Sleep apnea and vascular disease risk

A wealth of observational data collected over more than two decades from both clinic- and population-based cohorts demonstrates a strong association of obstructive sleep apnea (OSA) with incident or recurrent major cardiovascular and cerebrovascular disease events. Physiologic studies in humans and animal models provide a strong basis that OSA is a cause of these events, through the impact of intermittent hypercapnic hypoxia and recurrent arousal from sleep on autonomic nervous system activity, metabolic dysregulation, oxidative stress, and systemic inflammation. ¹ In the absence of randomized trials demonstrating a reduction in vascular event risk with OSA treatment, however, it remains plausible that the association of OSA with vascular events reflects shared vascular disease risk factors such as adiposity and sedentary behavior. Quasi-experimental studies in patients diagnosed with OSA in the clinical setting have almost uniformly found that patients treated with continuous positive airway pressure (CPAP) have a lower risk of vascular events than patients who decline or are nonadherent to CPAP. ² Such studies provide a very weak basis for causal inference, however, due to the likelihood of a strong “healthy adherer” effect: patients who are adherent to therapy generally have much better health outcomes than those who are nonadherent, even when the therapy to which they adhere is a placebo. ³

The publication in 2016 of two randomized clinical trials of CPAP therapy for secondary prevention of cardiovascular events was therefore an important step forward. The Randomized Intervention with CPAP in Coronary Artery Disease and Sleep Apnoea (RICCADSA) trial screened for the presence of OSA in patients who had recently undergone coronary revascularization procedures. ⁴ In this study, 244 patients with minimally symptomatic, moderate to severe OSA were randomized to treatment with CPAP or to usual care without specific treatment of OSA. Patients were followed for a median of 57 months for a composite outcome of repeat revascularization, myocardial infarction, stroke, or cardiovascular mortality. The study was powered to detect a 52% relative risk reduction for the primary end point (from 25% to 12%), an unrealistic expectation that exceeds the cardiovascular risk reduction achieved with aspirin ⁵ or statins. ⁶ Ultimately 20% of randomized patients reached the combined end point, and CPAP therapy was associated with a nonsignificant 20% relative risk reduction (hazard ratio 0.80, 95% CI 0.46-1.41). That such a potential clinically important reduction in risk was not statistically significant attests the large numbers of patients needed to provide adequate statistical power in cardiovascular outcomes studies.

The Sleep Apnea Cardiovascular Endpoint (SAVE) study met this need. With a sample of 2687 patients followed up for a mean of 3.7 years, the study had 85% power to detect a 25% relative reduction in risk of the composite outcome of myocardial infarction, stroke, cardiovascular death, or hospitalization for heart failure, acute coronary syndrome, or transient ischemic attack. ⁷ This multinational study screened adults with a diagnosis of coronary or cerebral vascular disease for the presence of OSA, defined as a 4% desaturation index >12 events/h on respiratory polygraphy. To improve adherence, the study randomized only those patients who were adherent to sham CPAP for at least 3 h per night during a 1-week run-in period. Adherence to CPAP therapy was 5.2 h per night during the run-in period and averaged 3.3 (SD 2.3) h per night during the entire follow-up period. Good control of OSA was achieved with CPAP, with a device-estimated residual apnea–hypopnea index of 3.7 events/h. The primary end point was reached in 16% of patients, with no significant difference between the CPAP and usual care groups and a trend slightly favoring usual care (hazard ratio 1.10, 95% CI 0.91-1.32). No significant differences were observed for any of the individual components of the primary composite outcome or for any prespecified secondary outcome. Prespecified subgroup analyses showed no significant effect modification by region (China vs. other), age, sex, obesity, OSA severity, or sleepiness.

