Author + information
- Received December 17, 2018
- Revision received February 25, 2019
- Accepted March 13, 2019
- Published online June 17, 2019.
- Dominik Linz, MD, PhDa,
- Anthony G. Brooks, PhDb,
- Adrian D. Elliott, PhDa,
- Chrishan J. Nalliah, MBBSc,
- Jeroen M.L. Hendriks, PhDa,
- Melissa E. Middeldorp, PhDa,
- Celine Gallagher, RNa,
- Rajiv Mahajan, MD, PhDa,
- Jonathan M. Kalman, MBBS, PhDc,
- R. Doug McEvoy, MDd,
- Dennis H. Lau, MBBS, PhDa and
- Prashanthan Sanders, MBBS, PhDa,∗ ()
- aCentre for Heart Rhythm Disorders, South Australian Health and Medical Research Institute, University of Adelaide and Royal Adelaide Hospital, Adelaide, South Australia, Australia
- bCardiac Rhythm Management, MicroPort, Scoresby, Victoria, Australia
- cDepartment of Cardiology, Royal Melbourne Hospital and Department of Medicine, University of Melbourne, Melbourne, Victoria, Australia
- dAdelaide Institute for Sleep Health, College of Medicine and Public Health, Flinders University, and Sleep Health Service, Respiratory and Sleep Services, Southern Adelaide Local Health Network, Adelaide, South Australia, Australia
- ↵∗Address for correspondence:
Dr. Prashanthan Sanders, Centre for Heart Rhythm Disorders, Department of Cardiology, Royal Adelaide Hospital, Port Road, Adelaide 5000, Australia.
Objectives This study sought to determine night-to-night variability in the severity of sleep-disordered breathing (SDB) and the dynamic intraindividual relationship to daily risk of incident atrial fibrillation (AF) by using simultaneous long-term day-by-day SDB and AF monitoring.
Background Night-to-night variability in SDB severity may result in a dynamic exposure to SDB related conditions impacting the timing and extent of cardiovascular responses.
Methods This study was an observational cohort study. Daily data for AF burden and average respiratory disturbance index (RDI) were extracted from pacemakers capable of monitoring nightly SDB and daily AF burden in 72 patients. Nightly RDI values were grouped into quartiles of severity within each patient. AF burdens of >5 min, >1 h, and >12 h were the outcome variables.
Results A total of 32% of patients had a mean RDI of ≥20/h, indicative of overall severe SDB. There was significant night-to-night variation in RDI reflected by an absolute SD of ±6.3 events/h (range 2 to 14 events/h) within any given patient. Within each patient, the nights with the highest RDI (in their highest quartile) conferred a 1.7-fold (1.2 to 2.2; p < 0.001), 2.3-fold (1.6 to 3.5; p < 0.001), and 10.2-fold (3.5 to 29.9; p < 0.001) increase risk of having at least 5 min, 1 h, and 12 h, respectively, of AF during the same day compared with the best sleep nights (in their lowest quartiles).
Conclusions There is considerable night-to-night variability in SDB severity which cannot be detected by 1 single overnight sleep study. SDB burden may be a better metric with which to assess the extent of dynamic SDB related cardiovascular responses such as daily AF risk than the categorical diagnosis of SDB. (Night-to-Night Variability in Severity of Sleep Apnea and Daily Dynamic Atrial Fibrillation Risk [VARIOSA-AF]; ACTRN 12618000757213)
- atrial fibrillation
- dynamic substrate
- night-to-night variability
- sleep-disordered breathing
The number of apnea and hypopnea episodes per hour (apnea-hypopnea-index [AHI]) determined during a single overnight sleep study is clinically used to diagnose and assess the severity of sleep-disordered breathing (SDB) (1). However, recent studies suggest that SDB severity in an individual patient is not stable over time but exhibits considerable night-to-night variability which cannot be detected by 1 overnight sleep study (2,3). This night-to-night variability in SDB severity may result in a variable and dynamic exposure to SDB related conditions which may impact the timing and extent of cardiovascular responses (4).
