Sleep as restitution: an introduction


Professor Torbjörn Åkerstedt, Institute of Psychosocial Medicine and Karolinska Institutet, PO Box 230, S-171 77 Stockholm, Sweden. (fax: +46 8 3320125; e-mail:


Abstract.  Åkerstedt T, Nilsson PM (Karolinska Institutet, Stockholm, University Hospital, Malmö, Sweden). Sleep as restitution: an introduction (Minisymposium). J Intern Med 2003; 254: 6–12.

This paper summarizes a symposium that has shown that sleep is a state of altered metabolism, and that disturbances and curtailments of sleep have far-reaching effects on endocrinology, immunology and metabolism, changes that may be linked to disease. As yet, the entire causal chain is weak but there are indications that, in particular, the risk of type 2 diabetes and cardiovascular disease may result from disturbed sleep. It is hypothesized that both insulin resistance and chronic low-grade inflammation may be involved.


Disturbed sleep has become prevalent in society [1–3] and the estimated costs in terms of sick days, treatment and other impacts on society are considerable [4, 5]. Yet, research interest has been very modest during most of the last century. The first physiological description of sleep was made in the 1930s [6] and the area received a considerable boost by the identification of rapid eye movement (REM) sleep in the 1950s [7] and by the dramatic performance impairment during total sleep deprivation [8]. Still, the interest was mainly limited to those engaged in psychiatry and psychology. The peculiar remitting effects of sleep deprivation on depression also seldom extended outside psychiatry [9]. Relatively recently, however, research results have started to describe sleep as the key factor in physiological restitution, with far-reaching medical implications. To discuss these developments an international symposium on Sleep as Restitution was organized in August 2002 by the Swedish Society of Medicine, the Royal Society of Medicine (UK) and the Karolinska Institute, Stockholm.

The main focus of the symposium was metabolism in a wide sense – the new findings that sleep not only ‘merely’ provides central nervous system (CNS) restitution but also seems to provide the same service for the entire physiological system. Four papers from the session on metabolism were selected for detailed presentation in this issue. The present paper serves as an introduction and framework, with highlights from the other sessions on related topics. First, however, a short summary of the basic characteristics of sleep physiology is presented.

Sleep characteristics

Sleep is defined from the combined impression from the electroencephalogram (EEG), the electrooculogram (EOG) and electromyogram (EMG). The resulting polysomnogram identifies the stages of sleep [10].

Thus, stage 1 shows 6–8 Hz EEG frequency and low amplitude, relatively high muscle tonus and often the presence of slow rolling eye movements. The recuperative value seems negligible.

Stage 2 is identified by the presence of so-called sleep spindles in the EEG (14–16 Hz short bursts) and occasional K-complexes against a background of 4–8 Hz activity. Muscle tonus falls further. This stage provides basic recovery and occupies 50% of the sleep period.

Stages 3 and 4, often grouped together under the label slow wave sleep (SWS), show large amounts of 0.5–4 Hz high amplitude waves (present 20% of the time for stage 3 and 50% of the time for stage 4). Muscle tonus decreases further. SWS is considered to represent the daily process of restitution, responds in a quantitative way to the time spent awake and shows a large increase of growth hormone secretion, together with a suppression of cortisol secretion. Metabolism falls with the sleep stages and SWS is characterized by slow breathing, low heart rate and low cerebral blood flow.

Rapid eye movement sleep is a completely different sleep stage characterized by an EEG similar to that of stage 1, but with REMs, a virtual absence of muscle tonus in antigravity muscles and a largely awake brain, particularly the hippocampus, amygdala and occipital projection areas. Interestingly, the prefrontal areas are not involved in this awakened brain. It is evident that we normally dream in REM sleep (although dream reports may be elicited from other stages), and to prevent the acting-out of dreams the efferent signals to the muscles are blocked.

