins (43) concluded that any relationship between BMI and sleep was largely environmental.

Figure 3.

Body mass index (BMI) by sleep duration in men (n = 3619) and women (n = 1422) – mean and SD. From Stranges et al. (2008) –(39). For both groups <5 h sleep was associated with statistically significant but small increases in BMI and waist circumference – see text for details.

Of further note are the two most recent prospective studies (44,45). Hairston et al.'s (44) 5 years follow-up of 522 men and women, found that for those originally aged <40 years who slept <5 h, and compared with 6 h–7 h sleepers, BMI had increased by 1.8 kg m−2 over the 5 years period, and by 0.9 kg m−2 compared with >8 h sleepers. However, for people >40 years no such effects were found. Wanatabe et al. (45) monitored 31 206 men and 3770 women over a year, and reported no BMI change for those habitually sleeping 7–8 h, and a near zero change for each of the three male groups sleeping 5–6 h, 6–7 h and 8–9 h. However, for men sleeping <5 h there was a significant, slight BMI gain of 0.07 kg m−2, which can be seen for men, in Fig. 4 here. None of the female groups changed significantly.

Figure 4.

Body mass index (BMI) by sleep duration at baseline and after 1 year, in 31 206 men – mean and SD. From Watanabe et al. (2010) –(45). There were small, significant BMI increases for the <5 h, 5 h–6 h and >9 h groups only. There were no BMI changes for the women's sleep duration groups (not shown) – see text for further details.

Finally, the prospective birth cohort study of 1037 men and women, by Landhuis et al. (46), found that higher BMIs at 32 years were linked to estimated bed-cum-sleep times previously obtained at 5 years, 7 years, 9 years and 11 years, categorized as ‘short’ (<11 h in bed), ‘moderate’ (11–11.5 h) or ‘long’ sleepers (>11 h) over this 6 years period. Short sleep was significantly associated with higher adult BMI values, with the suggestion that sleep restriction in childhood increases later obesity. However, the BMI difference at 32 years between short sleepers and the other two (similar BMIs) groups was only about one BMI unit (cf. their Fig. 1).

Type 2 diabetes and metabolic syndrome

There are few, if any, epidemiological findings pointing to any real, increased risk of type 2 diabetes in those sleeping 6 h (27), with the great majority having slept like this for many years. Ayas et al. (47) findings show that 96.8% of 6 h sleepers were asymptomatic (compared with 95.7% and 97.5% for ≤5 h and 7 h sleepers, respectively). Although the prospective study by Yaggi et al. (48) of diabetes prevalence in 6 h sleepers was double (10%) that of 7 h of sleepers (5%) this necessitated up to 14 years of sleeping like this, and that any causal relationship could not be presumed.

Also, there is little epidemiological evidence showing that healthy, habitual 6 h sleepers are at any real risk for developing metabolic syndrome (27). In Hall et al.'s (49) recent study of 1214 participants (30–54 years old), of whom 268 had the syndrome, the authors presented the findings mostly as ORs. From what can be gleaned from their Fig. 1, the syndrome was evident in approximately 18% of the 7–8 h sleepers, compared with 24% for 6–7 h sleepers, 28% for <6 h sleepers, and 24% for >8 h sleepers; all of which seem to be similarly high incidences. Nevertheless, in terms of who is symptom-free, then the figures look less persuasive (i.e. between 76% and 82%).

It could be argued that as the syndrome and type 2 diabetes may be inflammatory responses (50), inadequate sleep might impair immunity; hence this effect. However, it now seems (51) that sleep loss itself, leads only to equivocal changes in human immune function, not necessarily detrimental but only indicative of greater immunological activity. Evidence pointing to adverse effects of prolonged sleep loss on immunity derives from extreme studies on rats (cf. (52)).

Timing of sleep – not all of sleep is equivalent

Humans naturally have an early afternoon circadian ‘dip’, worsened by inadequate night-time sleep. For example, a short (15–20 min) nap at this time is as effective in maintaining alertness throughout the rest of the day as is extending night-time sleep by an hour (53); i.e. one should not judge the efficiency of sleep on its duration alone. Thus, gauging sleep duration only in terms of night sleep overlooks the benefit of short sleepers also having a short afternoon nap. As also noted by Nielsen et al. (1), much of the epidemiological findings rely on sleep duration either in terms of a typical night's sleep or as a total per 24 h. However, adding to their point, here, the former overlooks naps, and the latter wrongly assumes naps to be of equal value, time-wise, to night sleep.


