Abstract
Study Objectives
To examine the unclear, inconsistent role of sleep architectural disruption in atrial fibrillation (AF) development.
Methods
Patients (age ≥ 18 years) who underwent in-laboratory sleep studies at Cleveland Clinic 2000–2015 were examined (follow-up: 7.8 ± 3.5 years). Primary predictors were arousal index and total sleep time. Secondary predictors included sleep efficiency, wakefulness after sleep onset, sleep and REM latency, and percentage of each sleep stage. Predictors were fit to Cox proportional hazard models predicting time from sleep study to AF by diagnosis code. Covariates included demographics, anthropometrics, tobacco use, sleepiness, apnea–hypopnea index, sleep apnea-specific hypoxic impact, cardiovascular risk factors and disease, mood disorders, medications, and positive airway pressure.
Results
In our cohort (n = 27 232, age: 49.4 ± 14.5 years, 43.7% male, 73.9% white), 2077 (7.6%) developed incident AF. Arousal index was not associated with AF incidence. For every hour of decreased total sleep time, AF incidence increased 8% (HR = 1.08, 95% CI = 1.04 to 1.11). For every 10-unit decrease in sleep efficiency, AF incidence increased 6% (HR = 1.06, 95% CI = 1.04 to 1.09). For every hour of increased wakefulness after sleep onset, AF incidence increased 11% (HR = 1.11, 95% CI = 1.05 to 1.16). For every 10-unit increase in percent N1, AF incidence increased 6% (HR = 1.06, 95% CI = 1.01 to 1.10).
Conclusions
Less sleep time and greater sleep disruption were associated with increased incident AF in this large clinical cohort. These results suggest that sleep macro-architecture can influence AF development. Mechanistic and prospective studies are needed to verify whether sleep disruption is a novel target for AF prevention.
Keywords: atrial fibrillation, sleep duration, sleep stages
Graphical Abstract
Graphical Abstract.
Statement of Significance.
Results of our large cohort study suggest that sleep deprivation (total sleep time), and sleep disruption (sleep efficiency, wakefulness after sleep onset, and amount of sleep stage N1), but not sleep fragmentation (arousal index) may have a role in the development of atrial fibrillation (AF). Our findings clarify cross-sectional or self-reported inconsistencies identified in other large cohort studies, setting the stage for future prospective studies designed to assess clinical risk stratification, integrate sleep architecture in endophenotyping, identify the patients at greatest AF risk for clinical trials, and mitigate sleep disruption with the goal of reducing atrial fibrillation burden and improving outcomes. Sleep deprivation and disruption, but not arousal index, may be modifiable risk factors for AF development.
Atrial fibrillation (AF) is the most common clinically significant arrhythmia with 50 million cases worldwide [1]. AF is associated with considerable morbidity and mortality [2], including ischemic stroke [3], resulting in substantial public health impact. The prevalence of AF has increased 3-fold over the last 50 years [1], and efforts have been initiated to identify effective strategies to address targetable risk factors [4]. As such, the role of sleep disorders has emerged as an important risk factor in the pathophysiology of incident AF, as highlighted in a recent American Heart Association Scientific Statement [5]. Most studies investigating sleep disorders as risk factors for AF have focused on the role of sleep disordered breathing at the crux of several deterministic pathophysiologic mechanisms such as intrathoracic pressure alterations and intermittent hypoxia, which contribute to autonomic fluctuations.
Existing knowledge gaps specific to the association between disruption of sleep and AF include (1) inconsistent existing studies of sleep duration as an AF predictor [6–8], (2) limited longitudinal data of sleep architectural alterations in AF development, and (3) incomplete understanding of the impact of arousals on AF. Polysomnographic-based sleep architectural disruptions, including curtailment of sleep, shorter N3 sleep, lower sleep efficiency, and reduced arousal index, have been identified as associated factors of AF in a cross-sectional investigation [9]; however, longitudinal studies are limited. One longitudinal study showed that more frequent nighttime awakening and an insomnia diagnosis were associated with increased AF development [10]. Despite the relationships between these sleep architecture measures and AF cross-sectionally, whether these measures represent longitudinal AF risk factors is unknown.
Given these knowledge gaps, we leveraged a large clinic-based registry of individuals with objective sleep indices from overnight sleep testing to examine the association of sleep disruption with incident AF. We hypothesized that sleep fragmentation (arousal index), decreased total sleep time (TST), and sleep architectural disruption (increased wakefulness after sleep onset, decreased sleep efficiency, and changes in sleep and rapid eye movement (REM) latency and sleep stage proportions) are significantly associated with incident AF.
Methods
Study methods and results are reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement for cohort studies [11]. A waiver of informed consent was granted by the Cleveland Clinic Institutional Review Board (IRB 21-474) for the following analyses.
Study design
We performed a retrospective cohort study to investigate the association of sleep architectural disruption with AF development. We used the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits (STARLIT) Registry, a repository of 320 607 sleep studies on 215 017 patients conducted at the Cleveland Clinic since 1998. We sought to elucidate inter-relationships of our predictors (primary: arousal index and TST, secondary: wakefulness after sleep onset, sleep efficiency, sleep latency, REM latency, and percentage of each sleep stage) with incidence of AF.
Study population
Adult patients (≥18 years old) who had full night diagnostic polysomnography between January 2, 2000 and December 30, 2015 were included in analyses. Split studies with a titration portion, titration studies, studies with less than 3 hours of recording time, patients who never followed up after their sleep study, and patients with baseline AF at the time of the sleep study were excluded from all models. If a patient underwent multiple sleep studies, data from their first study were used. A waiver of informed consent from study patients in the STARLIT Registry was granted by the Cleveland Clinic institutional review board (IRB number 19-864).
