Two sleep recordings linked deep-sleep loss with later dementia

A night of sleep is not only a count of hours. In a 2023 Framingham Heart Study analysis, older adults generally kept a similar total amount of sleep as they aged, while the share spent in slow-wave sleep gradually fell. Slow-wave sleep is commonly described as deep sleep, the stage identified here through an overnight recording of brain and other body signals.

The study associated a larger decline in that deep-sleep share with a greater chance of developing dementia during later follow-up. It did not test a sleep treatment, and it does not show that restoring deep sleep would prevent dementia. Its contribution was more specific: it followed change in the same people across two detailed sleep assessments, then examined what happened afterward.

The study followed change, not a single night

The researchers analyzed 346 adults who were at least 60 years old and free of dementia at their second sleep assessment. Their mean age was 69, and 179 participants were women. The participants came from the Framingham Heart Study and completed overnight polysomnography, or PSG, in two periods: 1995 to 1998 and 2001 to 2003.

PSG is a laboratory-style overnight recording that includes brain activity along with other measurements used to identify sleep stages. The two assessments were an average of 5.2 years apart, with individual intervals ranging from 4.8 to 7.1 years. The researchers calculated each person’s annualized change from those two observations. Sleep was not measured continuously or once every year.

After the second assessment, the cohort was followed for dementia for an average of 12 years, with follow-up reaching 17 years. During that period, 52 participants developed incident dementia. Forty-four of those cases were consistent with Alzheimer disease dementia.

The sequence matters. First came a measurement of sleep, then a second measurement years later, followed by surveillance for dementia. This design provides a stronger view of long-term change than a single sleep score, while still leaving open questions about why the change occurred and what it means for an individual.

What the annual decline and 27% estimate mean

The percentage of sleep spent in slow-wave sleep fell by an average of about 0.57 percentage points per year between the two recordings. At the cohort level, total sleep time stayed stable. Participants were not simply sleeping fewer hours overall; the composition of those hours changed as deep sleep occupied a smaller share.

The dementia analysis used an annualized measure of that change. After adjustment for age, sex, study cohort, APOE ε4 status, smoking, and use of sleep, antidepressant, and anxiolytic medications, each additional one-percentage-point decrease in slow-wave sleep per year was associated with a 27% higher relative hazard of all-cause dementia. The reported hazard ratio was 1.27, with a 95% confidence interval from 1.06 to 1.54.

A simple hypothetical example helps clarify the unit. Suppose a person’s deep-sleep share fell from 20% to 15% over five years. That would be a five-percentage-point drop across the interval, or one percentage point per year on average. The study’s estimate describes how that annualized difference related to dementia hazard when comparing people in the cohort. It does not say that the person’s risk rises by 27% every calendar year, nor does it convert the result into an individual probability or an absolute increase in cases.

The researchers also tested whether the association was sensitive to different analytical choices. The result did not meaningfully change when analyses accounted for the first two years of dementia follow-up, the earlier deep-sleep measurement, sleep apnea and low oxygen levels, vascular risk factors and physical activity, daytime sleepiness, or total sleep time. These checks make the association less dependent on any one of those measured factors. They do not turn a relative cohort estimate into a personal forecast.

The genetic result narrowed one part of the question

The study also examined APOE ε4, a genetic variant associated with increased Alzheimer disease risk. Compared with participants carrying the ε3/ε3 reference combination, APOE ε4 carriers had greater declines in slow-wave sleep over the interval between the two PSG recordings.

That finding raised a more specific question: could the loss of deep sleep help explain why APOE ε4 is associated with dementia? A mediation analysis estimated that 17% of the APOE ε4 association with dementia operated through changes in slow-wave sleep. However, the indirect effect was not statistically significant. The study therefore did not establish that deep-sleep loss mediates APOE ε4-related dementia risk.

Hippocampal volume supplied another prospective test. Lower volume in the measured comparison was not significantly associated with subsequent slow-wave sleep change. That result does not rule out early neurodegenerative processes. It only says that the hippocampal measure used in this analysis did not show the reported association with later deep-sleep decline.

Why the association is still hard to interpret

The authors proposed that slow-wave sleep could support the brain’s clearance of proteins involved in Alzheimer disease. In that account, losing more deep sleep might reduce processes sometimes described as glymphatic clearance and contribute over time to protein accumulation. This was a proposed explanation, not a mechanism demonstrated by the cohort’s measurements.

The study did not have gold-standard Alzheimer disease biomarkers available. Adjustment for APOE status, vascular factors, medications, and other measured characteristics can reduce the influence of some alternative explanations, but it cannot settle the direction of the relationship. Deep-sleep loss could contribute to later disease, developing disease could affect sleep before dementia is diagnosed, or both could reflect other biological or health-related causes.

The population also sets the boundaries for interpretation. These were older adults with a mean age of 69, and 84% identified as White. The findings may not apply in the same way to younger people or to populations with different health, genetic, social, and demographic profiles. Broader applicability requires separate evidence.

The most useful conclusion from this study is therefore not that deep sleep has a universal target or that a measured decline identifies an impending diagnosis. It is that a longitudinal change in one sleep stage was associated with later dementia in this cohort, while the causal pathway remained unresolved.

A consumer sleep score cannot provide a dementia forecast

The research recordings and a consumer sleep score answer different questions. PSG measured sleep overnight with brain activity and other physiological signals, allowing researchers to classify sleep stages in a standardized study protocol. A watch or other consumer device may provide an estimate of deep sleep from its own sensors and algorithms.

The paper’s discussion cautioned that wearable estimates of N3 sleep, another term for deep sleep, can be inaccurate compared with PSG. A single reading also lacks the repeated measurement and long dementia follow-up that defined this study. Even a sequence of consumer readings would not automatically reproduce the study’s annualized exposure or its statistical comparison.

For that reason, the 27% figure should not be attached to a person’s watch display. The study did not validate a wearable-based dementia score, establish a diagnostic threshold, or identify a device that can predict who will develop dementia. Its result belongs to a carefully defined comparison within an older research cohort.

What this finding could mean for dementia prevention

Preserving deep sleep could eventually give older adults another way to reduce dementia risk and maintain independence, but that benefit depends on two findings that this study did not provide. Researchers would first need to establish that the relationship is causal and modifiable. They would then need to show that a safe intervention that preserves or improves slow-wave sleep also improves dementia outcomes.

That possibility was reasonable to investigate because the authors identified slow-wave sleep loss as a potentially modifiable risk factor. At the publication date, however, the human payoff remained conditional. The study supplied a longitudinal association and a direction for further research; it did not yet show that changing sleep changes the course of dementia.

Key Sources

• Primary source: Association Between Slow-Wave Sleep Loss and Incident Dementia.
  DOI or Journal Page · Peer-Reviewed Paper · 2023-12-01