Leveraging research and clinical cohorts to improve Obstructive Sleep Apnea subtyping - Obstructive Sleep Apnea (OSA), characterized by recurrent pharyngeal obstructive events during sleep, affects as many as 38% of adults, most of whom are not diagnosed and treated, and thus at potential risk for daytime impairment and long-term health problems. Efforts to implement evidence-based approaches for improving diagnosis and treatment have been impeded by the lack of data from longitudinal studies and clinical trials that address which patients are at greatest risk for cardiovascular disease and other health problems, and who is most likely to benefit from intervention. Notably, the standard disease-defining metric, the Apnea Hypopnea Index (AHI), the number of breathing pauses during sleep, does not characterize the heterogeneity of OSA mechanisms, the physiological impact of the breathing disturbances, risk for adverse outcomes, and likelihood of responding to treatment. Our multi-disciplinary team will address this gap by integrating multiple data sources to derive physiologically informative individual and composite metrics of OSA and characterize their associations with mechanistic traits, symptoms, risk factors, molecular markers, and clinical outcomes. We will cost-effectively utilize data from: a) existing large prospective research cohorts (including data within the National Sleep Research Resource and Trans-Omics in Precision Medicine; n≈29,000; b) newly ingested sleep and clinical outcome data from a large clinical biobank; n≈30,000; and c) a prospective study of patients with OSA recruited to examine the short-term reproducibility of OSA metrics. In Aim1, we will quantify the reproducibility of OSA severity markers (i.e., that quantify ventilatory reduction and hypoxia, heart rate, and sleep fragmentation responses to obstructive events) in a diverse sample of patients with a range of AHI studied twice over ≈ 2 weeks with polysomnography. Using the most reproducible metrics, we will derive OSA composites that are potentially more informative than individual metrics and describe OSA subtypes. In Aim 2, we aim to further characterize the clinical and physiological heterogeneity of individual OSA metrics and composites by describing their associations with symptoms and socio-demographic, health-related, metabolomic, and endotypic markers. We will use genetic instruments and Mendelian Randomization to study associations of OSA metrics/composites with each other and with common co-morbidities (e.g., obesity, diabetes, hypertension, etc) and candidate biological pathways (inflammation, lung function, etc) - clarifying which associations reflect shared genetics, indirect effects, and/or direct causal effects. In Aim 3, we will use existing longitudinal and clinical trial data to evaluate which OSA metrics/composites predict long term health outcomes and OSA treatment response. Scalability will be addressed by identifying parsimonious sets of metrics that can be most readily incorporated into clinical practice. With professional society and patient input, we will develop and disseminate a foundational OSA classification (taxonomy), annotated for relevance by sex and social determinants of health, with the goal of improving the understanding of the heterogeneity of OSA, providing a foundation for OSA precision medicine.