Data collection
Sleep testing and registry
Electroencephalography (EEG), electrocardiography (ECG), electromyography of the chin and bilateral anterior tibialis muscle, electrooculography, oral and nasal airflow, respiratory effort, end-tidal carbon dioxide (Salter Labs, Arvin, CA; sampling rate 4 Hz), and pulse oximetry (Nihon Kohden, Japan, sampling rate 25 Hz) were recorded. Polysmith® software was used for data collection. The American Academy of Sleep Medicine guidelines [12] were used to stage sleep. An arousal was defined as ≥3 seconds of abrupt shift in EEG frequency. The arousal index was defined as the number of arousals per hour of sleep. Total sleep time was the time asleep during the sleep study. Sleep efficiency was expressed as a percentage and defined as the time asleep compared to time in bed. Wakefulness after sleep onset was defined by amount of wake activity in hours occurring after initial onset of sleep. Sleep latency was the time in minutes from lights out to initial onset of sleep. REM latency was the time in minutes from initial onset of sleep to onset of REM. Apnea–hypopnea index (AHI) was defined as the number of respiratory events per hour of sleep, with hypopnea definition of 3% or 4% desaturation based on insurance guidelines, given this was a clinical cohort. Sleep apnea-specific hypoxic burden (SAHB) was defined as the area under the curve of desaturations related to respiratory events.
Discrete variables were extracted directly from the electronic health record (EHR; Epic®, Madison, WI) at the patient level. Demographics, comorbidities, and positive airway pressure (PAP) prescriptions were collected from the EHR and sleep study variables were collected from the sleep study reports. The ability to collect these data from Epic employed natural language processing methods such as parsing with regular expressions using structured query language (SQL) for database creation.
Study outcome and validation
AF was identified by International Classification of Diseases codes 427.31, 427.32, and I48, and Current Procedural Terminology codes 93650, 93653, 93656, and 93657 in clinical and procedural encounter notes. We developed parsing programs to validate this approach. Using this method, 90% of patients with these diagnosis codes were confirmed by ECG, sleep study report, or echocardiography. Of those who could not be confirmed internally, 80% were confirmed external to the Cleveland Clinic, however, those data are not available for our use. Those with AF prior to or at the time of the sleep study were excluded. The study outcome was incident AF developed after the sleep study, defined as the first occurrence of AF diagnosis code following the sleep study date. Patients who were not diagnosed with AF were censored at their last follow-up date.
Statistical analyses
The primary predictors were arousal index and TST (hours). Secondary predictors were sleep efficiency (%), wakefulness after sleep onset (hours), sleep latency (minutes), REM latency (minutes), and percentage of time spent in each sleep stage (percentage of non-REM 1 (N1), N2, N3 and REM). We examined TST continuously and categorically. We chose category thresholds based on the median TST for our cohort (5.7 hours), based upon precedence from prior studies [13–15], the consensus statement from the American Academy of Sleep Medicine that adults should sleep 7 or more hours per night [16], and the size of each group (≤5 hours, n = 7745; 5–7 hours, n = 16 272; and ≥ 7 hours, n = 2925). We used the 5–7 hour TST group as the reference, because other cohorts have shown a J- or U-shaped relationship of TST with cardiovascular outcomes [17–19].
Hazard ratios (HR) with 95% confidence intervals (CI) were computed using Cox proportional hazards models that were fit with the dependent variable: time from sleep study to AF diagnosis. Predictors were examined in individual models. Multi-collinearity was assessed with variance inflation factors and deemed substantial if greater than 5 [20]. The proportional hazards assumption was assessed with plots of Schoenfeld residuals. Missingness was handled with multiple imputation by chained equations. Covariates included age, sex, race, median income by zip code, marital status, body mass index (BMI, kg/m2), neck circumference, tobacco use, Epworth Sleepiness Scale score, AHI as a measure of sleep disordered breathing severity, SAHB, cardiovascular risk factors (hypertension, diabetes, dyslipidemia), cardiovascular disease (coronary artery disease, heart failure, myocardial infarction, coronary artery bypass grafting, stroke), depression, anxiety, anti-depressant use, benzodiazepine use, non-benzodiazepine benzodiazepine receptor agonist use, anti-arrhythmic medication use, and PAP therapy, which was treated as a time-varying covariate. Diagnoses were ascertained via diagnosis code from office visits and procedural encounters in the EHR using the same diagnosis codes as published previously [21, 22] (Table S1).
In secondary analyses, interactions were created between each significant predictor and demographic variable (age, sex, and race), as well as between each significant predictor and the presence of obstructive sleep apnea (OSA; AHI ≥ 5) and moderate-severe OSA (AHI ≥ 15). Sensitivity analyses were performed (1) excluding those with OSA (AHI ≥ 5), (2) excluding those on antiarrhythmic medications in case these patients had AF at baseline and did not have a diagnosis code, and (3) stratified by hypopnea definition. Tests were two-sided and p-values <.05 were considered statistically significant. Computations were done on R, version 4.3.1.
Results
Overall cohort
Our registry-derived analytic cohort included 44 594 adult patients (≥18 years old). Of these, 881 (2.0%) were excluded for recording time <3 hours, 14 681 (33.6%) were excluded for split night/PAP titration sleep study type, 1761 (6.1%) were excluded for having baseline AF at the time of the sleep study, and 39 (0.1%) were excluded for lack of follow-up after the sleep study, leaving 27 232 in our final analytic cohort (Figure 1). Of these, 2077 (7.6%) went on to develop incident AF after their sleep study during the mean follow-up period of 7.8 ± 3.5 years.
Figure 1.
