Relationship between Positional Obstructive Sleep Apnea and Excessive Daytime Sleepiness in Patients with Atrial Fibrillation

Abstract

Purpose:

Obstructive sleep apnea (OSA) is highly prevalent in patients with atrial fibrillation (AF). Prior studies suggest that excessive daytime sleepiness (EDS), a common symptom of OSA, is less pronounced in patients with AF. This study focused on a subtype of OSA, positional OSA (POSA), and examined its association with EDS overall and separately in patients with and without AF.

Methods:

This study included 222 OSA patients with and without AF history who underwent in-lab polysomnography. POSA was defined using three criteria: conventional POSA (cPOSA; supine/non-supine apnea-hypopnea index [AHI] ratio ≥ 2), intermediately conservative POSA (iPOSA; supine/non-supine AHI ratio ≥ 3), and exclusive POSA (ePOSA; supine/non-supine AHI ratio ≥ 2 and non-supine AHI < 5 events/h). EDS was determined by an Epworth Sleepiness Scale score ≥ 11. Logistic regression was performed to assess the associations.

Results:

Overall, 55.4% of patients had cPOSA, and 43.6% reported EDS. Analysis of the entire cohort showed that patients with cPOSA (adjusted odds ratio [aOR]: 0.51, 95% confidence interval [CI]: 0.26–0.98) or iPOSA (aOR: 0.37, 95% CI: 0.18–0.74) had lower odds of EDS than their non-POSA counterparts. Among patients with AF, cPOSA was associated with lower odds of EDS (aOR: 0.32, 95% CI: 0.11–0.90). Among patients without AF, iPOSA was associated with lower odds of EDS (aOR: 0.28, 95% CI: 0.10–0.75).

Conclusion:

The association between POSA and EDS varied by POSA definitions and AF status. The findings suggest that EDS manifests differently by cPOSA status in patients with AF.

INTRODUCTION

Atrial fibrillation (AF) is the most common cardiac arrhythmia in clinical practice [1]. Beyond traditional cardiovascular comorbidities, patients with AF are at substantially higher risk for sleep-related breathing disorders [2,3]. One study found that the prevalence of high-risk obstructive sleep apnea (OSA) was 2.5 times higher in AF patients compared to non-AF patients [4]. OSA, characterized by recurrent episodes of complete or partial collapse of the upper airway leading to oxygen desaturation or arousal from sleep, affects an estimated 40%−80% of patients with cardiovascular disease (CVD), including those who have AF [5,6].

Patients with AF often report fatigue and low energy, partly attributed to poor sleep quality [7,8]. When AF coexists with OSA, the disrupted sleep caused by repeated apnea events would be expected to further intensify poor sleep quality. Thus, excessive daytime sleepiness (EDS) may represent a hallmark symptom shared by AF and OSA [9,10]. However, available evidence shows a more complex relationship. In Starkey’s study of 188 adults with AF, 86% were diagnosed with OSA, and nearly half had moderate or severe OSA. Despite this high prevalence, the average daytime sleepiness severity in this cohort did not reach the threshold for EDS [11]. This suggests that OSA-related EDS symptoms may be less pronounced in patients with AF.

While OSA occurs in any position, a considerable proportion of patients experience supine-predominant OSA – also known as positional OSA (POSA). POSA is characterized by significantly greater breathing disruptions in the supine position compared with other sleep positions [12]. Data from the Sleep Heart Health Study (N = 1,719) showed that POSA was present in more than half of participants, and it accounted for 75% of those diagnosed with OSA [13]. Despite its high prevalence, whether the POSA subtype has implications in EDS manifestation remains limited. Patients with POSA experience intermittent periods of relatively unobstructed breathing that may attenuate cumulative sleep fragmentation and the arousal burden in comparison to patients with position-independent OSA of comparable severity [14,15]. This positional buffer may translate into a lower symptomatic burden of daytime sleepiness [16]. Additionally, patients with AF exhibit altered autonomic regulation, elevated arousal thresholds, and a symptom profile more often dominated by fatigue than by classical hypersomnolence [17,18]. These AF-specific characteristics may modify how the positional phenotype manifests as daytime sleepiness. Clarifying the POSA-EDS relationship separately in AF and non-AF patients may potentially help explain why EDS appears less pronounced in AF populations despite a high OSA burden and inform how clinicians interpret sleepiness symptoms in this group. Therefore, in this study, we examined the relationship between POSA and EDS in the full cohort and assessed whether POSA status was associated with EDS differently among patients with AF compared to those without AF.

