Hospital-diagnosed sleep disorders and incident dementia: a nationwide observational cohort study

Abstract

Background and purpose: Several smaller, community-based studies have suggested a link between sleep disorders and dementia with a focus on sleep as a modifiable risk factor for dementia. Studies on neurodegenerative diseases are prone to reverse causation, and few studies have examined the association with long follow-up time. Our aim was to explore the possible association between sleep disorders and late-onset dementia in an entire population.

Methods: In a nationwide cohort with 40-year follow-up, associations between hospital-based sleep disorder diagnoses and late-onset dementia were assessed. Incidence rate ratios (IRR) were calculated using Poisson regression.

Results: The cohort consisted of 1,491,276 people. Those with any sleep disorder had a 17% higher risk of dementia (IRR 1.17, 95% confidence interval [CI] 1.11-1.24) compared to people with no sleep disorder, adjusted for age, sex, calendar year, highest attained educational level at age 50, and somatic and psychiatric comorbidity. The risk of dementia was significantly increased 0-5 years after sleep disorder diagnosis (IRR 1.35, 95% CI 1.25-1.47), whilst the association after 5 years or more was non-significant (1.05, 95% CI 0.97-1.13).

Conclusions: Our findings show an increased short-term risk of dementia following a hospital-based sleep disorder diagnosis, whilst weaker evidence of a long-term risk was found. This could potentially point towards sleep disorders as an early symptom of dementia. Further research is needed to distinguish sleep disorders as an early symptom of dementia, a risk factor, or both.

Keywords: dementia; epidemiology; risk factors; sleep apnea syndromes; sleep-wake disorders.

FIGURE 1

Flowchart of study population.

FIGURE 2

IRRs of dementia in individuals by sleep disorder category and sex. For the reference group (people with no diagnosis of the sleep disorder[s] analyzed), the IRR is equal to 1 (as indicated by the dotted vertical line). Error bars represent 95% confidence intervals (CIs). IRRs were estimated using Poisson regression. A two‐sided type 1 error of 5% was considered statistically significant. Total person‐years were 180,328 in people with sleep disorder and 13,899,476 in people with no sleep disorder. The IRRs presented are adjusted for age, sex, calendar year, highest attained educational level at age 50, somatic comorbidity (CCI) and psychiatric comorbidity (model 3). Adjustment models 1 and 2 are presented in Table S6. There were too few patients with narcolepsy and dementia to facilitate sub‐analyses.

FIGURE 3

IRRs of dementia in individuals by sleep disorder category and age group. For the reference group (people with no diagnosis of the sleep disorder[s] analyzed in that age group), the IRR is equal to 1 (as indicated by the dotted vertical line). Error bars represent 95% confidence intervals (CIs). IRRs were estimated using Poisson regression. A two‐sided type 1 error of 5% was considered statistically significant. Total person‐years were 100,866 in people with sleep disorder age 50–64 and 13,978,938 in people with no sleep disorder age 50–64. Total person‐years was 98,885 in people with sleep disorders age ≥65 and 13,980,920 in people with no sleep disorder age ≥65. The IRRs presented are adjusted for age, sex, calendar year, highest attained educational level at age 50, somatic comorbidity (CCI) and psychiatric comorbidity (model 3). Adjustment models 1 and 2 are presented in Table S7. There were too few patients with narcolepsy and dementia to facilitate sub‐analyses.