Flow chart of cohort selection. Of 44 594 adult patients who underwent sleep studies between January 2, 2000 and December 20, 2015, 881 were excluded for short recording time, 14 681 were excluded for split night/PAP titration sleep study type, 1761 were excluded for baseline AF, and 39 were excluded for lack of follow-up. There were 27 232 patients in the final analytic subset. Of these, 2077 patients (7.6%) developed incident AF (outcome of interest). Created in BioRender. Heinzinger, C. (2025) https://BioRender.com/h39p557
The analytic cohort had a mean age of 49.4 ± 14.5 years, 43.7% were male, 73.9% were White, 19.8% were Black, 5.5% were other race, and median BMI was 31.9 kg/m2 with interquartile range [27.4–37.9] (Table 1, Table S2 for supplemental characteristics, Table S3 for missingness). Older age, male sex, white race, being widowed, higher BMI, larger neck size, less sleepiness, higher AHI, all comorbidities, and use of anti-arrhythmic medications, PAP, and benzodiazepines were associated with those patients who developed AF after their sleep study compared to those who did not.
Table 1.
Patient Characteristics
| All patients | Atrial fibrillation |
No atrial fibrillation |
P-value | |
|---|---|---|---|---|
| N | 27 232 | 2077 | 25 155 | |
| Age, mean (SD) | 49.4 (14.5) | 61.4 (12.9) | 48.4 (14.2) | <.001 |
| Sex | ||||
| Female | 15 323 (56.3%) | 976 (47.0%) | 14 347 (57.0%) | <.001 |
| Male | 11 907 (43.7%) | 1101 (53.0%) | 10 806 (43.0%) | <.001 |
| Race | ||||
| White | 20 137 (73.9%) | 1589 (76.5%) | 18 548 (73.7%) | <.001 |
| Black | 5394 (19.8%) | 403 (19.4%) | 4991 (19.8%) | <.001 |
| Other | 1491 (5.5%) | 85 (4.1%) | 1406 (5.6%) | <.001 |
| BMI | 31.9 (27.4, 37.9) | 32.9 (28.5, 38.4) | 31.8 (27.3, 37.8) | <.001 |
| Apnea–hypopnea index | 11.1 (5.1, 22.6) | 16.3 (7.2, 30.7) | 10.7 (5, 21.9) | <.001 |
| Central apnea index | 0.0 (0.0, 0.2) | 0.0 (0.0, 0.4) | 0.0 (0.0, 0.2) | <.001 |
| Obstructive apnea index | 0.4 (0, 2.0) | 0.9 (0.2, 3.3) | 0.4 (0.0, 1.9) | <.001 |
| Hypopnea index | 10.1 (4.5, 19.5) | 13.5 (6.2, 24.3) | 9.8 (4.4, 19.1) | <.001 |
| Sleep apnea-specific hypoxic burden | 20.2 (7.4, 50.6) | 32.9 (13.1, 70.1) | 19.6 (7.1, 48.9) | <.001 |
| Percent time SaO2 < 90% | 0.6 (0.0, 4.3) | 2.5 (0.3, 13.2) | 0.6 (0, 3.8) | <.001 |
| Minimum SaO2 | 86 (82, 90) | 84 (79, 88) | 87 (82, 90) | <.001 |
| Comorbidities | ||||
| Hypertension | 12 552 (46.1%) | 1460 (70.3%) | 11 092 (44.1%) | <.001 |
| Diabetes | 4555 (16.7%) | 643 (31.0%) | 3912 (15.6%) | <.001 |
| Dyslipidemia | 13 292 (48.8%) | 1371 (66.0%) | 11 921 (47.4%) | <.001 |
| Coronary artery disease | 2491 (9.1%) | 527 (25.4%) | 1964 (7.8%) | <.001 |
| Coronary artery bypass grafting | 559 (2.1%) | 173 (8.3%) | 386 (1.5%) | <.001 |
| Heart failure | 944 (3.5%) | 287 (13.8%) | 657 (2.6%) | <.001 |
| Myocardial infarction | 471 (1.7%) | 95 (4.6%) | 376 (1.5%) | <.001 |
| Stroke | 2448 (9.0%) | 374 (18.0%) | 2074 (8.2%) | <.001 |
| Depression | 14 992 (55.1%) | 1203 (57.9%) | 13 789 (54.8%) | .007 |
| Anxiety | 13 641 (50.1%) | 1110 (53.4%) | 12 531 (49.8%) | .002 |
| Anti-arrhythmia use | 73 (0.3%) | 30 (1.4%) | 43 (0.2%) | <.001 |
| Antidepressant use | 11 390 (41.8%) | 860 (41.4%) | 10 530 (41.9%) | .704 |
| Benzodiazepine use | 5374 (19.7%) | 526 (25.3%) | 4848 (19.3%) | <.001 |
| Non-benzodiazepine benzodiazepine receptor agonist use | 2047 (7.5%) | 163 (7.8%) | 1884 (7.5%) | .581 |
| Predictors | ||||
| Arousal index | 23.9 (15.9, 35.3) | 28.4 (18.3, 43.1) | 23.5 (15.7, 34.8) | <.001 |
| Total sleep time (hours) | 343 (290, 386) | 310 (248, 356) | 345 (294, 387) | <.001 |
| Sleep efficiency (%) | 80.4 (69, 88.2) | 72.9 (59.1, 83.2) | 81 (69.9, 88.4) | <.001 |
| Wakefulness after sleep onset (hours) | 53 (28.5, 91.5) | 78.8 (44.5, 124.5) | 51 (27.5, 88.5) | <.001 |
| Sleep latency (minutes) | 18 (8.5, 35.5) | 19.5 (9, 37.5) | 18 (8.5, 35.5) | .005 |
| REM latency (minutes) | 114 (75, 190) | 121 (73, 202) | 113 (75, 189) | .184 |
| Percent stage N1 | 7.3 (4.3, 12) | 9.5 (5.5, 15.8) | 7.2 (4.3, 11.7) | <.001 |
| Percent stage N2 | 65.3 (57.2, 73) | 67.4 (59.25, 75.1) | 65.1 (57, 72.8) | <.001 |
| Percent stage N3 | 7.7 (0.6, 15.8) | 2.2 (0, 11.3) | 8 (0.8, 16) | <.001 |
| Percent REM | 16.7 (10.9, 21.75) | 14.7 (8.4, 20.4) | 16.9 (11.1, 21.9) | <.001 |
Variables presented as median (interquartile range) for continuous variables unless otherwise noted and n (%) for categorical variables. BMI, body mass index; PAP, positive airway pressure; REM, rapid eye movement sleep stage; SaO2, oxygen saturation; SD, standard deviation.
