Repeated sleep-medication reporting patterns and cognitive change in HRS and SHARE
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Self-reported sleep-medication use has been inconsistently linked to cognitive decline, largely because sleep disturbances, affective symptoms, and treatment transitions co-evolve over time. Moreover, whether repeated sleep-medication reports across consecutive survey waves follow a monotonic dose-response gradient with subsequent cognitive aging remains unclear. We evaluated baseline, two-wave, and three-wave longitudinal patterns of sleep-medication reporting in relation to cognitive trajectories across two large aging cohorts in the United States and Europe.
Methods
We analyzed population-based data from adults aged ≥50 years in the US Health and Retirement Study (HRS) and the Survey of Health, Ageing and Retirement in Europe (SHARE). First, we examined associations between baseline sleep-medication use and longitudinal changes in standardized global cognition and memory scores (Analysis 1: HRS n = 16,323; SHARE n = 41,453). Second, we conducted a prespecified two-wave landmark analysis classifying participants by reported sleep-medication use across two consecutive visits (HRS waves 11–12; SHARE waves 4–5) as 0, 1, or 2 use-positive visits, and by reporting sequence (00, 01, 10, and 11), with cognitive trajectories modeled from the second visit onward (Analysis 2: HRS n = 13,915; SHARE n = 29,214). Linear mixed-effects models with random intercepts and slopes adjusted for baseline sociodemographic, lifestyle, and clinical covariates and all covariate-by-time interactions. Third, we extended the landmark window to three consecutive visits (HRS waves 11–13, 2012–2016; SHARE waves 6–8, 2015–2020) to separate sustained from transient reporting, adjusting for interview mode and testing equivalence of sustained use to non-use (Analysis 3: HRS n = 12,310; SHARE n = 24,692).
Results
In baseline analyses, sleep-medication use was associated with modestly faster decline in global cognition (HRS: β = −0.0060 SD/year, 95% CI −0.0103 to −0.0017; SHARE: β = −0.0034 SD/year, 95% CI −0.0070 to 0.0002) and memory (HRS: β = −0.0062 SD/year, 95% CI −0.0115 to −0.0009; SHARE: β = −0.0037 SD/year, 95% CI −0.0074 to 0.0001). In the two-wave landmark analysis, reporting sleep-medication use at both visits (2 vs 0 visits) was not associated with accelerated cognitive decline in either cohort, refuting a monotonic consistency gradient. Instead, reporting use at a single visit (1 vs 0 visits) was associated with steeper decline in Western Europe (global cognition: β = −0.0087 SD/year, 95% CI −0.0157 to −0.0016; memory: β = −0.0083 SD/year, 95% CI −0.0158 to −0.0007) and among SHARE participants with secondary/vocational or tertiary education (interaction P = 0.041 for global cognition and P = 0.017 for memory). In the three-wave extension, no exposure contrast survived familywise correction; sustained three-visit reporting was statistically equivalent to persistent non-use within ±0.05 SD/year in both cohorts (equivalence P ≤ 0.0012).
Conclusions
Baseline sleep-medication reporting was associated with small negative differences in cognitive trajectories, whereas a fixed two-wave landmark design did not support a monotonic cumulative-exposure gradient. The localization of steeper cognitive decline to intermittent (one-visit) reporting in Western Europe and higher educational strata points to the role of treatment transitions and symptom instability rather than sustained medication exposure. Three-wave, mode-adjusted and equivalence-tested analyses indicate that sustained sleep-medication reporting is not accompanied by accelerated cognitive decline at the population level, whereas transient reporting continues to mark periods of elevated risk.