In patients with atrial fibrillation (AF), the prevalence of categorically moderate-to-severe SDB ranges between 21% and 72% due to variable selected cohorts, applied scoring rules, and thresholds used (5–8). Observational studies suggest that SDB reduces the efficacy of catheter-based and pharmacological antiarrhythmic therapy and that treatment with continuous positive airway pressure (CPAP) lowers the rate of AF recurrence after electrical cardioversion and improves catheter ablation success rates in AF patients (9,10). Despite evolving evidence supporting an important role for the categorical SDB diagnosis in the management of AF patients, the pathophysiological importance of night-to-night variability in SDB severity for AF risk is unclear. Pre-clinical and clinical observations suggest that SDB together with chronic concomitant conditions such as hypertension and obesity may promote an arrhythmogenic progressive structural remodeling progress in SDB (11–14). In addition to the chronic structural alterations in the atria, transient apnea-associated arrhythmogenic changes may further enhance AF susceptibility, creating a dynamic arrhythmogenic substrate in the atrium (11,15–17). Theoretically, rather than the categorical diagnosis of SDB obtained from a single diagnostic overnight sleep study, AF risk may be mediated through dynamic night-to-night changes in SDB severity.
Although chronic structural alterations induced by SDB have already been described in AF patients with SDB (13,14), clinical evidence for a dynamic AF substrate in patients related to nightly SDB severity is lacking. Using simultaneous long-term night-by-night SDB and AF monitoring, this study characterized night-to-night variability in SDB severity and examined the relationship between SDB and AF, with each patient acting as their own control. Previously, preliminary findings highlighted the nightly variation in SDB (18); the present study describes the complete and expanded analysis of the VARIOSA-AF (VARIability in severity Of Sleep Apnea and daily dynamic Atrial Fibrillation) risk study.
This study was an observational cohort study. The study was approved by the Clinical Research Ethics Committee of the Royal Adelaide Hospital, Adelaide, Australia, and was registered in Australian New Zealand Clinical Trials Registry (ACTRN identifier 12618000757213).
The cohort was extracted from a de-identified home monitoring database of patients with dual-chamber pacemakers implanted according to guideline-directed indications (19). A data dump was performed in a sample of 191 patients with Reply 200 or Kora 100 DR pacemakers (MicroPort, Clamart, France) with implemented SDB monitoring measuring a respiratory disturbance index (RDI) (20,21). Daily SDB and AF burdens of all downloaded patient data were extracted by the MicroPort research department and provided in a de-identified file. Demographic or clinical descriptive data could not be retrieved beyond that recorded on the pacemaker, given the de-identified nature of the dataset. Figure 1 provides a CONSORT Figure of the study population.
Night-to-night SDB monitoring
Regardless of the diagnosis of SDB, daily SDB severity was assessed by the pacemaker-based sleep apnea monitoring (SAM) (MicroPort) algorithm that was described and validated previously against polysomnography (20,21). Briefly, transthoracic impedance measurements are measured at 6 Hz (through micropulses) and are processed to determine the amplitude and period of respiratory cycles (i.e., minute ventilation). The SAM algorithm defines ventilation pause as an absence of respiratory cycle for >10 s and ventilation reduction as a decrease of respiratory amplitude by at least 50% for >10 s. An RDI for the respective day is then calculated as the total number of ventilation pauses and reductions per hour during a 5-h programmable monitoring period between 12:00 am and 5:00 am. In the DREAM study, an RDI ≥20/h identified patients with severe SDB (AHI ≥30/h) by polysomnography performed on the same night with 89% sensitivity and 85% specificity (NCT01537718) (20,21). The RDI is determined on a nightly basis.
Night-to-night variability in RDI was determined using a nightly coefficient of variation and/or SD within the patient monitoring period and then averaged across the sample. To characterize the SDB pattern within each patient, SDB category changes between severe (mean RDI: ≥20/h) or nonsevere SDB (mean RDI: <20/h) based on the average RDI during consecutive, nonoverlapping, 1-week blocks were determined. A paroxysmal pattern was present if consecutive weeks of 1 SDB category were separated by at least 1 week with a different SDB category. In contrast, a persistent SDB patterns was present if all consecutive weeks displayed a mean RDI of either ≥20/h or <20/h. To standardize across patients, each patient's daily RDIs were grouped into quartiles, with the highest quartile representing the 25% of nights with the highest RDI within each patient.
