Wake, NREM, and REM sleep measures predict incident dementia

Multiple features of sleep–wake disorders have been associated with major neurodegenerative diseases (NDD) including Alzheimer’s disease (AD) and Parkinson’s Spectrum Disease. Nuclei associated with sleep–wake regulation are within the brainstem and hypothalamus, in areas earliest affected in many NDDs. Thus, neurodegenerative processes may produce disruptions to both sleep and wake behavior prior to the onset of cognitive or motor symptoms. There has been a growth of evidence demonstrating that arousals and wake activity during the main sleep period may increase the release of pathological proteins associated with neurodegeneration [1–3]. Furthermore, reductions in sleep time, particularly slow wave sleep (SWS), may reduce clearance of proteins via glymphatic fluid flow in the interstitial spaces of the brain [4–8]. Confirmation of the primary role of sleep–wake behavior on prospective risk for NDD has been limited by the small number of studies that have measured sleep EEG in large sample-sized at-risk cohorts.

Ibrahim et al. examined whether sleep architecture and wake after sleep onset from laboratory polysomnography (PSG) in a neurology-based sleep center predicted incident NDDs [9]. They retrospectively analyzed the data of 999 participants who were studied with PSG during 2004–2007 who had complete neurologic examinations and were found to have no evidence for NDD at baseline or in the 5-year window post-sleep assessment. The cohort was followed for a median of 12.8 years, and 75 patients developed a NDD which included AD, amnestic mild cognitive impairment, and vascular dementia, grouped as suspected amyloid pathology; and Parkinson’s disease, Parkinson’s Dementia, and idiopathic REM sleep behavior disorder, grouped as suspected α-synucleinopathy. They found that reduced N3 and reduced REM sleep, as well as increased wake after sleep onset, each predicted increased risk for developing the broad grouping of NDDs. These results held up after controlling for age, sex, body mass index, apnea–hypopnea index (AHI), periodic limb movements of sleep, and antidepressants. Multivariate random survival forest models showed that combining all three predictors optimized performance, and that wake, followed by N3, and then REM sleep were the most important single predictors for predicting all causes of NDD. When separating out into separate suspected amyloid versus α-synucleinopathy groups, only sleep efficiency and wake after sleep onset predicted amyloid-related disorders. No individual sleep or wake feature specifically predicted any of the synucleinopathies, most likely due to insufficient sample size. Surprisingly, in obstructive sleep apnea (OSA) patients with AHI > 15, adherence to positive airway pressure (PAP) was not protective, and there was no relationship between the continuous AHI score and NDD risk. The only OSA group at increased risk for NDD were patients with AHI above 30 who were not treated with PAP.

The study has many strengths including laboratory quality PSG, large sample size, and careful neurologic assessment at baseline, and a lengthy follow-up period to capture incident NDDs. The report also demonstrates the importance of reduced sleep efficiency as a predictor for NDD risk which is perhaps easier to measure at scale in the general population. Related to the issue of scalability is their interesting and discrepant finding that longer subjective sleep duration and shorter objective sleep duration predicted NDD risk, highlighting the importance of objective measures. Limitations to their study pertain to a cross-sectional predictor, retrospective design, and reliance on electronic health records to capture individual NDD diagnoses. Other limitations include insufficient power to examine individual sleep–wake predictors to individual NDD diagnoses and lack of other NDDs outside of amyloid and α-synucleinopathies.

The critical importance of all three major brain states; wake, NREM, and REM sleep, in predicting NDDs is not surprising given the shared anatomy of sleep–wake regulating nuclei and areas affected by different NDD diagnoses. The prototypical example of this is the relationship of REM behavior disorder with α-synucleinopathy [10]. In the earliest stages of disease, prior to motor symptoms, aggregates of α-synuclein affect medullary, and pontine sites regulating REM sleep atonia, prior to involvement with the substantia nigra [11, 12]. The importance of N3 sleep in the study by Ibrahim et al. is consistent with the glymphatic hypothesis given that slow oscillations (SO) associated with this stage are associated with the greatest flow of extracellular fluid [13]. However, the discovery of a brainstem SWS generator [14–17] in a parvalbumin-positive neuron group in the medullary parafacial zone in mice [16, 18–25] points to a prominent role of the brainstem in SWS regulation. Thus, reduced N3 as a predictor of future NDD is also consistent with the hypothesis that initial brainstem involvement of NDD produces an early prodrome of reduced SWS. A recent report from the Framingham Heart Study showed that in participants who completed PSGs over two timepoints, in the time periods 1995–1998 and 2001–2003, the decline in SWS across this interval predicted development of all causes of dementia in the subsequent 17 years [26]. Finally, there is less data pertaining to REM sleep and it is less clear that this stage is relevant to the glymphatic hypothesis. There are several studies showing reduced REM as being a predictor for future NDD [27–29]. However, as previously noted, most NDDs, including AD, start in the brainstem prior to involvement in cortical structures. The fact that the site of early disease onset is critical for regulating features of REM sleep, in addition to arousal and NREM sleep illustrates the importance of a multivariate sleep–wake predictors for predicting future NDDs.

OSA syndrome is associated with increased risk for cognitive decline and progression to dementia [30–32]. Interestingly, the report by Ibrahim et al. did not show that risk for NDD was appreciable across the continuum of severity of OSA. The only increased risk was found in the small subgroup with AHI above 30 who were not in treatment with PAP. Across the full range of apnea severity, adherence to PAP, was not found to be protective for incident NDD disease. They conclude that the relationship of OSA to future NDD is non-linear and only evident at the upper end of disease severity. If confirmed, this has practical implications for future clinical trials that will test the efficacy of treating sleep disturbances to slow progression to NDDs.

Sleep macro- and microarchitecture (δ, θ, α power, SO, spindles, and SO/spindle coupling) can discriminate among healthy controls, people with mild cognitive impairment, and those with a heterogeneous group of dementia-associated disorders [33], but the sleep/dementia field has been hamstrung by inadequate attention to NDD heterogeneity. Thus, the field needs to move beyond predicting the broad category of all causes of dementia to focus on specific diseases that are characterized by having their own sites of disease onset and progression. In addition, future work will be strengthened using disease-specific biomarkers (plasma, CSF, and PET). As plasma biomarkers are proving to be robust and scalable, this presents the opportunity to improve the rigor and impact of future prospective studies. Ultimately this will lead to the development of disease-specific therapeutics (e.g. targeting SWS, REM, or wake hyper-arousal) in future prevention trials.

Funding

Supported by grants from the Rainwater Charitable Foundation and the National Institute of Aging: 5R01AG060477-05 and 5R01AG064314-05.

Disclosure Statement

Financial disclosure: Drs. Neylan and Walsh do not have financial disclosure. Nonfinancial disclosure: Drs. Neylan and Walsh do not have any potential conflicts of interest.