From a metabolic point of view it is also of interest that metabolic rate (including heart rate, respiratory rate, blood pressure, body temperature, etc.) is increased above resting waking levels, whilst at the same time temperature regulation is inhibited – the dreamer no longer responds to changes in environmental temperature [11].

The normal development across time involves a rapid descent from waking to stage 4 sleep in 15–25 min, 30–50 min of SWS, followed by a short (5–10 min) period of REM sleep (Fig. 1). This cycle is repeated another three to four times during the night but with decreasing amounts of SWS and increasing amounts of REM sleep. The last two sleep cycles usually lack stages 3 and 4.

Sleep and mortality

The medical interest in sleep may have begun to increase with the systematic demonstrations of the lethal effects of sleep deprivation in rats [12]. The mechanism, however, is still unclear. A ‘vascular collapse’ was indicated, although some authors found an apparent inability of the immune system to protect against endotoxins [13].

At approximately the same time Lugaresi's group in Bologna demonstrated the first cases of fatal familial insomnia [14]. This is an autosomal dominant prion disease that involves degeneration of medio-dorsal and antero-ventral nuclei of the thalamus. These overlap with the medial thalamic areas, which are engaged in the synchronization of CNS neurones, particularly during SWS, that are the hallmark of sleep. Death ensues in about 7 months to 7 years and the intervening period is characterized by apathy, drowsiness, hallucinations, dream enactment (oneiric stupor), sympathetic overactivation, ataxia and spontaneous myoclonus. Remarkably, only stage 1 and REM remain.

Another development is the long-term follow-up of individuals with different amounts of habitual sleep [15]. Large (3–4 h) deviations from the median (8 h) in apparently healthy individuals yielded an increased mortality. Kripke et al. [16] reported similar results, even if the median with lowest risk turned out to be 7 h. This is supported by data from Nilsson et al. [17] demonstrating effects on cardiovascular events and total mortality in middle-aged males followed prospectively for 17 years.


Poor sleep is also prospectively associated with an increased prospective risk of myocardial infarction, particularly when combined with increased resting heart rate – a marker of sympathetic overactivity [17]. In women under rehabilitation from a myocardial infarction, the risk of recurrent myocardial events is increased in self-reported poor sleepers [18]. In addition, frequent events of waking-up exhausted in the morning are a predictor of subsequent myocardial infarction [19]. The exhausted state is also associated with reduced amounts of sleep stages 3 and 4 [20].

One new and interesting aspect of sleep loss is the impact on glucose metabolism and diabetes. Previous epidemiological studies have shown that patients with type 2 diabetes report more sleep problems than nondiabetic subjects [21]. This finding could be confounded by obesity or obstructive sleep apnoea (OSA) to some degree. However, in a prospective follow-up study of healthy middle-aged men from Malmö, Sweden, it was recently shown that the 12-year risk of developing type 2 diabetes was independently predicted by self-reported difficulties in falling asleep and by elevated resting heart rate, after full adjustment for obesity, lifestyle factors and other risk factors. What could be the mechanism behind this association? One candidate is OSA, which was not measured in the Malmö study, but another possibility would be chronic low-grade inflammation, both linked to insomnia and risk of type 2 diabetes.

Poor sleep is prospectively related to fatal accidents at work and accident risk is considerably increased in relation to irregular work hours [22]. In particular, as a consequence of sleep apnoea [23], the risk is especially evident in road transport [24], and several EEG studies of truck drivers, train drivers and process operators demonstrate marked intrusion of sleep in the EEG pattern of the actively working night worker [25–27]. It is highly probable that this also concerns on-call work in health care although there is a remarkable absence of accident analysis in this sector of work. However, it is well known that at least self-reported major medical mistakes are related to fatigue on call [28], and (simulated) medical procedures involve more mistakes or take a longer time to carry out [29, 30].