Atrial fibrillation/atrial tachycardia events
The outcome variables were classified according to pacemaker-defined “mode switch events” (MicroPort) (22). In a MicroPort dual-chamber device, a mode switch requires a sustained pathological acceleration of the atrial channel (>25% acceleration compared to the mean of 8 preceding “normal” atrial cycles) >120 beats/min. At least 28 of 32 atrial accelerated beats (87.5%) (primary criterion) or secondarily at least 36 of 64 accelerated atrial beats (56.3%) are required for mode switch declaration. The duration of mode switch for each day was determined to range from 1 to 86,400 s (24 h) after the primary or secondary criterion were met. To assess the relationship between SDB severity and AF burden, analyses of 3 different total cumulative mode switch episodes were recorded: >5 min per day; >1 h per day; and >12 h per day were included. Thus, 5 × 1-min mode switch episodes were effectively treated the same as a 1 × 5-min episode of AF. Electrograms of up to 11 mode switch events were available, and all mode switch recordings were manually adjudicated to ensure genuine AF, atrial flutter, or atrial tachycardia.
Direction of relationship
To examine the directionality of relationship “SDB begets AF” or “AF begets SDB,” the prediction analysis for RDI data was run (e.g., from Friday night to Saturday morning) and the AF burden (e.g., from Saturday) of the same day and reran the analysis with a 1-day offset between the AF episode and the RDI, such that the AF burden during 1 day (e.g., during Saturday) then corresponded to the SDB severity determined by RDI of the following morning (e.g., from Saturday night to Sunday morning). We also examined the 6-patient subset with higher overall AF burden was also examined (>25% AF burden) and means tested the nocturnal RDI on days with 100% AF burden compared to days without AF within each patient.
Data are mean ± SD when normally distributed and median (interquartile range [IQR]) if data followed a skewed distribution. Categorical data are proportion with numerator and denominator in brackets. A binary logistic generalized estimating equation was used to develop a prediction model for the dichotomous AF outcome (>5 min odds ratio [OR]: >1 h; OR >12 h of cumulative mode switch per day) using the within-patient standardized (quartiles) SDB severity data. A subject variable of device serial number was introduced as the clustering variable to account for the nesting of outcome events within the individual. The SDB severity reference category was always the lowest RDI quartile. We performed our analyses in patient subsets that had at least 1 day of cumulative daily AF >5 min (n = 32), >1 h (n = 29), and >12 h (n = 11). Patients who did not have a suitable “AF event” were excluded because their within-patient OR were null. Mean changes in RDI in the subgroup with higher AF burden (n = 6) were assessed using mixed effects models with device serial number as the random factor. All data were analyzed using SPSS version 24 software (IBM, Armonk, New York) and significance was set at a p value of <0.05.
The sample consisted of 72 de-identified SAM-monitored dual-chamber pacemaker patients with lower overall AF burden (<25% AF burden). The median percent atrial pace was 55% (range 20% to 85%), and the percent of ventricular pace was 2.4% (range 0% to 41%). Ninety percent of the sample was programed to Safe-R (65 of 72) a minimized ventricular pacing algorithm; 58% of patients (42 of 72) had rate-response activated. The mean follow-up per patient was 21 ± 8 weeks, resulting in a total daily monitoring time for the sample of >28 years (10,383 days).
Sleep-disordered breathing severity: burden and night-to-night variability
The SAM algorithm provided a valid RDI estimate in 10,315 of 10,383 days (99%). The average RDI across the sample was 17.9 ± 11.5 events/h. In 32% of patients (23 of 72) the mean RDI was ≥20/h, indicative of overall severe SDB.
Figure 2 shows the night-to-night RDI variability in 3 representative patients with different levels of underlying SDB severity. Fifty-six percent of patients (40 of 72) had at least 1 week with a mean RDI ≥20/h, of whom 73% (29 of 40) demonstrated paroxysmal weekly SDB patterns (i.e., weeks of RDI ≥20/h separated by at least 1 week with an RDI <20/h) in contrast to persistent SDB patterns in the minority of 28% (11 of 40) where all consecutive weeks displayed a mean RDI ≥20/h. An RDI ≥20/h during at least 1 night was observed in 85% of patients (61 of 72). The individual mean night-to-night RDI coefficient of variation was 41 ± 16%, which reflected an absolute SD of ±6.3 events/h within each patient (range 2 to 14 events/h). The median RDI for each within-patient quartile (Q1 to 4) was 8.2 (IQR: 4.6 to 15.2), 12.6 (IQR: 7.8 to 21.6), 16.4 (IQR: 10.4 to 26.8), and 23.2 (IQR: 15.4 to 35.5). From the lowest to the highest quartile for the entire sample.
