Implications of Sleep Restriction and Recovery on Metabolic Outcomes

Abstract

Context:

Alongside the growing epidemics of obesity and diabetes mellitus, chronic partial sleep restriction is also increasingly common in modern society, and the metabolic implications of this have not been fully illustrated as yet. Whether recovery sleep is sufficient to offset these detriments is an area of ongoing research.

Objective:

This review seeks to summarize the relevant epidemiological and experimental data in the areas of altered metabolic consequences of both shortened sleep and subsequent recovery sleep.

Data Acquisition:

The medical literature from 1970 to March 2012 was reviewed for key articles.

Data Synthesis:

Epidemiological studies suggest associations between shortened sleep and future obesity and diabetes. Experimental data thus far show a probable link between shortened sleep and altered glucose metabolism as well as appetite dysregulation.

Conclusion:

Sleep often seems undervalued in modern society, but this may have widespread metabolic consequences as described in this review. Acute sleep loss is often unavoidable, but chronic sleep restriction ideally should not be. Understanding the implications of both sleep restriction and recovery on metabolic outcomes will guide public health policy and allow clinical recommendations to be prescribed.

Sleep is a fundamental biological requirement, and lack of sleep has increasingly recognized metabolic consequences. All mammalian species sleep, but precisely why we do so remains a mystery for which putative functions have been ascribed (1). Modern living increasingly devalues the importance of sleep, obtaining enough sleep, and planning to sleep. These fundamental changes in societal values are contributing to our modern, busy, “24/7” lifestyle and are leading to a widespread reduction in sleep duration in some countries (2). Available research has largely investigated the effects of total sleep deprivation (TSD) on metabolic outcomes rather than the impact of chronic partial sleep restriction, which is more relevant. Epidemiological studies of often self-reported chronic partial sleep restriction are available; however, few well-controlled, laboratory-based, experimental studies have been conducted (3). This review will summarize the available data regarding the metabolic consequences of chronic partial sleep restriction and the evidence regarding recovery sleep and will highlight gaps in our current knowledge. We will review data that show that adequate sleep is a necessity, not a luxury, due to the metabolic costs of insufficient sleep and that defining adequate sleep requires attention to individual differences. Sleep restriction confounded by circadian misalignment (4–6), including shift-work, seasonality (7), or illnesses such as obstructive sleep apnea are beyond the scope of this review and will not be addressed.

Epidemiology

Trends in sleep patterns

Population representative, national surveys from geographically diverse locations show that occupational demands increasingly curtail the sleep of both men and women (8–13). The American Time Use survey, conducted between 2003 and 2005, found that time at work, followed by time for travel, were the most apparent reciprocal influences on sleep time (9). The survey (9) also demonstrated that less sleep is observed in men compared with women, and in old and young individuals compared with the middle-aged. A recent U.S. National Sleep Foundation (NSF) survey found that 25% of the 1007 respondents between 25 and 60 yr of age had work schedules that did not allow for adequate sleep (8). Additionally, bedtimes were found to be earlier and sleep duration shorter on workdays. This pattern has also been observed in other populations (10, 11, 13). In parallel, “catch-up” recovery sleep during non-workdays has also been observed across many nations over the past decade (8–11). For example, data from Bartlett et al. (11) demonstrated that among Australian men who reported sleeping less than 6.5 h/night on weekdays, 25% subsequently slept more than 8 h/night on weekends, whereas 35% slept 6.5–8 h/night, and the remaining 40% continued to sleep less than 6.5 h/night over the weekend.

Leisure, as well as vocational activities, can also influence sleep duration. Regular internet/computer use for recreation or work in the hour before bed has increased from 12% in 2005 (9) to 28% in 2010 (8) in the United States. On the other hand, watching television remained essentially unchanged (69% in 2005 to 65% in 2010). This demonstrates the significant effect these societal influences can have at a population level: for instance, the increasing availability of the internet and scheduling of television programs. A recent policy statement about the use of modern media by the American Academy of Pediatrics warned against its overuse by children and adolescents and suggested that increased screen time may promote obesity and reduce the time available for sleep (14). Additionally, light exposure from these sources may influence circadian rhythm (15).

Seemingly nonmodifiable factors also influence sleep. Both sleep habits and evening activities show ethnic variation, and either may affect sleep; for example, in the 2010 U.S. NSF survey (8), African-Americans self-reported the least amount of weekday or weekend sleep compared with Asians, Hispanics, or Caucasians, whereas Asians were more likely to use a computer or access the internet in the hour before going to bed (52%) (8). However, ethnicity may not be directly responsible for these differences, particularly because the U.S. NSF survey did not adjust for socioeconomic status and this may influence both sleep duration and sleep quality (12, 16–18). On the other hand, another large (n = 33,000) cross-sectional study in the United States has shown that ethnic differences in sleep duration remain even after correction for socioeconomic factors (12).

In summary, employment and specific sedentary leisure activities before bed increasingly curtail sleep opportunities in modern society. Changes in work practices and societal values will be required to address this, but must be informed by rigorous epidemiological and clinical research.

Sleep duration and associated trends in mortality, obesity, and diabetes

The U-shaped relationship linking extremes of sleep duration to increased mortality was first recognized at least five decades ago (19) and has now been confirmed in many different populations (20–32). Mortality risk (33, 34) and cardiovascular outcomes (35) are probably lowest for individuals reporting sleep durations in the 7–8 h/night range. Although both extremes are associated with increased mortality, only part of the association with longer sleep can be explained by reverse causality, whereby serious and ultimately life-shortening illnesses cause some individuals to spend more time in bed (TIB). This is because both long and short sleep durations have also been associated with the future development of life-shortening metabolic diseases such as obesity, and serious illnesses do not generally result in weight gain.

Some have postulated that societal reductions in sleep duration may have contributed to worldwide increases in obesity and type 2 diabetes mellitus. A study by Kelly et al. (36) estimated that 23.2% of the global population was overweight and 9.8% obese, which equated to 937 million and 396 million people, respectively. The authors estimated that by the year 2030 there will be 2.16 billion overweight and 1.12 billion obese individuals worldwide. The healthcare implications are enormous. Being overweight with a body mass index (BMI) over 25 kg/m² is linearly associated with all-cause mortality (37), as well as comorbidities such as hypertension, dyslipidemia, heart disease, stroke, hepatobiliary disease, and several cancers (38–42). Visceral abdominal fat may be the more appropriate measurement because certain hormonal changes that may occur with sleep restriction, such as increases in cortisol, may favor accumulation of visceral abdominal fat. The relative utility of indirect measures of visceral abdominal fat such as waist circumference over measures of general obesity such as BMI in epidemiological studies of sleep restriction requires further evaluation.