Changes to children's sleep with age are seen in Fig. 2. Albeit 100 years old, the point of showing these findings is not only that they remain applicable today (54), but this illustrates that attempting to relate sleep duration with obesity across this age range is clearly going to be difficult, given the natural decline in sleep duration and individual differences in age of puberty. Moreover, within the normal distributions of sleep duration for any age, there are cultural differences in child rearing practices in relation to sleep and daytime napping (55). Nevertheless, many children do obtain inadequate sleep, which impairs their behaviour, mood and learning ability (56) and, arguably, has a much greater and immediate effect on a child's health and mental well-being than does the prospect of sleep-related obesity.

Concerning sleep and obesity in children, there have been further reviews and findings since that of Nielsen et al. (1), with perhaps the most notable being by Monasta et al. (57) who concluded that short sleep duration was a factor in obesity, among several others, but like Nielsen et al. noted it was difficult to establish a causal association. Chen et al.'s (58) meta-analysis is more forthright in concluding that ‘the prevalence of childhood obesity may be decreased by increasing sleep duration independent of other risk factors’ (p. 272), which is probably overstated, being based on standardized sleep durations and heavily reliant on statistically significant and pooled ORs.

As with adults, the epidemiological evidence, here, is weaker than at first it might seem, especially as sleep durations are usually based on parental responses to a single question about their child's sleep duration, with few studies defining ‘sleep’ vs. ‘time in bed’. Schooldays vs. weekends/holidays are often overlooked. Again, links between sleep duration and obesity usually entail some years of having slept in this manner, with small BMI increases (BMI as an index of child obesity remains debatable –(59)), accompanied by numerous hours of implied ‘lost sleep’. These studies tend to dichotomize sleepers into those who sleep more or less than (usually) 10 h a night, incorporating those sleeping far below this threshold, who might be at a greater risk of obesity, etc.

Before turning to the most recent reports, there are aspects of older studies, not covered by Nielsen et al. that can be raised. Although the prevalence of obesity in <10 h sleepers has been reported to be double that for those sleeping >10 h (60,61), this can be seen from another perspective. In one study (60), about double (7.7%) the proportion of ≤10 h sleepers were obese compared with 3.6% for ≥10 h sleepers. However, 92.3% of the ≤10 h sleepers were of normal weight vs. 96.4% for ≤10 h sleepers. Similarly, another finding (61), that 5.4% of ≤10 h sleepers were obese, which was about double of (2.8%) 10.5–11.5 h sleepers, can also be seen from this obverse, arguably less impressive perspective.

Lumeng et al.'s study (62), culminating with 785 children (equal boys and girls), monitored from ages 9 to 12, also identified short sleep linked to becoming overweight. Covariates included: ‘level of chaos at home, the quality of the home environment, and lax-parenting’. At age 12, 17.7% of the children were overweight, whose sleep duration averaged 8.8 h and significantly different from the 9.02 h for normal weight children. However, this comprises only 14.4 min, including a 7-min average later bed-time for those who were overweight. Nevertheless, the authors extrapolated the findings to propose that for every additional 1 h of sleep beyond 9 h at age 12, the child was 20% less likely to be overweight, and for every extra 1 h at age 9, 40% less likely to be overweight by age 12; both of which estimates seem to go beyond the data. Other findings (63), of a significant negative correlation between sleep duration and body weight in boys, only accounted for 10% of the total variance, and was not significant for girls.

The impressive and large (‘Avon’) prospective study by Reilly et al. (64) of UK children, from birth, reported on 7758 (equal boys and girls) at age 7.9–8.5. BMI data revealed 9.2% of the boys and 8.1% of the girls to be obese. BMIs were also retrospectively compared with data at age 3, for sleep length obtained from parental questionnaires. Eight factors at age 3 were associated with the risk of subsequent obesity, including: parental obesity, watching television ≥8 h per week and short (≤10.5 h) sleep. Even so, over the 5 years, and at around age 8, 89.7% of these short sleepers had normal BMIs, compared with 93.2% for ≥12 h sleepers.