METHODS

Sample and Design

We conducted a cross-sectional study at a single academic sleep center. The cohort consists of both AF and non-AF patients who underwent diagnostic in-laboratory polysomnography (PSG) for evaluation of OSA [19]. We initially identified AF using ICD-10 diagnoses recorded in the EMR. Our team then reviewed electrocardiograms to confirm AF and performed targeted chart reviews to establish that AF occurred before the PSG. For patients without AF, we reviewed charts to ensure no AF diagnosis was present by ICD-10 code, electrocardiogram interpretation, or medical comorbidity list. We enrolled consecutive patients with AF who were referred between January 2011 and January 2019. For comparison, we retrospectively identified patients without AF who underwent PSG during a 6-month window (July-December 2017) to yield a sample size of non-AF patients comparable to the AF cohort. Patients met the study criteria if they were aged ≥18 years old, had an apnea-hypopnea index [AHI] ≥ 5 events/h, had a total sleep time (TST) of at least 120 minutes, and spent at least 30 minutes in both supine and non-supine positions. For patients with multiple PSG studies, only the first was included. Exclusion criteria included predominant central sleep apnea, split-night studies, positive airway pressure titration studies, and home sleep studies. Data on sociodemographics, comorbidities, and polysomnography were extracted from electronic medical records. The University of Virginia Institutional Review Board approved the study (#19394).

MEASURES

PSG Processing and POSA

Overnight PSG was performed and scored in accordance with the American Academy of Sleep Medicine (AASM) guidelines [20,21]. Body position during sleep was recorded through an automated position sensor. The PSG recordings were processed using the Embla Sandman Elite software (Natus Medical Incorporated, Pleasanton, CA) and scored by certified PSG technologists. We extracted sleep metrics from PSG reports in an electronic medical record system using natural language processing, as described previously [22]. A thermocouple system was utilized to measure episodes of apneas and hypopneas. We defined an apnea event as a reduction in airflow exceeding 90% of the baseline for at least 10 seconds [21]. We defined a hypopnea event as the nasal pressure flow signal decreasing by more than 30% of baseline for at least 10 seconds. For hypopnea scoring, we applied the criterion requiring a minimum of 4% oxygen desaturation and followed the 2013 AASM 1B guidelines, or equivalent standards for studies conducted before 2013 [21]. The AHI was defined as the number of apnea and hypopnea events divided by the total sleep time.

We defined POSA using three criteria. Conventional POSA (cPOSA) was defined as an AHI ≥ 5 events/h and a supine/non-supine AHI ratio ≥ 2. Intermediately conservative POSA (iPOSA) was defined as an AHI ≥ 5 events/h and a supine/non-supine AHI ratio ≥ 3. Exclusive POSA (ePOSA), the most conservative definition, was defined as an AHI ≥ 5 events/h and a supine/non-supine AHI ratio ≥ 2 with normalization of breathing during non-supine sleep (non-supine AHI < 5 events/h).

Daytime Sleepiness

Daytime sleepiness was measured using the Epworth Sleepiness Scale (ESS) [23]. This 8-item, 4-point Likert-type scale evaluates the likelihood of dozing off or falling asleep in eight common life situations. Respondents who had not experienced some of the situations were asked to estimate how each might affect them. Scores for individual items are summed to derive a total score, with values >10 indicating EDS [24].

Covariates

We included the following variables as covariates in this study: age (< 65 or ≥ 65 years), biological sex (male or female), body mass index (BMI), history of hypertension (yes or no), history of type 2 diabetes (yes or no), TST, and OSA severity (mild [5 ≤ AHI < 15], moderate [15.0 ≤ AHI < 30], and severe [AHI ≥ 30.0]) ascertained from the PSG reports. We selected those variables because they are important confounding factors that may impact the relationship between POSA, EDS, and AF status.