Primary analyses
Arousal index was not associated with incident AF (omnibus p-value = .148). Every hour of decreased TST was associated with a 8% increased incidence of AF (HR = 1.08, 95% CI = 1.04 to 1.11). The short sleepers (≤5 hours) had 18% greater incidence of AF compared to the 5–7-hour group (HR = 1.18, 95% CI = 1.08 to 1.30). The difference between longer sleepers (≥ 7 hours) and the 5–7-hour group was not significant (HR = 0.89, 95% CI = 0.73 to 1.09), possibly due to the smaller sample size of the longer sleeper (≥7 hours) group. Every 10-percentage point decrease in sleep efficiency (e.g. 80% to 70%) was associated with 6% increase in AF incidence (HR = 1.06, 95% CI = 1.04 to 1.09). Every hour of increased wakefulness after sleep onset was associated with an 11% increase in AF incidence (HR = 1.11, 95% CI = 1.05 to 1.16). Every 10-percentage point increase in percent sleep stage N1 (e.g. 5% to 15%) was associated with a 6% increase in AF incidence (HR = 1.06, 95% CI = 1.01 to 1.10; Table 2, Figure 2). Sleep latency, REM latency, and percentage of sleep stages N2, N3, and REM were not associated with incidence of AF.
Table 2.
Cox Proportional Hazards Models of Incident Atrial Fibrillation
| Hazard ratio (95% CI) | P-value | |
|---|---|---|
| Arousal Index (vs. 0–15.9) | ||
| 16.0–23.9 | 0.87 (0.76, 1.00) | .049 |
| 24.0–35.3 | 0.97 (0.85, 1.10) | .633 |
| 35.4+ | 1.00 (0.88, 1.15) | .964 |
| Total sleep time (per 1-hour decrease) | 1.08 (1.04, 1.11) | <.001 |
| Sleep efficiency (per 10% decrease) | 1.06 (1.04, 1.09) | <.001 |
| Wakefulness after sleep onset (per 1 hour) | 1.11 (1.05, 1.16) | <.001 |
| Sleep latency (per 10 minutes) | 1.01 (0.99, 1.02) | .366 |
| REM latency (per 10 minutes) | 1.00 (0.99, 1.00) | .594 |
| Percent stage N1 (per 10%) | 1.06 (1.01, 1.10) | .016 |
| Percent stage N2 (per 10%) | 0.98 (0.94, 1.02) | .265 |
| Percent stage N3 (per 10%) | 1.01 (0.95, 1.08) | .647 |
| Percent stage REM (per 10%) | 0.96 (0.91, 1.01) | .151 |
Predictors are per increment increase unless otherwise noted. The omnibus p-value for categorized arousal index is 0.148. Each model included the following covariates: age, sex, race, median income by ZIP code, marital status, BMI, neck circumference, Epworth Sleepiness Scale score, apnea–hypopnea index, sleep apnea-specific hypoxic burden, hypertension, diabetes, dyslipidemia, coronary artery disease, coronary artery bypass grafting, heart failure, myocardial infarction, stroke, depression, anxiety, anti-arrhythmia medication use, antidepressant use, benzodiazepine use, non-benzodiazepine benzodiazepine receptor agonist use, and positive airway pressure therapy (treated as a time-varying covariate). REM, rapid eye movement sleep stage.
Figure 2.
Cumulative incidence plots of significant variables. Models were adjusted for age, sex, race, median income by zip code, marital status, body mass index, neck circumference, tobacco use, Epworth Sleepiness Scale score, apnea–hypopnea index as a marker of sleep disordered breathing severity, sleep apnea-specific hypoxic impact, coronary artery disease, heart failure, myocardial infarction, coronary artery bypass grafting, hypertension, diabetes, dyslipidemia, stroke, depression, anxiety, anti-depressant use, benzodiazepine use, non-benzodiazepine benzodiazepine receptor agonist use, anti-arrhythmic use, and positive airway pressure (PAP) therapy, which was treated as a time-varying covariate. A 1-hour decrease in total sleep time was associated with an 8% increase in incidence of AF (HR = 1.08, 95% CI = 1.04 to 1.11; Panel A). A 10-percentage point decrease in sleep efficiency was associated with a 6% increase in incidence of AF (HR = 1.06, 95% CI = 1.04 to 1.09; Panel B). A 1-hour increase in wakefulness after sleep onset was associated with an 11% increase in incidence of AF (HR = 1.11, 95% CI = 1.05 to 1.16; Panel C). A 10-percentage point increase in percent sleep stage 1 was associated with a 6% increase in incidence of AF (HR = 1.06, 95% CI = 1.01 to 1.10; Panel D).
There was no evidence of multi-collinearity and the proportional hazards assumption was met in all Cox proportional hazards models.