A recent systematic review showed cross-sectional and longitudinal associations between short, habitual sleep duration and obesity in adults and children, but only a cross-sectional association between long habitual sleep duration and development of obesity in adults (43). However, four longitudinal (44–47) as well as three cross-sectional (48–50) studies published subsequent to that review confirmed that long, as well as short, habitual sleep are both associated with obesity in adults. In children, long sleep has not been associated with obesity in cross-sectional or longitudinal analyses, and this might be explained by their greater sleep requirements, so that biologically excessive sleep does not often occur in children. Indeed, systematic reviews have only consistently described an inversely linear relationship between short sleep and higher BMI in children (51).

Studies of both adults and children have generally relied upon self and parental report to determine sleep duration, rather than objectively measured sleep, which is more robust. However, objectively measured sleep can only feasibly be measured for periods of days to weeks, rather than months, and therefore may not be representative of longer-term sleep. Actigraphy is widely used to measure sleep in field studies, but it does not correspond exactly with sleep determined by polysomnography (52). Available studies have used actigraphy nested within established cohorts recruited for other purposes. In the first study, the CARDIA study of coronary artery risk development in young adults (53, 54), BMI was not an independent linear determinant of sleep duration (measured over a 72-h period) in a cross-sectional analysis of 669 adults (53). However, a cross-sectional reanalysis subsequently showed the converse, namely that increasing strata of sleep duration was an inverse determinate of BMI (54). The discrepant cross-sectional findings likely reflect three issues: the linear vs. categorized analyses conducted, the recategorization of sleep duration using combined baseline and follow-up data, and the covariates used for adjustment. Individuals were also reassessed after 5 yr for a further 72 h of actigraphy, and no longitudinal relationship between sleep duration and future obesity was found (54). Another study collected actigraphic data from two established cohorts examining osteoporotic fractures in 3052 women (the Study of Osteoporotic Fractures) and 3055 men (the Osteoporotic Fractures in Men Study) for four or five nights on average, respectively (55). Sleep disordered breathing was additionally verified by polysomnography in a subgroup of 2862 men and 455 women. This large study clearly showed that short sleep is cross-sectionally associated with increased obesity, and this relationship, albeit diminished, persisted even after adjustment for apnea-hypopnea index. Limitations of all these studies include the lack of findings in longitudinal analyses and the use of cohorts not specifically recruited to investigate sleep.

One notable study (56) examined both subjective and objective sleep duration but this time measured by polysomnography. The authors reported an association between increased BMI and shorter subjective sleep duration but no association with one night of polysomnographically measured sleep in 1300 men and women randomly selected specifically to assess sleep-disordered breathing. The authors concluded that the differences observed between objective and subjective measurements arose because an individual's perception of sleep duration is influenced by factors other than sleep, including sleep complaints and stress. Other explanations might be that a single night of polysomnography, particularly without an acclimatization night, may not be the same as longer-term sleep measured actigraphically, or that self-reported subjective sleep captures an element of “satisfaction” with sleep that may be important. Longitudinal studies examining all three of polysomnographically, actigraphically, and self-reported sleep simultaneously will be required to resolve these possibilities.

Longitudinal studies examining the relationship between baseline sleep and future development of obesity are summarized in Table 1. Eight (20, 44–47, 57–59) of the 10 studies identified reported an association between short sleep and future obesity. However, only one of the 10 studies assessed sleep with actigraphy (54), and, as mentioned above, no relationship with the development of obesity was detected. The remaining studies have generally only reported that sleep of less than 5 h/night on average is associated with future obesity. Nevertheless, the overwhelming conclusion is that short sleep is associated with future obesity.

Table 1.

Studies examining sleep duration and longitudinal risk of developing obesity

First author, year (Ref.)	Population	n (% men)	Baseline age (yr)	Follow-up (yr)	Sleep assessed by	Normal sleep reference (h)	Obesity definition	Short sleep reference (h)	Adjusted risk of future obesity	Average weight gain per yr follow-up in short sleepers	Longitudinal effect
Hasler, 2004 (57)	Zurich Cohort study, Switzerlanda	496 (50%)	27	13	Self-report at each timepoint	n/a	BMI ≥30 kg/m2	<6	OR (95% CI) = 8.2 (1.9, 36.3); P < 0.01	Not reported	Short sleep predicted weight gain
Gangwisch, 2005 (20)	NHANES I population study, United States	3,682 (32%)	32–49	8–10	Self-report at baseline	7	BMI ≥30 kg/m2	2–4	2–4 h, OR (95% CI), men = 2.51 (0.83, 7.53); women = 2.34 (1.24, 4.41)	Approximate ↑ BMI in 2–4 h = 0.16 kg/m2/yr (sd, 0.38)	Small and statistically insignificant effect of short sleep on weight gain
Patel, 2006 (58)	Nurses Health study, United States	68,183 (0%)	39–65	16 or to age 65	Self-report at baseline	7	BMI ≥ 30 kg/m2 or weight gain ≥15 kg	≤5	Obesity, HR (95% CI) = 1.15 (1.04, 1.27); weight gain, HR (95% CI) = 1.28 (1.15, 1.42)	↑ Weight gain in ≤5 h = 0.05 kg/yr (95% CI, 0.008, 0.09)	Short sleep predicted weight gain
Lopez-Garcia, 2008 (44)	Spanish population cohort, >60 yr of age	2,335 (46%)	Mean, 70.7 ± 7.2	2	Self-report at baseline	7	≥5 kg weight gain	≤5	Weight gain, OR (95% CI), men = 0.58 (0.20, 1.62); P = 0.34; women = 3.41 (1.34, 8.69); P = 0.02	No overall association between sleep duration and weight gain	Short and long sleep predicted weight gain in females only
Stranges, 2008 (115)	Whitehall II study, British white civil servants	4,378 (∼70%)	47–67	7	Self-report at each timepoint	7	BMI ≥30 kg/m2 or overweight (≥25 kg/m2 )	≤5	Obesity, OR (95% CI) = 1.05 (0.60, 1.82), P = ns; overweight, OR (95% CI) = 1.28 (0.80, 2.06), P = ns	Nonsignificant ↓ ΔBMI = 0.06 kg/m2/yr (95% CI, −0.26, 0.14; P = ns)	None
Chaput, 2008 (47)	Quebec Family Study, Canada	276 (42%)	21–64	6	Self-report at each timepoint	7–8	BMI ≥30 kg/m2	5–6	Obesity, ↑ by 27% (P < 0.05); 5 kg weight gain, ↑ by 35% (P < 0.05)	↑ Weight gain in 5–6 h compared to 7–8 h, 0.31 kg/yr (95% CI, 0.18–0.44); P < 0.05	Short and long sleep predicted weight gain
Lauderdale, 2009 (54)	CARDIA sleep study, Chicago, IL	612 (45%)	33–45	5	72-h actigraphy at baseline and follow-up	n/a	n/a	<4.5	Not reported	Approximate ↑ BMI in <4.5 h = 0.23 kg/m2/yr	None
Watanabe, 2010 (46)	Japanese electric power company employees	35,247 (89%)	Mean, 40 ± 9.6 (M), 38 ± 9.0 (F)	1	Self-report at each timepoint	7–8	BMI ≥25 kg/m2	<5	OR (95% CI), men = 1.91 (1.36, 2.67), P < 0.001; women = 0.35 (0.05, 2.69), P = 0.31	ΔBMI, men = ↑ 0.016 kg/m2/yr (95% CI 0.024, 0.146; P < 0.01); women = ↑ 0.010 kg/m2/yr (95% CI, −0.108, 0.194; P = 0.57)	Short and long sleep predicted weight gain in males only
Hairston, 2010 (45)	IRAS family study, 3 U.S. communities of African-Americans and Hispanics	1,107 (38%)	18–81 (mean 41.7)	5	Self-report at baseline	6–7	n/a	≤5	Not reported	↑ BMI in ≤5 h, <40 yr = 0.52 kg/m2/yr; P < 0.0001; >40 yr = 0.15 kg/m2/yr; P = 0.38	Short and long sleep predicted weight gain in <40 yr only
Nishiura, 2010 (59)	Japanese gas company employees	2,632 (100%)	40–59	4	Self-report at each timepoint	7–7.9	BMI ≥25 kg/m2	<6	OR (95% CI) = 2.46 (1.41, 4.31); P = 0.011	Not reported	Short sleep predicted obesity