Statistical Analysis

Descriptive statistics were presented as means ± standard deviations for continuous variables and as frequencies (proportions) for categorical variables. Patients’ sociodemographic and clinical characteristics were described for the overall cohort and stratified by POSA and AF status for comparison using either the t-test or the chi-square test. We first conducted logistic regression with the entire cohort to examine the association of three POSA definitions with EDS. Model 1 assessed the association between POSA and EDS. Model 2 added covariates, including age, sex, OSA severity, history of hypertension, history of diabetes, and TST. To assess whether the association between POSA and EDS was modified by AF status, we included an interaction term POSA × AF to the models described previously, along with the same covariates. The interaction analysis was repeated for each of the three POSA definitions (cPOSA, iPOSA, and ePOSA). Next, we examined the relationship between POSA and EDS for the AF and non-AF subgroups, separately. Since cPOSA is the most commonly used definition, we used cPOSA as the definition of the primary analysis and iPOSA and ePOSA for the secondary analysis. In addition, sensitivity analyses were performed to examine whether the inclusion of supine time or the supine time ratio (supine time/total sleep time) might change the relationship between POSA and EDS. All statistical analyses were performed using R (version 4.5.1). The α-level was set a priori for all statistical procedures at 0.05.

RESULTS

The analytic sample consisted of 222 patients after record review and application of the study criteria (Figure 1). Table 1 summarizes the sample’s sociodemographic and clinical characteristics. The patients’ mean age was 62 ± 14, and 55.2% were male. On average, they slept for 314.5 ± 98.4 minutes during the night of PSG. The average AHI was 23.3 ± 17.9 events/h. More than half (56.3%) had moderate to severe OSA. The average ESS score was 9.9 ± 5.2, with 43.6% of patients reporting EDS. As expected, the prevalence of POSA was highest for cPOSA (55.4%), followed by iPOSA (41.4%) and ePOSA (23%). When stratified by cPOSA, the cPOSA group had a significantly lower AHI (p = 0.001) and a lower proportion of patients reporting EDS (p = 0.017) than the non-POSA group. We additionally found significant differences in age, sex, BMI, history of hypertension, supine AHI, and non-supine AHI between the cPOSA and non-POSA groups. The cohort was subsequently stratified by AF status. There were 114 (51.4%) patients diagnosed with AF, of whom 91 (79.8%) had paroxysmal AF and 23 (20.2%) had persistent/permanent AF. Although the AF group had a slightly higher mean ESS score than the non-AF group, the difference was not statistically significant (p = 0.238), nor was there a significant difference in the prevalence of EDS (p = 0.809). Significant differences between AF and non-AF groups were found in age, history of hypertension, AHI, and non-supine AHI (all p < 0.05).

Figure 1.

Flowchart of Study Participants

Table 1.

