Investigators reported that associations with obstructive sleep apnea concentrated most consistently in cardiometabolic domains. Short, direct fact: hypertension and metabolic dysfunction appeared repeatedly. The cohort-level signals included impaired glucose metabolism, hyperlipidemia, obesity, and heart failure, and these diagnostic associations had concordant laboratory and medication signals, as well as cardiac-measurement correlates described by the authors. The manuscript lists specific phecodes and aggregate results in its supplementary materials; individual-level data are protected and not public.
The study team analyzed 222,109 adults who underwent diagnostic sleep testing within Kaiser Permanente Southern California. They used the apnea-hypopnea index as the primary exposure and examined cross-sectional associations across multiple electronic health record domains: diagnoses, laboratory tests, medications, procedures, healthcare utilization, behaviors, and symptoms. The investigators additionally evaluated associations with incident diseases and created data-driven phenotypes that combined signals across domains. The preprint describes the broad analytic strategy, but the source did not provide detailed model specifications or variable-by-variable covariate lists in this abstract; those details are located in the full manuscript and supplementary files.
After full adjustment and application of Bonferroni correction, the authors reported 169 prevalent phecodes and 16 incident phecodes that remained statistically significant. The report emphasizes that these retained associations reflect a broad set of clinical endpoints across the phenome. The investigators noted that many positive associations converged on cardiometabolic disease clusters. The manuscript provides the corrected results and the full phecode lists in the main text and supplementary materials for replication and follow-up.
Data indicated concordant signals across domains: laboratory results, medication records, and cardiac measurements mirrored diagnostic associations. Briefly: medication and lab patterns supported the diagnosis-level associations rather than contradicting them. The authors presented these concordant cross-domain signals as reinforcing the cardiometabolic link to obstructive sleep apnea, but the preprint itself should be consulted for the specific lab panels, medication classes, and cardiac metrics analyzed; the abstract does not enumerate those variable names.
The study team derived multidimensional physiologic, symptom, and comorbidity phenotypes using a data-driven approach. These composite phenotypes produced heterogeneous patterns of disease incidence even after statistical adjustment; importantly, those incidence patterns were not aligned solely with apnea-hypopnea index severity. Short statement: AHI alone did not capture the full heterogeneity observed. The authors argue that combining physiologic and symptom domains yields different risk profiles than AHI-defined severity classes, and they make the resulting atlas available to facilitate replication and hypothesis generation.
The Institutional Review Board of Beth Israel Deaconess Medical Center provided ethical approval for the work, and the authors confirm that participant consent and oversight procedures were followed. Individual-level data underlying the study are not publicly available because they contain protected health information; however, the manuscript and supplementary materials provide aggregate results, and the authors point readers to https://www.osaatlas.com/ for atlas resources. The preprint lists funding from multiple sources including NIH grants (R01 HL161253-01A1, P01HL160471, R01HL170675), the American Heart Association (20CDA35310360), and NCATS CTSA Award UL1TR002366. The report is a medRxiv preprint and has not been peer-reviewed; the authors caution that the findings should not be used to guide clinical practice without further validation.
Limitations and outstanding questions documented by the authors include that individual-level data remain protected and are not publicly shared, and that readers should consult the full manuscript and supplementary material for complete methodological details and the full list of significant phecodes. The preprint did not present peer-review validation at the time of posting, so replication in external cohorts and detailed methodological transparency remain specific next steps documented in the source.