Secondary analyses
Interactions of each of continuous and categorical TST, sleep efficiency, wakefulness after sleep onset, and percent stage N1 were created with age, sex, and race to examine effect modification by these demographic variables, as well as with OSA (AHI ≥ 5) and moderate-severe OSA (AHI ≥ 15). No interactions with demographic characteristics were significant (Table S4). Interactions of moderate-severe OSA (AHI ≥ 15) with TST (p = .046), sleep efficiency (p = .032), and wakefulness after sleep onset (p = .040) were significant (Figure 3, Table S5). The relationship between shorter TST, lower sleep efficiency, and greater wakefulness after sleep onset with increased AF incidence was stronger in those with moderate-severe OSA (n = 10 813) than in those with no/mild OSA (n = 16 286). A sensitivity analysis excluding those with OSA showed findings were no longer significant, perhaps due to decreased sample size (n = 6348; Table S6).
Figure 3.
Cumulative incidence plots of significant interactions of moderate-severe obstructive sleep apnea with predictors. Models were adjusted for age, sex, race, median income by zip code, marital status, body mass index, neck circumference, tobacco use, Epworth Sleepiness Scale score, sleep apnea-specific hypoxic impact, coronary artery disease, heart failure, myocardial infarction, coronary artery bypass grafting, hypertension, diabetes, dyslipidemia, stroke, depression, anxiety, anti-depressant use, benzodiazepine use, non-benzodiazepine benzodiazepine receptor agonist use, anti-arrhythmic use, and positive airway pressure (PAP) therapy which was treated as a time-varying covariate. The interactions of moderate-severe OSA with TST (p = .046), sleep efficiency (p = .032), and wakefulness after sleep onset (p = .040) were significant.
The patients on anti-arrhythmic medications who did not have AF by diagnosis code at baseline (n = 77) underwent manual chart review by a content expert (CH). There were 4 patients who had AF; these patients were excluded from all analyses. We performed a sensitivity analysis excluding the remaining patients on anti-arrhythmic medications (n = 73 on antiarrhythmic medications, n = 27 159 included in sensitivity analysis) in case any of these patients had AF at baseline, did not have an AF diagnosis code, and were missed by manual review. Results were unchanged (Table S7).
Models stratified by hypopnea definition showed that results persisted in studies with hypopneas defined by 3% desaturation or arousal, but no longer in studies with hypopneas defined by 4% desaturation. The sleep architectural changes in studies with hypopneas defined by 4% desaturation may be due to decreased sample size (n = 2731 vs. n = 20 802). In the models for studies with hypopneas defined by 3% desaturation or arousal, arousal index was significantly associated with AF incidence (omnibus p = .044; Table S8).
Discussion
This large observational cohort (n = 27 232) study leveraged a clinical registry to elucidate the relationships between objective polysomnographic measures of sleep architecture and the development of newly diagnosed AF, with a mean follow-up of 7.8 ± 3.5 years. We identified important novel findings from a single night of recording that address prior knowledge gaps and confirm findings of a previous study [23]. These key findings include: (1) decreased TST is associated with increased AF incidence, (2) reduced sleep efficiency and increased wakefulness after sleep onset are associated with incident AF, (3) arousal index is not associated with incident AF, (4) increased sleep stage N1 is associated with incident AF, possibly as an indirect measure of sleep fragmentation, and (5) interactions of moderate-severe OSA (AHI ≥ 15) with TST, sleep efficiency, and wakefulness after sleep onset were statistically significant. Our novel findings, which suggest increased vulnerability of atrial arrhythmogenesis in those with curtailed or disrupted sleep, are translationally and clinically relevant.
Our clinically significant finding of reduced TST and increased AF incidence is consistent with a large clinic-based cohort with a similar magnitude of association (8% increased incidence of AF in our cohort versus 9% increased incidence in a Pittsburgh-based cohort) [23]. Although our findings are similar to this prior work, we attempted to further enhance rigor by considering confounding influences of factors such as race, anthropometrics (neck circumference), sleepiness, sedative hypnotics, etc. Furthermore, we adjusted for SAHB in an effort to account for sleep apnea-related hypoxia with rigor and considered PAP as a time-varying covariate. We examined additional sleep architecture measures including sleep efficiency and wakefulness after sleep onset. To test the robustness of our findings, we performed several sensitivity analyses including comparison of OSA versus no OSA and varying degree of OSA. We also excluded those with OSA and stratified by hypopnea definition.
Our work is also within range of prior reports investigating the association of self-report of sleep duration and AF (6%–7%) [7, 8]. In contrast, an epidemiological study found no association between TST and incident AF [10]. This difference in findings may be due to a higher prevalence of AF in clinical cohorts compared to population-based studies, since sleep studies may be ordered to elucidate an underlying cause for arrhythmia. Sleep deprivation increases sympathetic nervous system activation and is associated with elevated levels of systemic pro-inflammatory biomarkers, thereby supporting biological plausibility for our observations and potentially serving as intermediates in the causal pathway(s) of AF [24–26]. Although sympathetic nervous system surges accompany EEG micro-arousals, cross-sectional results of the Multi-Ethnic Study of Atherosclerosis (MESA) study implicate an inverse association of arousal frequency and AF suggesting that arousals serve as protective indicators of termination of respiratory events [9]. While we did not find an association of arousal index and increased AF risk in the full cohort, we did observe this association in sleep studies with hypopneas defined by 3% desaturation or arousal, possibly due to increased inherent consideration of arousals in the 3% hypopnea definition. The lack of association in the full cohort may be due to a greater burden of hypoxia in a clinical cohort, and hypoxic mechanisms may serve as a strong driving factor in atrial arrhythmogenic risk related to respiratory events. Furthermore, we considered all arousal subtypes and not only those specifically associated with respiratory events.