Sociodemographic and clinical characteristics of study sample

	Total	POSA			AF
Variables	N = 222	POSA (n = 123)	Non-POSA (n = 99)	p	AF (n = 114)	Non-AF (n = 108)	p
Age in years (mean ± SD)	62 ± 14	64 ± 13	60 ± 16	0.049	66 ± 12	58 ± 15	< 0.001
< 65 years old	107 (48.4%)	52 (42.3%)	55 (55.6%)	0.067	42 (36.8%)	65 (60.2%)	0.001
≥ 65 years old	115 (51.6%)	71 (57.7%)	44 (44.4%)		72 (63.2%)	43 (39.8%)
Sex
Male	123 (55.2%)	76 (61.8%)	47 (47.5%)	0.046	65 (57.0%)	58 (53.7%)	0.662
Female	99 (44.6%)	47 (38.2%)	52 (52.5%)		49 (43.0%)	50 (46.3%)
Body mass index (mean ± SD)	32.3 ± 8.2	29.9 ± 6.7	35.2 ± 8.9	< 0.001	32.4 ± 7.7	32.1 ± 8.7	0.792
History of hypertension	135 (60.8%)	83 (67.5%)	52 (52.5%)	0.033	59 (51.8%)	76 (70.4%)	0.007
History of diabetes	73 (32.9%)	41 (33.3%)	32 (32.3%)	0.988	39 (34.2%)	34 (31.5%)	0.772
Total sleep time (mean ± SD)	314.5 ± 98.4	322.0 ± 89.9	306.0 ± 108.0	0.238	302.0 ± 100.1	328.0 ± 95.3	0.051
Total supine time (mean ± SD)	118.6 ± 82.4	115.0 ± 74.8	123.0 ± 91.1	0.481	108.4 ± 85.7	129.0 ± 77.6	0.058
Total non-supine time (mean ± SD)	195.5 ± 97.0	207.0 ± 92.2	183.0 ± 102.0	0.069	194.0 ± 104.0	198.0 ± 89.5	0.715
AHI (mean ± SD)	23.3 ± 17.9	19.4 ± 12.9	28.0 ± 21.8	0.001	26.0 ± 19.3	20.4 ± 15.8	0.018
Mild (5.0–14.9)	97 (43.7%)	60 (48.8%)	37 (37.4%)	0.008	42 (36.8%)	55 (50.9%)	0.043
Moderate (15.0–29.9)	69 (31.1%)	42 (34.1%)	27 (27.3%)		36 (31.6%)	33 (30.6%)
Severe (≥ 30.0)	56 (25.2%)	21 (17.1%)	35 (35.4%)		36 (31.6%)	20 (18.5%)
Supine AHI (mean ± SD)	35.4 ± 26.1	40.6 ± 24.9	29.0 ± 26.2	0.001	38.1 ± 27.0	32.6 ± 24.8	0.119
Non-supine AHI (mean ± SD)	18.2 ± 20.0	8.7 ± 8.4	29.9 ± 23.6	< 0.001	20.9 ± 21.4	15.2 ± 18.0	0.032
cPOSA (conventional)	123 (55.4%)	-	-	-	59 (51.8%)	64 (59.3%)	0.323
iPOSA (intermediately conservative)	92 (41.4%)	-	-	-	41 (36.0%)	51 (47.2%)	0.118
ePOSA (exclusive)	51 (23.0%)	-	-	-	21 (18.4%)	30 (27.8%)	0.134
ESS score (mean ± SD)	9.9 ± 5.2	9.4 ± 5.2	10.6 ± 5.1	0.085	10.4 ± 5.1	9.5 ± 5.3	0.238
Excessive daytime sleepiness	85 (43.6%)	37 (35.2%)	48 (53.3%)	0.017	44 (42.3%)	41 (45.1%)	0.809

Abbreviations: AF = atrial fibrillation; AHI = apnea-hypopnea index; ePOSA = exclusive positional obstructive sleep apnea; ESS = Epworth Sleepiness Scale; iPOSA = intermediately conservative positional obstructive sleep apnea; POSA = positional obstructive sleep apnea; SD = standard deviation

Note: Stratification of POSA was based on the conventional definition of a supine/non-supine AHI ratio ≥ 2. Counts may not sum to column totals due to missingness; percentages are based on non-missing observations.

Table 2 presents the relationship between POSA and EDS for the entire cohort. Patients with cPOSA had significantly lower odds of reporting EDS than non-POSA patients in both unadjusted (odds ratio [OR]: 0.48, 95% confidence interval [CI]: 0.27–0.84, p = 0.012) and adjusted models (adjusted OR [aOR]: 0.51, 95% CI: 0.26–0.98, p = 0.044). The same relationships were found when using the iPOSA definition in the unadjusted (OR: 0.32, 95% CI: 0.17–0.59, p <0.001) and adjusted models (aOR: 0.37, 95% CI: 0.18–0.74, p = 0.005). In contrast, when the most conservative definition, ePOSA, was applied, its association with EDS was observed only in the unadjusted model (OR: 0.48, 95% CI: 0.23–0.98, p = 0.048) and was no longer significant after adjusting for the covariates (aOR: 0.69, 95% CI: 0.30–1.55, p = 0.378). In addition to POSA, older age was consistently associated with lower odds of EDS across all models (all p < 0.05). Sensitivity analyses revealed no association between POSA and EDS after including supine time or supine time ratio (supine time/total sleep time) as a covariate in the models (Supplementary Tables 1a and 1b).

Table 2.

