Positional Obstructive Sleep Apnea in Patients with Atrial Fibrillation

Abstract

Purpose

Obstructive sleep apnea (OSA) is a common, potentially modifiable condition implicated in the pathogenesis of atrial fibrillation (AF). The presence and severity of OSA is largely sleep position dependent, yet there is high variability in positional dependence among OSA patients. We investigated the prevalence of positional OSA (POSA) and examined associated factors in AF patients.

Methods

We recruited an equal number of patients with and without AF who underwent diagnostic polysomnography. Patients included had ≥120 minutes of total sleep time with 30 minutes of sleep in both supine and lateral positions. POSA was defined as an overall apnea hypopnea index (AHI) ≥5/hr, supine AHI (sAHI) ≥5/hr, and sAHI greater than twice the non-supine AHI. POSA prevalence was compared in patients with and without AF adjusting for age, sex, OSA severity, and heart failure.

Results

215 total patients (male: 55.8%, mean age 62 years) were included. POSA prevalence was similar between the two groups (46.4% vs. 39.1%; p=0.33). Obesity and severe OSA (AHI≥30/hr) were associated with low likelihood of POSA (OR [CI] of 0.17 [0.09 – 0.32] and 0.28 [0.12 – 0.62]). In patients with AF, male sex was associated with a higher likelihood of POSA (OR [CI] of 3.16 [1.06 −10.4]).

Conclusion

POSA is common, affecting more than half of AF patients, but the prevalence was similar in those without AF. Obesity and more severe OSA are associated with lower odds of POSA. Positional therapy should be considered in patients with AF, mild OSA, and POSA as a potential treatment.

Background

Obstructive sleep apnea (OSA) is a common comorbid and potentially modifiable condition that has been implicated in the pathogenesis of atrial fibrillation (AF) [1]. AF is the most common chronic cardiac arrhythmia, and is associated with increased cardiovascular morbidity and mortality [2]. As such, screening and treatment of OSA in patients with AF is recommended [3]. Observational studies have shown that treatment of OSA with continuous positive airway pressure (CPAP) is associated with a lower risk of AF recurrence following cardioversion or ablation and makes pharmacological rhythm control more durable when pursuing rhythm control [4, 5]. Similarly, multiple epidemiological studies have previously proposed that OSA is associated with an increased risk of heart failure (HF) [6, 7] and is a common comorbidity in patients with HF [8, 9]. Considering the well-established association of HF and AF [9, 10], OSA may have an important implication in the interplay between HF and AF as well [11, 12].

Body habitus including obesity, large neck circumference, craniofacial morphology, male sex, and older age are some of the well-established risk factors for OSA [13, 14]. In a given night, however, sleep position is one of the most critical variables that determines susceptibility to OSA events [15-17]. In most patients, the occurrence of obstructive respiratory events increases in supine sleep and diminishes in non-supine sleep position, regardless of sleep state [18]. Moreover, OSA severity, in terms of desaturation and apnea event duration, worsens in supine sleep [19]. In fact, a large portion of patients have positional OSA (POSA), wherein OSA is predominantly seen in supine sleep only [20]. Given the association of OSA and AF, as well as the therapeutic implication of OSA in AF management, information on POSA may be useful in better characterization of OSA and selection of therapy in patients with AF.

One previous study investigated POSA prevalence in patients with various forms of documented or suspected arrhythmias [21]. In this study, the AF population had more prevalent and more severe POSA compared to the ventricular ectopy and symptoms suggesting arrhythmia group. We sought to perform a more detailed analysis on an enriched AF group in comparison to a control. The purpose of this study was to investigate the prevalence of POSA and to examine its associated factors in patients with AF.