Mechanistic studies have shown evidence of biologic plausibility that arousal index may be associated with cardiovascular disorders, but epidemiologic studies have been less consistent. Specifically, an experimental study of healthy subjects (n = 20) implicated endothelial dysfunction and increased arterial stiffness as a result of repetitive arousals [27]. Results of two cross-sectional studies showed that arousal index may be a biomarker of sympathetic nervous system overactivity, one via measuring heart rate variability [28] and the other by measures of nocturnal blood pressure elevation [29]. Larger cross-sectional studies of Sleep Heart Health Study data have found weak to null associations of arousal index and cardiovascular disease [30]. Informing our hypothesis, a longitudinal study (n~6000) over an 11-year follow-up period showed that low arousal threshold predicted mortality [31]. In that study, arousal threshold was estimated by respiratory event duration, since arousals often truncate respiratory events. Therefore, increased arousals may represent an indicator of frequent, shorter respiratory events. More studies are needed to determine the longitudinal implications of these objective biomarkers of sleep disruption and cardiovascular outcomes.
We adjusted all models for SAHB (area under the respiratory event-related desaturation curve) to account for sleep apnea-related hypoxia. Findings suggest that the sleep architectural changes observed are associated with AF incidence independently of sleep apnea-specific hypoxia. This is important because hypoxia has been shown to be associated with AF incidence. Within this same cohort, although using a different analytic sample, a 10-unit increase in percentage time oxygen saturation <90%, a 10-unit decrease in mean oxygen saturation, and a 10-unit decrease in minimum oxygen saturation were associated with 6%, 30%, and 9% increased AF incidence, respectively [22].
While interactions with OSA presence (AHI ≥ 5) were not significant, interactions with moderate-severe OSA (AHI ≥ 15) were statistically significant and potentially clinically meaningful. In particular, the relationship between shorter TST, reduced sleep efficiency, and greater wakefulness after sleep onset with increased AF incidence was stronger in those with moderate-severe OSA compared to those with no/mild OSA. Findings suggest that the observed relationship between sleep architecture and incident AF may be driven by the presence of moderate-severe OSA, which is a novel finding not previously described. The attenuation of the primary findings in the sub-analysis of those without OSA may be due to decreased sample size (n = 6348 vs. 27 232). OSA is a risk factor for AF and disrupts sleep architecture. Our prior work shows that every 10-unit increase in AHI was associated with 2% higher AF incidence [22]. Another study found a 4-fold higher likelihood of AF in those with severe sleep apnea than in those without [32]. PAP therapy has been shown to play a role in mitigating AF recurrence post-ablation [33]. Therefore, we conducted secondary analyses elucidating the role of OSA and carefully accounted for AHI, SAHB, and PAP prescription (time-varying) via covariate adjustment.
A strength of this study is the large sample size, which allowed us to: (1) examine effect modification by demographic factors and sleep disordered breathing, and (2) more accurately estimate effect sizes across sleep architectural measures to better inform mechanistic or prospective studies that may then inform clinical trials. Half of our cohort were women, which results in broader generalizability of findings compared to similar studies that only examined [7] or primarily examined [8] men. The consideration of objective sleep measures gleaned directly from polysomnography enhances rigor compared to reports of self-reported sleep measures, the latter of which are more likely to have ascertainment bias. We validated our AF ascertainment with ECG to enhance rigor. Careful consideration of the impact of OSA on the findings via covariate adjustment and secondary analyses is a strength.
Several limitations should be noted. The retrospective nature of this study precludes causal inference. Selection bias may exist given this is a clinical referral sample, which may influence generalizability of the findings. Residual confounding is possible despite our attempts to optimize rigor by adjustment for potential confounding through medication use, sleepiness, sleep disordered breathing, psychiatric conditions, comorbidities, and demographics. Some patients may not have had any medical records prior to the sleep study. Therefore, we could not rule out preexisting AF. Our study population may have had a higher prevalence of OSA, limiting generalizability to individuals without OSA. We adjusted our models for PAP prescription but lacked data on PAP adherence. Adaptations in sleep study staging and scoring, updates in data acquisition software, and modernized sleep laboratory hardware may introduce temporal variation, however, this would be expected to bias toward the null. While single night sleep studies align with clinical guidelines, night-to-night variation in sleep stages, TST, timing, and disruption may exist. Due to first night effect, these sleep features may not align with those of their habitual sleep in the domiciliary environment. Finally, patients may have followed up outside our hospital system. This would be expected to cause non-differential misclassification and underestimate the magnitude of association.
Our findings implicate sleep deprivation and disruption as clinically relevant exposures in terms of AF risk, possibly driven in part by OSA. Future studies should investigate the effects of sleep architectural disruption on AF in individuals without OSA specifically, or implement meticulous control for OSA severity. Findings are relevant to inform future mechanistic or prospective observational studies, which may then advise risk-stratification for clinical trials. In addition, findings may inform studies to identify endophenotypes characteristic of sleep architecture associated with different arrhythmia risk, which may be more responsive to interventions focused on enhancing sleep quantity and quality. Future research should consider using actigraphy data to provide a more reliable understanding of our findings or focus on interventional studies that mitigate sleep disruption with the goal of reducing AF impact and improving outcomes. Our results can be applied to studies to examine novel measures with enhanced accuracy capturing specific pathophysiologic consequences of disrupted sleep architecture, such as heart rate variability in relation to occurrence of micro-arousals.
Supplementary material
Supplementary material is available at SLEEP online.
Acknowledgments
John Fredieu, medical writer. The work was performed at Cleveland Clinic.
Contributor Information
Catherine M Heinzinger, Sleep Disorders Center, Neurological Institute, Cleveland Clinic, Cleveland, OH, USA.
Nicolas R Thompson, Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH, USA; Neurological Institute Center for Outcomes Research & Evaluation, Cleveland Clinic, Cleveland, OH, USA.
