Evaluating the nonlinear effects of sleep duration on biological aging across phenotypic, genomic, and epigenomic data

Short and long sleep durations have been inconsistently linked to aging and health outcomes, potentially due to underexplored nonlinear associations. Using phenotypic and genomic data from the UK Biobank (n=442,664), we applied multivariable linear regression, restricted cubic splines, and Mendelian randomization (MR) to analyze nonlinear relationships between self-reported sleep duration and biomarkers of accelerated aging: PhenoAge acceleration (PhenoAgeAccel), BioAge acceleration (BioAgeAccel), and leukocyte telomere length (LTL). Functional annotation analyses were performed to assess potential shared biological pathways using epigenomic profiles. Observational analyses supported U-shaped phenotypic associations between sleep duration and PhenoAgeAccel/BioAgeAccel, with optimal sleep around 7 h/d. For LTL, linear models suggested a U-shape, while spline models indicated an inverted reverse J-pattern. MR analyses corroborated the deleterious impacts of insufficient, but not excessive, sleep, by revealing a threshold nonlinear relationship between increasing genetically-predicted sleep duration up to 7 h/d and lower PhenoAgeAccel/BioAgeAccel, and a linear relationship with longer LTL. Cell-type enrichment analyses connected short sleep to BioAgeAccel/LTL through pathways related to muscle maintenance and immune function. These findings suggest that extending sleep may mitigate accelerated aging, though further research is needed to clarify the underlying biological mechanisms and whether excessive sleep also contributes causally to biological aging.