How are we to make sense of this result, which is so contrary to the expectations engendered by observational, physiological, and quasi-experimental studies? Certainly we must seriously consider the possibility that CPAP treatment of OSA does not, in fact, reduce vascular risk. The SAVE trial was well designed, well executed, and well powered to detect a clinically important effect of CPAP treatment, yet found none. A common counterargument, made in an accompanying editorial, ⁸ is that the average adherence of 3.3 h per night may be insufficient for a clinical benefit. Although it reflects only about half of the expected total sleep time, it was a greater adherence than posited in the initial study power calculations; one might expect to see at least a trend in favor of CPAP therapy if it were indeed efficacious. Analyses comparing a CPAP-adherent subgroup to the entire control group are often used to assess the effect of CPAP under conditions of optimal use. This was done in the RICCADSA study and showed an apparently significant benefit of treatment. ⁴ Such analyses are subject to the same healthy adherer bias as are quasi-experimental studies, however, and should be largely discounted. The SAVE trial took a more sophisticated approach, deriving a propensity score predicting the likelihood of CPAP adherence within the group randomized to CPAP, and applying this propensity score to the usual care group in order to compare the 561 patients using CPAP ≥4 h per night to a “propensity-matched” group of control patients. ⁷ This analysis also showed no significant difference between CPAP and control groups in the primary composite end point, albeit with a clinically important estimated effect size (hazard ratio 0.80, 95% CI 0.60-1.07). The performance of the propensity score is not reported, however, and a comparison of actually adherent patients to those with a similar adherence propensity score is likely to have residual bias. Indeed, when all CPAP patients were matched to usual care patients with similar propensity score, no meaningful difference was observed in any stratum of adherence.

Another argument against low adherence explaining the null result is that CPAP therapy in the SAVE trial was sufficient to result in substantial and significant improvements in daytime sleepiness, as well as significant improvements in symptoms of anxiety and depression, in several measures of health-related quality of life, and in missed work days due to poor health. Of course, different outcome domains may be responsive to different degrees of OSA control. This concern is raised in the SAVE trial by the lack of a clear impact of CPAP on blood pressure. The effect size of <1 mmHg for both systolic and diastolic pressure is considerably smaller than the effects seen in prior multicenter studies with somewhat greater mean CPAP adherence of 3.5–5.1 h per night, ^{9 –11} although this may reflect the reliance on office blood pressure measurements rather than 24-h ambulatory monitoring.

Several other potential reasons for the null findings of the SAVE study should be considered, including patient selection and concomitant therapy. While ancestry, adiposity, and comorbid illness did not appear to be responsible, in this trial the overall OSA was not severe, with a median apnea–hypopnea index of 25 events/h and with almost one-quarter of patients having only mild OSA. Moreover, patients with severe hypoxemia or severe daytime sleepiness were excluded and, unlike typical OSA patients, study participants had normal levels of sleepiness based on the Epworth Sleepiness Scale. While subgroup analyses did not suggest differences in CPAP effect by OSA severity, the relatively mild OSA severity and possible exclusion of patients at highest risk may have limited the ability to detect a treatment effect. The use of antihypertensive, lipid-lowering, and antithrombotic therapy was naturally quite common in these patients with established vascular disease. It is therefore possible that much of the cardiovascular risk associated with OSA was ameliorated by treatment of its downstream physiologic effects.

While the results of the SAVE study argue for equipoise regarding the role of OSA therapy in management of vascular disease risk, a single negative clinical trial is hardly determinative. In addition to the chance of type 2 statistical error (estimated at 15% for this trial), the epidemiologic and biologic rationale for an effect of OSA on vascular risk mandates continued exploration of this association. Future epidemiologic studies should focus on cardiovascular risk stratification with respect to OSA, using physiologic, genomic, and proteomic markers in addition to novel metrics derived from polysomnography. Clinical trials would be well advised to focus on groups at highest risk, which pending improved risk stratification means those with severe OSA or OSA-related hypoxemia. A multifaceted treatment approach, in which CPAP nonadherent individuals are offered alternative treatments such as mandibular repositioning devices, may address the problem of low CPAP adherence. Consideration should also be given to primary prevention trials, as treatment of OSA may be less beneficial once severe vascular disease is established, although this speculation is not supported by prior observational studies. While much remains to be elucidated, it should be acknowledged that at the present time, the predominant indication for treatment of OSA is to provide symptomatic relief. Although CPAP may be a useful adjunctive therapy for asymptomatic patients in specific clinical situations, such as resistant hypertension, the current evidence base does not indicate that CPAP treatment reduces the risk of major adverse cardiovascular and cerebrovascular events.