Wake up, sleepyhead—your risk of dementia may have just doubled. According to a study published February 22 in Neurology, people who slumber for more than nine hours per night have twice the risk of dementia as people who sleep for just six to nine hours. And for people without a high school diploma, the risk was three times higher again. Drawing on multigenerational data from the Framingham Heart Study, researchers led by Matthew Pase and Sudha Seshadri of Boston University in Massachusetts also reported that a gradual transition from normal to prolonged sleep may be a harbinger of incipient dementia. However, setting the alarm an hour earlier is unlikely to do much good, as the researchers proposed that prolonged sleep is a marker of ongoing neurodegenerative processes, rather than their cause.

Sleep is essential for brain health—that much is clear. During each rejuvenating bout, synapses are neatly pruned, the day’s highlights enshrined into memory, and the glymphatic system flushes unwanted detritus, including Aβ, out of the brain. Acute sleep deprivation taxes cognitive function, as everyone knows. However, the relationship between long-term sleep patterns and chronic processes such as neurodegeneration is hazy. Some studies have reported that both shortened and/or prolonged sleep predict dementia, but findings were limited by short follow-up periods, making it difficult to determine whether early symptoms of dementia could be the cause of sleep disturbances or vice versa (see Benito-León et al., 2014; Potvin et al., 2012; Chen et al., 2016). Other findings using polysomnography to assess sleep quality and duration found a link between sleep apnea, but not sleep duration, and dementia (see Aug 2011 news). Nearly in line with the current study, questionnaires and telephone interviews of 15,000 nurses identified a link between shortened or prolonged sleep duration in mid- or late life with poor cognitive performance later in life (see Aug 2012 conference news; Devore et al., 2014). 

Against this complex backdrop, Pace, Seshadri, and colleagues turned to the Framingham Heart Study, an ongoing longitudinal health study initiated in 1948. Drawing on data from its more than 2,000 participants, the researchers asked whether how long a participant slept at baseline related to his or her risk of dementia within the following decade. They found that compared to people who slept between six and nine hours per night, those who reported sleeping more than nine hours at baseline had about double the incidence of dementia within the next 10 years. In a subsequent analysis, the researchers found that this association was primarily driven by people who had mild cognitive impairment at baseline. Furthermore, long sleepers who never graduated from high school had about a sixfold increase in dementia risk. Short sleepers (less than six hours) did not have an elevated dementia risk.

The researchers next asked whether a transition from short to longer sleep affected dementia risk. Referencing data from questionnaires given about 13 years before baseline, the researchers also found that people who transitioned from sleeping less than nine hours to more had nearly 2.5 times the risk of developing dementia or AD as those who remained below nine hours at both time points. This association held regardless of MCI status at baseline. Interestingly, for people who reported sleeping long hours at both time points, dementia risk was only slightly elevated. This suggested that a change in sleep status, rather than simply sleeping long hours, was the most important predictor of cognitive downslide in cognitively normal adults.

The participants also underwent cognitive testing and magnetic resonance imaging at baseline. Both short and long sleepers suffered from deficits in visual attention and executive function as measured by the Trail Making Test. Strikingly, long sleepers also tended to have smaller brains than those who slept normal hours, while short sleepers had larger brains. While ongoing neurodegeneration could explain the smaller brains of prolonged sleepers, the researchers were unsure what to make of the larger brain volumes in short sleepers.

Given the known importance of sleep for brain health, Seshadri told Alzforum she was initially surprised to find that prolonged sleep predicted dementia. She proposed that a shift to sleeping longer hours could be an early disease marker, perhaps triggered by the brain’s need to compensate for emerging neurodegeneration.

Jeffrey Iliff of Oregon Health Sciences University in Portland agreed. He commented that the most striking finding of the study was not that prolonged sleep at baseline predicted dementia in people with MCI, but rather that a conversion from normal sleep to longer sleep predicted it across the whole sample. “It’s the change in status that is most important,” he said.

While working at Maiken Nedergaard’s lab at the University of Rochester Medical Center, New York, Iliff helped reveal the existence of the glymphatic system, which clears solutes, including Aβ, from the mouse brain during sleep (see Oct 2013 news). Others have reported that sleep deprivation in humans promotes Aβ build-up, and that Aβ clears from the human brain in a diurnal rhythm (see Research Timeline 2009Jun 2014 news). Iliff proposed that sluggish brain clearance may lead to an accumulation of Aβ or other solutes that hasten neurodegeneration, which subsequently leads to changes in sleep patterns observed in the current paper.