Methods

Study Subjects and Design

This is a single center retrospective study. The study design has been previously described [22]. In short, we recruited consecutive patients with AF referred to an academic sleep medicine center who then underwent diagnostic polysomnography (PSG) from January 2011 through January 2019. Consecutive patients without AF who underwent PSG were identified retrospectively from July 2017 through December 2017 (a 6-month period) and were included. This was done in order to have a similar number of non-AF patients to compared to the AF group. The indication for the diagnostic PSG was for OSA evaluation in all patients. Inclusion criteria were age ≥18 years and complete diagnostic PSG data. Patients with predominant central sleep apnea, defined as ≥5 central apneas and/or central hypopneas per hour with the number of central apneas and/or central hypopneas >50% of the total number of apneas and hypopneas, were excluded. Further, those with incomplete PSG data, and patients with split-night sleep studies or CPAP titration studies were excluded. Patients with insufficient total sleep time (<120 minutes), insufficient lateral sleep time (<30 minutes), insufficient supine sleep time (<30 minutes), apnea-hypopnea index (AHI) <5/hr, or a supine AHI (sAHI) that was not recorded or that was equal to 0 were also excluded.

Patient demographics, medical conditions, medication prescriptions, and PSG data were acquired from the electronic medical record (EMR). Age was modeled as a binary variable by median value. The diagnosis of AF was first identified using ICD-10 coding in the EMR. Individual review of electrocardiograms then confirmed the AF diagnosis. The timing of AF diagnosis preceding the PSG was confirmed by individual chart review by investigators. In patients without a diagnosis of AF, patient charts were individually reviewed to ensure the lack of AF diagnosis as an ICD-10 code, electrocardiogram finding, or problem in provider documentation. Information about comorbid conditions and medications were extracted from EMR. HF was defined as either 1) a left ventricular ejection fraction ≤40% (i.e. HF with reduced ejection fraction) or 2) a left ventricular ejection fraction >40% (i.e. HF with preserved ejection fraction) with a prior admission for decompensated HF as confirmed by clinician EMR review. This study was approved by the University of Virginia Institutional Review Board.

Polysomnography Processing and Scoring

Each PSG was scored using the American Academy of Sleep Medicine guidelines [23, 24]. PSGs were scored by a certified PSG technologist and verified by a board-certified sleep physician. PSG data were processed using Embla Sandman Elite software (Natus Medical Incorporated, Pleasanton, CA, USA). A thermocouple system was used to measure episodes of apneas and hypopneas. Using standard definitions, apneas were scored as a nasal pressure flow signal decrease of > 90% of baseline for ≥10 seconds. Hypopneas were scored as pressure flow signal decrease of >30% of baseline for ≥10 seconds with a ≥4% oxygen saturation decrease. The AHI was defined as the number of apnea and hypopnea events divided by the total sleep time, which was expressed as events per hour. Mild OSA was AHI ≥5/hr to <15/hr, moderate OSA was ≥15/hr to <30/hr, and severe OSA was AHI ≥30/hr for this study. POSA was defined if overall AHI ≥5/hr, sAHI ≥5/hr, and supine AHI greater than twice the non-supine (nsAHI) [15]. We also used an alternative more strict definition of POSA, “exclusive POSA (ePOSA),” if meeting the above with the additional criteria of nsAHI <5/hr [25].

Statistical Analysis

Categorical variables were analyzed using the Chi-squared test and detailed as frequencies with percentages. Continuous variables that were normally distributed were analyzed by the Student’s t-test. Continuous variables that were non-normally distributed were analyzed by the Mann-Whitney U-test for between group comparisons or Wilcoxon signed-rank test for within group comparisons. These data were detailed as mean with standard deviation (SD). Using multivariable logistic regression, the prevalence of POSA between the AF and no AF groups was compared adjusting for age, sex, obesity, OSA severity, AF, and HF. Only variables that were deemed relevant and contributory to both AF and OSA were included in the model. The statistical interaction between AF, age, sex, and obesity were tested by serially including the product of the two variables of interest. Factors associated with POSA in AF cohort were determined by multivariable model including the same covariates except AF. This data was reported as Odds Ratio with 95% confidence intervals (CI). All statistical tests were two-tailed. A p-value of <0.05 was determined to be statistically significant. Analyses were performed using Statistical Analysis System (SAS) version 9.2 (SAS Institute Incorporated, Cary, NC, USA).

