Currently, we try to phenotype patients and characterize them with anatomical, biochemical, and functional characteristics. This is applied to normal and disordered sleep. Phenotyping disordered sleep is of particular interest for insomnia patients and patients with sleep-disordered breathing. For functions of sleep-disordered breathing, we are phenotyping patients according to their dominance of central apnea, obstructive apnea, or hypoventilation during sleep. The results may be the result of impaired respiratory control with different sensitivity to CO2 and pH changes, or anatomical preconditions with narrow and collapsible upper airways or obesity, or sleep regulation conditions with a low arousal threshold. For sleep regulation conditions, in contrast to anatomical or respiratory control conditions, we do not have good and well-qualified biomarkers. Therefore, we need reliable sleep electroencephalogram (EEG) biomarkers.
Biomarkers need to be reliable, stable, and should not show night-to-night variability. Otherwise, the parameters cannot be regarded as “markers” to be used for phenotyping. For sleep parameters, such as percentages of sleep stages, or latencies, we see a lot of night-to-night variability. We know that external conditions, nutrition, light, and noise modify these parameters. Still the EEG itself, the sleep EEG, and the hypnogram can be seen as a characteristic for an individual, such as a fingerprint. The individual EEG had even been considered as a personal metric to be used to encode individual and nonexchangeable biometric characteristics, similar to an individual fingerprint. The study by Levendowski et al. tries to identify biomarkers derived from a frontopolar sleep EEG and investigates their accuracy and stability.1 More precisely, the group tries to introduce and investigate sleep architecture biomarkers. To identify biomarkers, it makes sense to use a broad variety of disorders, if possible, to cover the entire range of disorders. Only then is it possible to determine the sensitivity, specificity, and predictive value of a biomarker. A good sensitivity, specificity, and predictive value are key to biomarkers. Indeed, a broad range of sleep disorders was investigated by Levendowski et al. Data were collected in a central database for processing in the same way. The first bench test for a biomarker is accuracy. In order to ensure accuracy, all sleep recordings were scored by five technologists and one computer-based autoscoring algorithm. This is very important in view of the high variability observed between different scorers and different sleep centers worldwide.2,3 Only with this kind of procedure, the scorer variability, or here the biomarker test reliability, can be controlled. This is a prerequisite to identify biomarkers, called “biomarker accuracy” by Levendowski et al. The second bench test for a biomarker is test repeatability. For functional sleep biomarkers, this corresponds to night-to-night variability. In the study presented, 2 nights were recorded and compared. Levendowski et al. called this biomarker variability and stability. Maybe 2 nights are not really enough for this type of repeatability testing. However, we understand the limitations of practicality. With these two prerequisite bench tests, the study by Levendowski et al. is directing toward the future identification and validation of sleep architecture biomarkers derived from sleep EEG.
On the search for biomarkers for sleep disorders on the basis of functional characteristics, we should not remain at the identification of sleep architecture biomarkers introduced here. We need to move beyond these established metrics. Levendowski et al. already mention that sleep spindles, autonomic activation index, and slow wave sleep have more consistent patterns compared to usual architecture variables such as sleep time and latencies reflecting sleep macrostructure. This is really promising because it directs us to derive new parameters such as sleep spindle duration, or sleep spindle frequency, or alpha wave frequency recently investigated.4 In the age of computer-processed sleep EEG, we are able to derive new parameters, such as wave characteristics or new pattern descriptors, which really allow us to go beyond established sleep architecture. Engineers can make use of the sleep EEG to derive new parameters that can be used as true sleep EEG biomarkers. This is often called sleep microstructure.5
Once we have candidate biomarkers, whether derived from sleep architecture or novel parameters derived from the sleep EEG itself (eg, wave characteristics), or parameters similar to speech processing, (formants), then we can procced to the ultimate step in finding biomarkers. This final step is the determination of sensitivity, specificity, and predictive value for specific sleep disorders. This step was demonstrated by Levendowski et al. as they show a framework how their sleep architecture biomarkers associate with some groups of disorders, such as insomnia, cardiovascular, diabetes, obesity, hypersomnia, and others. However, this is not really what we want as sleep physicians. As sleep physicians, we would be most happy to have biomarkers, which allow the diagnosis of each sleep disorder as defined according to the International Classification of Sleep Disorders6 with sufficient sensitivity and specificity by a sleep EEG-derived marker, or by a blood test-derived marker.
Figure 1.
Sixteen hypnograms from patients with varying degree of sleep apnea show profound difference in sleep regulation. Not only percentages of sleep stages vary, but the time course of sleep stages varies. These hypnograms during a diagnostic night, and hypnograms during treated nights (not shown here), can serve as non-interchangeable individual markers similar to fingerprints. Even with very similar apnea index (AI), we see totally different time courses of sleep stages.
Usually biomarkers are taken from clinical chemistry. However, in sleep medicine, we rely on physiological recordings and therefore our approach to identify biomarkers is clearly related to the signals recorded. This does not mean that biochemical chemistry biomarkers should be replaced, but they can be complemented by physiological biomarkers, which may be more appropriate to characterize functional characteristics, as we see them in the variety of sleep disorders.
DISCLOSURE STATEMENT
TP has received grants from Cidelec, Heinen+Löwenstein, Itamar, Philips, ResMed, and Weinmann not related to this article. Research was partially supported by a grant from the German Israel Foundation (GIF) no. I-1372-303.7/2016. IF has received grants from ResMed, Weinmann, and Fisher & Paykel, he is a board member of ASR GmbH and Somnico GmbH, a consultant for Respicardia, Vanda, Stada, and Wickholds, and holds the Speaker's Bureau for ResMed and Somnomed; none of these relationships is related to this article. CV reports no conflicts of interest.
CITATION
Penzel T, Fietze I, Veauthier C. The need for a reliable sleep EEG biomarker. J Clin Sleep Med. 2017;13(6):771–772.
REFERENCES
- 1.Levendowski DJ, Ferini-Strambi L, Gamaldo C, Cetel M, Rosenberg R, Westbrook PR. The accuracy, night-to-night variability, and stability of frontopolar sleep electroencephalography biomarkers. J Clin Sleep Med. 2017;13(6):791–803. doi: 10.5664/jcsm.6618. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zhang X, Dong X, Kantelhardt JW, et al. Process and outcome for international reliability in sleep scoring. Sleep Breath. 2015;19(1):191–195. doi: 10.1007/s11325-014-0990-0. [DOI] [PubMed] [Google Scholar]
- 3.Danker-Hopfe H, Kunz D, Gruber G, et al. Interrater reliability between scorers from eight European sleep laboratories in subjects with different sleep disorders. J Sleep Res. 2004;13(1):63–69. doi: 10.1046/j.1365-2869.2003.00375.x. [DOI] [PubMed] [Google Scholar]
- 4.Schwabedal JTC, Riedl M, Penzel T, Wessel N. Alpha-wave frequency characteristics in health and insomnia during sleep. J Sleep Res. 2016;25(3):278–286. doi: 10.1111/jsr.12372. [DOI] [PubMed] [Google Scholar]
- 5.Rodenbeck A, Danker-Hopfe H, Geisler P, et al. Ergänzende Regeln zu Frequenzen und Graphoelementen der Schlafstadienanalyse. Update 2014. Somnologie. 2015;19:51–60. [Google Scholar]
- 6.American Academy of Sleep Medicine. International Classification of Sleep Disorders. 3rd ed. Darien, IL: American Academy of Sleep Medicine; 2014. [Google Scholar]