Alex Milinovich, Department of Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH, USA.
Matheus Araujo, Sleep Disorders Center, Neurological Institute, Cleveland Clinic, Cleveland, OH, USA; Lerner Research Institute, Cleveland Clinic, Cleveland, OH, USA.
Cinthya Pena Orbea, Sleep Disorders Center, Neurological Institute, Cleveland Clinic, Cleveland, OH, USA.
Nancy Foldvary-Schaefer, Sleep Disorders Center, Neurological Institute, Cleveland Clinic, Cleveland, OH, USA.
Michael Faulx, Heart, Vascular, and Thoracic Institute, Cleveland Clinic, Cleveland, OH, USA.
David R Van Wagoner, Lerner Research Institute, Cleveland Clinic, Cleveland, OH, USA.
Mina K Chung, Lerner Research Institute, Cleveland Clinic, Cleveland, OH, USA; Heart, Vascular, and Thoracic Institute, Cleveland Clinic, Cleveland, OH, USA.
Reena Mehra, Department of Pulmonary, Critical Care, and Sleep Medicine, Department of Medicine, University of Washington, Seattle, WA, USA.
Disclosure Statements
Financial disclosure: Catherine M Heinzinger: Supported by American Academy of Sleep Medicine grant 302-PA-23. Mina K Chung: Supported by NIH grants P01HL158502, R01HL158071, American Heart Association grants 18SFRN34110067 and 18SFRN 34170013. Matheus Lima Diniz Araujo: Supported by NIH grant R21HL170206-01. Reena Mehra: Received an honorarium from the American Academy of Sleep Medicine; has received funds for service on the American Board of Internal Medicine and as associate editor of the American Journal of Respiratory and Critical Care Medicine; has received National Institutes of Health funding (R21HL170206); funds from Eli Lilly; and has received royalties from UpToDate. Nonfinancial disclosure: none.
Data Availability
The data underlying this article may be shared upon reasonable request to the corresponding author.
References
- 1. Kornej J, Borschel CS, Benjamin EJ, Schnabel RB. Epidemiology of atrial fibrillation in the 21st century: novel methods and new insights. Circ Res. 2020;127(1):4–20. doi: https://doi.org/ 10.1161/CIRCRESAHA.120.316340 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Lippi G, Sanchis-Gomar F, Cervellin G. Global epidemiology of atrial fibrillation: an increasing epidemic and public health challenge. Int J Stroke. 2021;16(2):217–221. doi: https://doi.org/ 10.1177/1747493019897870 [DOI] [PubMed] [Google Scholar]
- 3. Kannel WB, Wolf PA, Benjamin EJ, Levy D. Prevalence, incidence, prognosis, and predisposing conditions for atrial fibrillation: population-based estimates. Am J Cardiol. 1998;82(8A):2N–9N. doi: https://doi.org/ 10.1016/s0002-9149(98)00583-9 [DOI] [PubMed] [Google Scholar]
- 4. Benjamin EJ, Chen P-S, Bild DE, et al. Prevention of atrial fibrillation. Circulation. 2009;119(4):606–618. doi: https://doi.org/ 10.1161/CIRCULATIONAHA.108.825380 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Mehra R, Chung MK, Olshansky B, et al. ; on behalf of the American Heart Association Electrocardiography and Arrhythmias Committee of the Council on Clinical Cardiology; and Stroke Council. Sleep-disordered breathing and cardiac arrhythmias in adults: mechanistic insights and clinical implications: a scientific statement from the American heart association. Circulation. 2022;146:CIR0000000000001082. doi: https://doi.org/ 10.1161/CIR.0000000000001082 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Chokesuwattanaskul R, Thongprayoon C, Sharma K, Congrete S, Tanawuttiwat T, Cheungpasitporn W. Associations of sleep quality with incident atrial fibrillation: a meta‐analysis. Intern Med J. 2018;48(8):964–972. doi: https://doi.org/ 10.1111/imj.13764 [DOI] [PubMed] [Google Scholar]
- 7. Khawaja O, Sarwar A, Albert CM, Gaziano JM, Djoussé L. Sleep duration and risk of atrial fibrillation (from the Physicians’ health study). Am J Cardiol. 2013;111(4):547–551. doi: https://doi.org/ 10.1016/j.amjcard.2012.10.038 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Song Q, Liu X, Hu W, et al. Long sleep duration is an independent risk factor for incident atrial fibrillation in a Chinese population: a prospective cohort study. Sci Rep. 2017;7(1):1–7. doi: https://doi.org/ 10.1038/s41598-017-04034-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Kwon Y, Gharib SA, Biggs ML, et al. Association of sleep characteristics with atrial fibrillation: the multi-ethnic study of atherosclerosis. Thorax. 2015;70(9):873–879. doi: https://doi.org/ 10.1136/thoraxjnl-2014-206655 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Christensen MA, Dixit S, Dewland TA, et al. Sleep characteristics that predict atrial fibrillation. Heart Rhythm. 2018;15(9):1289–1295. doi: https://doi.org/ 10.1016/j.hrthm.2018.05.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Von EE, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370(9596):1453–1457. doi: https://doi.org/ 10.1016/s0140-6736(07)61602-x [DOI] [PubMed] [Google Scholar]
- 12. Troester MM, Quan SF, Berry RB, et al. The AASM manual for the scoring of sleep and associated events. Am Acad Sleep Med. 2023;3:18–71. [Google Scholar]
- 13. Cappuccio FP, Stranges S, Kandala NB, et al. Gender-specific associations of short sleep duration with prevalent and incident hypertension: the Whitehall II Study. Hypertension. 2007;50(4):693–700. doi: https://doi.org/ 10.1161/HYPERTENSIONAHA.107.095471 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Gangwisch JE, Heymsfield SB, Boden-Albala B, et al. Short sleep duration as a risk factor for hypertension: analyses of the first national health and nutrition examination survey. Hypertension. 2006;47(5):833–839. doi: https://doi.org/ 10.1161/01.HYP.0000217362.34748.e0 [DOI] [PubMed] [Google Scholar]