Relationship between positional obstructive sleep apnea and excessive daytime sleepiness (N = 222)

Study Variables	Excessive Daytime Sleepiness				Excessive Daytime Sleepiness				Excessive Daytime Sleepiness
	cPOSA (conventional)				iPOSA (intermediately conservative)				ePOSA (exclusive)
	Model 1		Model 2		Model 1		Model 2		Model 1		Model 2
	OR (95%CI)	p	aOR (95%CI)	p	OR (95%CI)	p	aOR (95%CI)	p	OR (95%CI)	p	aOR (95%CI)	p
POSA (ref: no)	0.48 (0.27, 0.84)	0.012	0.51 (0.26, 0.98)	0.044	0.32 (0.17, 0.59)	< 0.001	0.37 (0.18, 0.74)	0.005	0.48 (0.23, 0.98)	0.048	0.69 (0.30, 1.55)	0.378
Age (ref: < 65 years)			0.38 (0.19, 0.74)	0.005			0.43 (0.22, 0.85)	0.015			0.40 (0.20, 0.77)	0.007
Sex (ref: female)			0.95 (0.50, 1.82)	0.874			0.99 (0.52, 1.91)	0.977			0.87 (0.46, 1.64)	0.656
Body mass index			1.01 (0.97, 1.05)	0.684			1.01 (0.97, 1.05)	0.708			1.02 (0.98, 1.06)	0.386
History of hypertension (ref: no)			1.37 (0.70, 2.72)	0.367			1.34 (0.68, 2.66)	0.397			1.24 (0.64, 2.42)	0.529
History of diabetes (ref: no)			1.56 (0.79, 3.08)	0.196			1.46 (0.74, 2.89)	0.279			1.47 (0.75, 2.88)	0.260
Total sleep time			1.14 (0.93, 1.42)	0.216			1.17 (0.95, 1.46)	0.150			1.12 (0.91, 1.39)	0.287
OSA severity (ref: mild)
Moderate			1.48 (0.70, 3.15)	0.306			1.67 (0.78, 3.64)	0.191			1.35 (0.63, 2.90)	0.437
Severe			1.75 (0.74, 4.23)	0.205			1.82 (0.76, 4.42)	0.179			1.75 (0.72, 4.34)	0.218

Abbreviations: AHI = apnea-hypopnea index; aOR = adjusted odds ratio; CI: confidence interval; cPOSA = conventional positional obstructive sleep apnea; ePOSA = exclusive positional obstructive sleep apnea; iPOSA = intermediately conservative positional obstructive sleep apnea; OR = odds ratio

Note: Results are shown separately for three definitions of POSA: cPOSA, iPOSA, and ePOSA. Model 1 represents unadjusted analyses; Model 2 adjusted for covariates.

To assess whether AF modified the association between POSA and EDS, we added the POSA x AF interaction term to the adjusted models. No significant interaction was observed for any POSA definitions: cPOSA × AF: aOR 0.60, 95% CI 0.17–2.12, p = 0.431; iPOSA × AF: aOR 1.74, 95% CI 0.48–6.44, p = 0.400; ePOSA × AF: aOR 2.03, 95% CI 0.42–9.61, p = 0.370. Please see Table S2.

Next, we stratified the cohort by AF status (Tables 3 and 4). AF patients with cPOSA had 68% lower odds of experiencing EDS than AF patients without cPOSA (aOR: 0.32, 95% CI: 0.11–0.90, p = 0.034). iPOSA was associated with EDS only in the unadjusted model (OR: 0.41, 95% CI: 0.17–0.94, p = 0.039), while ePOSA showed no association with EDS in any models. Among non-AF patients, iPOSA was associated with lower odds of EDS in both unadjusted (OR: 0.24, 95% CI: 0.10–0.58, p = 0.002) and adjusted (aOR: 0.28, 95% CI: 0.10–0.75, p = 0.013) models. Although ePOSA was associated with lower odds of EDS in the unadjusted model (OR: 0.34, 95% CI: 0.12–0.88, p = 0.031), this association was attenuated after including the covariates (aOR: 0.46, 95% CI: 0.15–1.36, p = 0.169).

Table 3.