Results

Records of a total of 465 patients were consecutively reviewed. After pre-specified exclusions, a total of 215 patients were included for this analysis, in which all had a diagnosis of OSA and approximately half of the cohort (N=110 [51%]) had a diagnosis of AF as by design (Figure 1). The mean age of the population was 62.3 (14.4) years old and 56% were male. Patient demographics and clinical data are presented in Table 1. Paroxysmal AF (80.9% of AF population) was the most common subtype of AF, followed by persistent AF or permanent AF (19.1% of AF population), with 37.2% of AF patients having a history of cardioversion and 37.2% of AF patients having undergone AF ablation. The prevalence of any severity OSA was similar between the AF and no AF groups, as all patients included had OSA (p=1.0) There was a trend of patients with AF having more severe OSA compared with those without AF (N of AF vs. no AF groups; mild: 40 vs. 53, moderate: 35 vs. 32, severe: 35 vs. 20; [p=0.052]).

Figure 1.

Patient group studied with exclusion criteria.

Table 1.

Patient Characteristics Stratified by Atrial Fibrillation.

Patient Characteristic	Overall (N=215) [N (SD)]	Atrial Fibrillation (N=110) [N (SD)]	No Atrial Fibrillation (N=105) [N (SD)]	p value
Age (years) ± SD	62.3 ± 14.4	66.2 ± 12.5	58.1 ± 15.0	<0.001
Male Sex	120 (55.8)	63 (29.3)	57 (26.5)	0.68
Body Mass Index (kg/m2) ± SD	32.1 ± 7.9	32.5 ± 7.6	31.7 ± 8.2	0.46
Coronary Artery Disease	51 (23.7)	33 (15.3)	18 (8.4)	0.04
Hypertension	129 (60.0)	56 (26.0)	73 (34.0)	0.006
Diabetes	70 (32.5)	37 (17.2)	33 (15.3)	0.77
Heart Failure	39 (18.1)	32 (14.9)	7 (3.3)	<0.001
HFrEF	25 (11.6)	20 (9.3)	5 (2.3)	0.003
HFpEF	14 (6.5)	12 (5.6)	2 (0.9)	0.007
Stroke / Transient Ischemic Attack	9 (4.2)	7 (3.3)	2 (0.9)	0.17
Medications
Beta Blocker	83 (38.6)	56 (26.0)	27 (12.6)	<0.001
Calcium Channel Blocker	26 (12.0)	12 (5.6)	14 (6.5)	0.59
Digoxin	14 (6.5)	12 (5.6)	2 (0.9)	0.007
Benzodiazepine	69 (32.0)	39 (18.1)	30 (14.0)	0.28
SSRI/SNRI	34 (15.8)	15 (7.0)	19 (8.8)	0.37

Sleep study characteristics stratified by AF are presented in Table 2. The overall average severity of OSA was moderate with a mean AHI of 26/hr. The sAHI in both the AF and no AF groups was similar (p=0.11). However, the nsAHI was significantly higher in the AF population (p=0.04). AHI in both groups was lower in the nonsupine vs. supine position (Figure 2). The overall difference in AHI between the supine and non-supine positions was comparable between the AF and no AF groups (−13.1 vs. −16.4, p=0.99). POSA was present in 57.2%, whereas ePOSA was present in 23.7% of the entire cohort. There was no difference in POSA prevalence between the AF vs. no AF groups (46.4% vs. 39.1%; p=0.33). For the entire cohort (i.e., combining the AF and no AF groups), POSA was more common in patients with mild OSA compared to moderate OSA or severe OSA (27.9% vs. 19.5% vs. 17.0%, respectively; p=0.004). The prevalence of ePOSA was lower in the AF group (vs. no AF group) [9.8% vs. 14.0%, respectively (p=0.03)]. Similar to POSA, for the entire cohort, ePOSA was more common in mild OSA (17.2% vs. 5.1% vs. 1.4%, mild, moderate and severe respectively; p=<0.0001).

