Multidimensional physiological burden of nocturnal respiratory events improves the prediction of cardiovascular mortality in obstructive sleep apnoea

Background Obstructive sleep apnoea (OSA) is a prevalent cardiovascular (CV) risk factor, yet the clinical reliance on the apnoea-hypopnoea index (AHI) is increasingly questioned. The AHI is a crude frequency count that ignores the physiological heterogeneity of respiratory events, such as their duration, intensity, and temporal organization across the night. This "metric failure" likely contributes to the neutral findings of randomized trials regarding CV outcome improvements through AHI reduction. There is an urgent need for multidimensional metrics that capture the actual "physiological burden" of OSA. Purpose We aimed to evaluate whether a novel composite metric, the event-weighted desaturation burden (EWDB)—which integrates event intensity, temporal distribution, and cumulative hypoxic dose—improves CV risk stratification and mortality prediction compared to established severity measures in a large prospective cohort. Methods Data from 4,904 community-dwelling adults in the Sleep Heart Health Study were analysed. EWDB was derived from baseline polysomnography by combining three physiological dimensions: (1) the Respiratory Weighted Index (RWI), assigning higher weights to apnoeas (1.0) than hypopnoeas (0.3) to reflect hemodynamic impact; (2) the Distribution Factor (DF), quantifying the temporal uniformity of events to distinguish sustained stress from clustered patterns; and (3) the Hypoxic Burden (HB), measuring the cumulative event-linked desaturation area. The primary outcome was adjudicated CV mortality over a median follow-up of 11.0 years. Multivariable Cox models were adjusted for demographics, cardiometabolic risk factors, sleep architecture, and AHI. Results A total of 341 CV deaths occurred during follow-up. In fully adjusted models, EWDB demonstrated a robust, dose-response association with mortality; participants in the highest tertile faced a 72% higher risk of CV death compared to those in the lowest (HR 1.72; 95% CI 1.28–2.33; P-trend < 0.001). Conversely, the independent associations for AHI (HR 1.21; 95% CI 0.85–1.72; P=0.128) and HB (HR 1.25; 95% CI 0.91–1.73; P=0.167) lost statistical significance after full adjustment. Each 1-SD increase in EWDB was associated with a 13% risk elevation (HR 1.13; 95% CI 1.05–1.22). Incorporating EWDB into a comprehensive clinical model significantly improved risk stratification, yielding a continuous net reclassification improvement (NRI) of 0.507 and a superior C-index (0.807 vs 0.803) compared to models using AHI. Conclusions EWDB captures distinct physiological risk signatures that simple event counts neglect. Transitioning from frequency-based diagnostics to multidimensional physiological burden metrics provides superior CV risk stratification and may refine candidate selection for targeted prevention strategies in patients with OSA.