The current findings demonstrate an association between high night-to-night variability in OSA severity with increased uncontrolled hypertension risk and blood pressure variability. Even after accounting for the OSA severity category, OSA variability itself was a strong independent predictor of both uncontrolled hypertension and blood pressure variability. These new findings provide important insight into the different manifestations of OSA and its consequences and highlight the need to consider the importance of night-to-night variability in disease severity. The current multi-night recordings to quantify OSA with a simple non-invasive under-mattress sensor in the home, also support previously documented associations between OSA severity and hypertension risk from conventional single-night polysomnography recordings3,24,25. Accordingly, there is considerable potential to incorporate new simplified monitoring approaches to aid current single-night diagnostics.
The current findings are in accordance with emerging evidence that indicates an association between cardiovascular disease and night-to-night variability in OSA severity13,14. A key mechanism through which OSA and cardiovascular risk may be related is via increased day-to-day blood pressure variability20,21,26. Blood pressure variability has been associated with an increased risk of cardiovascular events, all-cause mortality, vascular organ damage, atrial fibrillation, and dementia15,16,17,18,19,27. Consistent with the current findings of increased blood pressure in those with severe or variable OSA, day-to-day blood pressure variation of similar magnitude to the current study is associated with increased all-cause mortality and non-fatal and fatal cardiovascular events28. The increased risk of hypertension and blood pressure variability may reflect a direct physiological increased risk from people with a “variable OSA phenotype”. The multifaceted nature of OSA pathophysiology may place some people more prone to night-to-night variation in OSA severity. These concepts, and the potential underlying mechanisms, warrant further investigation.
Our findings also highlight the important new information that multi-night monitoring of OSA and blood pressure can yield. Current one-size-fits-all clinical care approaches based on a single-night diagnostic study may not be appropriate for people with high internight variability and may explain, at least in part, the heterogeneity in prior treatment trials12,29,30. More in-depth assessment of hypoxemia, OSA endotypes, insomnia, and sleep fragmentation may be valuable in more complex manifestations of OSA31,32,33. The findings of this study also warrant future prospective trials to investigate the effects of different sleep patterns and night-to-night variability in OSA severity and the effects of different therapies on other key cardiovascular and health consequences such as mental health, sleepiness, workplace and traffic accident, and cognitive impairment and their potential interactions.
The volume of physiological data available from this study is substantially greater than previous cohort studies that have investigated OSA and its consequences3,24,25. The large sample size allows for greater precision around estimates of OSA severity and uncontrolled hypertension risk for increased power to detect relationships with available exposure variables. The non-invasive monitoring technology used in this study allowed us to identify relationships between variability in OSA severity and blood pressure in a large number of participants over a prolonged period, which is simply not feasible with conventional sleep monitoring approaches. Data were also collected in the participant’s home environment, more directly relevant to real-world risk exposure conditions compared to clinical laboratory sleep study settings.
However, AHI derived from the under-mattress sensor includes fewer input variables in which to detect respiratory events compared to conventional polysomnography. Accordingly, the possible contribution of other physiological aspects of OSA (e.g., hypoxia) that may contribute to hypertension risk and blood pressure variability requires further consideration. However, validation studies versus gold standard polysomnography in over 150 participants2,22 support the device performance characteristics. Nonetheless, the potential effects of comorbid insomnia, medications, and other clinical covariates on device accuracy remain to be investigated. Despite these potential unknown influences, OSA prevalence estimates using, non-contact multi-night data yield very similar findings to previously published literature1,2. Similarly, misclassification rates and AHI variability are comparable to data derived from multi-night in-laboratory polysomnography and other home sleep apnea tests2,8,9. The effect size of the association between the estimated AHI and uncontrolled hypertension detected in the current study is also similar to existing epidemiological trials such as the Sleep Heart Health Study, the Wisconsin cohort, and the HypnoLaus cohort25,34,35. While the magnitude of the effect size was comparable between studies, it is important to note that, unlike the current investigation, prior epidemiological studies also included the use of antihypertensive medications in the definition of hypertension. Thus, these findings provide support that the multi-night mean AHI estimates derived in the current study provide comparable, and potentially superior insight, into key health outcomes such as hypertension versus traditional single-night but more complex polysomnography approaches.
Instructions on the timing of blood pressure measurements were not given to participants. Thus, some of the variability in blood pressure may reflect circadian changes across the day28. However, the main study findings remained in sensitivity analyses when limited to morning blood pressure entries. Accordingly, circadian influences appear unlikely to have been a major confounder in the current study. Indeed, consistent with previous findings, variability in systolic blood pressure is a significant predictor of death when blood pressure is measured in the morning or evening, or both16. The number of clinical covariates available in this study was also somewhat limited. Thus, the potential impact of uncontrolled confounding behavioral and lifestyle factors (diet, exercise, alcohol, caffeine, tobacco use, and medications)36,37 and treatment status to influence the current findings remain to be investigated. Like many digital health innovations and consumer-data research, there is a balance between data volume and the ability to capture all clinically relevant variables. Conceptually, these data could also be synchronized via electronic health records databases, and via the development of new digital health tools to automatically capture more diverse aspects of a person’s health. Multimodal inputs to predictive algorithms (i.e., using health data from different sources) have been recently shown to better predict health outcomes compared to single-source approaches across 12 predictive tasks, including 10 distinct chest pathology diagnoses, hospital length-of-stay, and 48 h mortality predictions38. Similarly, the field of sleep medicine may benefit from multi-input clinical data where daytime symptoms, overall clinical history, and multi-sensor recordings could be used to better predict health outcomes and treatment response.
The decision of the users to purchase a WSA and monitor their blood pressure may represent a selection bias towards those who had preexisting sleep and/or cardiovascular problems. Indeed, users were predominantly male, clearly indicating a sex-specific participation bias. Nonetheless, while no sex differences were detected in the current study in relation to the detected associations with uncontrolled hypertension, the pathophysiology of OSA differs in women versus men39. Thus, it will be important to investigate whether the potential consequences of night-to-night variability in OSA severity are comparable between sexes in a larger cohort with more comparable numbers of women.
Despite these methodological considerations, the identification of strong associations between blood pressure variability, uncontrolled hypertension, and high night-to-night variability in OSA severity provides further support for the unique and important insights that multi-night non-invasive monitoring can yield regarding clinical end-points and potential health risks.