Investigators concluded there were no viable candidates for CVD risk phenotyping in sleep clinics because there were no strong links between any individual or type of oximetry pattern and incident cardiovascular disease.
New data from the Sleep Heart Health Study analyzing baseline oximetry patterns of patients with obstructive sleep apnea (OSA) found no evidence that specific patterns were related to the development of cardiovascular disease (CVD). Some individual patterns showed marginal relationships with CVD risk in women, but none of them stood out above the rest.1
A total of 2878 individuals with OSA who had Apnea Hypopnea Index (AHI) of at least 5 events per hour and no preexisting CVD at baseline or within the first 2 hours of follow-up were included in the analysis. Investigators calculated a total of 31 oximetry patterns across 4 analysis types: desaturation characteristics, time series analysis, power spectral density, and nonlinear analysis.
Senior investigator Philip de Chazal, PhD, ResMed Chair in Biomedical Engineering, University of Sydney, and colleagues compared the associations of each pattern with incident CVD using a series of Cox proportional hazard regression models. Models were adjusted for covariates such as age, race, smoking status, body mass index, and sex. As AHI was highly correlated with several oximetry patterns, investigators were not able to adjust for OSA severity by AHI in the models due to collinearity.
Incident CVD was defined as first episode of angina, stroke, myocardial infarction, percutaneous transcutaneous angioplasty, coronary stent placement, coronary artery bypassing grafting, chronic heart failure, coronary heart disease, or death due to CVD 2 years or more after baseline polysomnogram. The length of study follow-up was a median of 11.5 (IQR, 2.8) years, with incident CVD occurring in 17.2% (n = 495) of patients.
All told, not only did AHI not show any significant association with incident CVD as a predictor variable, but there were no significant associations between any of the oximetry patterns of the 4 analysis types in the total sample. Similar to the total sample, there were no significant associations between any of the oximetry patterns and incident CVD in the models of solely men.
"In an era where sleep medicine is looking to extract more from the multitude of sleep signals routinely collected for better prognostic tools, understanding aspects of the oximetry signal which could fag future CVD in the sleep clinic is highly desirable," de Chazal et al wrote. "Further work is needed to understand the complex relationship between OSA-related intermittent hypoxia patterns and development of CVD to further understanding of OSA phenotypes at risk of cardiovascular consequences."
Although none of the oximetry patterns reached statistical significance in women, some were suggestive, including within the desaturation characteristics OD15, which was significant for incident CVD (hazard ratio [HR], 0.77; 95% CI, 0.64-0.83; P = .007). Higher OD15 was associated with lower hazard for CVD. For time series analysis, the standard deviation of the oximetry distribution was nominally significant (HR 0.81; 95% CI 0.68-0.97; P = .014), but not by the pre-specified significance level (P <.013). Similarly, for the cumulative distribution, the nadir (HR 1.2; 95% CI 1.04, 1.40; P = .01) was associated with CVD incidence.
There were several limitations to the study noted by the investigators, including the fact that there was a lack of OSA treatment information, which could have introduced error around the severity and duration of OSA experienced by patients. Additionally, the study may have been underpowered to observe differences in oximetry patterns, as there were fewer than 500 cases of incident CVD in the analysis. "In future studies it could be possible to examine much larger sample with polysomnography data from those who have attended clinical sleep laboratories and obtain follow-up data on CVD from data linkage and electronic medical records," they concluded.