Mild obstructive sleep apnea increases hypertension risk, challenging traditional severity classification

INTRODUCTION

Mild obstructive sleep apnea (OSA) is a highly prevalent disorder in adults with potential adverse neurocognitive, metabolic, and cardiovascular complications.¹ Changes in sleep recording technology, apnea-hypopnea scoring criteria, and the definition of sleep apnea according to the recent classification system (ICSD-3) have all contributed to an increased prevalence of OSA. In particular, mild OSA appears to provide the largest subgroup of this condition, accounting indifferent reports for up to 50–70% of all patients with OSA.²

Because mild OSA may progress to more severe disease, it is important to know whether this category is associated with adverse health outcomes, such as systemic arterial hypertension (SAH) and whether treatment improves these outcomes. However, there are conflicting data on this question. A recent research statement by the American Thoracic Society sought to quantify the impact of mild OSA in adults³ and concluded that there were no significant differences in the risk of SAH among patients with mild OSA when stratified by age, sex, body mass index (BMI), or daytime sleepiness. In contrast, a recent meta-analysis concluded that mild OSA is associated with an increased risk of SAH.⁴ Previous analysis from the European Sleep Apnoea Database (ESADA) cohort showed an independent association between OSA and SAH, possibly mediated by intermittent hypoxia during sleep.⁵ If mild OSA predicts prevalent SAH, then early identification and treatment of mild OSA assumes greater importance. Therefore, the aim of the current study was to investigate the relationship of mild OSA with SAH in the ESADA cohort and the role of various confounding factors in this relationship.

METHODS

The ESADA cohort⁶ constitutes a collaborative European network project between mostly academic sleep centers. The ESADA cohort was established in 2007 with the objective to recruit a large prospective international cohort of treatment naïve participants with OSA. From the ESADA cohort, data recorded between March 2007 and May 2016 were analyzed. A cohort of 19,556 participants suspected of having OSA was assembled in 30 centers distributed across 20 countries in Europe and Israel. Inclusion criteria were as follows: referral because of suspected OSA, age of at least 18 years, and the ability to read, speak, and understand the local language. Exclusion criteria were as follows: previous treatment with continuous positive airway pressure or oral device, severe comorbidity unrelated to OSA with limited life expectancy, and alcohol or drug abuse within 1 year of inclusion. The flow chart of participants recruited is shown in Figure 1. Of 19,556 participants, 4,732 were analyzed. All participants signed an informed consent, and the study was approved by the local ethics committee of each participating center.

RESULTS

Role of sleep studies used (PG vs PGS) on SAH prevalence

The type of sleep study used for the diagnosis of mild OSA (PSG or PG) was found to influence the predictive value of the 2 mild OSA groups on prevalent SAH. Mild_{AHI 5-<11/h} OSA was a significant predictor in the PG group (P < .001) but not in the PSG group (P = .679; Figure 2). However, mild_{AHI 11-<15/h} OSA remained a significant predictor in the PG (P < .001) and PSG groups (P = .025; Table 5).

Table 5 Crude and adjusted odds ratios for prevalent systemic arterial hypertension in models containing the mild OSA groups, according to sleep study used.

	Total	B	OR (95% CI)	P
PG (n = 3,135)
Model 0
MildAHI 5-<11/h OSA	3,135	0.827	2.285 (1.939–2.694)	<.001
MildAHI 11-<15/h OSA	3,135	1.022	2.780 (2.261–3.418)	<.001
Model 1
MildAHI 5-<11/h OSA	2,869	0.563	1.757 (1.457–2.118)	<.001
MildAHI 11-<15/h OSA	2,869	0.576	1.779 (1.403–2.256)	<.001
Model 2
MildAHI 5-<11/h OSA	2,921	0.567	1.762 (1.470–2.124)	<.001
MildAHI 11-<15/h OSA	2,921	0.569	1.769 (1.395–2.226)	<.001
PSG (n = 1,596)
Model 0
MildAHI 5-<11/h OSA	1,596	0.306	1.358 (1.076–1.714)	.010
MildAHI 11-<15/h OSA	1,596	0.645	1.906 (1.474–2.465)	<.001
Model 1
MildAHI 5-<11/h OSA	1,330	0.058	1.060 (0.805–1.396)	.679
MildAHI 11-<15/h OSA	1,330	0.354	1.424 (1.047–1.939)	.025
Model 2
MildAHI 5-<11/h OSA	1,401	0.118	1.125 (0.862–1.489)	.387
MildAHI 11-<15/h OSA	1,401	0.389	1.476 (1.100–1.981)	.009

