Abstract Introduction The potential contributions of sleep apnea (SA) to the risk of developing many diseases remain unidentified due to survey limits in cohort studies. We aimed to identify novel associations between common comorbidities in participants with SA compared to matched controls in a clinical biobank by leveraging electronic health record (EHR) information informed by natural language processing (NLP)-based phenotyping. We also investigated the relationship between polysomnography (PSG) statistics and these diseases. Methods The sample from the Mass General Brigham Biobank was comprised of 4,876 NLP-defined SA adult cases and 12,314 controls matched on age, biological sex, BMI, race/ethnicity, and healthcare utilization with a “data floor” of minimal EHR information. Prospective (>1 year post SA diagnosis) and cross-sectional analyses considered 527 merged PheCode groups of NLP-defined diseases with ≥1% prevalence. Sex-stratified analyses were also performed. Associations with case/control status were further analyzed using rank-normalized Apnea-hypopnea Index (AHI3p) and the percentage of sleep with hypoxemia <88% (Per88) in 4,544 participants, adjusting for age, biological sex, BMI, and race/ethnicity in a smaller single-night clinic sample (57% with AHI3p ≥15). Results In prospective analyses, 170 diseases had significant odds ratios (OR) when comparing participants with SA versus controls following Bonferroni adjustment (p < 3.3 × 10-5), 8 of which were confirmed using PSG. Lead associations included a broad range of pathophysiology (e.g. chronic pulmonary heart disease, tension headache, and chronic fatigue syndrome). Diseases with higher ORs in women compared to men included chronic pulmonary heart disease, chronic renal failure, congestive heart failure not otherwise specified (NOS), and gout (p for sex interaction ≤3.2 × 10-3). Significant associations with PSG traits included hypertensive heart disease, fluid overload, and heart failure NOS (p ≤7.8 × 10-5). 281 diseases significantly differed in cross-sectional analyses, 41 of which were confirmed using PSG. 37 of these 41 associated diseases had lower p-values for Per88 compared to the AHI3p. Conclusion Sleep apnea may contribute to disease risk across a broad range of pathophysiology, with increased sex-specific risks for multiple common comorbidities. Future work should investigate the specific role of hypoxemia in these associations. Support (If Any) NIH (K01-HL135405, R35-HL135818, U01-HG008685 and P30-AR070253).
Read full abstract