Relationship between positional obstructive sleep apnea and excessive daytime sleepiness in patients with atrial fibrillation (n = 114)

Study Variables	Excessive Daytime Sleepiness				Excessive Daytime Sleepiness				Excessive Daytime Sleepiness
	cPOSA (conventional)				iPOSA (intermediately conservative)				ePOSA (exclusive)
	Model 1		Model 2		Model 1		Model 2		Model 1		Model 2
	OR (95%CI)	p	aOR (95%CI)	p	OR (95%CI)	p	aOR (95%CI)	p	OR (95%CI)	p	aOR (95%CI)	p
POSA (ref: no)	0.38 (0.17–0.84)	0.018	0.32 (0.11–0.90)	0.034	0.41 (0.17, 0.94)	0.039	0.53 (0.19, 1.43)	0.210	0.70 (0.22–2.02)	0.524	1.19 (0.32–4.31)	0.790
Age (ref: < 65 years)			0.25 (0.09–0.69)	0.009			0.30 (0.11, 0.82)	0.020			0.28 (0.10–0.75)	0.013
Sex (ref: female)			0.98 (0.38–2.59)	0.962			0.80 (0.32, 2.01)	0.626			0.70 (0.29–1.70)	0.429
Body mass index			1.00 (0.93–1.07)	0.935			1.01 (0.95, 1.08)	0.731			1.02 (0.96–1.09)	0.471
History of hypertension (ref: no)			2.28 (0.84–6.65)	0.117			1.95 (0.74, 5.43)	0.187			1.92 (0.73–5.30)	0.194
History of diabetes (ref: no)			1.69 (0.65–4.43)	0.281			1.51 (0.60, 3.83)	0.383			1.45 (0.58–3.62)	0.427
Total sleep time			1.23 (0.91, 1.69)	0.177			1.22 (0.91, 1.66)	0.189			1.20 (0.9, 1.64)	0.222
OSA severity (ref: mild)
Moderate			1.83 (0.59–5.94)	0.303			2.07 (0.69, 6.62)	0.204			2.10 (0.68–6.82)	0.202
Severe			1.69 (0.49–6.04)	0.409			2.15 (0.66, 7.45)	0.213			2.44 (0.71–9.00)	0.165

Abbreviations: AHI = apnea-hypopnea index; aOR = adjusted odds ratio; CI: confidence interval; cPOSA = conventional positional obstructive sleep apnea; ePOSA = exclusive positional obstructive sleep apnea; iPOSA = intermediately conservative positional obstructive sleep apnea; OR = odds ratio

Note: Results are shown separately for three definitions of POSA: cPOSA, iPOSA, and ePOSA. Model 1 represents unadjusted analyses; Model 2 adjusted for covariates.

Table 4.

Relationship between positional obstructive sleep apnea and excessive daytime sleepiness in patients without atrial fibrillation (n = 108)

Study Variables	Excessive Daytime Sleepiness				Excessive Daytime Sleepiness				Excessive Daytime Sleepiness
	cPOSA (conventional)				iPOSA (intermediately conservative)				ePOSA (exclusive)
	Model 1		Model 2		Model 1		Model 2		Model 1		Model 2
	OR (95%CI)	p	aOR (95%CI)	p	OR (95%CI)	p	aOR (95%CI)	p	OR (95%CI)	p	aOR (95%CI)	p
POSA (ref: no)	0.59 (0.25, 1.37)	0.220	0.72 (0.29, 1.80)	0.478	0.24 (0.10, 0.58)	0.002	0.28 (0.10, 0.75)	0.013	0.34 (0.12, 0.88)	0.031	0.46 (0.15, 1.36)	0.169
Age (ref: < 65 years)			0.54 (0.19, 1.55)	0.254			0.66 (0.22, 1.99)	0.457			0.59 (0.20, 1.72)	0.337
Sex (ref: female)			1.04 (0.41, 2.67)	0.934			1.14 (0.44, 3.07)	0.787			1.06 (0.41, 2.73)	0.909
Body mass index			1.03 (0.97, 1.09)	0.324			1.02 (0.96, 1.08)	0.482			1.03 (0.98, 1.09)	0.294
History of hypertension (ref: no)			0.76 (0.24, 2.43)	0.644			0.82 (0.25, 2.73)	0.742			0.69 (0.21, 2.23)	0.540
History of diabetes (ref: no)			1.77 (0.63, 5.14)	0.282			1.50 (0.51, 4.49)	0.463			1.67 (0.58, 4.88)	0.340
Total sleep time			1.08 (0.79, 1.49)	0.612			1.15 (0.83, 1.61)	0.416			1.07 (0.79, 1.47)	0.668
OSA severity (ref: mild)
Moderate			1.14 (0.39, 3.28)	0.806			1.51 (0.49, 4.72)	0.474			0.97 (0.33, 2.83)	0.956
Severe			1.93 (0.50, 7.79)	0.339			1.82 (0.45, 7.78)	0.404			1.46 (0.36, 6.12)	0.598