Atrial fibrillation/atrial tachycardia burden
Of patients with lower overall AF burden (<25% AF burden), 57% of patients (41 of 72) had at least a single mode switch episode recorded on their device. Cumulative AF on any single day of their follow-up of >5 min, >1 h, and >12 h was observed in 44% (32 of 72), 40% (29 of 72), and 15% of patients (11 of 72), respectively. The total mode switch burden ranged from 0% to 24.2% with the median daily episode duration of 11.6 min (range 1.3 to 84 min). The total number of days satisfying >5 min, >1 h, and >12 h AF events within the monitoring period of each patient subset was 661 of 4,543 days (14.5%), 472 of 3,994 days (11.8%), and 130 of 1,748 days (7.4%), respectively.
Relationship between SDB severity and AF burden
Figure 3 summarizes the proportion of days with at least >5 min, >1 h, and >12 h AF for each within-patient quartile of SDB severity. Within each patient, the nights (12:00 am to 5:00 am) with the highest RDI (in their highest quartile) conferred a 1.7-fold (range 1.2- to 2.2-fold; p < 0.001), a 2.3-fold (range 1.6- to 3.5-fold; p < 0.001), and a 10.2-fold (range 3.5- to 29.9-fold; p < 0.001) increased risk of having at least 5 min, 1 h, and 12 h of AF during the same day compared with the quartile of nights with the lowest RDI (Figure 4).
To test for a potential bidirectional relationship with AF begetting SDB, the dataset of patients with lower overall AF burden was reanalyzed (AF burden: <25%) with a 1-day offset between the AF episode and the RDI. It was found that all relationships dissolved (Figure 4). Six patients with a higher AF burden (AF burden: >25%) were analyzed separately to examine the potential relationship between longer AF episodes and increased RDI. In this subgroup, the number of days with 24 h of AF, 0% AF, and paroxysmal AF (>0%, but <24 h) was 422 (50%), 139 (17%), and 279 (33%) days, respectively. It was found that days with 100% AF burden were associated with significantly higher RDIs (p < 0.001) than those days without AF.
Long-term SDB monitoring by chronically implanted devices revealed a considerable night-to-night variability in SDB severity, which cannot be detected by a single overnight sleep study. Long-term SDB and simultaneous AF monitoring showed for the first time that, in individual patients, the nights with the highest SDB severity conferred a more than 2.3-fold increase risk of having at least 1 h of AF the same day compared to nights with the lowest SDB severity. By contrast, there was no increased risk of AF in the 24 h before the nights with the highest SDB severity. These 2 contrasting observations provide strong evidence for a feed-forward mechanistic link between SDB and AF, whereby the increased cardiac stress induced by even a single night of more severe SDB can establish conditions conducive to AF.
These findings have important implications for the assessment of SDB severity and the guidance of SDB treatment in patients. The result of a single overnight sleep study may not be representative of SDB severity during the remaining 364 days of a year in a specific patient and could thus lead to frequent misclassification of SDB status in patients if just 1 overnight sleep study is performed (4). It may also help explain the highly variable prevalence of SDB (2–4) and may have contributed to the negative results of recent randomized controlled SDB treatment outcome studies (23–25). The proportion of nights with high RDI (SDB burden), rather than a categorical diagnosis of SDB per se, may better reflect the exposure to SDB-related factors and may be a more useful metric to determine SDB severity in the management of concomitant cardiovascular diseases such as AF (4). Importantly, the relationship between nights with more severe SDB and increased AF risk held for all individuals, regardless of the clinical diagnosis of SDB and mean RDI. This suggests that it may be the relative rather than the absolute increase in SDB severity on a single night that increases AF risk in the individual patient. Different mechanisms such as longer time spent sleeping supine, more alcohol consumption before bed, and variable compliance or withdrawal from CPAP treatment may contribute to variable SDB severity. SDB burden determined by continuous long-term SDB monitoring incorporates the considerable night-to-night variability in SDB severity documented in this study and previously described in patients with and without co-occurring cardiovascular disease (2,3,20,21).