How might early neurodegenerative processes lead to longer sleep? Iliff commented that neurodegeneration could alter the neurocircuitry that governs sleep homeostasis, a process already known to falter with age. The authors proposed a similar hypothesis, noting that levels of the wake-promoting neuropeptide hypocretin drop in AD patients. Indirect factors such as depression and anxiety could also influence sleep, they proposed.

Ricardo Osorio of New York University School of Medicine pointed out that in addition to the circadian rhythms that sync sleep with sunlight, there is also a force called sleep pressure. It gradually builds throughout the day as more sleep-promoting molecules, such as adenosine, accumulate in the brain. Therefore, if malfunctions in brain clearance mechanisms coincide with neurodegeneration, perhaps they increase solutes that promote sleep pressure. He added that neuroinflammation could also beckon sleep, as pro-inflammatory cytokines promote drowsiness.

Osorio, the authors, and other commentators all agreed on limitations of the study. The researchers relied on participants’ self-report of how long they slept, and Osorio said people are notoriously bad at assessing how long they sleep. Furthermore, the strong influence of low education could be confounded by many variables, including socioeconomic status and lifestyle.

As studies seek to tease out the relationship between sleep and dementia, researchers need to interpret findings cautiously, commented Jurgen Claassen of Radboud University Nijmegen Medical Centre in the Netherlands. “We need polysomnography to provide objective measures of sleep time, sleep stages, sleep efficiency, arousals, apnea, restless legs, etc.,” he said.

Seshadri agreed, and told Alzforum that she is currently using available polysomnography data from about 3,000 participants in the Framingham Heart Study to link sleep quality and duration to dementia risk.—Jessica Shugart


  1. By definition, large studies like these are inherently limited to very simple measurements. In this study, something as physiologically complex as sleep is reduced to participants ticking a single box as part of a larger general questionnaire at two single time points several years apart. This is then further simplified by dichotomizing: If you sleep 8 hours and 59 minutes you are fine, but if you sleep 9 hours and one minute you are at risk of dementia. Added to this is a reduction of “completed high school education, yes-no” to a measure of education, ignoring the related differences in socio-economic status and all associated confounders.

    As we progress in this field of the relationship between sleep and dementia, we need to become more cautious interpreting data. We need polysomnography to provide objective measures of sleep time, sleep stages, sleep efficiency, arousals, apnea, restless legs, etc. A second best is actigraphy to separate time asleep from time lying in bed awake.

    In essence, the present study is not so much about sleep as about what people have indicated as their estimated sleep time at the time of the questionnaire. We don’t know if this reflects their actual sleeping patterns, or whether it is more a reflection of sedentary behavior (people who are so inactive they stay in bed until late in the morning may report this as sleeping time). The strong relationship in people without high school education suggests to me that we may actually be looking at something that has nothing to do with sleep. People with poor education are more likely to be unemployed. Employment forces regular sleep/wake patterns, whereas unemployment does not. So we may be looking at relationships between poor socio-economic status and all associated health issues (unhealthy lifestyle, lack of insurance, poor health care, unhealthy eating habits, substance abuse) and it is impossible in this study to truly correct for this.

    So, while the authors acknowledge the study’s limitations, I think there is a risk of over-interpreting the data. Nonetheless, there are some positive things to note. The study clearly demonstrates the complexity of dementia. This is highly relevant in the light of the discussion on the amyloid hypothesis and the failing anti-amyloid trials. In this discussion, AD is often presented as a simple, linear disease: There is amyloid accumulation and this leads to dementia; removing amyloid will reduce dementia. The present study identifies several factors (reported sleep time, socio-economic class) that strongly influence the risk of developing AD dementia that have not been taken into account in the design of anti-amyloid trials. This again to me indicates that late onset AD is a multifactorial disease and explains why the simplified single factor approach has failed repeatedly.