Table 2.

Sleep Study Characteristics Stratified by Atrial Fibrillation.

Characteristic	Overall [Mean (SD)]	Atrial Fibrillation [Mean (SD)]	No Atrial Fibrillation [Mean (SD)]	p value
Total Sleep Time (min)	313.2 (97.4)	301.6 (98.4)	325.4 (94.8)	0.07
Total Supine Sleep Time (min)	116.4 (77.8)	107.6 (82.6)	125.6 (71.3)	0.09
Total Non-supine Sleep Time (min)	196.8 (96.9)	194.0 (104.2)	200.0 (88.6)	0.66
Total Apnea-Hypopnea Index (1/hour)	23.6 (18.0)	26.4 (19.4)	20.7 (15.8)	0.02
Supine Apnea-Hypopnea Index (1/hour)	36.5 (25.7)	39.2 (26.6)	33.6 (24.4)	0.11
Non-supine Apnea-Hypopnea Index (1/hour)	17.8 (19.9)	20.6 (21.1)	14.9 (18.0)	0.04
Central Sleep Apnea-Hypopnea Index (1/hour)	1.7 (5.5)	2.38 (7.0)	1.1 (3.1)	0.07
Sleep Efficiency (%)	74.3 (15.8)	72.2 (15.7)	76.6 (15.6)	0.048
N3 Slow Wave Sleep Quantity (min)	10.0 (9.9)	9.1 (9.4)	10.9 (10.3)	0.19
Mean Oxygen Saturation (%)	94.0 (2.3)	94.0 (2.3)	93.9 (2.4)	0.79
Nadir Oxygen Saturation (%)	83.5 (9.5)	83.4 (7.7)	83.5 (11.0)	0.93

Figure 2.

Apnea-hypopnea index (AHI) in the supine (dark grey) and lateral positions (white) in patients with and without atrial fibrillation (AF). The light grey bar marks the difference between the lateral and supine AHI. The overall height of the bar is the median value and the whisker bars demonstrate the lower and upper quartiles. There was a statistically significant difference between supine and lateral AHI in both the AF and no AF populations. When this difference was compared as stratified by AF, there was no significant difference (p=0.99).

The results of the multivariable analysis are described in Table 3. Obesity and severe OSA were associated with low odds of POSA. There was no association between AF and POSA or AF and ePOSA. Further analysis was performed given the interaction between AF and sex on POSA (p = 0.036). For AF patients, males had higher likelihood of POSA when compared to females (Odds Ratio 3.16, 95% Confidence Interval 1.06 – 10.43, p=0.046). For ePOSA, obesity and severe OSA were associated with low odds, and older age with the higher odds.

Table 3a.

Multivariable Analysis of POSA.

Characteristic	Odds Ratio (95% CI)	Chi Squared	p Value
Age	1.26 (0.64 – 2.47)	0.46	0.50
Male Sex	1.45 (0.74 – 2.82)	1.17	0.28
Obesity	0.17 (0.09 – 0.32)	28.2	<0.0001
Atrial Fibrillation	1.13 (0.57 – 2.26)	0.12	0.73
Heart Failure	0.60 (0.25 – 1.40)	1.35	0.25
Moderate OSA	0.96 (0.46 – 2.01)	2.86	0.09
Severe OSA	0.28 (0.12 – 0.62)	10.8	0.001

Discussion

We found that POSA is common, affecting about half of AF patients, but the prevalence was similar in those without AF. POSA was more common in patients with mild OSA as compared to patients with more severe OSA. Milder OSA was associated with a higher chance of having POSA in both group of patients with and without AF. In patients with AF, but not without AF, being male was associated with higher likelihood of POSA.