CI = confidence interval, OR = odds ratio, OSA = obstructive sleep apnea, PG = polygraphy, PSG = polysomnography.

Subgroup analysis

Further analysis demonstrated that mild OSA predicted prevalent SAH in all age, BMI, and ESS groups and sexes (all P < .05), with slightly stronger associations observed in younger adults (<60 years), males, participants who were sleepier (ESS ≥ 10), and participants with obesity (BMI ≥ 30 kg/m²; Figure 3).

Because multiple regression models cannot completely remove confounding effects, we performed a subanalysis excluding participants with preexisting cardiovascular or cerebrovascular diseases (ie, stroke, angina), diabetes, and hyperlipidemia, confirming the previous results in the whole group and in PG and PSG groups separately (Table S1, Table S2, and Table S3 in the supplemental material).

Finally, to explore the role of sleep fragmentation on SAH risk, we performed a subanalysis in 685 participants who underwent PSG with available data on AI. Although mild_{AHI 5-<11/h} and mild_{AHI 11-<15/h} OSA groups showed increased AI (Table S4) compared with nonapneic snorers (19 vs 14 and 15 vs 14, respectively; P < .001), AI was not a significant predictor of SAH (OR, 0.989; 95% CI, 0.973–1.004; P = .155) after adjustment for confounders. Similar results were obtained when different AI cutoffs (>10, >20) were used as categorical variables (all P > .05).

DISCUSSION

This study suggests that mild OSA was associated with a 56% (AHI 5 to <11 events/h) and 79% (AHI 11 to <15 events/h) risk increase of SAH, and the association was slightly stronger in participants who were younger, obese, sleepy, and male. Importantly, the OSA-mediated risk for SAH remained after extensive adjustment for potential confounders including measures of intermittent hypoxia.

It is known that approximately half of all patients with OSA present with comorbid SAH.¹⁴ However, most published studies on OSA have focused on patients with moderate or severe disease, and there is a paucity of studies selectively assessing patients with the mild form of the disease. The possible association between mild OSA and SAH has attracted considerable attention in recent years.^3,4 Indeed, the risk for incident hypertension was significantly increased in mild OSA in prospective cohort studies,^15,16 and there was an increased prevalence of a nondipping blood pressure pattern in patients with mild OSA.¹⁷ A recent meta-analysis also found that across all OSA severities, there was an association with SAH in a dose-dependent manner.⁴ A statement from the American Thoracic Society³ suggested that despite the compelling evidence of a causal relationship between these 2 conditions,^18,19 part of this association may be attributed to confounding factors.^20–24 We therefore performed extensive adjustment in our statistical models including comorbidities, several measures of obesity, and different indices of OSA severity.

In clinical research, mild OSA has almost universally been defined as an AHI in the range of 5–15 events/h. In our study, 2 cutoff limits of AHI were applied, attempting to further identify subphenotypes of patients at risk within the mild OSA population. Based on these AHI severity criteria, the mild_{AHI 5-<11/h} OSA group (AHI 5 to <11 events/h) remained a significant predictor for SAH when the sleep study was performed with PG. This information is of particular importance as a large proportion of European patients are diagnosed using PG without application of AHI thresholds for OSA severity derived from PG cohorts.²⁵ Furthermore, the mild_{AHI 11-<15/h} OSA group with AHI ≥11 to <15 events/h was the stronger predictor of SAH in both the PG and PSG groups. Therefore, this study provided novel data that may challenge the conventional threshold defining mild OSA independently of the sleep study methodology used as AHI >5 to <15 events/h in relation to SAH risk.

