Sleep Problems and Sleeping Pills Add Little to Dementia Risk Models for Cognition

Sleep disturbances and the medications used to treat them have long been suspected of quietly eroding thinking skills in older adults, but a new study suggests that, at least in one group of patients, they may add surprisingly little explanatory power once the better-known risk factors for dementia are taken into account. In a cross-sectional analysis published in Ageing International, researchers examined whether insomnia symptom severity, the dose of benzodiazepine receptor agonist hypnotics people were taking, and the risk of obstructive sleep apnea predicted performance on standard cognitive tests above and beyond a set of established dementia risk factors. The answer, across four different cognitive measures, was no: none of the sleep-related variables significantly improved the fit of statistical models that already included age, education, vascular risk, depression, and other classic contributors.

The work was carried out by a team led by Lydia Ann and Constance H. Fung, drawing on baseline data from a hypnotic deprescribing clinical trial registered as NCT03687086. Participants were patients without dementia recruited from one Veterans Affairs site and one academic site, all of whom were already taking at least one of five common benzodiazepine or z-drugs: alprazolam, clonazepam, lorazepam, temazepam, or zolpidem. A total of 348 participants enrolled, with a mean age of 68.8 years, 28 percent women, and an average of 15.8 years of education.

The motivation for the study rests on a substantial body of prior research. The number of people living with cognitive impairment or dementia is projected to rise from 55 million in 2019 to 139 million in 2050, and the Lancet Commission’s 2024 update lists a series of modifiable risk factors, including less education, hearing loss, untreated vision loss, high LDL cholesterol, smoking, obesity, depression, physical inactivity, diabetes, excessive alcohol consumption, traumatic brain injury, air pollution, and social isolation. Sleep disorders, meanwhile, have been linked to cognitive decline and Alzheimer’s disease through pathways that include increased beta-amyloid burden, amyloid-driven tau phosphorylation, systemic inflammation, oxidative stress, and reduced glymphatic clearance of metabolic waste from the brain.

Each component of the sleep picture carries its own literature. Obstructive sleep apnea, a common cause of sleep disturbance, has been shown to impair attention, verbal memory, executive function, and learning. Insomnia is associated in some but not all studies with deficits in working memory, episodic memory, and aspects of executive functioning. Hypnotic medications themselves, including benzodiazepines and non-benzodiazepine receptor agonists, have been associated with impairments in memory, learning, attention, and visuospatial ability, as well as with major cognitive decline and dementia, although mild memory impairment from short-term, low-dose benzodiazepine use appears dose-dependent and reversible after discontinuation. What has been missing, the authors argue, is a way to weigh these factors against one another and against conventional dementia risks in the same population.

The results were strikingly uniform. The base model explained an R-squared of 0.139 of the variance in Mini-Mental State Examination scores among 313 participants, 0.241 for Trail Making Test A among 298 participants, 0.229 for Trail Making Test B among 287 participants, and 0.189 for the Digit Symbol Substitution Test among 285 participants. Adding the insomnia index, the hypnotic dose, the apnea risk category, and finally the interaction term changed those figures only marginally: the full Model 5 reached R-squared values of 0.142 for the mental state examination, 0.246 for Trail A, 0.238 for Trail B, and 0.190 for the substitution test. Every F-test comparing the nested models returned a p-value above 0.145, well short of the 0.05 threshold for statistical significance.

In practical terms, the change in explained variance attributable to each sleep variable ranged from essentially zero to about 0.006 depending on the outcome, and none of the four hypotheses the team set out to test was supported. Higher hypnotic dose did not predict lower cognition beyond the base factors, higher apnea risk did not either, and the combination of the two showed no special interaction effect. The sample’s own characteristics help frame these null results: the mean Insomnia Severity Index score was 14.2 with a standard deviation of 6.6, the mean seven-day hypnotic dose was 29.5 diazepam equivalents with a standard deviation of 31.4, and 44 percent of participants were classified as high risk for obstructive sleep apnea. Mean cognitive scores reflected a generally intact group, with an average mental state examination score of 28.2 out of 30, an average Trail A time of 48.2 seconds, and an average Trail B time of 130.2 seconds.

The authors interpret the findings cautiously rather than as a refutation of sleep-cognition research. They note that the sleep-related variables may not exert independent, additive effects because they overlap with the pathways already captured by the base model, and that the population studied was relatively well educated and cognitively intact, leaving limited room for additional variation to be detected. They also point out that no prior study, to their knowledge, has examined the incremental contribution of sleep-related factors relative to other dementia risk factors and to one another in this nested fashion, an approach they say offers an empirical basis for clinical scenarios where addressing every risk factor simultaneously is not feasible.

Several limitations temper the conclusions. The sample consisted entirely of people already prescribed hypnotics, only 28 percent were women, positive airway pressure adherence information was unavailable, and the Mini-Mental State Examination is less sensitive than tools such as the Montreal Cognitive Assessment for detecting early impairment. The authors call for future studies comparing people who do and do not take insomnia medications, and for work in populations with more varied cognitive status, including younger patients aged 45 to 65 with mild cognitive impairment, where any reversible effects of sleep treatments on cognition might be easier to observe. For now, the message is a measured one: in this group of older hypnotic users, the well-established dementia risk factors carried the explanatory weight, and the sleep variables added none that could be statistically detected.

Subject of Research: A cross-sectional study of whether insomnia symptoms, hypnotic dose, and obstructive sleep apnea risk predict cognitive performance beyond established dementia risk factors.

Article Title: Sleep and Cognitive Performance: Exploring Benzodiazepine Receptor Agonist Use, Insomnia Symptoms, and Obstructive Sleep Apnea

Article References: Ann, L., Te, T., Mitchell, M., Alessi, C., Martin, J., Dzierzewski, J. M., Josephson, K., Naeem, S., Der-Mcleod, E., Ghadimi, S., Zeidler, M., Badr, M. S., Moore, A., & Fung, C. H. (2026). Sleep and Cognitive Performance: Exploring Benzodiazepine Receptor Agonist Use, Insomnia Symptoms, and Obstructive Sleep Apnea. Ageing International, 51(4), Article 48. https://doi.org/10.1007/s12126-026-09684-6

Image Credits: AI Generated

DOI: 10.1007/s12126-026-09684-6

Keywords: sleep, cognition, insomnia, benzodiazepines, z-drugs, obstructive sleep apnea, dementia risk factors, older adults, MMSE, nested regression, hypnotics, veterans