Evaluating routine blood and cerebrospinal fluid samples in narcolepsy patients.

Cerebrospinal fluid hypocretin-1 is proven to be a precise diagnostic marker of narcolepsy Type 1 (NT1). However other characteristics of cerebrospinal fluid and blood parameters have not yet been described. The objective of this study was to evaluate the differences in routine blood and cerebrospinal fluid analyses between NT1 patients and patients suspected of hypersomnia. We collected retrospectively all measures of cerebrospinal fluid hypocretin-1 between 2019 and 2022. This yielded 612 patients out of which 146 were diagnosed with NT1 and the rest (466 patients) were used as a control group. We selected the most relevant routine samples from both blood, plasma and cerebrospinal fluid and compared the two groups. The only significantly different analytes were plasma lactate dehydrogenase and cerebrospinal fluid hypocretin-1. No other differences were found between the groups including thyroid markers, markers of neuroendocrine function, inflammatory markers in blood or cerebrospinal fluid, markers of permeability of the blood brain barrier or metabolic markers in blood samples. We found no significant differences in routine blood or cerebrospinal fluid components, neuroendocrine function, neuroinflammation and metabolic markers. The results reflect that the hypocretin system does not seem to play a chronic major role in regulation of these markers. None of the parameters routinely measured in blood in these patients could differentiate between NT1 and non-NT1 disorders besides CSF-hcrt-1.

Sleep disorders and sleep disturbances in persons with multiple sclerosis: A population-based matched case-control study in Denmark.

Introduction Adverse sleep is common in multiple sclerosis (MS). Population-based studies including adequate control groups are lacking. We hypothesized that the prevalence of sleep disorders and other sleep disturbances would be higher in persons with MS than in controls. Methods We conducted a population-based study linking individual-level data from the Danish MS Registry (n=21,943 persons with MS) and the Danish Population Registry (n=109,715 matched controls) with information on sleep disorders from the Danish National Patient Registry and other sleep disturbances assessed by dispensed prescription drugs from the Danish National Prescription Registry. Results Prevalence of diagnosed sleep disorders in terms of central hypersomnia (0.15% vs. 0.06%), sleep disturbances (1.05% vs. 0.70%), and sleep movements (0.22% vs. 0.13%) and other sleep disturbances identified by dispensed central acting (10.73% vs. 1.10%) and hypnotic use (30.65% vs. 20.13%) medication was statistically significantly higher among persons with MS when compared to controls. We found no statistically significant difference in the prevalence of sleep apnea and parasomnia between groups. Stratified by sex and age at MS diagnosis, results for differences between persons with MS and controls were similar. Conclusion In this registry-based study, we found that the prevalence of several diagnosed sleep disorders was higher persons with MS than in controls, that is those reflecting insomnia and daytime symptoms including hypersomnia. Other sleep disturbances identified by dispensed prescription medication was markedly higher in persons with MS than controls.

The Role of Obstructive Sleep Apnea in Vision-Threatening Diabetic Retinopathy-A National Register-Based Study.

INTRODUCTION: we investigated the association between OSA and vision-threatening diabetic retinopathy (VTDR). METHODS: we used three nationwide registers to identify subjects with and without OSA and patients with type 1 (T1DM) or type 2 diabetes mellitus (T2DM). The Danish Civil Registration System was used to link OSA with diabetes diagnosis. The primary outcome was the occurrence of VTDR in diabetic patients with and without OSA. The secondary outcome was the prevalence of diabetes mellitus in patients with and without OSA. RESULTS: we included 532,828 diabetic subjects comprising 13,279 patients with OSA (2.5%) and 519,549 without OSA (97.5%). Diabetic patients with OSA had a 57% lower risk of VTDR compared to diabetic patients without OSA (OR 0.43, 95% CI 0.38-0.50, p < 0.0001). CONCLUSIONS: our findings indicate that OSA is associated with a lower risk of VTDR. Since we did not adjust our outcomes for diabetes duration, hypertension control and hemoglobin A1c, future studies are needed to confirm our findings.