The present study provides the first clinical evidence for the pathophysiological relevance of the night-to-night variability in SDB severity which contributes to a dynamic AF substrate. Pre-clinical studies suggest that intermittent hypoxia, autonomic dysregulation, and intrathoracic pressure swings during obstructive respiratory events, together with chronic concomitant conditions like hypertension and obesity, promote an arrhythmogenic progressive structural remodeling process in the atrium (4,11). This structural remodeling, characterized by more scar formation in the atrium, could be documented by electroanatomical mapping studies in AF patients with SDB (13,14). In animal models of SDB, simulated obstructive respiratory events reproducibly and reversibly shortened the atrial refractory period and transiently enhanced the inducibility of AF (15–17) and occurrence AF triggers (26). These short-term transient SDB-related arrhythmogenic changes may further contribute to the AF substrate. The present study provides clinical evidence supporting this hypothesized dynamic relationship between nocturnal SDB severity and the same day AF risk, which has not been reported in humans before (11).
Until recently, the causal link between SDB and cardiovascular disease has been believed to be mainly unidirectional, with SDB having a detrimental effect on the cardiovascular system and the heart (“SDB begets AF”) (4,11). Interestingly, emerging evidence points toward a crucial involvement of cardiovascular hemodynamics, nocturnal fluid shifts, and cardiac performance in the genesis of particularly central but also obstructive respiratory events in AF individuals (“AF begets SDB”) (4,11). To test for this potential bidirectional relationship, the dataset with a 1-day offset between the AF episode and SDB severity was reanalyzed, which showed no significant relationship. However, more persistent AF episodes, which were associated with significantly higher SDB severity in our study, may impact SDB severity in longer term follow-up, which needs to be evaluated in future studies.
Using this de-identified sample of patients provided clear evidence that the preceding night’s (12:00 am to 5:00 am) RDI can modify the risk of experiencing AF during the same day within each patient. Attention was focused exclusively on the intraindividual analysis of the interaction between SDB and AF burden, and comparisons between patients were not performed. Clinical risk factors are likely to remain constant in the maximum 6-month window that was analyzed. Further studies in larger well-characterized patient cohorts are needed. The SAM algorithm cannot distinguish between obstructive and central apneas or hypopneas. However, a detailed characterization of SDB events is not necessary to determine SDB severity assessed by the AHI (4). Additionally, detection of SDB by the SAM algorithm might not have always been accurate (20,21).
Long-term SDB monitoring demonstrates considerable night-to-night variability in SDB severity, which cannot be detected by 1 single overnight sleep study. The present study provides evidence for a dynamic substrate for AF, where SDB severity on a specific night directly relates to AF risk during the same day (Central Illustration). Instead of the current practice of establishing a categorical diagnosis of SDB from a single overnight sleep study, SDB burden determined as the proportion of nights with higher RDI may be a better metric to assess the extent of dynamic SDB-related cardiovascular responses such as daily AF risk and cardiovascular outcomes (27,28).
COMPETENCY IN MEDICAL KNOWLEDGE: Given the documented night-to-night variability in SDB severity, repeated sleep studies should be considered in patients with high clinical suspicion of SDB, especially in the setting of treatment-resistant hypertension or AF recurrence after catheter ablation. Many patients use CPAP intermittently which results in an artificial night-to-night variability in SDB severity; long-term SDB monitoring could help to document adherence to CPAP and response in respect to reduction in SDB and AF burden upon treatment.
TRANSLATIONAL OUTLOOK: Technologies implemented in implanted pacemaker home monitoring modules built in CPAP machines as well as radar-based or ballistic noncontact biomotion sensors with actimetry for remote monitoring of breathing during sleep can provide valuable data on short- and long-term variability in SDB. Future prospective and randomized intervention studies in well characterized patient cohorts are warranted A) to identify the best technology to assess daily SDB severity, B) to determine feasibility, accuracy and cost effectiveness of the implementation of long-term SDB monitoring as a component of risk factor modification programs, and C) to determine, whether SDB severity assessed in terms of SDB burden is a better indicator of cardiovascular outcomes than the categorical diagnosis of SDB.