  2. Linking Sleep Duration with Pathological Brain Aging: A Novel Risk or Warning Prodrome?

    Scientific evidence connecting sleep disturbance and impaired cognitive function date back more than 120 years when the first experimental sleep deprivation study on humans was reported (Patrick and Gilbert, 1896). The focus of previous clinical studies has been on short-term sleep deprivation. Although the epidemiologic data linking sleep with dementia risk remain limited, a small number of longitudinal studies have begun to examine whether habitual short or long sleep duration may impose long-term consequences for the aging brain by increasing the risk for dementia. Recently, we demonstrated a V-shaped association between sleep duration and dementia risk in older women (Chen et al., 2015). Two other previous studies also suggested a similar pattern, but did not reach statistical significance (Benito-León et al., 2009; Yaffe et al., 2011). The epidemiologic link between short sleep and increased dementia risk has received strong support for its neurobiological plausibility, and some neuroscientists even advocate the inclusion of sleep deprivation in developing translational models (Babiloni et al., 2013) for drug discovery in Alzheimer’s disease (AD). However, the reported associations (Chen et al., 2015; Hahn et al., 2014) between prolonged sleep and increased risk for dementia, including AD, raise two possible and somewhat competing hypotheses: Whether one supports long sleep duration as a novel risk factor or considers it simply a neurobehavioral manifestation in the prodromal phase of dementia, we need strong data from high-quality studies testing both hypotheses, as each has obviously different clinical implications.

    This new report by Westwood et al. represents one important step toward better understanding the link between long sleep duration and increased dementia risk. The powerful approach with longitudinal design was clearly illustrated in this study with repeated assessments of self-reported sleep duration on two occasions, separated by more than 10 years, but both before the clinical follow-up for dementia ascertainment. Given the extensive neurocognitive outcome and brain MRI data already collected in their study cohort, this powerful design allowed the investigators to look at the change in sleep duration, how it relates to brain structure and cognitive status at base, and how it affects the subsequent risk for dementia, including AD. In the presented multivariable analyses, prolonged sleep duration (>9h vs. 6-9h) was associated with an increased risk of incident dementia (hazard ratio [HR] 2.01; 95 percent confidence interval [CI] 1.24-3.26). Change from ≤9 to >9 h of sleep was associated with an increased risk of all-cause dementia (HR 2.43; 1.44-4.11) and clinical AD (HR 2.20; 1.17-4.13). In their “sensitivity analyses,” the authors showed that the observed increase in dementia risk associated with baseline sleep >9h (vs. ≤9h) was primarily driven among those with mild cognitive impairment (MCI) at baseline. These led to the authors’ argument that long sleep duration would more likely be an early marker of neurodegeneration. Another important finding from their sensitivity analyses was that long sleep elevated dementia risk sixfold among study participants with less than high school education. This was an interesting observation, because many sleep researchers found that self-reported long sleep was more common in racial/ethnic minorities and those of low socioeconomic status. If long-duration sleep is a prodromal manifestation of early dementia, then what kind of early neurodegenerative pathologies could affect the subpopulations with presumed social adversities?  Future studies designed to answer this question may help us better understand if, and how, sleep behavior shapes the social disparities in AD and related dementias. 

    Although Westwood et al.’s is the first study designed to look into the competing hypothesis (a novel risk factor versus a warning prodrome), debates are likely to continue before we better understand the connection between long sleep and dementia risk. In large population surveys, poor performance in cognitive tests was also found in individuals who were largely cognitively intact but self-reported with long sleep duration. Unless everyone subscribes to the belief of two different neuropathological processes (linking long sleep with cognitive aging versus pathological brain aging), extant evidence from longitudinal cohorts has not completely ruled out the possibility that long sleep duration imparts novel risk. For instance, as reported by Westwood et al., even among those without MCI at baseline, the relative risk for dementia associated with sleep >9h (vs. ≤9h) was still elevated (HR=1.66; 0.89-3.08]; Table 3), and transitioning to sleep >9h was associated with an increased risk for all-cause dementia (HR=2.43; 1.31-4.54) and clinical AD (HR= 2.52; 1.26-5.04). It is important for future studies to also examine whether long-sleep duration increases the subsequent risk for MCI or even preclinical AD. If that is the case, then the evidence will favor long sleep duration as a risk factor, although it is not clear whether it is possible for Westwood et al. to pursue the analyses in this reported study population. In their post hoc analysis, the authors did adjust for prevalent conditions (cardiovascular disease, diabetes, atrial fibrillation, depression) and the observed pattern of associations remained. These adjusted results reduce the legitimate concern that statistical associations were confounded by these prior medical co-morbidities, which often influence an individual’s sleep duration and increase the risk for dementia. Considering these clinical comorbidities is important because of the extant epidemiologic evidence linking long sleep duration with increased risks for these conditions (especially stroke; see Leng et al., 2015). If long sleep duration predisposes individuals to the new occurrence of conditions that contribute to neurodegeneration, then it will become another line of supportive evidence for long sleep duration as a novel risk factor for dementia.