Usage and impact of patient-reported outcomes in epilepsy.

BACKGROUND: The use of patient-reported outcomes (PRO) in clinical practice is gaining increasing attention. This study aimed to provide a critical assessment of the current state-of-the-art and beliefs about the use of PRO in the management of people with epilepsy across some European countries. METHODS: Structured interviews were conducted with European experts to collect insights about (I) the personal experience with PRO; (II) the value and impact of PRO in the decision-making process at the national level; and (III) the interest for and use of PRO by national health authorities. RESULTS: Nine neurologists (Austria, Belgium, Czechia, Denmark, France, Greece, Italy, Poland, and United Kingdom), three health economists (Portugal, Romania, and Sweden), and one epidemiologist (Slovakia) participated. They all stated that PRO are collected at their own countries in the context of clinical trials and/or specific projects. During everyday clinical practice, PRO are collected routinely/almost routinely in Austria and Sweden and only at the discretion of the treating physicians in Czechia, Denmark, France, Greece, and Portugal. There was complete consensus about the favorable impact that the PRO can have in terms of clinical outcomes, healthcare resources utilization, and general patient satisfaction. Only participants from Portugal and Sweden answered that the PRO are perceived as very important by the National Health Authorities of their respective countries. CONCLUSIONS: Differences exist in attitudes and perspectives about PRO in epilepsy across Europe. An active plan is warranted to harmonize the measurement of PRO and ensure they can be relevant to people with epilepsy and health services.

The European Insomnia Guideline: An update on the diagnosis and treatment of insomnia 2023.

Progress in the field of insomnia since 2017 necessitated this update of the European Insomnia Guideline. Recommendations for the diagnostic procedure for insomnia and its comorbidities are: clinical interview (encompassing sleep and medical history); the use of sleep questionnaires and diaries (and physical examination and additional measures where indicated) (A). Actigraphy is not recommended for the routine evaluation of insomnia (C), but may be useful for differential-diagnostic purposes (A). Polysomnography should be used to evaluate other sleep disorders if suspected (i.e. periodic limb movement disorder, sleep-related breathing disorders, etc.), treatment-resistant insomnia (A) and for other indications (B). Cognitive-behavioural therapy for insomnia is recommended as the first-line treatment for chronic insomnia in adults of any age (including patients with comorbidities), either applied in-person or digitally (A). When cognitive-behavioural therapy for insomnia is not sufficiently effective, a pharmacological intervention can be offered (A). Benzodiazepines (A), benzodiazepine receptor agonists (A), daridorexant (A) and low-dose sedating antidepressants (B) can be used for the short-term treatment of insomnia (≤ 4 weeks). Longer-term treatment with these substances may be initiated in some cases, considering advantages and disadvantages (B). Orexin receptor antagonists can be used for periods of up to 3 months or longer in some cases (A). Prolonged-release melatonin can be used for up to 3 months in patients ≥ 55 years (B). Antihistaminergic drugs, antipsychotics, fast-release melatonin, ramelteon and phytotherapeutics are not recommended for insomnia treatment (A). Light therapy and exercise interventions may be useful as adjunct therapies to cognitive-behavioural therapy for insomnia (B).

Repeated polysomnography and multiple sleep latency test in narcolepsy type 1 and other hypersomnolence disorders.

