0484 The Effect of Statistical Analysis Method on the Trend of Obstructive Sleep Apnea-Induced Heart Rate Change: The Multi-Ethnic Study of Atherosclerosis

Abstract Introduction Obstructive sleep apnea (OSA) causes intermittent hypoxia and autonomic activation, producing heart rate (HR) changes that reflect cardiovascular stress. Although HR changes during OSA events are widely studied as noninvasive biomarkers, there is no standardized analysis for quantifying these changes, which has led to inconsistent findings. A systematic evaluation is needed to elucidate how analysis choices influence observed HR changes. Methods We analyzed polysomnography data from 2,035 participants with validated OSA annotations in the Multi-Ethnic Study of Atherosclerosis Sleep Ancillary Study. ECG-derived HR changes in pre-, intra-, and post-OSA events were evaluated across various (1) HR metrics (mean, quartile, and max–min), (2) analysis window durations for pre- and post-OSA events, (3) OSA segmentations, and (4) OSA episode definitions (combined vs. separate pre-/intra-OSA periods, and using the initial, full, or final portion of the intra-OSA events). Results Quartile and max–min metrics produced about 5 times and 8 times larger pre–intra-OSA and intra–post-OSA HR changes than mean metric on average. However, max–min metric showed twice larger variability than quartile metric due to its noise sensitivity. 10–15s analysis window estimated 25–85% larger pre–intra- and intra–post-OSA HR changes than 5s analysis window for all metrics. When analyzing all OSA events, pre-OSA HR may be contaminated by previous post-OSA and thus pre–intra-OSA HR changes were larger relative to when considering OSA events separated by ≥60s with nearly identical results for even larger separations, indicating that 60s is sufficient to exclude potential interference between adjacent OSA events. Combining pre- and intra-OSA periods as intra-OSA period vs. treating them separately affected intra–post-OSA HR changes (mean metric: 15% ↓, max–min metric: 10% ↑, quartile metric: no change). Finally, intra-OSA HR was more interpretable when initial portion or full intra-OSA events were used, whereas using only the final portion often produced patterns contradicting physiological expectations. Conclusion Analysis choices influence observed HR changes induced by OSA. Optimizing analyses will improve reproducibility and physiological interpretability of HR-based biomarkers in response to OSA. Support (if any)