Accelerometer-derived 24-hour behavior composition and incident clinically diagnosed sleep apnoea: a prospective cohort study

Physical activity, sedentary behavior, and sleep form a finite 24-hour behavioral composition. Whether accelerometer-derived composition and specific sedentary-time reallocation contrasts are associated with incident clinically diagnosed sleep apnoea remains unclear. We analyzed 93,262 UK Biobank participants without clinically diagnosed sleep apnoea before accelerometer assessment. Daily time in sleep, sedentary behavior, light physical activity, and moderate-to-vigorous physical activity (MVPA) was derived from wrist-worn accelerometry and analyzed using isometric log-ratio transformation. The primary outcome was incident hospital-recorded sleep apnoea (ICD-10 G47.3). Cox models used attained age as the timescale, with missing covariates handled using multiple imputation. Compositional analyses estimated hazard ratios (HRs) for reallocating 30 min/day between behaviors. Sensitivity analyses included the broader linked-record outcome, competing-risk models, obesity-stratified analyses, and other robustness analyses. A metabolomic risk score (MRS) was cross-fitted using the hospital outcome among participants with at least one observed candidate metabolite; full-cohort cross-fitting was a sensitivity analysis. During a median follow-up of 8.15 years, 816 participants had a first hospital-recorded sleep apnoea diagnosis. A model-based 30-min/day reallocation from sedentary time to MVPA showed the strongest inverse association (HR, 0.87; 95% CI, 0.83–0.90), followed by reallocation to sleep (HR, 0.94; 95% CI, 0.91–0.96). Reallocation to light physical activity showed little association (HR, 0.99; 95% CI, 0.97–1.01). The broader outcome and age-adjusted competing-risk analyses yielded similar association patterns. Among 91,197 participants with observed metabolite measurements (801 events), each 1-SD higher MRS was associated with a higher hazard (HR, 1.17; 95% CI, 1.06–1.29). Accelerometer-derived 24-hour behavior composition was associated with incident hospital-recorded sleep apnoea. The sedentary-to-MVPA contrast showed the clearest inverse association. The cross-fitted MRS was also associated with hospital-recorded sleep apnoea after covariate adjustment.