IRAS, Insulin Resistance Atherosclerosis Study; NHANES, National Health and Nutrition Examination Survey; OR, odds ratio; CI, confidence interval; HR, hazard ratio; ns, not significant; n/a, not available; ↑, increase; ↓, decrease.

^aStudy population oversampled for risk cases for psychiatric disorders.

Longitudinal studies (Table 2) and a recent meta-analysis (60) also confirm a reproducible relationship between self-reported short sleep and the development of type 2 diabetes, even after adjustment for factors known to be associated with the incidence of type 2 diabetes, such as obesity (61–69). In the largest study of 174,542 men and women from a U.S. cohort, Xu et al. (69) showed that the incidence of diabetes was consistently increased in habitual, nocturnal short sleepers of less than 6 h/night compared with 7–8 h/night. Furthermore, gender appears to be important because the increased risk of diabetes with short sleep is more apparent in men, not women, even after adjusting for obesity (Table 2) (61–67). This suggests that other factors necessary for the development of diabetes are present, particularly in men. One factor could be sleep apnea, whose association with the male gender and diabetes is well established. Indeed, none of these studies corrected for this likely confounder, although two studies (62, 64) did assess snoring.

Table 2.

Studies examining sleep duration and longitudinal risk of developing type 2 diabetes mellitus

First author, year (Ref.)	Populationb	Gender	n	Baseline age (yr)	Follow-up (yr)	Short sleep definition (h)	Adjusted diabetes risk (95% CI)a	Unadjusted diabetes risk (95% CI)
Yaggi, 2006 (61)	Massachusetts Male Aging Studyb	Men	1564	40–70	15	<5	1.95 (0.95–4.01)	2.60 (1.28–5.27)
						6	1.95 (1.06–3.58)	1.93 (1.06–3.50)
Mallon, 2005 (62)	Swedish random population sampleb	Men	550	45–65	12	<5	2.8 (1.1–7.3)	3.1 (1.3–7.2)a
		Women	620	45–65	12	<5	1.8 (0.5–6.8)	1.5 (0.5–4.5)
Bjorkelund, 2005 (63)	Gothenburg population study in womenb	Women	1622	38–60	32	<6	1.35 (0.89–2.1)	1.22 (0.82–1.08)
Ayas, 2003 (64)	Nurses Health Study, United States	Women	70026	30–55	10	<5	1.18 (0.96–1.58)	1.57 (1.28–1.92)
Gangswisch, 2007 (82)	NHANES I population study, United Statesb	Both (37% men)	8992	32–86	10	<5	OR, 1.47 (1.03–2.09)	OR, 1.91 (1.37–2.67)
Chaput, 2009 (67)	Quebec Family Studyb	Both (42% men)	276	21–64	6	<6	2.42 (1.49–3.33)	Not described
Hayashino, 2007 (66)	Japan work site trial (HIPOP-OHP study)	Both (74% men)	6509	19–69	4.2	<6	HR, 1.15 (0.76–1.74)	HR, 1.19 (0.81–1.76)
Beihl, 2009 (68)	IRAS study, U.S. multiethnic community studyb	Both (43% men)	900	40–69	5	≤7	OR, 2.15 (1.21–3.79)c	Not described
Xu, 2010 (69)	NIH-AARP Diet and Health Cohort, United States	Both (57% men)	174 542	50–71	3–10	≤5	OR, 1.33 (1.19–1.49)	Not described

Data are expressed as relative risk, unless stated otherwise. Bold indicates significant data. OR, Odds ratio; HR, hazard ratio; CI, confidence intervals.

^aAdjusted for obesity and other variables.

^bRepresentative population.

^cIn non-Hispanic whites and Hispanics only, not African-Americans.

Metabolic Consequences of Sleep Restriction

Several decades ago, experiments in rodents (70–72) showed that chronic, inadequate sleep produced vivid changes in metabolism resulting in ill health alongside significant reductions in body weight, leading to major organ dysfunction and in some cases death. The nocturnal rodent is, however, much different from the human—physiologically, experimentally, and from a circadian clock perspective. Hence, whether this animal model translates to similar changes in humans requires direct evaluation. Such studies in humans, however, require intensive, experimental, in-laboratory studies to control for the many confounders of human lifestyle. To date, few such studies have been conducted, and these have used varying study designs, making it difficult to compare findings. Additionally, the degree to which the data collected can be extrapolated to a public health platform remains to be determined.

