Obstructive sleep apnea phenotypes and cardiovascular risk: Is there a role for heart rate variability in risk stratification?

The global prevalence of obstructive sleep apnea (OSA) is estimated to be nearly 1 billion people [1], and this is likely to only increase in the coming years and decades given the strong association between OSA and obesity [2]. Although OSA is associated with adverse cardiovascular health, the utilization of continuous positive airway pressure (CPAP) to treat OSA has not consistently yielded significant and/or clinically meaningful improvements in certain cardiovascular outcomes [3, 4]. For example, several meta-analyses of randomized controlled trials report significant, yet marginal (i.e. 2–3 mmHg), reductions of blood pressure after CPAP in participants with OSA [5–9]. Moreover, three prospective community-based studies reporting on the association between OSA severity and incident hypertension (i.e. Wisconsin Sleep Cohort Study, Sleep Heart Health Study, and Victoria Sleep Cohort Study) have yielded conflicting findings [10–12]. We have previously posited that a number of experimental design differences (i.e. subject demographics, length of treatment, CPAP adherence, severity of disease, definition of hypertension, etc.) likely contribute to the heterogeneity of findings [13]. More recently, efforts have been made to classify OSA into various subtypes based upon clinical symptoms at diagnosis, including (1) OSA with self-reported disrupted sleep symptoms (e.g. insomnia), (2) OSA with excessive daytime sleepiness, (3) OSA with moderate sleepiness, and (4) OSA without these traditional symptoms. Mazzotti et al. [14] recently examined these various subtypes within the Sleep Health Heart Study and demonstrated the highest cardiovascular risk in patients with OSA with excessive daytime sleepiness. More specifically, OSA with excessive daytime sleepiness was more strongly associated with incident cardiovascular disease, coronary heart disease, and heart failure when compared to the other OSA subtypes [14]. This suggests that factors beyond the apnea–hypopnea index (AHI) that defines traditional OSA diagnosis (i.e. mild, moderate, and severe), such as excessive daytime sleepiness, may provide key insights into cardiovascular risk. In this issue of SLEEP, Qin et al. [15] report that certain heart rate variability (HRV) indices taken during 5 min of wakefulness prior to turning off the lights during a polysomnogram were associated with OSA severity, with the general finding that severe OSA (i.e. AHI ≥30 events per hour) was more likely to have reduced or less complex HRV. A major strength of the study was the large sample size (n = 1,247) that drew from multiple laboratories across the world and led to statistically robust groups that included mild, moderate, and severe OSA and a control group without OSA. Moreover, the investigators evaluated HRV using three well-defined approaches, including time-domain, frequency-domain, and nonlinear analyses. The results suggest that HRV indices could provide a level of risk stratification, much like the work of Mazzotti et al. [14] using excessive daytime sleepiness. It would be of substantial clinical importance if wake resting heart rate provided information about the state of cardiovascular regulation in relation to OSA. It would be fairly trivial to introduce a standard brief measurement of heart rate into laboratory polysomnogram and even home sleep apnea testing protocols. Acquisition of this type of data is also now increasingly at hand, with consumer wearable devices. More daunting is how this information might be used at the individual level, as HRV measures do not have a clear cutoff for disease versus health. One concern is that alterations in HRV across the sleep-wake continuum could be a nonspecific marker of sleep disruption and not specific to sleep apnea [16]. The study by Qin et al. [15] needed hundreds of individuals to show statistical significance. Age and physical fitness will inevitably add variance. However, measures of HRV could become clinically useful if nomograms considering age/gender/race/ethnicity were available (much like osteoporosis bone density scores) in order to facilitate risk stratification—which patients with sleep apnea, regardless of severity, are at risk for adverse cardiovascular outcomes? This information may skew treatment decisions in relatively clinically asymptomatic patients with evidence of severe disease. For example, abnormal HRV can predict both sudden and non-sudden cardiac death in patients with acute myocardial infarction [17]. High nocturnal heart rate and low nocturnal