BACKGROUND: The diagnosis of narcolepsy is based on clinical information, combined with polysomnography (PSG) and the Multiple Sleep Latency Test (MSLT). PSG and the MSLT are moderately reliable at diagnosing narcolepsy type 1 (NT1) but unreliable for diagnosing narcolepsy type 2 (NT2). This is a problem, especially given the increased risk of a false-positive MSLT in the context of circadian misalignment or sleep deprivation, both of which commonly occur in the general population. AIM: We aimed to clarify the accuracy of PSG/MSLT testing in diagnosing NT1 versus controls without sleep disorders. Repeatability and reliability of PSG/MSLT testing and temporal changes in clinical findings of patients with NT1 versus patients with hypersomnolence with normal hypocretin-1 were compared. METHOD: 84 patients with NT1 and 100 patients with non-NT1-hypersomnolence disorders, all with congruent cerebrospinal fluid hypocretin-1 (CSF-hcrt-1) levels, were included. Twenty-five of the 84 NT1 patients and all the hypersomnolence disorder patients underwent a follow-up evaluation consisting of clinical assessment, PSG, and a modified MSLT. An additional 68 controls with no sleep disorders were assessed at baseline. CONCLUSION: Confirming results from previous studies, we found that PSG and our modified MSLT accurately and reliably diagnosed hypocretin-deficient NT1 (accuracy = 0.88, reliability = 0.80). Patients with NT1 had stable clinical and electrophysiological presentations over time that suggested a stable phenotype. In contrast, the PSG/MSLT results of patients with hypersomnolence, and normal CSF-hcrt-1 had poor reliability (0.32) and low repeatability.

Investigation of sleep and circadian rhythm disorders.

Diagnosis of a sleep disorders is multimodal. An overview is presented in this review. The medical history leads to a tentative diagnosis supported by questionnaires, sleep diary and objective methods. Examination may reveal upper airway problems in a patient suspected to have obstructive sleep apnoea or rigidity in an elderly patient with shouting during sleep, suggestive of rapid eye movement sleep behaviour disorder. The choice of diagnostic sleep test is based on the tentative diagnosis. Other tests (e.g., lumbar puncture, brain scan) may be indicated. Wearables pose the advantage of documenting the patients' habitual sleep and circadian rhythm.

Effect of common medication classes on sleep patterns.

This review summarises the effect of common medication on sleep patterns. Evaluation of current medication status is an important part of the assessment in case of complaints of disturbed sleep. Medication may affect sleep continuity and sleep architecture directly via effects on wake or sleep promoting neurotransmitter systems and indirectly via beneficial therapeutic effects or unwanted side effects. It is important for clinicians to be aware of potentially sleep disturbing effects of prescribed medication, especially in case of polypharmacy, and to adjust treatment accordingly to avoid disrupted sleep patterns and the resulting impairment of daytime functioning.

Effect of psychotropics on sleep continuity and sleep architecture.

Psychotropics affect the regulation of sleep and wake by influencing brain stem neurotransmitter systems. The monoaminergic systems are active during wake but diminish their activity when transitioning to sleep in response to increased gamma-aminobutyric acid activity. The cholinergic system is active both during wake and during rapid eye movement sleep. According to their mode of action, different classes of psychotropics can be understood to act differently on sleep continuity and sleep architecture. This review outlines the differences. Increased knowledge of the detailed sleep effects of psychotropics may improve the perceived subjective quality of sleep.

Neurological sleep disorders.

This review summarises the current knowledge of neurological sleep disorders. These disorders are frequent and involve a number of serious diseases associated with complications or precede other serious brain diseases. There is an underdiagnosis of neurological sleep disorders in Denmark. Several of these disorders can be treated, and some of them express a marker for later disease which can be important to diagnose when effective preventive treatment is available.

Ageing and sleep.

This review focuses on ageing and sleep. A key focus in aging is to improve senescence by extending good health, optimal cognitive function, and medical and social assistance into later life. Given that one third of the human lifespan is spent sleeping, the importance of maintaining deep, stable, and consistent sleep is self-evident for the good life quality expected and optimal daytime functioning, which the aging process always curtails. For this reason, employees in the health system should know and focus on the expected changes in sleep patterns and disturbances from young adults to old age, including possible disorders and treatments.

Melatonin use among children, adolescents, and young adults: a Danish nationwide drug utilization study.

