By Alan Z. Segal, MD

Associate Professor of Clinical Neurology, Weill Cornell Medical College

Dr. Segal reports no financial relationships relevant to this field of study.

SYNOPSIS: The relationship between quality of sleep and the development of dementia is controversial and not yet clearly elucidated nor understood.

SOURCE: Suh SW, Han JW, Lee JR, et al. Sleep and cognitive decline: A prospective nondemented elderly cohort study. Ann Neurol 2018;83:472-482.

Quality sleep is known to facilitate the consolidation of memory and enhance daytime cognitive function. Over the long term, sleep also may protect against the development of dementia. Restorative sleep, both stage three (slow-wave sleep) and rapid eye movement sleep, have been shown to reduce the deposition of amyloid beta (Aβ) in the brain and facilitate Aβ clearance. Sleep also has been shown to enhance the so-called “glymphatic” system of the brain, widening gap junctions and allowing drainage of metabolic toxins.

Most studies of sleep and cognition, including this work by Suh et al, rely on patient questionnaires rather than hard data from polysomnography. This introduces a significant degree of subjectivity into the analysis. Furthermore, it can be difficult to distinguish whether poor sleep is a cause or effect of cognitive decline (since effects may be “bidirectional”), and there may be pathological effects at the extremes of sleep, which may be protective in a more moderate range (with U- or J- shaped relationships). For example, exceedingly short or long sleep duration may both be more damaging than moderate sleep times, and such extremes of sleep duration may be more of an effect of dementia rather than a truly causal factor. Further ambiguity is created by the plethora of sleep-related variables used. These include latency to sleep onset (falling asleep), sleep disruption (poor sleep architecture or frequent waking after sleep onset), overall sleep duration (which does not itself account for fragmentation), advance or delay of sleep phase (so-called “night owls” or “larks”), use of hypnotics/other psychotropic medications, and several other subjective patient perceptions of sleep quality.

Suh et al studied 2,238 Korean, non-demented individuals prior to the development of significant cognitive change. A smaller cohort of 655 subjects with mild cognitive impairment also was included. There was an association found between long sleep latency (> 30 minutes) and cognitive decline (odds ratio [OR], 1.4). Additionally, long sleep duration (> 8 hours) showed a similar risk for dementia (OR, 1.67), and there appeared to be a protective effect for subjects with an overall delay in their sleep phase. The odds of cognitive decline was 0.61 in subjects with a “mid-sleep” time later than 3 a.m. (maintaining an overall sleep schedule generally later than 11 p.m. to 7 a.m.).

Among the participants with established mild cognitive impairment, a proportion (approximately 30%) reverted to “normal cognition” at four-year follow-up, but this was significantly less common among subjects with long sleep latency times.


These results, particularly those regarding long sleep latencies, can be considered a valuable supplement to the large existing literature exploring sleep and cognition. However, these findings conflict with large cross-sectional epidemiological analyses, which not only associate prolonged sleep latency with dementia, but also implicate a deleterious effect of truncated sleep duration. Suh et al found the opposite, with longer sleep times associated with cognitive decline. Additionally, because Suh et al’s data are prospective, conclusions can be drawn regarding cause and effect relationships that cannot be drawn from other studies. Although prior studies have suggested an association between cognitive compromise and “advanced sleep phase” (with early bed and wake times), Suh et al’s data suggested that this circadian shift may not be the result of dementia, but rather, may be a risk factor for dementia or at least a predictive, pre-existing condition. When studied prospectively, starting at a younger, normal cognitive state, “larks” appear more likely than “night owls” to subsequently develop cognitive impairment.