Insulin resistance

Several laboratory-based studies (Table 3) have examined the role of sleep restriction in normal, healthy sleepers and its relationship with glucose metabolism and insulin sensitivity. Changes in fasting as well as dynamic blood glucose and insulin concentrations reflect changes in glucose metabolism. These have provided convincing evidence that shortened or altered sleep impairs insulin sensitivity and glucose homeostasis in individuals without preexisting diabetes mellitus (73–77). Impaired insulin sensitivity occurs when a higher systemic level of insulin is required to maintain glucose homeostasis. Many of these experiments were performed in younger, healthy, predominantly male subjects in controlled laboratory conditions and showed significant impairments with sleep restriction in either or both of insulin sensitivity and glucose utilization. Those studies that used a dynamic test of insulin sensitivity, assessed after a glucose or meal challenge, have generally shown that sleep restriction impairs peripheral insulin sensitivity, meaning that the rate of glucose clearance peripherally, particularly by muscle, is reduced. These studies assessed insulin sensitivity by “gold standard” euglycemic hyperinsulinemic clamp or by minimal modeling of glucose or meal challenges after many different sleep restriction or disruption protocols and are summarized in Table 3 (upper). In particular, these data show that one (78), six (73), and 14 (74) nights of sleep restriction can each reduce peripheral insulin sensitivity by 20–30%, showing a rapid (1 d) and persistent (up to 14 d) effect (Table 3, upper). Furthermore, a single night of sleep restriction did impair hepatic as well as peripheral insulin sensitivity when gold standard assessment of endogenous (hepatic) glucose production was measured with infused glucose tracers during the euglycemic clamp (78). Impaired hepatic insulin sensitivity results in greater gluconeogenesis. This is in contrast with those studies that assessed hepatic insulin sensitivity from fasting glucose and insulin measurements (77, 79–81) and did not show any change with sleep restriction after one (81) or more (77, 79, 80) nights (Table 3, lower). The discrepancy in findings likely reflects the different methodologies used, particularly because assessment of insulin resistance from single fasting measurements is imprecise; however, one unanswered question is whether hepatic insulin sensitivity remains impaired in the longer term. Understanding the impact of sleep restriction on both peripheral and hepatic insulin sensitivity is important because both contribute to diabetes mellitus (73).

Table 3.

Studies examining sleep manipulation and insulin sensitivity by methodology

Study	No. and sex of subjects	Age, mean ± sd (yr)	BMI, mean ± sd (kg/m2)	Design	Comparator condition	Intervention	In-lab	Insulin sensitivity measurement	Effect of intervention on
									Insulin sensitivity	Other glucose parameters
Donga, 2010 (78)	5 M, 4 F	44.6 ± 14.7	23.8 ± 2.4	Randomized, 2 period, crossover	1 × 8.5-h TIB	1 × 4-h TIB SR	Yes	3-h hyperinsulinemic euglycemic clamp	↓ 25%	↓gluc disposal 20%, ↑gluc production 22%, ⇆ fasting gluc, fasting ins
Buxton, 2010 (83)	20 M	26.8 ± 5.2	23.3 ± 3.1	Fixed ordera	3 × 10- h TIB BLa	7 × 5-h TIB SRa	Yes	3-h hyperinsulinemic euglycemic clamp on BL 3, SR 7 MINMOD from 3 h ivGTT (insulin modified) on BL 2, SR 6	Clamp, ↓11%; ivGTT, ↓20%	↓disposition index, ↓gluc tolerance. ⇆ acute ins response, gluc effectiveness, fasting gluc, fasting ins. No effect of modafinil
Nedeltcheva, 2009 (74)	6 M, 5 F	39 ± 5	26.5 ± 1.5	Randomized, 2 period, crossover	14 × 8.5-h TIB	14 × 5.5-h TIB SR	Partial- <30 min/d outside	MINMOD from 3-h ivGTT on d 15 AUC from 3-h OGTT on d 14 (30-min sampling)	↓18%	↑gluc effectiveness, ↑2 h gluc, ↑gluc AUC. ⇆ fasting gluc, fasting ins, ins AUC, acute ins response, disposition index
Tasali, 2008 (75)	5 M, 4 F	Range, 20–31	Range, 19–24	Fixed order	2 × 8.5-h TIB BL	3 × 8.5-h TIB INT with SWS reduction by acoustic stimuli	Yes	MINMOD from 3 h ivGTT on BL 2 and INT 3	↓25%	↓disposition index 20%, ↓gluc tolerance 23%
Stamatakis, 2010 (76)	9 M, 2 F	23.2	24.3 ± 3.0	Fixed order	Habitual sleep at home (mean >7-h TIB week before study)	3 × habitual TIB INT with sleep fragmentation by acoustic stimuli on nights 2 and 3	Yes	MINMOD from 3-h ivGTT on day prior to BL sleep and on INT 2	↓25%	↓gluc effectiveness 21%, ⇆ disposition index
Spiegel, 1999 (73) and	11 M	22 ± 3.3	23.4 ± 1.7	Fixed order	6 × 12-h TIB REC	6 × 4-h TIB SR	Partial-day release on BL 1, SR 1–4, REC 1–4	MINMOD from 3-h ivGTT on SR 5 and REC 5	⇆	↓gluc clearance 40%, ↓gluc effectiveness 30%, ↓acute ins response 30%
Spiegel, 2004 (87)								HOMA-IR from 24-h sampling (10–30 min) on SR 6 and REC 6	↑ HOMA-IR 56%, post breakfast, ⇆ other meals	Response to identical meals showed ↑gluc and ↑ins production
Bosy-Westphal, 2008 (80)	14 F	27.5 ± 5.3	25.8 ± 5.8	Fixed order	2 × >8-h TIB BL	1 × 7-h, 2 × 6-h, 1 × 4-h TIB SR, then 2 × >8-h TIB REC	No	HOMA-IR on BL 2, SR 4, REC 2	⇆	⇆ fasting gluc, fasting ins, gluc AUC, insulin AUC
								Matsuda index from OGTT on BL 2, SR 4 (6 samples over 90 min)	⇆
Schmid, 2009 (81)	10 M	25.3 ± 4.4	23.8 ± 1.6	Randomized, 2 period, crossover	1 × 7-h TIB	1 × 4.5-h TIB SR	Yes	HOMA-IRb	⇆	⇆ fasting gluc, fasting ins, C-peptide
Zielinski, 2008 (77)	12 M, 28 F	60.4 ± 5.3	25.5 ± 3.2	Randomized, controlled, parallel group	56 × >8.5-h TIB (CON, n = 18)	56 × TIB ↓ by 1.5 h (INT, n = 24)	No	QUICKI	⇆	⇆ gluc, ins
Schmid, 2011 (111)	15 M	27.1 ± 5.0	22.9 ± 1.2	Randomized, 2 period, crossover	2 × 8.25-h TIB	2 × 4.25-h TIB SR	Partial-day release on d 1	AUC 3 h (60 min sampling) response to breakfast after night 2 of condition		⇆ fasting gluc, fasting ins, ↑peak gluc 11% + AUC, ↑peak ins 40% + AUC, ↑C-peptide AUC, ↓glucagon AUC. ⇆ during rest of day
van Leeuwen, 2010 (79)	23 M	23.1 ± 2.5	23.2 ± 2.6	Randomized, controlled, parallel group	10 × 8-h TIB (CON, n = 8)	2 × 8-h TIB BL, 5 × 4-h TIB SR, 3 × 8-h TIB REC (INT, n = 15)	Yes	Fasting morning sample on BL 2, SR 5, REC 2		⇆ gluc (but ↓ after REC cf BL), ↑ins , ↑ins:gluc ratio (returned to BL after REC)

