Pooling HCHS/SOL, MESA and ARIC (N=11,563), long sleep (9+ hours) tracked with lower baseline global cognitive function (β=-0.19, p=3.8x10-14) and 77% higher odds of poor cognition (OR=1.77, 95% CI 1.49-2.09), with consistent results across cohorts. Insomnia showed a weaker, suggestive link (OR=1.22; FDR p=0.05, with notable heterogeneity). Short sleep, daytime sleepiness and obstructive sleep apnea showed no consistent association, and no sleep measure predicted cognitive change over follow-up.
The lack of association with cognitive change is the key caveat. Because sleep and cognition were linked only cross-sectionally, long sleep may be a marker of existing brain changes rather than a cause; earlier cohort work has similarly suggested that prolonged sleep can accompany preclinical neurodegeneration, depression or poor health. Reverse causation therefore looms large, and the findings do not show that sleeping less would protect cognition. Sleep was also self-reported and dichotomized at crude thresholds, and the insomnia result sits at the edge of statistical significance. The strength lies in harmonized definitions across diverse populations, which addresses a real inconsistency in the literature. For adults, the practical takeaway is modest: persistently long sleep, or insomnia, is worth mentioning to a clinician as a possible signal rather than something to self-correct. This is a preprint that has not been peer reviewed, and results may change. Editorially, the work is confirmatory and incremental, strengthening the long-sleep signal without clarifying its direction.