A strength of our study is the large number of participants included and the multicentric, multinational setting, which increases the generalizability of our findings. In contrast, certain limitations of our study need to be considered. First, participants were enrolled based on a clinical referral to a sleep center, and they do not represent the general population. Second, given that our study was cross-sectional, we cannot assign directionality to the associations between mild OSA and SAH. Also, the duration of OSA is difficult to ascertain in this population, and the cumulative exposure to OSA may vary considerably among participants. Another potential limitation is the lack of 24-hour ambulatory blood pressure monitoring measurements, which could provide nocturnal blood pressure measurements as an additional variable of assessing SAH and its control in these patients. Finally, study participants represented a mixture of individuals undergoing PG or PSG investigations according to clinical routine in the participating centers, and no central scoring of events has been applied. We showed previously in the ESADA cohort,¹¹ as have other investigators elsewhere,^26,27 that PG sleep studies will underestimate AHI, and this discrepancy is likely to be most evident in mild OSA, whereas in severe disease (AHI > 30 events/h), small discrepancies in the AHI will have marginal practical implications for disease classification and management. Furthermore, this underdiagnosis and possible misclassification of OSA need to be taken into account because a large proportion of European patients, as in our study, are diagnosed using PG without application of different AHI thresholds for OSA severity.²⁵ We therefore performed sensitivity analysis separately in the PG and PSG groups, which confirmed the main study results and adds important new insights.

Our findings have multiple clinical implications. First, we illustrated an AHI level of ≥11 events/h was associated with a steep increase in the odds for SAH. This may lead to early identification of patients at risk and treatment despite the lack of other daytime or sleep-related symptoms. Second, considering that OSA has a tendency to progress from mild sleep-disordered breathing to more severe OSA over a varying period of time,^28,29 early intervention with personalized treatment and reduction of risk factors should be initiated.^30–32 Third, our data clearly emphasize the need for outcome-based thresholds of OSA severity, which take the sleep diagnostic method into account. The risk impact at a lower AHI value was shown in the mild OSA group assessed by PG. The result needs to be verified by other studies before new definitions of mild, moderate, and severe sleep apnea based on PG recordings should be applied.

In conclusion, our findings suggest that mild OSA is an independent risk factor for SAH, mainly in the 11- to <15 events/h range of AHI. These findings potentially introduce an important challenge to the traditional thresholds of OSA severity and may also add to early identification of patients at risk.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. L. Grote received grants from Bayer, Resmed, Respironics/Philips, and from the European Respiratory Society during the conduct of the study; nonfinancial support from Itamar Medical, Resmed, Philips, and Astra Zeneca, outside the submitted work; and has been issued a patent on sleep apnea therapy. J. Hedner received grants from ResMed, Respironics, and Bayer on behalf of the European Sleep Apnoea Database International Study Group; received speaker fees from ResMed, Itamar, Merck, and Jazz Pharmaceuticals outside the submitted work; and has been issued a patent on sleep apnea therapy. J. Verbraecken received grants and personal fees from ResMed, personal fees from Philips, personal fees from Sanofi, personal fees from Agfa-Gevaert, grants and personal fees from Bioprojet, grants and personal fees from Jazz Pharmaceutics, grants from AirLiquide, personal fees from Springer, grants from Westfalen Medical, grants from SomnoMed, grants from Vivisol, grants from Total Care, grants from Medidis, grants from Fisher & Paykel, grants from Wave Medical, grants from OSG, grants from Mediq Tefa, grants from NightBalance, grants from Heinen & Löwenstein, grants from AstraZen, grants from Accuramed, grants from Bekaert Deslee Academy, and grants from UCB Pharma, outside the submitted work. All other authors report no conflicts of interest.

REFERENCES