We aimed to provide a detailed description of the use of melatonin in Danish children, adolescents, and young adults during 2012-2019. We identified melatonin users 0-24 years of age (n = 43,652; median age 16 years) via the Danish nationwide health registers. Melatonin is a prescription drug in Denmark. The incidence of melatonin use increased from 2.4 to 3.9/1000 person-years during 2012 to 2019. Among 6,557 incident users in 2019, 53% filled only a single prescription within the first 6 months. Long-term use was most common among the younger age groups, with 17% of 5-9-year-olds and 14% of 10-13-year-olds being in continued treatment (no treatment breaks) 12 months after their first melatonin prescription. Disregarding treatment breaks, 3 in 10 were using melatonin 12 months after their first melatonin prescription and this proportion was also highest among 5-9-year-olds (63%) and 10-13-year-olds (51%). Psychopathology was common among melatonin users with 75% registered with either a psychiatric disorder diagnosis (54%), a filled prescription for another psychotropic (58%), or a contact to a private practice psychiatrist (15%) within ± 12 months of treatment initiation. General practitioners authorized melatonin prescriptions to almost half of all new users (48%), while psychiatric specialists authorized 37% of first prescriptions. In conclusion, the incidence of melatonin use increased in Denmark from 2012 to 2019. A substantial proportion of users had concurrent psychopathology most likely explaining their use of melatonin. Long-term melatonin use was more common among the youngest age groups, which should be a focus of interest due to limited safety data.

Estimates of epilepsy prevalence, psychiatric co-morbidity and cost.

PURPOSE: This study estimated epilepsy prevalence, psychiatric co-morbidity and annual costs associated with epilepsy. METHODS: We used Danish national health registers to identify persons diagnosed with epilepsy and psychiatric disorders, and persons using antiseizure medication and persons using drugs for psychiatric disorders. We calculated the prevalence of epilepsy and co-morbid psychiatric disorders in Denmark on December 31, 2016, using information on epilepsy and psychiatric disorders based on combinations of hospital contacts and use of antiseizure and psychoactive medication. Further, direct and indirect annual costs associated with epilepsy were calculated using individual-level data from a range of socioeconomic registers. RESULTS: There were 5,044,367 persons alive and living in Denmark on December 31, 2016, including 33,628 persons with at least one hospital contact with epilepsy in the previous five years (epilepsy prevalence 0.67% (0.69% males; 0.65% females)). Among these persons with epilepsy, we identified 12,562 (37.4%) persons with a psychiatric disorder or use of drugs used for psychiatric disorders as compared with 801,052 (15.9%) persons in the general population. The estimated total annual individual net costs associated with epilepsy was €30,683. Compared with prevalence estimates on December 31, 2006, the prevalence of epilepsy on December 31, 2016, was slightly higher in the older population and slightly lower in children CONCLUSIONS: Population estimates from national registers provide epilepsy prevalence estimates of approximately 0.6-0.7% - similar to previous reviews of epilepsy prevalence. In addition, the national sample allowed idenitfication of high prevalence of psychiatric disorders and high societal costs associated with epielspy.

Subjective sleep assessment compared to polysomnography in mechanically ventilated critically ill ICU patients.