Abbreviations: AHI = apnea-hypopnea index; aOR = adjusted odds ratio; CI: confidence interval; cPOSA = conventional positional obstructive sleep apnea; ePOSA = exclusive positional obstructive sleep apnea; iPOSA = intermediately conservative positional obstructive sleep apnea; OR = odds ratio

Note: Results are shown separately for three definitions of POSA: cPOSA, iPOSA, and ePOSA. Model 1 represents unadjusted analyses; Model 2 adjusted for covariates.

DISCUSSION

In this clinical cohort of patients referred to a sleep clinic for OSA evaluation, we observed that patients with POSA had lower odds of reporting EDS. This finding was more pronounced in patients with AF than in those without AF, although the interaction between POSA and AF was not statistically significant. In patients with AF, the cPOSA definition was associated with a lower likelihood of EDS, whereas in patients without AF, this association was observed only when a more conservative definition of POSA was applied. These findings suggest that the POSA-EDS relationship may depend on AF status as well as the definition of POSA used.

In this cohort, EDS was highly prevalent in both AF and non-AF groups, with no significant difference between the groups. Specifically, 42% of patients with AF reported EDS. This prevalence is much higher than that reported in most prior studies of AF patients with OSA. For example, in a study of AF patients who underwent PSG, two-thirds were found to have OSA, of whom only 11.9% reported similarly defined EDS [25]. Another study including patients undergoing direct current cardioversion reported EDS in 37.2% of AF patients with OSA, which is similar to our study [26]. The overall higher prevalence of EDS observed in our study may be attributable to the clinical setting, as patients referred to sleep clinics are more symptomatic, and they are more likely to be systematically screened for OSA than in non-referred settings.

One of our main study findings was that OSA patients who exhibited positional dependence were less likely to report EDS than those without POSA, regardless of AF status. Several mechanisms may explain the lower likelihood of EDS observed among patients with POSA. First, because POSA is defined by relatively patent upper airway during non-supine sleep, these patients experience intermittent recovery periods when in non-supine positions, which likely reduce cumulative sleep fragmentation and arousal load compared with patients whose respiratory events occur across positions [14,27]. Second, POSA has been associated with a milder overall physiologic phenotype, such as shorter and less severe apneas, less oxygen desaturation, and better preserved sleep architecture, all of which may be linked to reduced daytime sleepiness [28,29]. Third, it is important to note that POSA is often uncovered in the sleep lab because PSG setup (multiple wires) may force patients to sleep more in supine sleep position and patients are also at times asked to attempt a supine position with hope to capture worst-case scenario [30]. This implies that the severity of OSA based on in-lab PSG may be artificially inflated in patients with POSA when in reality, their OSA is less severe at home. This may also explain the attenuated OSA related symptoms in these patients compared with patients with a similar OSA severity. Fourth, patients with POSA may adapt behaviorally by favoring non-supine sleep postures, which could further alleviate symptoms [30]. Importantly, our study was not designed to disentangle these mechanisms. Further work is needed to investigate the underlying mechanism that drive the lower EDS burden observed in POSA.

Our analyses suggest that this association was not explained by the difference in OSA severity or other relevant factors included in the models, with the exception of age. POSA represents a phenotype of positional dependence, and time spent in supine or non-supine positions can vary substantially within any given POSA phenotype. In sensitivity analyses, the association between POSA and EDS was attenuated after adjusting for supine sleep time or the supine ratio. This attenuation may reflect the overlap between these covariates and the POSA classification itself. Because POSA is derived from supine and non-supine AHI and each position-specific AHI is scaled by time spent in that position, adding supine time could introduce a highly correlated predictor and may therefore dilute significance without changing the underlying association. Because these analyses were exploratory, they should be interpreted as complementary to, rather than contradicting, the primary findings. As such, replication in independent cohorts with detailed positional sleep data will be needed to clarify the relative contributions of positional dependence and supine sleep duration to daytime sleepiness.