M, Male; F, female; BL, baseline; SR, sleep restriction; REC, recovery; CON, control group; INT, intervention; gluc, glucose; ins, insulin; ins sens, insulin sensitivity; ivGTT, iv glucose tolerance test; OGTT, oral glucose tolerance test; MINMOD, minimal model analysis; QUICKI, quantitative insulin-sensitivity check index; HOMA-IR, homeostatic model assessment of insulin resistance; AUC, area under curve; ⇆, no change; ↑, increased; ↓, decreased; cf, compared to.

^aThen randomized, double-blind, placebo-controlled ± 300 mg modafinil/d during SR.

^b4-h hypoglycemic clamp also performed, no change in gluc infusion rate, ins concentration or C-peptide levels, ↓fasting glucagon, and ↓glucagon during clamp.

The mechanisms that underpin the metabolic effects of sleep are likely to be related to sleep architecture, specifically slow wave sleep (SWS), because selectively (75) or nonselectively (76) reducing SWS with repetitive auditory stimuli impairs peripheral insulin sensitivity. Arousals from sleep would be expected to be induced by this method in both studies, and hence, it is possible that arousals rather than altered sleep architecture are responsible for the changes in insulin sensitivity. However, in one study, SWS duration was correlated with insulin sensitivity (75), whereas neither study reported correlations with arousals (75, 76). Wakefulness per se is also unlikely to be a major contributing factor because wakefulness promoters such as modafinil do not alter peripheral insulin sensitivity (83). Exactly how reduced sleep or SWS impairs insulin sensitivity is not known, but hormonal mechanisms, in particular changes in appetitic hormones, have been postulated.

Appetite regulation—leptin

Changes in the hormones that control appetite, in particular leptin and ghrelin, can alter eating behavior and energy balance, leading to changes in weight over time (6, 84). Leptin is an important satiety signal (in addition to other hormones such as insulin and cholecystokinin), whereas ghrelin is the only hormonal hunger signal so far identified, and both are important for pancreatic insulin secretion (85). Leptin has been postulated as a possible mechanistic link between habitual sleep duration, obesity, and insulin resistance (48, 84, 86–92). Four cross-sectional epidemiological studies have examined the relationship between sleep duration, leptin, and obesity (48, 50, 93, 94). The two larger studies both relied on self-reported sleep duration (48, 50) and showed in the population-based Wisconsin Sleep Cohort Study (48) of 1024 individuals (54% male), as well as the community-based Quebec Family Study (50) of 740 individuals (44% male), that short sleep was associated with reduced blood leptin. Specifically, the larger Wisconsin study showed that 5-h sleep/night was associated with a 15.5% lower morning serum leptin level and 14.9% higher ghrelin level compared with 8-h sleep/night, independent of BMI and gender. The third study did not show any association between sleep duration measured actigraphically and morning leptin levels (relative to body fat percentage), but in contrast to the previous two reports, it studied a smaller (n = 80), predominantly female (75%), sleep-restricted and obese (mean BMI = 38 kg/m²) cohort (93). Although sleep was measured more precisely in this last study, it seems unlikely that this increased accuracy would have compensated for the 10-fold smaller sample size. In contrast, the fourth study showed increased leptin levels with decreasing sleep time, even after adjustment for obesity and degree of sleep apnea in 561 male and female subjects (94). Important distinguishing factors of this cohort vs. the other three cohorts are that sleep was assessed by a single night of polysomnography, and individuals were recruited by first identifying siblings with extremes in apnea-hypopnea index to examine the genetic basis of sleep phenotypes. Whether the increase in leptin observed in this study, in contrast with the other studies, is due to the objective method by which sleep was assessed, underlying differences among the populations studied, or to some other factor will require other large epidemiological cohorts to be established.

Longitudinal studies examining the effect of sleep duration on future leptin concentrations are not available, and hence, in-laboratory sleep manipulation studies are critical. Because leptin secretion varies diurnally, with lowest levels in the morning and highest in the first part of nighttime sleep (95–98), elucidating all the changes caused by modulating sleep time on circulating leptin requires assessment of full 24-h blood profiles; however, few such studies are available (Table 4). Furthermore, leptin secretion is modulated by obesity (98) and gender (99, 100), and both must be considered as potential confounders. It is known to be higher in females because it tends to be dependent on fat stores, and females generally have proportionally more adipose tissue compared with males (100). However, in more obese individuals, leptin resistance may occur, curtailing its actions; hence, a within-group study design is preferable to control for potential confounders. Indeed, changes in women may therefore be more subtle than in men. Both dietary intake (101) and exercise (102) can also acutely alter leptin and therefore also need to be controlled in experimental conditions.

Table 4.