Sleep deprivation is expected in the intensive care unit (ICU) and is associated with delirium and increased mortality. Polysomnography (PSG) is the gold standard for sleep assessment, but practical issues limit the method. Hence, many ICUs worldwide use subjective sleep assessment (SSA) for sleep monitoring, but the agreement between SSA and PSG is unknown. The hypothesis was that the level of agreement between SSA and PSG was low and that total sleep time (TST) assessed with SSA would be overestimated compared to PSG in this existing cohort database. In this sub-analysis, 30 consecutive study participants underwent 15-h PSG recordings during two consecutive nights. The attending nurse performed an hourly subjective observer rating of sleep quantity during both nights, and the agreement between SSA and PSG was determined along with mean TST. Primary outcome: The level of agreement between SSA and PSG determined by Bland-Altman analysis. Secondary outcome: (1) The overall mean TST estimated by SSA compared to PSG in all study participants enrolled in the main study during both study nights, (2) TST for all study participants evaluated hourly during both study nights, (3) TST assessed with SSA compared to PSG in study participants sedated with dexmedetomidine during the second night and for study participants treated with placebo or non-sedation the first and second nights. The level of agreement between SSA and PSG was low. Mean TST estimated by SSA during the time interval 4.00 p.m. to 7.00 a.m. was 481 min (428;534, 95% CI) vs. PSG at 437 min (386;488, 95% CI) (p = .05). When sedated with dexmedetomidine, TST estimated using SSA was 650 min (571;729, 95% CI) versus PSG which was 588 min (531;645, 95% CI) (p = 0.56). For participants treated with placebo or non-sedation TST estimated with SSA was 397 min (343;450, 95% CI) versus PSG at 362 min (302;422, 95% CI) versus (p = 0.17). In mechanically ventilated critically ill ICU patients, the level of agreement between SSA and PSG was low, and there was a significant overestimation of mean TST. SSA should only be used under awareness that it is imprecise and overestimates TST.

Repeated measures of hypocretin-1 in Danish and Italian patients with narcolepsy and in controls.

STUDY OBJECTIVES: The assay currently used worldwide to measure cerebrospinal fluid hypocretin-1 (CSF-hcrt-1) for diagnosing narcolepsy uses a competitive radioimmunoassay with polyclonal anti-hcrt-1 antibodies. This assay detects multiple hypocretin-1 immunoreactive species in the CSF that are all derived from full-length hcrt-1. We aimed to revalidate CSF-hcrt-1 cut-offs for narcolepsy type 1 (NT1) diagnosis and to evaluate temporal changes in CSF-hcrt-1 levels in patients suspected of having central hypersomnia. METHOD: We carried out a repeat lumbar puncture with a mean follow-up of 4.0 years, to measure CSF-hcrt-1 in patients suspected of having central hypersomnia in a follow-up study. Data from CSF samples of patients with NT1 and of controls without known hypersomnia, from the Italian-Stanford and Danish populations, were examined using a receiver-operating characteristic analysis. RESULTS: The optimal CSF-hcrt-1 cut-offs for identifying NT1 were 129 pg/ml and 179 pg/ml for the Italian-Stanford and Danish populations, respectively. The sensitivity was 0.93-0.99 and the specificity was 1. Follow-up lumbar puncture measurements of CSF-hcrt-1 were obtained from 73 patients. 30 of 32 patients with low CSF-hcrt-1 levels continued to be categorized as low, with an unaltered diagnosis; two patients showed a marked increase in CSF-hcrt-1, attaining normal values at follow-up. One of these patients relapsed to low CSF-hcrt-1 after follow-up. All 41 patients with normal CSF-hcrt-1 at baseline had normal CSF-hcrt-1 at follow-up. CONCLUSION: CSF-hcrt-1 measurement can provide an accurate test for diagnosing NT1, although it is important to validate the CSF-hcrt-1 cut-off for specific testing locations. Stable CSF-hcrt-1 levels support the already established prognosis of narcolepsy as permanent once the disorder has fully developed.

Two nights of home polysomnography in healthy 7-14-year-old children - Feasibility and intraindividual variability.