The goal of the study was to investigate the prevalence of POSA and examine the associated factors in patients with AF. Consistent with prior studies reporting the high prevalence of POSA, we found that a significant portion (57%) of the patients within both the AF and no AF groups included in our study had POSA. In a recent large multi-site clinic-based study conducted in France (Pays de la Loire sleep cohort), POSA was present in 53% of patients [26]. Similarly, the HypnoLaus study included a community cohort from Switzerland and showed that 75% of participants had POSA using the same definition as our study [27].

The results on milder OSA having a higher prevalence of POSA are consistent with prior studies [15, 16, 25, 26]. Although the symptoms for patients with OSA remains heterogeneous [28], patients with milder OSA are generally less symptomatic. Consequently, motivation to adhere to CPAP tends to be low. Coupling this with the finding from the Pays de la Loire sleep cohort, which linked POSA with a lower CPAP adherence [26], alternative therapy to CPAP should be considered for patients with mild OSA in the form of POSA. As many home sleep apnea tests do not record sleep position, it is challenging to determine the presence of POSA. In view of our finding, one should consider high possibility of POSA in AF patients with mild OSA without sleep position recording. We also found that patients with obesity and severe OSA have a lower likelihood of having POSA (using either POSA or ePOSA definitions in our study). This is understandable as body habitus is an important determinant of upper airway obstruction. Thus, those with a higher BMI may likely exhibit more “position resistant” OSA, in other words, gaining less benefit of being in non-supine position. This suggests that shifting from supine to non-supine sleep position (i.e., PT) is more effective for the non-obese patient compared to the obese patient, similar to prior studies [29]. The negative association of severity of OSA with POSA has also been demonstrated in prior studies [26, 27]. This can be explained in a similar context with the BMI. Although the severity of OSA is determined by AHI (i.e., count of the respiratory events), it is likely that those with high AHI will also likely exhibit more severe OSA (i.e., more severe individual OSA event) that is more position resistant.

In our analysis, there was an apparent interaction between sex and AF in their relation to POSA. In patients with AF, male (vs. female) patients had a higher odd of having POSA. This was not demonstrated in patients without AF. Prior work has demonstrated the relationship between the positive association of POSA and male sex [26]. There is a well-established relationship between AF and sex in general [2] and in patients with OSA [5]. However, the relationship between POSA, AF, and sex is less well described. While this finding needs to be verified in other studies, this finding suggests that in male AF patients diagnosed with mild OSA based on home sleep study without sleep position recording, presence of POSA needs to be strongly considered.

Our findings need to be interpreted in the context of the various POSA definitions that have been suggested. Indeed, POSA remains without a universally accepted definition for definitive diagnosis, and as such, there exists heterogeneity in POSA definition among studies. In our current analysis, we used the original, less strict Cartwright et. al definition (POSA in this study) and more strict definition (ePOSA in this study) by Mador et. al. By the definition of Cartwright et. al [15], POSA is diagnosed with an AHI ≥5/hr with a sAHI twice the nsAHI. Later work by Mador et. al. added to Cartwright’s definition to classify POSA patients by nsAHI, where patients had an AHI ≥5/hr with a sAHI twice the nsAHI and nsAHI normalized to AHI less than 5/hr. [25]. Marklund et. al defined POSA as sAHI ≥10/hr with nsAHI <10/hr in her study on mandibular advancement device use in POSA [30]. Most recently, Frank et. al put forth the Amsterdam Positional OSA Classification (APOC) to identify patients who would benefit from PT [16]. In this system to be diagnosed with POSA, patients with OSA have both a best sleep position (BSP) and worst sleep position (WSP) and spend more than 10% of total sleep time in each the BSP and WSP. Patients are then stratified further into 3 classes: 1) APOC I – BSP AHI <5/hr, 2) APOC II – lower OSA severity category in BSP than overall OSA severity category, and 3) APOC III – overall AHI ≥40/hr and BSP AHI improves ≥ 25% to overall AHI. As the definitions by Cartwright et. al and Mador et. al were the most used in other similar studies, these were employed for the current analysis. Of course, one challenge is that it is uncertain how reliable the diagnosis of POSA is due to these variable definitions.