Studies examining sleep restriction and leptin, ghrelin, hunger and cortisol

Study	No. and sex of subjects	Age, mean ± sd (yr)	BMI, mean ± sd (kg/m2)	Design	Comparator condition	Intervention	In-lab conditions	Caloric intake	Blood sampling regime	Effect of intervention on
										Leptin	Ghrelin	Hunger	Cortisol
Intensive sampling ≥24-h duration
Mullington, 2003 (90)	10 M	27.2 (22–37)	26.1 (20–34.5)	Fixed order	3 × 8-h TIB BL	88-h TSD, then 3 × 7-h or 14-h TIB REC	Yes	Standardized meals to maintain body weight, but optional to complete them	90 min over 120 h from 2100 h BL 3 to REC 1	↓ Diurnal amplitude, then ↑ in REC cf BL
Guilleminault, 2003 (92)	8 M	19.6 (18–25)	22.9 ± 2.0	Fixed order for sleep variablesa	2 × 8.5-h TIB REC	7 × 4-h TIB SRa	Partial- day release	Standardized meals at 0800, 1200, 1800 h, glucide rich on blood testing days	5 samples over 24 h on SR 7 and REC 2 (0800, 1200, 1800, 2200, 0215 h)	↓ Peak and rhythm amplitude
Spiegel, 2004 (87)	11 M	22 ± 3.3	23.4 ± 1.7	Fixed order	7 × 12-h TIB REC	6 × 4-h TIB SR	Yes, bed rest for last 60 h of condition	Identical between conditions. Day 1 of testing = ivGTT in morning, then identical CBH-rich meals (30 kcal/kg) at 1400, 1900 h, then on d 2 at 0900, 1400, 1900 h	10–30 min over 24 h during last 24 h of SR and REC	↓ Mean 19%, ↓ rhythm amplitude 20%, ↓ acrophase 26% and earlier (2 h)			⇆ Mean, curve shape changed
Nedeltcheva, 2009 (88 and 74)	5 F, 6 M	39 ± 5	26.5 ± 1.5	Randomized, 2 period, crossover	14 × 8.5-h TIB	14 × 5.5-h TIB SR	Partial, <30 min/d outside. Bed rest 48 h pre/post condition	Ad libitum, but intake identical during bed rest periods	30 min over 24 h before BL and 24 h before INT	⇆	⇆	↑ Calories (snacks)	⇆ AUC, curve shape changed
Pejovic, 2010 (95)	11 F, 10 M	23.5 ± 3.3 (nap), 24.6 ± 2.9 (no nap)	23.2 ± 2.8 (nap), 25.0 ± 2.1 (no nap)	Fixed order ± randomized nap	4 × 8-h TIB	1 × TSD ± 2-h nap in pm after TSD, then 2 × REC nights	Yes	Ad libitum, consistent portion sizes, but free choice from menu	30 min over 24 h on BL 4 and after TSD	↑ Mean. Flattening of diurnal rhythm after TSD. No effect of nap.		⇆ Hunger	⇆
Nedeltcheva, 2010 (107)	3 F, 7 M	41 ± 5	27.4 ± 2	Randomized, 2 period, crossover	14 × 8.5-h TIB	14 × 5.5-h TIB	Yes, bed rest 48 h pre/post condition	Individualized, weight loss diet; 90% of RMR. Same in both conditions	30 min over 24 h before BL and 24 h before INT	⇆	↑	↑ Hunger	⇆
Sampling <24-h duration
Spiegel, 2004 (84)	12 M	22 ± 2.0	23.6 ± 2.0	Randomized, 2 period, crossover	2 × 10-h TIB	2 × 4-h TIB SR	Partial- day release	Standard meals at 1900 and 0800 h, ad libitum at home during d 1. Continuous iv gluc infusion 5 g/kg on d 2	20 min from 0800–2100 h after second night of condition	↓ Mean 18%	↑ 28%	↑ Hunger 24%
Schmid, 2008 (104)	9 M	24.2 ± 3	23.8 ± 1.8	Randomized, 3 period, crossover	1 × 7-h TIB	1 × 4.5-h TIB SR/1 × TSD	Yes	Fasted	Fasting morning sample (0700 and 0730 h, results averaged)	⇆	↑ in TSD	↑ Hunger in TSD
Schmid, 2009 (106)	15 M	27.1 ± 5.0	22.9 ± 1.2	Randomized, 2 period, crossover	2 × 8.25-h TIB	2 × 4.25-h TIB SR	Partial-day release day 1	Standard breakfast, then ad libitum	Hourly from 0800 to 2300 h on d 2 of condition	⇆	⇆	⇆ Hunger, ↓ activity, ⇆ intake
Schmid, 2011 (111) and 2009 (81)	10 M	25.3 ± 4.4	23.8 ± 1.6	Randomized, 2 period, crossover	1 × 7-h TIB	1 × 4.5-h TIB SR	Yes	Fasted on morning of hypoglycemic clamp	30 min during 4 h hypoglycemic clamp			⇆ Basal	↓ Basal, ⇆ peak
Magee, 2009 (103)	10 M	20.4 ± 1.6	Not stated	Fixed order	1 × 8-h TIB BL	2 × 5-h TIB SR, then 1 × 8–10-h TIB REC	Partial-day release, REC night at home	Standardized breakfast and evening meal	Fasting daily morning samples, except SR 1	⇆	⇆	↓ Satiety
van Leeuwen, 2010 (79)	23 M	23.1 ± 2.5	23.2 ± 2.6	Randomized, controlled, parallel group	10 × 8-h TIB (CON n = 8)	2 × 8-h TIB BL, 5 × 4-h TIB SR, 3 × 8-h TIB REC (INT, n = 15)	Yes	Identical daily calories, except 50 kcal/d more in SR	Morning sample (0730 h) on BL 2, SR 5, REC 2	↑		⇆	∼ Peak delayed by 16 min
Simpson, 2010 (89)	67 F, 69 M	30.4	24.7	Randomized, controlled, parallel groupb	2 × 10-h TIB BL	5 × 4-h TIB SR, subset randomized to further 2 × SR (0, 2, 4-h TIB, n = 27) or 2 × REC (6, 8, 10-h TIB, n = 37)	Yes	3 meals plus ad libitum snacks	Fasting morning sample (between 1030 and 1200 h) on BL 2, SR 5, and FSR 2 or REC 2	↑, more in F. With FSR, further ↑ in F and M, and ↓ with REC
Donga, 2010 (78)	5 M, 4 F	44.6 ± 14.7	23.8 ± 2.4	Randomized, 2 period, crossover	1 × 8.5-h TIB	1 × 4-h TIB SR	Yes	Fasted from 2200 h	3-h hyperinsulinemic euglycemic clamp				⇆
Omisade, 2010 (105)	15 F	21.6 ± 2.2	24.5 ± 8.1	Fixed order	1 × 10-h TIB	1 × 3-h TIB SR	Partial, 1 h/d outside	Same intake for each condition	Salivary samples at 0830 and 2000 h	↑ in am, not in pm		⇆	↓ am, ↑ pmc
Bosy-Westphal, 2008 (80)	14 F	27.5 ± 5.3	25.8 ± 5.8	Fixed order	2 × >8-h TIB BL	1 × 7-h, 2 × 6-h, 1 × 4-h TIB SR, then 2 × >8-h TIB REC	No	Ad libitum	With OGTT on BL 2, SR 4 (0, 15, 30, 45, 90 min). Basal levels on REC 2.	↑ Basal level, 45, 90 min; ↑ L/fat mass 29%	⇆	⇆ Hunger, ↑ intake	⇆

M, Male; F, female; BL, baseline; SR, sleep restriction; REC, recovery; FSR, further sleep restriction; CON, control group; INT, intervention; CBH, carbohydrate; AUC, area under curve; ivGTT, iv glucose tolerance test; gluc, glucose; RMR, resting metabolic rate; ⇆, no change; ↑, increased; ↓, decreased.