Similarly small, significant reductions in habitual sleep in overweight children, were described in the birth to 9.5 years prospective study by Agras et al. (65). Five independent risk factors were identified: parent overweight, child temperament, low parental concern about child weight, food tantrums and daytime sleep at 3–5 years. Daily sleep duration at this earlier age was negatively related to becoming overweight, and those who did so, slept about 30 min less than those of normal weight. However, 25 min of this comprised shorter daytime sleep (unclear why), with only a 5-min difference in night sleep. The strongest factor relating to becoming overweight was parental BMI.

An overnight sleep electroencephalographic (EEG) study study (66) of 335 children found those who were overweight slept about 17 fewer minutes. After adjustments, the authors proposed that one hour less sleep was associated with an OR of 1.8 of being overweight, as was 1 h less of rapid eye movement (REM) sleep (OR = 2.9). Nevertheless, the actual difference in REM sleep between normal and overweight children was small, at 12 min.

Chaput et al.'s (63) cross-sectional study of 422 Canadian boys and girls, aged 5–10 years, centred on a telephone questionnaire to the parent, and on this basis, 20% of the boys and 24% of the girls were overweight or obese. Of the 50 short sleepers (8–10 h per day), six boys and five girls were overweight or obese, resulting in adjusted RRs for being overweight/obese of 5.6 for boys and 3.1 for girls. Clearly, this is derived from very small numbers. The overall significantly negative association between sleep duration and weight for all boys, of R2 = −0.11 (P < 0.01) was low. No significant correlation was found for girls. Parental obesity, low parental education and income, long hours watching TV/playing videogames, and physical inactivity were also significantly associated with being overweight or obese.

Turning to the most recent investigations involving children, few provide actual data. The longitudinal study of 519, 7 years children, by Nixon et al. (67), mostly presented multivariable analyses. Whereas sleep durations <9 h had an OR of 3.3 for being overweight/obese, the significant increase in body fat in this group compared with those sleeping >9 h, comprised only a 3.3% difference. Although short sleep was also associated with higher emotional lability and potential attention deficit/hyperactivity disorder, sleep duration was seen as an independent risk factor for becoming obese/overweight.

The Canadian study by Touchette et al. (68) assessed 1138 children from 2.5 years to 6 years, classified under four sleep durations: short (<10 h) persistent (5.2%), short increasing (4.7%), 10 h persistent (50.7%), 11 h persistent (39.4%). After controlling for confounds, the OR for becoming overweight or obese in short persistent sleepers vs. 11 h persistent sleepers, was 4.2. However, there was a very large difference in sample sizes between these two groups (59 vs. 448, comprising 13 vs. 45 obese, respectively), and a highly significant (OR = 7.13) of likelihood of short sleepers overeating.

Bayer et al. (69) measured 7767 German children aged 3–10 years for BMI and % body fat, calculated from weight, height, skin-fold thickness. Although significant BMI differences for parent-derived sleep durations were found for all age groups, these averaged only 0.235 of a BMI unit across the age ranges.

The remaining most recent studies on children, all report significant effects, based on parental reports of sleep duration: Danielsen et al. (70) found short sleep to be associated with obesity but not with being overweight. Tian et al.'s (71) study of 619 obese and 617 non-obese Chinese pre-school children, reported those sleeping for ≤8 h (vs. 9–10 h) had a greater likelihood of having hyperglycemia. In Turkish children, Ozturk et al. (72) found BMIs to be higher in boys sleeping ≤8 h vs. those sleeping ≥10 h. However, this (72) encompassed a wide age range (6–17 years), with the authors recommending that ≥10 h sleep should be an obesity prevention measure for all these ages, which is too broad a generalization considering the age range (cf Fig. 2, here). Interestingly, Westerlund et al. (73) reported that sleep duration during school nights was associated with higher consumption of energy-rich foods (more so for boys). Bayer et al.'s (69) findings with 7767, 3–10 years old German children, provide no sleep-related weight data, only statistical models, and concluded that short sleep was associated with BMIs mostly within children of varying obesity (comprising about 4% of the total population). Bell and Zimmerman's (74) study of almost 2000 US children found that low night-time sleep duration at a mean age of 32 month was linked to a greater BMI (to an unknown extent) 5 years later, whereas this was not the case for an older cohort (103 month at baseline). However, no actual sleep or BMI data are given, only statistical outcomes. Weight and height at baseline came from parental reports and the sleep data were derived from only two random days (one weekday and one from the weekend). Rutters et al. (75) provide sleep duration and BMI findings for the five Tanner stages, from 98 Dutch children, measured annually from 12 years to 16 years. Sleep durations were obtained from one question relating only to weekdays, asked of the children themselves. Although the authors concluded that BMI during puberty was inversely related to changes in sleep duration, the extent of these changes is unclear.