OBJECTIVE: Attended polysomnography (PSG) is the gold standard for childhood sleep evaluation. There is, however, only limited information regarding repeated ambulatory PSG in children. We aimed to test whether in hospital attached level 2 home PSG is feasible and reproducible in healthy children. METHODS: We recruited healthy children aged 7-14 years to undergo two nights of full level 2 PSG. The PSG equipment was attached at the hospital on the day of the sleep test and all recordings were performed at home. Subjective sleep quality, nocturnal urine production, sleep time and number of awakenings were documented for a week in connection to the first PSG night. RESULTS: Thirty-three children were recruited of whom 32 children (aged 11 ± 2.1 years) underwent two nights of PSG. All 64 PSGs were technically adequate for sleep evaluation. We found mean sleep efficiency of 94% and mean total sleep time of 8.4 h. Sleep stages distribution with 5.9% N1, 46.8% N2, 24.3% N3 and 22.8% REM sleep. We found poorer subjective sleep quality, more self-reported awakenings, and shorter total sleep time on nights with PSG compared to nights without PSG with no differences between PSG study nights. No differences in nocturnal urine production were found between nights with and without PSG. The comparison of PSG variables between the two PSG nights revealed no first night effect. CONCLUSIONS: Type 2 PSG recording is feasible for sleep evaluation in children 7-14 years of age producing good data quality. We found no first night effect on PSG variables. www. CLINICALTRIALS: gov Registration number: NCT03477812.

Sleep quality and quantity determined by polysomnography in mechanically ventilated critically ill patients randomized to dexmedetomidine or placebo.

BACKGROUND: Abnormal sleep is commonly observed in the ICU and is associated with delirium and increased mortality. If sedation is necessary, it is often performed with gamma-aminobutyric acid agonists such as propofol or midazolam leading to an absence of restorative sleep. We aim to evaluate the effect of dexmedetomidine on sleep quality and quantity. METHODS: Thirty consecutive patients were included. The study was conducted as a double-blinded, randomized, placebo-controlled trial with two parallel groups: 20 patients were treated with dexmedetomidine, and 10 with placebo. Two 16 h of polysomnography recordings were done for each patient on two consecutive nights. Patients were randomized to dexmedetomidine or placebo after the first recording, thus providing a control recording for all patients. Dexmedetomidine was administered during the second recording (6 p.m.-6 a.m.). OBJECTIVE: To compare the effect of dexmedetomidine versus. placebo on sleep - quality and quantity. PRIMARY OUTCOME: Sleep quality, total sleep time (TST), Sleep efficiency (SE), and Rapid Eye Movement (REM) sleep determined by Polysomnography (PSG). SECONDARY OUTCOME: Delirium and daytime function determined by Confusion Assessment Method of the Intensive Care Unit and physical activity. Alertness and wakefulness were determined by RASS (Richmond Agitation and Sedation Scale). RESULTS: SE were increased in the dexmedetomidine group by; 37.6% (29.7;45.6 95% CI) versus 3.7% (-11.4;18.8 95% CI) (p < .001) and TST were prolonged by 271 min. (210;324 95% CI) versus 27 min. (-82;135 95% CI), (p < .001). No significant difference in REM sleep, delirium physical activity, or RASS score was found except for RASS night two. CONCLUSION: Total sleep time and sleep efficiency were significantly increased, without elimination of REM sleep, in mechanically ventilated ICU patients randomized to dexmedetomidine, when compared to a control PSG recording performed during non-sedation/standard care.

Diagnostic value of actigraphy in hypersomnolence disorders.

OBJECTIVE: Differentiating between the central hypersomnias presents a challenge to the diagnosis of patients with hypersomnolence. Actitigraphy may support efforts to distinguish them. We aimed to evaluate: 1) the ability of actigraphy to quantify sleep continuity measures in comparison with polysomnography in patients with hypersomnolence; 2) whether actigraphy can distinguish patients with hypersomnolence with normal hypocretin-1 in cerebrospinal fluid from patients with narcolepsy type 1 and from sleep-healthy controls; and 3) the distinct activity profiles and circadian rhythms of patients with narcolepsy type 1, patients with hypersomnolence with normal hypocretin-1 in cerebrospinal fluid, and sleep-healthy controls. METHOD: Polysomnography, multiple sleep latency tests and actigraphy were conducted in 14 patients with narcolepsy type 1, 29 patients with hypersomnolence with normal hypocretin-1 in cerebrospinal fluid and 15 sleep-healthy controls. RESULTS: Actigraphy quantified several sleep continuity measures consistently with polysomnography in all the patients. Actigraphy distinguished patients with hypersomnolence with normal hypocretin-1 in cerebrospinal fluid from patients with narcolepsy type 1 and sleep-healthy controls. Patients with narcolepsy type 1 had poor sleep quality and altered circadian rest-activity rhythm compared with controls. CONCLUSION: Actigraphy is an adequate tool for establishing the amount of night sleep and supports the differential diagnosis of patients with hypersomnolence.