^aRandomized to early (from 2230 h) or late (from 0215 h) SR.

^bControl group of n = 9 with 10-h TIB throughout. No changes in leptin levels.

^cSeven cortisol samples between 0830 and 2100 h, approximately 10 salivary cortisol samples during the day.

Eight studies sampled leptin for less than 24 h, and of these, four assessed leptin by a single blood draw (79, 89, 103, 104); one (80) after oral ingestion of glucose, which may have confounded results; and another (105) examined two salivary (but no blood) samples for leptin (Table 4). Of these six studies, four report an increase in leptin with sleep restriction (79, 80, 89, 105), and the other two report no change (103, 104). Two of the four reporting an increase were solely in female populations (80, 105) and another was in a mixed gender population (89) (Table 4). The two remaining studies not yet discussed were both performed solely in young men, with one reporting that sleep restriction reduced leptin (84), whereas the other found no change (106). Interestingly, sleep restriction was found to reduce physical activity in the latter study (106), and this may have affected the leptin results.

Studies that assess 24-h leptin profiles exclusively in men consistently report that sleep restriction reduces diurnal leptin amplitude (87, 90, 92) and mean leptin levels (87), which should increase appetite, although appetite was not directly measured (Table 4). Two other 24-h sampling studies, which have not separated analyses according to gender, report no change in leptin (88, 107). The lack of decrease in leptin may have arisen putatively because of the small number of men examined [n = 6 (88) and n = 7 (107)]. The final 24-h sampling study reported an increase in leptin (95). The authors of the latter study postulate that the increase, not decrease, in leptin was due to a less stressful study protocol. The authors also found that gender did not modulate the effect of napping on leptin. However, it is not specified whether gender modulation of the effect of sleep restriction on leptin was examined for, and furthermore, the actual data showing what the effects of gender on this or any other relationship might be are not presented. As described in Table 4, laboratory study designs were significantly different in these studies, giving rise to varying factors that might allow for activation of the hypothalamic-pituitary axis and, as we will go on to discuss below, interaction between the effects of cortisol, leptin, and sleep. Nevertheless, it remains plausible that gender may modulate the discrepancies among studies, particularly because women have higher baseline levels than men.

In conclusion, there are discrepancies in the findings regarding the effect of sleep restriction on leptin. However, studies performed exclusively in men that fully account for leptin secretion across an entire 24-h period all consistently show that sleep restriction reduces diurnal leptin amplitude (87, 90, 92), although only one showed a decrease in mean leptin levels (87). A modulatory effect of gender is suggested; however, definitive conclusions cannot be drawn due to the differences in study designs.

Appetite regulation—ghrelin, hunger, and cortisol

Ghrelin is another important appetite-modulating hormone and is the only hormonal signal known to stimulate appetite. It is important for energy homeostasis and is suppressed by eating. There have been fewer studies examining the effect of sleep restriction on this appetite regulator, and these have reported either an increase (84, 104, 107) or no change (80, 88, 103, 106) with sleep restriction (Table 4).

Methodology is critically important in studies examining energy balance because changes in food intake and level of exercise may impact on, or alternatively be a consequence of, the sleep condition. Studies that maintained identical food intake and controlled exercise objectively by measurement (e.g. by actigraphy) or by restriction (e.g. with bed rest) are better designed to dissect the effect of sleep restriction on hunger and appetitic hormones. However, few studies have controlled for these important variables, and this has contributed to inconsistent conclusions. The most sophisticated studies thus far have randomized the order of condition and controlled food intake for up to 48 h before and at the time of blood testing (88, 107) or indeed replaced food intake with a continuous glucose infusion (84). This procedure ensures that food intake is controlled both at the time of and in the days before testing.

Spiegel et al. (84) showed that after two nights of sleep restriction, daytime ghrelin levels were in opposite phase to leptin and increased in parallel with hunger and appetite, in a mostly controlled energy intake environment. Energy expenditure was not assessed in this study. In a carefully controlled experiment, Nedeltcheva et al. (88) showed that sleep restriction increased both snacking and hunger without changing overall caloric intake (when controlled for baseline weight) or energy expenditure, measured by indirect calorimetry. This study did not show any change in leptin or ghrelin; however, weight increased equivalently in both groups. Subjects were overweight and included both genders, and these factors may have confounded changes, as previously discussed. Increases in food intake after one night of sleep restriction have also been observed by Brondel et al. (108) who examined 12 male subjects and found that subjects were more hungry and thus ate 22% more calories on the day after 4-h TIB compared with 8-h TIB under ad libitum food conditions; the differences were seen at breakfast and dinner, rather than lunch. Subjects were also more active after sleep restriction, despite feeling sleepier. This contradicts other work that found activity to be lower (106) or unchanged (80) with sleep loss. Schmid et al. (104, 106) have performed two experiments with conflicting results. First, after a single night of 7-h TIB vs. 4.5-h TIB vs. TSD, fasting ghrelin and hunger ratings were found to increase in TSD compared with 7-h TIB; however, this protocol only used a single fasting morning sample (104). In a further experiment using a more intensive sampling regimen over 15 daytime hours under ad libitum food intake conditions, no change in hunger, ghrelin, or energy intake was seen (106). So whether sleep itself impacts on ghrelin and hunger independently of food intake and energy expenditure remains unclear because other authors have also found no change in either or both of those with sleep restriction (79, 80, 103, 105).

Whether sleep quantity should therefore be routinely assessed when prescribing a weight-control program is a novel proposition. One study examined the effect of sleep on diet-induced weight loss (107) by prescribing a moderate hypocaloric diet during a randomized, two-period, crossover study in a mixed gender cohort of 10 overweight subjects exposed to 14 d of either 5.5-h or 8.5-h TIB. They found that proportionally less fat mass was lost in the short sleep group (0.6 vs. 1.4 kg), without an overall change in total weight loss (3.0 vs. 2.9 kg). Both ghrelin and hunger increased with sleep loss, but leptin was unchanged, suggesting that ghrelin, not leptin, was the metabolic hormone responsible for the body compositional changes. This study (107), in combination with a previous study performed in non-negative energy balance from the same group (88), shows that ghrelin levels increase only in the presence of negative energy balance. A randomized, prospective, controlled trial (109) is currently under way examining the feasibility of treating obesity with prescribed and objectively monitored sleep extension by nonpharmacological means in 150 obese short sleepers. The results of this shall be eagerly awaited because longitudinal changes in the appetitic hormones will also be assessed.