The medical consequences of insomnia, the inability to sleep per se, has been a focus of rather modest research despite the fact that it is a very prevalent medical problem [1]. In Europe around 38% suffer from insomnia DSM-IV symptoms and 6% fulfil the criteria for a diagnosis of insomnia. [31]. Excessive daytime sleepiness varies between 3.2 and 5.5%, whereas sleep apnoea varies between 1.1 and 1.9%. Generally, sleep problems increase with increasing age, female gender, stressful work and physical workload [1–3, 32].

A somewhat paradoxical finding in insomnia is its higher prevalence in women (as well as use of hypnotics) together with a lower risk for morbidity and mortality related to sleep problems [17]. This increased risk of negative biological consequences of sleep loss in men could be on account of their higher prevalence of, for example, sleep apnoea. Or, one might speculate that there is an evolutionary selection for sleep loss resilience in women, during pregnancy and early childhood years of offspring.


One of the major research areas that the symposium focused on was metabolism/endocrinology. Three key papers on this topic are presented in this volume.

The endocrine system is closely related to various aspects of sleep (Steiger, this volume). During sleep an interaction occurs between the electrophysiology and endocrinology. In adults the first part of sleep is characterized by increased growth hormone release (together with increased SWS and low levels of REM sleep) and suppressed secretion of the hormones of the hypothalamo-pituitary-adrenocortical (HPA) system, corticotropin (ACTH) and cortisol. During the second half of the night the HPA axis dominates and growth hormone (GH) secretion is essentially absent.

There are similarities between these changes during sleep and what is seen in connection with acute episodes of depression and with ageing. A corticotropin releasing hormone (CRH) antagonist will normalize sleep in depression. The pulsatile administration of certain peptides will affect sleep. Thus GH-releasing hormone (GHRH) causes increased SWS and GH and reduced cortisol in males (not females). CRH will exert the opposite effects and a somatostatin analogue will decrease SWS. Vasoactive intestinal polypeptide (VIP) decelerates the NREM–REM cycle. In older individuals GHRH is less effective but somatostatin retains its disturbing effects.

The quality of sleep appears to depend on an interaction of GHRH and CRH of sleep promotion, and changes in the GHRH/CRH balance in depression (CRH overactivity) and ageing seem to explain the reaction of sleep. In addition, somatostatin impairs sleep. Possibly, the differential response to GHRH in males and females can be related to the elevated risk for affective disorders in women. Elevated glucocorticoid levels may also contribute to the sleep EEG changes in depression and CRH antagonism normalizes sleep disturbances. One interesting implication of the gender differences in response to GHRH is whether this could be related to the increased risk of affective disorder in men and women.

It is of course well known that metabolic parameters such as rectal temperature, rate of breathing, heart rate, etc. are reduced during sleep. However, recent studies demonstrate that insulin and glucose levels are sensitive to manipulations of sleep [53]. In general, glucose levels during sleep are maintained at relatively normal levels and glucose infusion results in dramatically increased levels because the effects of insulin are impaired during sleep. Furthermore, sleep reduction down to 4 h for 6 days yields decreased glucose tolerance, increased evening cortisol, elevated sympatho-vagal balance, abnormal profiles of nocturnal growth hormone secretion and markedly decreased leptin levels, as well as a blunted response to influenza vaccination. In real life, short sleepers (<6 h) show results consistent with the experimental results – decreased insulin sensitivity, largely due to the increased GH secretion during sleep. Another observation in relation to experimental sleep reduction is that leptin levels are reduced and hunger markedly increased. The effects suggest links with the metabolic syndrome and may be related to (abdominal) obesity and poor lifestyle, often found in patients of lower socio-economic background.