Use of hypnotic drugs among children, adolescents, and young adults in Scandinavia.

BACKGROUND: Hypnotic use in children and adolescents is controversial. OBJECTIVE: To describe the use of hypnotic drugs (melatonin, z-drugs, and sedating antihistamines) among 5- to 24-year-old Scandinavians during 2012 to 2018. METHODS: Aggregate-level data were obtained from public data sources in Sweden, Norway, and Denmark. We calculated annual prevalence (users/1000 inhabitants) stratified by age group, sex, and country. Quantity of use (Defined Daily Dose (DDD)/user/day) was estimated for Norway and Denmark. RESULTS: Melatonin was the most commonly used hypnotic, and its use increased markedly from 2012 to 2018, particularly among females and 15- to 24-year-old individuals. Sweden had the highest increase in use (6.5 to 25/1000) compared with Norway (10-20/1000) and Denmark (5.7-12/1000). The annual prevalence of sedating antihistamine use was also highest in Sweden, reaching 13/1000 in 2018 in comparison to 7.5/1000 in Norway and 2.5/1000 in Denmark. Z-drug use decreased in all countries toward 2018, dropping to 3.5/1000 in Sweden, 4.4/1000 in Norway, and 1.7/1000 in Denmark. The quantity of hypnotic use in Norway and Denmark was 0.8-1.0 DDD/user/day for melatonin in 2018, as compared to 0.1-0.3 for z-drugs and antihistamines. CONCLUSION: The use of melatonin and sedating antihistamines increased among young Scandinavians during 2012-2018, and the increase was twice as high in Sweden compared with Norway and Denmark. In addition, Sweden had the highest use of sedating antihistamines. The Scandinavian variation of hypnotic use could reflect differences in frequency of sleep problems between populations or variation of healthcare access or clinical practice between countries.

U-Sleep: resilient high-frequency sleep staging.

Sleep disorders affect a large portion of the global population and are strong predictors of morbidity and all-cause mortality. Sleep staging segments a period of sleep into a sequence of phases providing the basis for most clinical decisions in sleep medicine. Manual sleep staging is difficult and time-consuming as experts must evaluate hours of polysomnography (PSG) recordings with electroencephalography (EEG) and electrooculography (EOG) data for each patient. Here, we present U-Sleep, a publicly available, ready-to-use deep-learning-based system for automated sleep staging ( sleep.ai.ku.dk ). U-Sleep is a fully convolutional neural network, which was trained and evaluated on PSG recordings from 15,660 participants of 16 clinical studies. It provides accurate segmentations across a wide range of patient cohorts and PSG protocols not considered when building the system. U-Sleep works for arbitrary combinations of typical EEG and EOG channels, and its special deep learning architecture can label sleep stages at shorter intervals than the typical 30 s periods used during training. We show that these labels can provide additional diagnostic information and lead to new ways of analyzing sleep. U-Sleep performs on par with state-of-the-art automatic sleep staging systems on multiple clinical datasets, even if the other systems were built specifically for the particular data. A comparison with consensus-scores from a previously unseen clinic shows that U-Sleep performs as accurately as the best of the human experts. U-Sleep can support the sleep staging workflow of medical experts, which decreases healthcare costs, and can provide highly accurate segmentations when human expertize is lacking.