Cortisol is the final hormone to be implicated by some, but not all, studies in the sleep-dependent neuroregulation of energy balance and glucose homeostasis. Cortisol itself is secreted diurnally in a phase that is opposite to that of leptin (110). Spiegel et al. (87) demonstrated that six nights of sleep restriction altered the wave shape of the diurnal variation, with a slower decline from acrophase to nadir, higher levels in the late afternoon and evening, and a later nadir by 1.5 h, but did not alter 24-h mean blood concentrations. One study in women (105) found a similar slowing of the decline in daytime salivary cortisol as well as higher levels in the afternoon and evening, but in addition reported lower morning levels with sleep loss. Another study showed that peak salivary cortisol was delayed slightly in sleep loss; however, it did not report overall mean levels (79). Schmid et al. (81) showed a decrease in cortisol and ACTH levels at baseline and during the start of a hypoglycemic clamp after one night of sleep restriction. The impact of these changes in cortisol dynamics on food intake is not clear. On the other hand, six other studies have found no change in frequently sampled blood cortisol levels with sleep loss (74, 78, 80, 95, 107, 111). Because salivary cortisol measurements primarily estimate an integrated unbound cortisol and may not be directly comparable with blood cortisol measurements, which primarily reflect bound cortisol, these discrepancies may reflect the measurements used as well as differences in study design (87, 95). These differences in study design may have also impacted upon stress and sympathetic balance, thereby confounding changes in cortisol (95).

Regulatory interaction between metabolic control centers and food habits make food intake, hunger, and energy expenditure important co-determinates that need to be controlled for. As we have discussed, differences in findings likely reflect modulation of the effects of sleep restriction by gender, obesity, and the magnitude and cumulative duration of restricted sleep. Differences in food intake may also contribute to changes in leptin; hence, the studies that allow ad libitum food intake may counteract effects of sleep loss alone. Circadian variation also needs to be considered and controlled for through 24-h sampling. Furthermore, the phase where the sleep restriction has been placed (in the early or latter part of the night) may also contribute to differences in results, especially when many of these studies were performed in small numbers. Seasonality may also be relevant. Larger, intensive, controlled studies are required.

Metabolic Recovery from Chronic Sleep Restriction

Much of our current knowledge regarding recovery from sleep loss has been derived from TSD studies (112, 113). The extent and amount of recovery sleep required to reverse the metabolic consequences after sleep restriction is an area of evolving work and discussion. To date, there has been no systematic examination of the metabolic time course of recovery from chronic sleep restriction. When, or if, recovery to baseline or in comparison to a control group occurs or if there is a critical point in the chronicity of detrimental changes, are important areas to investigate.

One of the few studies to examine sleep extension after a period of sleep restriction was by Spiegel et al. (73) (Table 3). Eleven healthy men underwent six nights of sleep restriction to 4-h TIB, and then seven recovery/sleep extension nights of 12-h TIB. Impairment of carbohydrate tolerance was found with sleep loss in a glucose tolerance test, with the rate of glucose clearance 40% of the recovery levels. Intensive blood sampling was performed for 24 h at the end of the restriction and recovery sections. Significant differences were seen between sleep restriction and recovery in post-breakfast glucose and homeostatic model assessment of insulin resistance (elevated by 56% in sleep restriction), however not in absolute insulin levels and not at other mealtimes (87). Sleep was measured during the last two nights of each section, and it seemed that the subjects had recovered their sleep debt because the sleep efficiency by d 5 of recovery was 76%. Baseline testing was not performed at the time, however, due to the intensive blood sampling regimen. This would have provided insight as to whether the improvements seen after five nights of recovery had truly returned the effects seen during sleep restriction to baseline or whether subtle changes remained. It is important to determine how plastic the metabolic control system is and whether sleep restriction does indeed contribute to the long-term development of insulin resistance and type 2 diabetes mellitus.

The study of van Leeuwen et al. (79) (Table 3) examined five nights of 4-h TIB sleep restriction followed by two nights of recovery sleep of 8-h TIB in healthy males. Only a single morning fasting blood sample was taken at baseline, after sleep restriction, and after recovery. Glucose levels did not significantly change after sleep restriction; however, both insulin and the glucose/insulin ratio increased. After recovery, glucose reduced significantly compared with baseline, and insulin decreased toward baseline. Leptin was significantly increased with sleep restriction and remained elevated after recovery. These data show that recovery sleep may repair glucose metabolism, but more data are needed from dynamic tests of insulin sensitivity to draw more specific conclusions. Another study by Bosy-Westphal et al. (80) examined the effect of two nights of over 8-h recovery sleep after four nights of increasing sleep restriction (7, 6, 6, and 4 h/night). No significant changes in glucose or insulin were seen after sleep restriction, so it was not possible to determine any recovery effects. However, sleep restriction did significantly increase leptin, which did not significantly decrease to baseline during the recovery period. As discussed previously, many variables may alter the effects of leptin, and so these two studies alone with single timepoint sampling are insufficient to draw conclusions from.

Napping is a common practice to counteract sleepiness caused by sleep restriction. Napping has the potential to recover lost sleep and offset metabolic changes. Very few studies have examined the metabolic outcomes from napping behavior; however, one study by Pejovic et al. (95) examined the effect of a 2-h midafternoon nap (from 1400 to 1600 h) after one night of TSD in a mixed gender cohort with an intensive blood sampling regimen (Table 4). They found that daytime napping in this setting did not influence the effects of sleep loss on leptin or hunger. This complements previous work by the same group, using the same study design (114), which showed that cortisol levels decrease during napping and increase upon waking, possibly due to the stress of awakening. Lower levels of cortisol were associated with sleepiness, and IL-6 also decreased upon awakening, leading the authors to postulate that the interaction among these factors may be responsible for fatigue and sleepiness.

In conclusion, these few studies together are not able to exactly determine when recovery occurs and to what degree, but they do illustrate that changes occur with increases in sleep after sleep restriction. More systematic experimentation is needed to properly understand how recovery sleep, including naps, may improve metabolic processes.

Conclusions

Chronic sleep restriction has well-established effects on cognitive performance, but its impact on all aspects of metabolism has been less thoroughly explored, although the evidence that sleep curtailment is detrimental to metabolic health is mounting. Appetitic hormones, such as leptin, are implicated in the metabolic effects of sleep restriction, but the data are somewhat contradictory due to varying study designs and possible gender differences. Preliminary evidence highlights the possibility that recovery sleep can improve glucose homeostasis and insulin action, but current studies do not show recovery of leptin with recovery sleep. Further dose-response studies are required. Acute sleep loss is often unavoidable, but chronic sleep restriction ideally should not be. Understanding the implications of both sleep restriction and recovery on metabolic outcomes will guide public health policy and allow clinical recommendations to be prescribed.