The impact of sleep fragmentation on health

Author: Bastien Lechat

Lechat, Bastien, 2021 The impact of sleep fragmentation on health, Flinders University, College of Science and Engineering

Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact with the details.


Inadequate sleep is highly prevalent in the population and leads to an increased risk of a broad range of adverse health impacts. These include traffic and other ac-cidents, mental health problems including depression, psychiatric disorders and sui-cidal tendencies and cardio-metabolic diseases such as diabetes, hypertension, heart disease and stroke. Assessments of sleep fragmentation and sleep quality are im-portant to evaluate and mitigate risks associated with poor sleep. However, sleep fragmentation has traditionally been defined according to conventional sleep scor-ing in 30-sec epochs and obvious electroencephalographic (EEG) changes associat-ed with arousals and awakenings, rather than more systematic and potentially more sensitive physiologically guided measurements derived from modern signal pro-cessing methods. Current markers of sleep fragmentation repeatedly fail to predict important clinical outcomes, such as sleepiness or cardiovascular events. Therefore, the aim of the work presented in this thesis was to develop new markers of sleep fragmentation based on key features of EEG changes during sleep. These bi-omarkers were subsequently tested for clinical utility in several population groups relevant to sleep fragmentation, including a sample of individuals exposed to exper-imental environmental noise manipulations and several large population samples in-cluding participants with sleep disorders.

Phasic sleep fragmentation due to experimental environmental noise was quantified using K-complexes, a subtle EEG marker of sensory processing during sleep. K-complexes were automatically detected and scored using a deep learning algorithm that was developed as part of this thesis. The effect of different types of environ-mental noise (traffic noise and wind farm noise) on sleep fragmentation was as-sessed in a pilot-study of 21 individuals exposed to a range of noises at different sound pressure levels throughout sleep. K-complexes were a more sensitive sensory disturbance marker of noise exposure during sleep than traditional metrics, such as arousals and awakenings. Statistically significant K-complex responses were ob-served at sound pressure levels as low as 33 dBA (75% more likely than control) and K-complex response probability further increased with sound pressure level. In contrast, arousals and awakenings were only detectable with noise exposures above 39 dBA. Overall, K-complexes were two times more likely to occur in response to noise than EEG arousals or awakenings, clearly indicating their superior sensitivity to noise exposure compared to traditional arousal scoring.

In a separate study and analysis, deep sleep fragmentation was assessed using a technique conceived during this thesis work, which combines power spectral analy-sis of the delta-frequency band (0.5 Hz to 4.5 Hz) with a measure of signal com-plexity via spectral entropy. The association between deep sleep fragmentation as-sessed with this new entropy metric and all-cause mortality was studied in the Sleep Heart Health Study (SHHS), a large US-based cohort (N = 5804). Delta sleep fragmentation was associated with a ~30% increased risk of all-cause mortality compared to no sleep fragmentation. This association was similar to a reduction in total sleep time from 6.5h to 4.25h. Conventional measures of sleep quality, includ-ing wake after sleep onset and arousal index were not predictive of all-cause mortal-ity.

Hyperarousal – a pathophysiological trait sometimes observed in patients with in-somnia, was quantified using the odds ratio product (ORP), a novel marker of sleep alertness. Association between the ORP during wake (hypothesised to reflect hy-perarousal) and sleepiness/poor sleep quality was assessed in two large cohort stud-ies (HypnoLaus N = 2162; MAILES N = 754). Hyperarousal was associated with around a 30% increased risk of self-reported poor sleep quality (Pittsburgh Sleep Quality Index score >5) in both HypnoLaus (28%) and MAILES (36%), but an approximately 20% decrease in excessive daytime sleepiness (Epworth sleepiness scale score >10) in the combined dataset. In contrast, no associations were detected using any traditional polysomnography markers.

The additive effect of multiple sleep disorders (co-occurrence of insomnia and ob-structive sleep apnoea (COMISA)) on all-cause mortality and sleep fragmentation was studied in the SHHS cohort (N = 5804). COMISA was associated with greater sleep fragmentation and COMISA patients were at higher risk of all-cause mortality (30%) and cardiovascular events (30%). Insomnia-alone and obstructive sleep ap-noea (OSA)-alone were not associated with all-cause mortality risk or cardiovascu-lar event risk.

The work presented in this thesis suggests that metrics designed to encapsulate core physiological and pathophysiological processes of sleep, sleep fragmentation and sleep disorders provide more informative markers that may be important pre-dictors of adverse health outcomes. Specifically, disrupted deep sleep and an in-creased state of hyperarousal were two pathways identified as potentially contributing to all-cause mortality, sleepiness and poor sleep quality. K-complexes were al-so established to be a more sensitive marker of sensory processing during sleep to environmental noise disturbances than conventional metrics. Together, these find-ings make an important contribution to understanding the impact of sleep frag-mentation on health and provide multiple EEG biomarkers with major potential to substantially improve clinical sleep medicine.

Keywords: slow wave sleep; sleep disruption; poor sleep; K-complex; environmental noise; all-cause mortality; cardiovascular disease; sleepiness; machine learning; artificial intelligence

Subject: Medical Biotechnology thesis

Thesis type: Doctor of Philosophy
Completed: 2021
School: College of Science and Engineering
Supervisor: Kristy Hansen