A 10-percentage point increase in percent time in REM sleep was associated with a 35% decreased hazard of myocardial infarction (95% CI 0.45-0.95; p=0.03).
Cohort (n=2,237)
Are polysomnography-derived markers beyond traditional OSA metrics associated with incident myocardial infarction and stroke risk?
Polysomnography metrics such as percent time in REM sleep and percent time with SaO2 <95% are associated with incident MI risk, highlighting the potential of non-traditional OSA metrics for cardiovascular risk assessment.
Effect estimate: 35% decreased hazard (95% CI 0.45-0.95)
p-value: p=0.03
Abstract Rationale The apnea-hypopnea index (AHI) is a widely used measure of obstructive sleep apnea (OSA) severity, but its ability to predict cardiovascular disease (CVD) risk is limited. Moreover, many epidemiologic studies of OSA have focused on composite cardiovascular outcomes, which may obscure specific directional associations with individual stroke and myocardial infarction (MI) events. We therefore utilized data from the Multi-Ethnic Study of Atherosclerosis (MESA) to investigate polysomnography-derived markers beyond traditional OSA metrics such as the AHI, in relation to incident MI and stroke risk. Methods We analyzed polysomnography variables from the MESA Exam 5 Sleep Study that were pre-screened for clinical relevance by two sleep medicine physicians. CVD events were tracked over the duration of Exam 5 (21 months). Univariate analyses identified variables that were significantly associated with MI, stroke, or both when controlling for multiple testing with false discovery rate. Collinearity was assessed using Pearson correlation coefficient (r), with variables with |r|0.7 and correlation p-value 0.05 considered collinear. Variables were grouped into collinear clusters from which representative metrics were chosen for multivariate modeling, with adjustments for age, sex, body mass index, history of hypertension/diabetes, and history of stroke or MI. Associations were estimated with Cox hazard models. Results 2,237 participants with sleep data were analyzed, with 69 adjudicated MI events and 70 stroke events. 224 polysomnography variables were deemed clinically relevant. In univariate analysis of these variables, 16 had statistically significant associations with MI (Fig 1a). No variables had significant associations with stroke. We found three clusters of correlated variables from which percent time in apnea/hypopnea with 3% desaturation, percent time in rapid eye movement (REM) sleep, and percent time with SaO2 95% were selected as representative variables. In adjusted analysis, a 10%-point increase in percent time in REM sleep was associated with a 35% decreased hazard of MI (95% CI = 0.45, 0.95, p = 0.03), and a 10%-point increase in percent time under 95% SaO2 was associated with a 9% increased hazard of MI (95% CI = 1.01, 1.19, p = 0.03, Fig 1b). Conclusions Our analysis found that two polysomnography variables, percent time in REM sleep and percent time with SaO2 under 95%, carried associations with MI but not stroke in MESA. The latter suggests that hypoxia cutoffs other than the more commonly used T90% may have clinical relevance. Overall, these findings are hypothesis-generating, suggesting heterogeneous effects of OSA metrics beyond the traditional AHI, and underscore the importance of examining event-specific associations in OSA. This abstract is funded by: NIH/NHLBI, NIH/NIA, NIH/NCATS
Xu et al. (Fri,) conducted a cohort in Obstructive sleep apnea (n=2,237). Polysomnography metrics (percent time in REM sleep and percent time with SaO2 <95%) was evaluated on Incident myocardial infarction (35% decreased hazard, 95% CI 0.45-0.95, p=0.03). A 10-percentage point increase in percent time in REM sleep was associated with a 35% decreased hazard of myocardial infarction (95% CI 0.45-0.95; p=0.03).