The metabolic changes in relation to sleep pathology received little early interest. Sleep apnoea was seen more as a local respiratory abnormality than a ‘systemic illness’. It now appears that the proinflammatory cytokines interleukin (IL)-6 and tumour necrosis factor alpha (TNF)-α are elevated in patients with disorders of excessive daytime sleepiness (EDS), one consequence of OSA. There seems to be a correlation with body mass index – and both cytokines are released by fat tissue (Vgontzas, this volume). The same cytokines and leptin are increased in sleep apnoea, independent of obesity, as is insulin resistance. Sleep apnoea itself is linked to both obesity and diabetes and it appears that insulin resistance, related to visceral obesity, may be partly responsible for sleep apnoea. Possibly, sleep apnoea, in turn, may accelerate such metabolic changes through elevation of stress hormones and cytokines (cortisol, IL-6 and TNF-α). Recently, similar changes have also been presented in insomniac patients.

Another aspect of insomnia and metabolism has been introduced by Bonnet et al. (this volume), who demonstrated increased oxygen use in insomnia patients and simulated insomnia by administering 400 mg of caffeine per day for 1 week. The results show increased VO2, disturbed sleep, increased fatigue and anxiety, but decreased sleepiness, as measured by the multiple sleep latency tests. The results suggest that insomnia may be more of a metabolic disturbance than a disturbance of the sleep mechanism per se. The same author has also shown that chronic insomnia may not necessarily imply chronically disturbed sleep. Rather, sleep disturbances seem to occur only on days with increased tension (metabolism). Similar disturbances occur in normals, again in the presence of tension.

The metabolic changes seen after sleep curtailment, in insomniacs and sleep apnoeics, are similar to those seen in connection with stress [33], and stress is usually considered the main causal factor behind primary insomnia [34]. However, there is very little data to connect real life stress with polysomnographical indicators of disturbed sleep. Most studies have used rather innocuous and artificial stressors in a laboratory environment. Field studies of stress are virtually lacking, with some exceptions [35], except for epidemiological approaches [32, 36]. The latter study demonstrated that stress (high work load) affected (self-reported) sleep only if there was an inability to turn off worries about work during leisure time. At present there are data (Ekstedt M, Åkerstedt T, Söderström M, Berglund G, Axelsson J, Nilsson J, submitted for publication) that suggest metabolic changes under conditions of poor quality sleep.

The immune system

Relatively speaking, recent research has also demonstrated close connections between sleep and the immune system [37]. Proinflammatory cytokines of the early nonspecific immune response, such as monocyte/macrophage derived cytokines IL-1, TNF-α and interferons (IFNs) can facilitate sleep. Injection of IL-6 and IFN (bacterial and viral infection) a few hours before sleep suppresses both SWS and REM sleep in the early part of sleep with a rebound in the second half, as well as increased cortisol (probably via IFN), and increased fatigue. The improvement of late sleep may be due to the effects of later stages in the immune cascade – but not IL-2 (from T cells).

During normal sleep circulating cell counts for most major white blood cells decrease (monocytes, natural killer cells, T and B cells). The latter seem to accumulate in lymphoid tissue during sleep, facilitating local immune responses. However, sleep does not seem to affect the production of proinflammatory cytokines like IL-1, IL-6 or TNF-α; but IL-2 is markedly increased (compared with wakefulness) as is also IFN-μ (also derived from T cells).

Sleep also shifts the T helper (Th)1/Th2 balance towards more Th1 activity (which is blocked by sleep deprivation). Together with increased graft reactions during sleep the results suggest that there is an increased activity in the adaptive immune functions during sleep. Indeed, antigen response to vaccination against hepatitis A is blunted if sleep deprivation follows vaccination. The reason for the activating effects of sleep on the immune system may be connected to the high plasma levels of GH, prolactin and melatonin, which support T-cell derived immune functions. At the same time immunosuppressive hormones such as cortisol are suppressed.

In recent years it has been demonstrated that plasma concentrations of inflammatory markers, especially C-reactive protein (CRP) are predictors of subsequent myocardial infarction and stroke. CRP is a serum marker of the acute phase response in humans. The synthesis of CRP in the liver is controlled by IL-6, TNF-α and IL-2 – all three cytokines are related to sleep and fatigue. Apparently, IL-6 and TNF-α receptor 1, as well as CRP, increase with sleep loss and so do most subpopulations of white blood cells. Thus, sleep loss would seem to enhance inflammatory responses that could, in the long run, contribute to cardiovascular risk.