With the use of these different POSA definitions, there were differences in multivariable analysis. Specifically, when using the ePOSA definition, age was independently associated with increased odds of POSA. However, the less strict definition did not demonstrate this association. While the other variables remained concordant, this highlights the importance of choosing an appropriate definition for investigating POSA.

Given that OSA is highly prevalent in patients with AF, the high proportion of POSA regardless of the AF status as shown in our study may have an implication in therapeutic decision making in OSA. While CPAP remains the gold-standard for OSA, adherence remains low [29]. With the strong association of OSA to cardiovascular disease, such as AF and HF, and with the high prevalence of POSA, patients who are unable to tolerate or unwilling to try CPAP should be considered for PT. In patients with POSA, PT has been shown to be as effective as CPAP in normalizing AHI, improving nocturnal oxygenation, and improving sleep quality [31]. However, despite the efficacy of PT, adherence over time has been generally poor [32]. Our findings prompt need for further development of PT that can be both effective and readily tolerated.

There are several strengths to our present study to discuss. One such strength is the thorough individual clinician review of the EMR for confirmation of AF status by ICD-10 code and electrocardiogram, clinically-relevant HF by echocardiogram review and confirmation of HF exacerbation admission for patients with HFpEF, and medication prescription data. Including HF in the analysis is important as the pathophysiology of OSA in patients with HF might differ from those without OSA. The PSG data was of high fidelity and no patients were excluded due to incomplete data or artifact. We employed rather restrictive inclusion criteria. Further, our multivariable analysis using two definitions of POSA, one more strict and one less, provided insight on how the POSA definition can affect results. Similarly, several limitations need to be considered in our study. Given that only patients referred to our center for diagnostic PSGs were included, selection bias cannot be avoided. However, this bias exists for the entire patient population. Only patients who underwent full night diagnostic PSGs were included, and those with split night studies were excluded. This is because the sleep time of the diagnostic portion of the study in split night study is often limited making it challenging to determine POSA. This may have introduced bias into the data. The patients with and without AF underwent their sleep studies during different time periods, though efforts to mitigate this were used in that the sleep studies and scoring of the studies were carried out in the same manner regardless of the time at which the sleep study was performed. The diagnosis and burden of AF is difficult to truly assess without repeated or continuous rhythm monitoring. However, we took steps to ensure that patients with AF were accurately diagnosed and those without a diagnosis of AF did not have mention of AF in the EMR, electrocardiograms, or have prior therapy for AF, such as cardioversion or AF ablation. The heterogeneity of the AF population likewise may have some bearing on the results, but due to the sample size, subgroup analysis was not possible.

Conclusion

POSA is common in AF patients and particularly in those with mild OSA. Body habitus and OSA severity are associated with POSA. PT should be considered in patients with AF who exhibit POSA. Future studies examining the effectiveness and adherence of PT and the impact of PT on clinical outcomes in patients with AF should be considered.

Table 3b.

Multivariable Analysis of POSA by Exclusive Definition (ePOSA).

Characteristic	Odds Ratio (95% CI)	Chi Squared	p Value
Age	2.78 (1.30 – 6.17)	6.68	0.01
Male Sex	0.97 (0.35 – 2.65)	0.004	0.95
Obesity	0.39 (0.17 – 0.85)	5.32	0.02
Atrial Fibrillation	0.79 (0.36 – 1.71)	0.35	0.55
Heart Failure	0.32 (0.07 – 1.09)	2.75	0.10
Moderate OSA	0.26 (0.11 – 0.58)	0.02	0.89
Severe OSA	0.08 (0.02 – 0.24)	8.63	0.003

Data Availability Statement:

Data will be made available on reasonable request.