    Researchers in the AD research community should not be intimidated by these complexities or feel alone, since stroke research communities are searching for similar answers (Ramos and Gangwisch, 2015). Putting it all together, one may start wondering what brain regions or neural networks are most vulnerable to occult cerebrovascular damage and early neurodegeneration that might lead to long sleep duration. Understanding this intriguing neurobiological basis of long sleep may provide important insights into how cerebrovascular dysfunction and neurodegenerative pathology interact in the preclinical pathogenesis of AD. For those in favor of novel risk hypothesis, it is perhaps a good idea to work with sleep-health scientists and chronobiologists to devise an experimental paradigm that simulates the behavioral phenotype of long sleep or mimics its neurophysiological underpinning and that could help design neurobiologically-based epidemiological investigations on long sleep and pathological brain aging. 


    . On the effect of loss of sleep. Psychol Rev 3, 469-486 (1896).

    . Sleep duration, cognitive decline, and dementia risk in older women. Alzheimers Dement. 2015 Jun 15; PubMed.

    . Total daily sleep duration and the risk of dementia: a prospective population-based study. Eur J Neurol. 2009 Sep;16(9):990-7. Epub 2009 Mar 31 PubMed.

    . Sleep-disordered breathing, hypoxia, and risk of mild cognitive impairment and dementia in older women. JAMA. 2011 Aug 10;306(6):613-9. PubMed.

    . Effects of pharmacological agents, sleep deprivation, hypoxia and transcranial magnetic stimulation on electroencephalographic rhythms in rodents: towards translational challenge models for drug discovery in Alzheimer's disease. Clin Neurophysiol. 2013 Mar;124(3):437-51. Epub 2012 Oct 4 PubMed.

    . A Change in Sleep Pattern May Predict Alzheimer Disease. Am J Geriatr Psychiatry. 2013 Aug 13; PubMed.

    . Sleep duration and risk of fatal and nonfatal stroke: a prospective study and meta-analysis. Neurology. 2015 Mar 17;84(11):1072-9. Epub 2015 Feb 25 PubMed.

    . Is sleep duration a risk factor for stroke?. Neurology. 2015 Mar 17;84(11):1066-7. Epub 2015 Feb 25 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Breathe Deep—Nighttime Oxygen Loss Linked to Dementia
  2. Night Owl? Early Bird? Good Night’s Sleep May Protect the Brain
  3. From ApoE to Zzz’s—Does Sleep Quality Affect Dementia Risk?
  4. While You Were Sleeping—Synapses Forged, Amyloid Purged

Paper Citations

  1. . Total daily sleep duration and the risk of dementia: a prospective population-based study. Eur J Neurol. 2009 Sep;16(9):990-7. Epub 2009 Mar 31 PubMed.
  2. . Sleep quality and 1-year incident cognitive impairment in community-dwelling older adults. Sleep. 2012 Apr 1;35(4):491-9. PubMed.
  3. . Sleep duration, cognitive decline, and dementia risk in older women. Alzheimers Dement. 2015 Jun 15; PubMed.
  4. . Sleep duration in midlife and later life in relation to cognition. J Am Geriatr Soc. 2014 Jun;62(6):1073-81. Epub 2014 May 1 PubMed.

Other Citations

  1. Research Timeline 2009

Further Reading

No Available Further Reading

Primary Papers

  1. . Prolonged sleep duration as a marker of early neurodegeneration predicting incident dementia. Neurology. 2017 Mar 21;88(12):1172-1179. Epub 2017 Feb 22 PubMed.