Brain metabolism

Several studies of brain metabolism have shown on the one hand that blood flow and metabolic activity are reduced in all areas of the brain during NREM sleep, and in particular in SWS. On the other hand, the transition to REM sleep is characterized by an increased blood flow in areas of the thalamus, hippocampus, amygdala and areas associated with the visual cortex. The frontal areas of the brain, however, remain quiescent [38].

Wakefulness is also associated with a breakdown of CNS adenosine triphosphate (ATP), leading to increased CNS levels of extracellular adenosine, which in turn is a potent sleep inducer [39]. This points to energy metabolism being closely related to sleep and wakefulness. Caffeine, of course, is a traditional antagonist of adenosine receptors.

Whether related to metabolism or not, sleep deprivation impairs psychomotor tasks requiring attention. However, recent work seems to indicate that higher ‘executive functioning’ is also very sensitive to sleep deprivation [40]. These are tasks that require handling of novelty, revising/updating plans, working memory, ignoring distraction and effective language communication.

A recent finding of major cognitive significance is the discovery of the role of REM in learning. Apparently, the learning of processes is strongly enhanced by REM sleep [41] and the same brain areas as were used in the learning process are activated in subsequent sleep [42]. It seems that REM sleep, which was discovered in 1953 [7], is finally being assigned an important biological role in brain functioning. Incidentally, the year 2003 marks the 50th anniversary of the discovery of REM sleep.

A new development in the CNS/sleep research is the identification of orexin or ‘hypocretin’, a peptide that was seen as a major regulator of appetite and metabolism, but turned out to be a very powerful wake(sleep) regulating agent [43]. Hypocretin cells project from the lateral hypothalamus to all monoaminergic cell groups and to many cholinergic neurones. It is altered in narcolepsy and hypersomnia.

Another substance that attracts attention presently is dopamine. It is not traditionally a sleep-related substance, but recent research suggests that emptying of dopamine vesicles in the CNS after, for example, amphetamine administration, will cause major sleep rebounds after the period of enhanced wakefulness [44]. However, inhibition of dopamine transporters, makes dopamine available without any marked loss in the vesicles. This will increase wakefulness without more than a marginal sleep rebound.

Sleep regulation

It should be emphasized that sleep/wakefulness and the processes involved are delicately regulated by homeostatic and circadian processes. Starting in the early 1960s the dependence of physiology on the circadian regulation was established in time-isolation units at the Max-Planck institute for Behavioral Physiology outside Munich. During a month of isolation from time cues the biological clock drove the sleep/wake pattern in a 25-h cycle [45]. The site of this clock is the suprachiasmatic nuclei of the hypothalamus [46].

To this was later added the homeostatic effect of the duration of waking time, increasing the need for sleep with increasing time awake [47]. Furthermore, any deviation from the 23.00 to 07.00 hour rhythm would disturb sleep and reduce its recuperative value [48]. This has implications for the present 24-h continuously active society [49]. At present, quantitative knowledge about sleep/wake regulation has been used to develop mathematical models for prediction of outcome of sleep attempts [50–52].

Final comments

The symposium brought together data to show that sleep is a state of altered metabolism and that disturbances and curtailments of sleep have far-reaching effects on endocrinology, immunology and metabolism, changes that may be linked to disease. As yet, the whole causal chain is weak but there are indications that, in particular, the risk of type 2 diabetes and cardiovascular disease may result from disturbed sleep. The whole chain of events involved have not, however, been described, but insulin resistance and chronic low-grade inflammation may be involved.

Conflict of interest statement

No conflict of interest was declared.


The work in this paper was supported by the Journal of Internal Medicine and by the Swedish research Council for Working life and Society.