When stratified by AF status, the POSA-EDS association depended on how POSA was defined. Among patients with AF, lower odds of EDS were observed with cPOSA, the most liberal definition. Among patients without AF, the association emerged only with iPOSA, a more stringent definition. This suggests that the POSA-EDS relationship is more pronounced in AF patients, in whom even a modest positional phenotype can produce difference in daytime sleepiness. Although ePOSA showed a similar direction of effect in the non-AF subgroup, its relatively low prevalence in our cohort (23.0%) likely limited statistical power to detect an association after adjusting for the covariates. One possible explanation of the study findings is that AF-specific physiology may amplify the fragmentation-sparing effect of positional dependence. AF is associated with increased sympathetic tone, elevated arousal thresholds, and a symptom profile more often dominated by fatigue than by sleepiness. In this context, the intermittent recovery periods afforded by non-supine sleep may translate more directly into preserved daytime alertness among AF patients, whereas in non-AF patients, other drivers of EDS may predominate and mask this effect. Nonetheless, these interpretations are speculative and require confirmation in future research.

Our findings inform the interpretation of sleep studies in patients with OSA and AF. Because a POSA phenotype cannot be identified until PSG is completed, it cannot serve as a pre-test screening criterion. However, once PSG has been performed and POSA is identified in an AF patient, a low symptom burden (including minimal or absent EDS) should not be interpreted as evidence of low disease burden. Rather, attenuated sleepiness may be a phenotypic feature of POSA in this population and should be considered in treatment decision making.

There are several study limitations that warrant consideration. First, our cohort comprised patients referred for diagnostic PSGs at a single medical center. This cohort may be subject to selection bias and limit generalizability. Second, the retrospective, cross-sectional design precludes causal inference. Third, the classification of POSA is highly sensitive to the quality and quantity of recorded supine and non-supine sleep based on the night of the sleep study. Similar to many research studies using PSG data, relying on a single-night recording could lead to misclassification of the POSA phenotype. In addition, sleep position captured during in-lab PSG may not accurately represent habitual sleep position at home, and patients classified as POSA in the laboratory may spend little or no time supine in their usual sleep environment, which could further complicate the interpretation of POSA and its symptomatic consequences, particularly in the AF population [30]. Fourth, daytime sleepiness was assessed only with the ESS, which measures sleeping propensity but not fatigue. Because AF patients frequently experience fatigue as the dominant sleep-related symptom, and fatigue is not readily differentiated from sleepiness, relying only on ESS may have failed to capture true symptom burden from OSA. Such misclassification may bias the association observed in this study. Future studies are necessary to incorporate fatigue-specific instruments (e.g., Fatigue Severity Scale) and objective measures of sleepiness to disentangle the overlapping constructs. Fifth, the AF and non-AF groups differed at baseline in age, hypertension, and OSA severity. Although these variables were adjusted for in all regression models, residual confounding cannot be fully excluded. Sixth, we lacked data on whether participants were in AF rhythm during the PSG. However, for the purpose of the study, AF was defined by a history of AF (e.g., chronic condition vs. presence of AF rhythm during sleep study) and EDS as a chronic symptomatology (vs. presence of EDS following the sleep study). Finally, the AF and non-AF groups were drawn from non-overlapping time periods (AF: January 2011-January 2019; non-AF: July-December 2017), which may introduce selection bias. Over this interval, growing clinician awareness of the AF-OSA link and expanding use of home sleep testing likely shifted the case-mix of patients referred for in-lab PSG. To minimize measurement variability, all studies were scored using uniform AASM criteria, but residual time-related confounding cannot be excluded.

In conclusion, our study identified POSA subtype as an associated factor for EDS manifestation in a cohort of patients with and without AF. In particular, this relationship was more pronounced in AF patients. Although EDS is an important factor in clinical decision-making for OSA treatment, our findings underscore the limitations of symptom-based screening in AF populations. Clinicians should consider OSA, in the form of POSA phenotype, in AF patients who do not report daytime sleepiness.