HRV predict a poor response to treatment of major depression in patients with stable ischemic heart disease [18]. We know very little about the recovery kinetics of various abnormalities associated with OSA, be it cognition, endothelial dysfunction, autonomic function, and even blood pressure over long periods of time. Autonomic recovery with therapy may be an important dimension to track, and heart rate measures both during wake and sleep [19] may provide readily accessible information, capably tracked with wearable devices [20, 21]. While sleep HRV measures may change in meaningful ways with continuous and bilevel positive airway pressure therapy [22–24], it remains unclear if altered wake measures will be clinically useful. Our interest in the findings by Qin et al. [15] must be balanced with significant concerns and reminders about the limitations of HRV. First and foremost, there is a preponderance of evidence that certain HRV analyses and indices are more reliable and reproducible than others. Specifically, both time-domain HRV and the high-frequency (HF) component of frequency-domain HRV consistently demonstrate high intraindividual reproducibility, while the low-frequency (LF) component of frequency-domain HRV analyses often has poor intraindividual reproducibility [25–27]. Moreover, Goldstein et al. [28] have presented evidence to suggest that the LF component of HRV may actually be more closely associated with cardiovagal baroreflex function than sympathetic activity as is often speculated [29]. Despite these limitations, the LF component of HRV continues to be a primary outcome variable in many psychophysiological studies [30]. Given the lack of intraindividual repeatability with the LF component of HRV, we were not surprised that adjusted comparisons of HRV and OSA were significant with the time-domain and nonlinear dynamic analyses, but not with the frequency-domain analyses [15]. Furthermore, while the supplementary table that stratified by obesity demonstrated some significant interactions with the frequency-domain analyses [15], the reported relationships for measures utilizing the LF component of HRV should be examined cautiously and should not be used to make inferences of “sympathetic predominance” given various methodological concerns [25–30]. Qin et al. [15] acknowledge a number of limitations and remaining questions that will require subsequent studies. Of the suggested future directions, we are most interested in how the reported HRV–OSA relationships are influenced by race and emerging sleep apnea phenotypes and endotypes that appear to be more susceptible to cardiovascular disease (i.e. excessive sleepiness vs. disrupted sleep vs. asymptomatic). Modulation by sleep apnea driver phenotypes, especially high loop gain, would be especially interesting, given the role of the carotid body in driving both sympathetic activation and respiratory control instability in OSA [31]. In the current pursuit of personalized medicine, it is intriguing to consider if a brief and objective measure such as HRV could add significant value in predicting which patients with OSA might be at highest cardiovascular risk and/or which would benefit the most from CPAP and other OSA therapies. In summary, Qin et al. [15] provide a robust data set from multiple laboratories across the world and make the case that HRV provides a potential window of insight into OSA severity. The frequency-domain data, and particularly indices incorporating the LF component of HRV, must be interpreted with extreme caution given the well-known but often ignored literature demonstrating the lack of intraindividual repeatability and lack of consensus that it accurately reflects sympathetic activity [25–30]. While HRV can serve as a powerful and objective tool for risk stratification, and may even have the potential to provide clinically relevant insights into the complex relationship between OSA and cardiovascular risk, it should not be used to estimate “sympathetic predominance” or for other interpretations that stretch beyond the analytical capabilities and limitations. This work was supported, in part, from the National Institute on Alcohol Abuse and Alcoholism (AA-024892; J.R.C.) and an American Academy of Sleep Medicine Foundation Award (#206-SR-19; R.J.T.). Conflict of interest statement. R.J.T. acknowledges serving as a consultant for Guidepoint Global, GLG Councils, and Jazz Pharmaceuticals, as well as patent, license, and/or royalties for MyCardio, LLC for ECG spectrogram (related to phenotyping sleep and sleep apnea) and DeVilbiss-Drive for Auto-CPAP algorithms. J.R.C. and B.M. have no conflicts to disclose.