The complexities involved in unravelling links between sleep and obesity in children is reflected by the discerning study by Kagamimori et al. (76), looking at lifestyle, social characteristics and obesity in a cohort of 9674, 3 years old Japanese children. Of those sleeping <10 h, 29% were obese. Adjustments for sex and various social variables showed the following to be related to obesity: irregular snacking, physical inactivity, expanded family with grandparents, mother as main caregiver but in full-time employment, attending a nursery and reduced sleeping hours. Another study of 1676 mother–infant pairs by Nevarez et al. (77) found during the first two years of life, that maternal depression, age at introduction of solids, attendance at child care outside the home, being black, Hispanic or Asian, all contributed to shorter sleep in infants. Benjamin et al.'s (78) investigation of adiposity in 1138 children (a non-sleep study), clearly pointed to the differential effects of daytime caregivers on adiposity, up to 3 years. Of course, for older children, TV exposure, especially to adult programmes, also has a marked effect on sleep (79).

Pluripotent peptides

Evidence of the phenotypic ‘give and take’ between wakefulness, sleep and exercise (locomotion), mentioned earlier, is reflected in the activities of several pluripotent neuropeptides, not only influencing sleep, but foraging, energy balance and locomotion. With other central nervous system mechanisms they facilitate the re-allocate sleep time to waking needs, without affecting sleepiness, or can institute the reverse process when demands on wakefulness are reduced. Although these findings largely emanate from research on non-human mammals, these time sharing mechanisms are also probably applicable to us, given that our largely accepted understanding of many other brain mechanisms underlying sleep and energy balance are derived from animals. Orexin-hypocretin (O-H) promotes wakefulness, locomotor activity, foraging and foraging, whereas melanin concentrating hormone promotes sleep and energy conservation (cf. (80,81)). O-H neurones, which are glucose-inhibited (melanin concentrating hormone neurones are glucose excited), project to central nervous system regions important in controlling feeding, sleep-wakefulness, neuroendocrine homeostasis and autonomic regulation (82). O-H activates waking-active monoaminergic and cholinergic neurones in the hypothalamus and brainstem regions to maintain long, consolidated waking periods (83,84). O-H neurones also have reciprocal links with the hypothalamic arcuate nucleus, which not only regulates feeding but has close connections with the suprachiasmatic nucleus controlling circadian feeding rhythms.

The responsiveness of O-H neurones to peripheral metabolic cues, such as leptin and glucose, further point to important links between energy homeostasis and vigilance states (84). O-H neurones are also linked to the dopaminergic reward system, interacting with mechanisms regulating emotion, reward and energy homeostasis to maintain appropriate vigilance states (84). In behavioural terms, during food shortages O-H facilitates extended wakefulness, greater locomotion but without loss of vigilance in order to forage efficiently (cf. (85,86)). Given that hypothalamic mechanisms are also strongly influenced by stress, sensory autonomic feedback, and by circulating hormones (87), this further points to the flexibility of sleep duration when there are competing demands on survival, without necessarily affecting sleepiness.

Added to these sleep systems are the recently identified food-entrainable circadian oscillators, presently found in rodents and probably evident in higher mammals, including humans (cf. (81)), that are linked to when daily food availability is at its most propitious (e.g. (88)), and the ability to anticipate this availability (89). Given that for humans, night-time sleep facilitates a physiological fasting state then, maybe, short sleepers, like ‘the early bird that catches the worm’, are more likely to be motivated to forage for food and eat after waking up.

