Abstract
Insomnia symptoms in combination with extreme habitual sleep durations may be associated with long-term dementia incidence, which we explored using multivariable-adjusted regression and genetic association analyses in Europeans. European-ancestry participants free of baseline dementia (N = 422,106; 5288 new cases over 12.7 years follow-up) from UK Biobank were stratified into six groups based on the combination of self-reported insomnia symptoms (no/yes) and short (4-6 h), average (7-8 h), and long (9-14 h) sleep duration. We assessed their associations with incident dementia and baseline cognitive function by multivariable-adjusted regression models. In addition, we grouped participants based on the combination of genetic scores for sleep duration and insomnia symptoms, and repeated the analyses accordingly. Compared to participants reporting no insomnia symptoms and average sleep duration at baseline, both long (hazard ratio 1.27, 95% confidence interval [1.15, 1.39]) and short (1.19 [1.09, 1.30]) sleep duration without concomitant insomnia symptoms were associated with increased dementia risk. Similarly, both short and long sleep duration, stratified by concomitantly reported insomnia symptoms, were associated with reduced function in different cognitive domains. Conversely, the genetic risk of insomnia symptoms, and not sleep duration, was associated with dementia (OR = 1.07 [1.02, 1.13]) and prospective memory result (OR = 1.07 [1.02, 1.14]). We observed no evidence for synergism between sleep duration and insomnia symptoms on the study outcomes. In conclusion, while long sleep duration was associated with both incident dementia and different cognitive domains, no such evidence was obtained from the genetic analyses, indicating the potential of disturbed sleep as a marker of future dementia rather than a causal risk factor.