The MicroPort research and development department exported the de-identified pacemaker records.
Dr. Linz is supported by a Beacon research fellowship, University of Adelaide; is an advisory board member for LivaNova/MicroPort and Medtronic; receives lecture and consulting fees from LivaNova/MicroPort, Medtronic, and ResMed; and has received funding from Sanofi, ResMed, and Medtronic. Drs Elliott and Hendriks are supported by Early Career Fellowships from the National Heart Foundation of Australia. Dr. Hendriks holds a Derek Frewin Lectureship, University of Adelaide; and is a consultant for Medtronic and Pfizer/BMS. Ms. Middeldorp is supported by a postgraduate scholarship from the National Health and Medical Research Council (NHMRC) of Australia. Ms. Gallagher is supported by a Leo J. Mahar scholarship, University of Adelaide. Dr. Mahajan is supported by an Early Career Fellowship, NHMRC, and a Leo J. Mahar lectureship, University of Adelaide; receives lecture and consulting fees from Medtronic, Abbott, Bayer, and Pfizer; and has received funding through his institution from Abbott, Medtronic, and Bayer. Drs. McEvoy, Kalman, and Sanders are supported by Practitioner Fellowships, NHMRC of Australia. Dr. McEvoy has received funding from Philips Respironics, ResMed, and Fisher and Paykel. Dr. Sanders is an advisory board member of Biosense-Webster, Medtronic, St. Jude Medical, Boston Scientific, CathRx; and has received funding through his institution from Biosense-Webster, Medtronic, St. Jude Medical, Boston Scientific and LivaNova/MicroPort. Dr. Lau is supported by a Robert J. Craig Lectureship, University of Adelaide; and has received lecture and consulting fees from Abbott, Biotronik, Boehringer Ingelheim, Bayer, and Pfizer. Dr. Brooks is an employee of LivaNova/MicroPort. Dr. Lim is supported by a Neil Hamilton Fairley Early Career Fellowship, NHMRC of Australia. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. William Stevenson, MD, served as Guest Editor for this paper.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- apnea-hypopnea index
- atrial fibrillation
- continuous positive airway pressure
- respiratory disturbance index
- sleep apnea monitoring
- sleep-disordered breathing
- Received December 17, 2018.
- Revision received February 25, 2019.
- Accepted March 13, 2019.
- Berry R.B.,
- Budhiraja R.,
- Gottlieb D.J.,
- et al.
- Stöberl A.S.,
- Schwarz E.I.,
- Haile S.R.,
- et al.
- Linz D.,
- Baumert M.,
- Catcheside P.,
- et al.
- Kohno T.,
- Kimura T.,
- Fukunaga K.,
- et al.
- Heinzer R.,
- Vat S.,
- Marques-Vidal P.,
- et al.
- Gami A.S.,
- Pressman G.,
- Caples S.M.,
- et al.
- Shukla A.,
- Aizer A.,
- Holmes D.,
- et al.
- Qureshi W.T.,
- Nasir U.B.,
- Alqalyoobi S.,
- et al.
- Linz D.,
- McEvoy R.D.,
- Cowie M.R.,
- et al.
- Iwasaki Y.K.,
- Kato T.,
- Xiong F.,
- et al.
- Anter E.,
- Di Biase L.,
- Contreras-Valdes F.M.,
- et al.
- Linz D.,
- Brooks A.G.,
- Elliott A.D.,
- Kalman J.,
- McEvoy R.D.,
- Lau D.H.,
- Sanders P.
- Epstein A.E.,
- DiMarco J.P.,
- Ellenbogen K.A.,
- et al.,
- for the American College of Cardiology Foundation; American Heart Association Task Force on Practice Guidelines; Heart Rhythm Society
- Géroux L.,
- Limousin M.,
- Cazeau S.
- Yu J.,
- Zhou Z.,
- McEvoy R.D.,
- et al.
- Linz D.,
- Hohl M.,
- Ukena C.,
- et al.
- McDonald K.,
- O'Hanlon R.,
- Savage H.O.,
- et al.
- Middeldorp M.E.,
- Pathak R.K.,
- Meredith M.,
- et al.