Whilst all these mechanisms suggest that shortened sleep would encourage eating behaviour, maybe to excess if there was inadequate locomotion, and eventual obesity, again it is emphasized that any such sleep-related weight gains are very slow to develop. Nevertheless, the relatively unexplored link between sleep (especially REM sleep) and energy balance in mammals of normal body weight is tantalizing, given the other peculiar thermoregulatory and motor phenomena of REM sleep (90). I have argued elsewhere (90) that loss of the last period of REM sleep, as happens with short sleepers may especially promote food seeking behaviour (‘optimal foraging’), and although the extra locomotion/energy expenditure required to do so should counterbalance any increase in energy intake, in normal weight people, something may have gone awry, in some obese short sleepers


Although often statistically significant, any association between short sleep and obesity for the majority of people is unlikely to be causal, and may not be of particular clinical importance in relation to other factors associated with obesity, even for those habitually sleeping around 5 h, who only comprise a small minority of the adult population. Of course, someone who has inadequate sleep such that it produces excessive daytime sleepiness, is more likely to be physically inactive and have an energy intake in excess of expenditure. But the point, here, is that irrespective of this possible cause, such sleep has to be endured for years, with a huge accumulation of ‘lost’ sleep, even for a relatively minor weight gain. Moreover, apart from dietary adjustment, rectification of this gain can be accomplished by relatively short exposures to exercise, whereas little is known about the comparable efficacy of extending sleep.

No study has compared in obese short sleepers, the interventions of sleep extension vs. another active intervention, especially exercise. Although, Neilsen et al. (1) drew attention to an ongoing study (91), there is only a single active intervention, here, of sleep extension in such sleepers, implicating other potential and relevant lifestyle changes.

The great majority of obese people are not habitual short sleepers and are as likely to be sleeping to excess (27), which is a topic not dealt with, here. For short sleepers, excessive daytime sleepiness is likely to be of greater concern than is obesity. Moreover, those who are obese can not attribute their obesity, metabolic syndrome or other illnesses solely to inadequate sleep. Parsimony indicates that short sleep as well as obesity are surrogates for lifestyle, psychological, socioeconomic and waking pressures. Claims that <7 h sleep (especially habitual 6–7 h sleep) is associated with increased obesity are mostly based on two factors: (i) Generalizations from 5 h sleepers and/or when the subdivisions of sleep duration are unclear and (ii) Laboratory settings of acute, atypical (4 h) sleep restrictions on healthy young adults of normal weight, i.e. conditions that are ostensibly stressful and certainly involve excessive sleepiness. Various hormonal explanations derived from these latter settings, claimed to link short sleep with obesity and metabolic syndrome, remain to be established as part of the pathophysiology of obesity. Thus, this review has argued for the Null hypothesis: that the case for short sleep as a cause rather than just a correlate of obesity, has yet to be established.

Study limitations and conclusions with children are similar to those for adults, also involving loose definitions of sleep, often a simple dichotomy of sleep durations, with both factors leading to over-generalizations, and any such sleep effect also being very small. Short, inadequate sleep in children is usually symptomatic of problems that can not be overcome simply by increasing sleep duration.

Obesity has many facets, of course, as elegantly argued by Childers and Allison (92) very recently in their paper on the ‘obesity paradox’, and the unexpected finding that obesity is often associated with increased survival time among people who have some serious injury or illness. Whilst nothing is known about the extent to which this paradox might apply to short sleepers whose obesity is accompanied by comorbidities, and who do not experience excessive daytime sleepiness, for some of them, their sleep-related obesity might even be protective.

Whereas sleep is intimately involved in the regulation of energy balance, to ensure normal body weight, the extent that these underlying mechanisms might have gone awry in some obese, short sleepers is unknown. Yet, others still advocate (e.g. (93) p. 11), ‘that chronic partial sleep loss may increase the risk of obesity and weight gain, and that sleep restriction results in metabolic and endocrine alterations. . . . Altogether, the evidence points to a possible role of decreased sleep duration in the current epidemic of obesity. Bedtime extension in short sleepers should be explored as a novel behavioral intervention that may prevent weight gain or facilitate weight loss’.

Such views imply that obese people may have sleep to blame for their condition, rather than diet or insufficient exercise, and that more sleep in front of the TV is indeed a worthwhile therapeutic solution.

Conflict of Interest Statement

The author has neither financial interest in, nor financial support for writing this review.