Research Square · Nov 18, 2026
Association of longitudinal hypoxic burden with cardiovascular mortality depends on apnea–hypopnea index adjustment: a Sleep Heart Health Study analysis
By Zhang B, SAU S, Yang M et al.
Background Hypoxic burden (HB) captures respiratory event-related hypoxic exposure, but its longitudinal association with mortality may depend on how changes in event frequency are accounted for. We examined whether the association between HB change and cardiovascular mortality depends on apnea–hypopnea index (AHI) definition and adjustment for concurrent AHI change. Methods This post hoc Sleep Heart Health Study reanalysis used paired polysomnography and follow-up from the second examination. Five Cox models compared no AHI adjustment, baseline AHI adjustment, and baseline-plus-change adjustment using 3% and 4% desaturation-threshold indices. Samples, clinical adjustment and HB scaling were fixed within each outcome. Cardiovascular models were stratified by race. Paired participant bootstrap resampling quantified the difference between HB-change coefficients under full AHI3 and AHI4 adjustment. Results Among 2,485 participants, 104 cardiovascular deaths occurred. Per fixed standard deviation of log-HB change, hazard ratios were 1.227 (95% confidence interval 0.987–1.525) without AHI adjustment, 1.685 (0.964–2.948) with full AHI3 adjustment, and 1.012 (0.611–1.679) with full AHI4 adjustment. The paired coefficient difference was 0.510 (1,000-replicate bootstrap interval 0.119–0.913). HB-change variance inflation factors were 7.66 and 5.31. Single-record omission and selected time-varying effects produced similar contrasts. The all-cause mortality contrast interval included zero. Conclusions The conditional HB-change estimate depended on AHI definition and concurrent-change adjustment. These exploratory findings support transparent specification of adjustment targets; they establish neither a causal HB effect nor superiority of either AHI definition.