- 15. Song Q, Liu X, Wang X, Wu S. Age- and gender-specific associations between sleep duration and incident hypertension in a Chinese population: the Kailuan study. J Hum Hypertens. 2016;30(8):503–507. doi: https://doi.org/ 10.1038/jhh.2015.118 [DOI] [PubMed] [Google Scholar]
- 16. Consensus Conference P, Watson NF, Badr MS, et al. Recommended amount of sleep for a healthy adult: a joint consensus statement of the American academy of sleep medicine and sleep research society. J Clin Sleep Med. 2015;11(6):591–592. doi: https://doi.org/ 10.5664/jcsm.4758 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Cui H, Xu R, Wan Y, et al. Relationship of sleep duration with incident cardiovascular outcomes: a prospective study of 33,883 adults in a general population. BMC Public Health. 2023;23(1):124. doi: https://doi.org/ 10.1186/s12889-023-15042-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Wang Z, Yang W, Li X, Qi X, Pan KY, Xu W. Association of sleep duration, napping, and sleep patterns with risk of cardiovascular diseases: a nationwide twin study. J Am Heart Assoc. 2022;11(15):e025969. doi: https://doi.org/ 10.1161/JAHA.122.025969 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Ayas NT, White DP, Manson JE, et al. A prospective study of sleep duration and coronary heart disease in women. Arch Intern Med. 2003;163(2):205–209. doi: https://doi.org/ 10.1001/archinte.163.2.205 [DOI] [PubMed] [Google Scholar]
- 20. Sheather S. A Modern Approach to Regression with R. New York, NY: Springer; 2009. [Google Scholar]
- 21. Heinzinger CM, Lapin B, Thompson NR, et al. Novel sleep phenotypic profiles associated with incident atrial fibrillation in a large clinical cohort. JACC Clin Electrophysiol. 2024;10:2074–2084. doi: https://doi.org/ 10.1016/j.jacep.2024.05.027 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Heinzinger CM, Thompson NR, Milinovich A, et al. Sleep-disordered breathing, hypoxia, and pulmonary physiologic influences in atrial fibrillation. J Am Heart Assoc. 2023;12:e031462. doi: https://doi.org/ 10.1161/JAHA.123.031462 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Genuardi MV, Ogilvie RP, Saand AR, et al. Association of short sleep duration and atrial fibrillation. Chest. 2019;156(3):544–552. doi: https://doi.org/ 10.1016/j.chest.2019.01.033 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Castro-Diehl C, Diez Roux AV, Redline S, et al. Sleep duration and quality in relation to autonomic nervous system measures: the multi-ethnic study of atherosclerosis (MESA). Sleep. 2016;39(11):1927–1940. doi: https://doi.org/ 10.5665/sleep.6218 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Kanagasabai T, Ardern CI. Contribution of inflammation, oxidative stress, and antioxidants to the relationship between sleep duration and cardiometabolic health. Sleep. 2015;38(12):1905–1912. doi: https://doi.org/ 10.5665/sleep.5238 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Guo Y, Lip GY, Apostolakis S. Inflammation in atrial fibrillation. J Am Coll Cardiol. 2012;60(22):2263–2270. doi: https://doi.org/ 10.1016/j.jacc.2012.04.063 [DOI] [PubMed] [Google Scholar]
- 27. Wu Y, Huang R, Zhong X, Xiao Y. Cardiovascular consequences of repetitive arousals over the entire sleep duration. Biomed Res Int. 2017;2017:4213861. doi: https://doi.org/ 10.1155/2017/4213861 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Kim JB, Seo BS, Kim JH. Effect of arousal on sympathetic overactivity in patients with obstructive sleep apnea. Sleep Med. 2019;62:86–91. doi: https://doi.org/ 10.1016/j.sleep.2019.01.044 [DOI] [PubMed] [Google Scholar]
- 29. Carrington MJ, Trinder J. Blood pressure and heart rate during continuous experimental sleep fragmentation in healthy adults. Sleep. 2008;31(12):1701–1712. doi: https://doi.org/ 10.1093/sleep/31.12.1701 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Shahar E, Whitney CW, Redline S, et al. Sleep-disordered breathing and cardiovascular disease: cross-sectional results of the sleep heart health study. Am J Respir Crit Care Med. 2001;163(1):19–25. doi: https://doi.org/ 10.1164/ajrccm.163.1.2001008 [DOI] [PubMed] [Google Scholar]
- 31. Butler MP, Emch JT, Rueschman M, et al. Apnea-hypopnea event duration predicts mortality in men and women in the sleep heart health study. Am J Respir Crit Care Med. 2019;199(7):903–912. doi: https://doi.org/ 10.1164/rccm.201804-0758OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Mehra R, Benjamin EJ, Shahar E, et al. ; Sleep Heart Health Study. Association of nocturnal arrhythmias with sleep-disordered breathing: the sleep heart health study. Am J Respir Crit Care Med. 2006;173(8):910–916. doi: https://doi.org/ 10.1164/rccm.200509-1442OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Kanagala R, Murali NS, Friedman PA, et al. Obstructive sleep apnea and the recurrence of atrial fibrillation. Circulation. 2003;107(20):2589–2594. doi: https://doi.org/ 10.1161/01.CIR.0000068337.25994.21 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data underlying this article may be shared upon reasonable request to the corresponding author.