[
Original Research Sleep Disorders
]
Severity of OSA Is an Independent Predictor of Incident Atrial Fibrillation Hospitalization in a Large Sleep-Clinic Cohort Gemma Cadby, PhD; Nigel McArdle, MD; Tom Briffa, PhD; David R. Hillman, MBBS; Laila Simpson, PhD; Matthew Knuiman, PhD; and Joseph Hung, MBBS (Hons)
BACKGROUND: OSA is a common condition that has been associated with atrial fibrillation (AF), but there is a paucity of data from large longitudinal cohorts to establish whether OSA is a risk factor for AF independent of obesity and other established risk factors. METHODS: We studied patients attending a sleep clinic referred for in-laboratory polysomnography for possible OSA between 1989 and 2001. Whole-population hospital data in Western Australia for 1970 to 2009 were linked to sleep study cases to determine incident AF hospitalization to 2009. Cox regression analyses were used to assess the independent association of OSA with incident AF.
Study case subjects (6,841) were predominantly middle aged (48.3 ⫾ 12.5 years old) and men (77%), and 455 developed AF during a median 11.9 years of follow-up. Univariate predictors of AF included age, BMI, hypertension, diabetes, valvular heart disease, coronary or peripheral artery disease, heart failure, and COPD (all P , .001). After multivariable adjustment, independent predictors of incident AF were an apnea/hypopnea index (AHI) . 5/h (hazard ratio [HR], 1.55; 95% CI, 1.21-2.00), log (AHI 1 1) (HR, 1.15; 95% CI, 1.06-1.26), and log (time with oxygen saturation , 90% 1 1) (HR, 1.12; 95% CI, 1.06-1.19). There were no interactions between age, sex, or BMI and AHI for AF development. RESULTS:
CONCLUSIONS: OSA diagnosis and severity are independently associated with incident AF. Clinical trials are required to determine if treatment of OSA will reduce the burden of AF.
CHEST 2015; 148(4):945-952
Manuscript received February 5, 2015; revision accepted April 1, 2015; originally published Online First April 30, 2015. ABBREVIATIONS: AF 5 atrial fibrillation; AHI 5 apnea/hypopnea index; HR 5 hazard ratio; IQR 5 interquartile range; PSG 5 polysomnography; Sao2 5 oxygen saturation; Sao2t , 90% 5 time spent with oxygen saturation , 90% AFFILIATIONS: From the Centre for Genetic Origins of Health and Disease (Drs Cadby and Simpson), the School of Anatomy, Physiology and Human Biology (Drs McArdle, Hillman, and Simpson), the School of Population Health (Drs Briffa and Knuiman), and the School of Medicine and Pharmacology (Dr Hung), Sir Charles Gairdner Hospital Unit, University of Western Australia, Crawley; and the West Australian Sleep Disorders Research Institute (Drs McArdle, Hillman, and Simpson), Department of Pulmonary Physiology and Sleep Medicine, Sir Charles Gairdner Hospital, Nedlands, WA, Australia.
journal.publications.chestnet.org
Drs Cadby and McArdle contributed equally to this manuscript. The study was supported by the National Health and Medical Research Council project [Grant 1020373] and a Ray Florence Shaw Award [2013/14-001]. CORRESPONDENCE TO: Nigel McArdle, MD, West Australian Sleep Disorders Research Institute, Department of Pulmonary Physiology and Sleep Medicine, Sir Charles Gairdner Hospital, Nedlands, WA, Australia 6009; e-mail: [email protected] © 2015 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.15-0229 FUNDING/SUPPORT:
945
Atrial fibrillation (AF) is the dominant chronic arrhythmia in adults, affecting 1% to 4% of the general population.1 AF is associated with significant morbidity and increased mortality.1,2 Furthermore, the prevalence and cost of AF is escalating, underpinned by an aging population and an increased prevalence of associated risk factors.1,2 Many of the established risk factors for AF, such as older age, male sex, heart failure, myocardial infarction, hypertension, diabetes, and obesity,1,3 are also associated with OSA. Whether the reported association of OSA with prevalent or incident AF4-6 is independent or is a consequence of comorbid obesity or other risk factors warrants further investigation. OSA is a sleep-related breathing disorder characterized by repetitive upper-airway obstruction during sleep that occurs to a clinically significant degree in at least 2% of women and 4% of men.7 The physiologic consequences of OSA include intermittent arterial hypoxemia, central nervous system arousal, and large swings in intrathoracic
Materials and Methods Study Population We studied a consecutive cohort of adults aged ⱖ 17 years referred for suspected sleep disorder to the Western Australian Sleep Disorders Research Institute sleep clinic between 1989 and 2001.19 All patients were clinically assessed by a sleep physician and, where indicated, were referred for an overnight laboratory-based PSG study. Among those who completed a PSG study (n 5 9,244), 6,841 patients were available in the study cohort with linked Western Australian hospital morbidity data, after appropriate exclusions (Fig 1).
pressure during sleep,8,9 which, in turn, are associated with sympathetic activation,8 endothelial dysfunction, and chronic inflammation.10 These same mechanisms have also been implicated in the pathogenesis of AF, either by triggering its initiation or by atrial remodeling so as to promote or maintain the arrhythmia.11-13 OSA is also a predictor of hypertension,14,15 myocardial infarction,16 and heart failure,17 and, in concert with obesity, these conditions can lead to cardiac remodeling and arrhythmias.3,12 Therefore, there is strong biologic plausibility that OSA may predispose toward the development of AF.13 To our knowledge, only one previous clinic-based study18 has demonstrated that OSA predicts the incidence of AF. Accordingly, the current study uses a large (N 5 6,841) sleep clinic sample and person-linked hospital data to determine whether the presence or severity of OSA independently predicts incident AF over 20 years following diagnostic polysomnography (PSG) study.
Western Australian Hospital Morbidity and Mortality Data: The records of all Western Australians who die or are admitted to a public or private hospital are available for high-quality data linkage. 22 The patient file data in this cohort were linked with mortality and hospital morbidity data from 1970 to 2009 (“linked data”). Linked data were used to establish baseline comorbid conditions among sleep
Measurement and Follow-up During the initial consultation, age, sex, and smoking history were recorded. Current medications were determined and BMI (kg/m2) measured using a stadiometer for standing height (m) and routinely calibrated scales for weight (kg) was tabulated on the night of PSG study. Diagnostic PSG study signals, data acquisition systems, and scoring rules used for apneas and hypopneas have been described previously.19 Notably, the hypopnea rule required . 4% desaturation prior to 200020 and subsequently required a ⱖ 3% desaturation or an associated cortical arousal.21 The apnea/hypopnea index (AHI) was the main index of OSA severity, defined as the number of apneas plus hypopneas per hour of sleep. Those with an AHI of , 5 events/h were considered not to have OSA. Predominant central sleep apnea was identified when the AHI was . 5/h, with . 50% of events determined as central. Cortical arousals were scored when there was an increase in EEG frequency for ⱖ 3 s. Overnight oxygen desaturation was assessed by time spent with oxygen saturation , 90% ([Sao2t , 90%], min) and lowest oxygen saturation ([lowest Sao2], %). Following PSG, patients were reviewed by their physicians, and treatment of OSA was initiated as clinically indicated. Treatment commonly consisted of CPAP or, infrequently, a mandibular advancement splint. Our practice was to initiate outpatient treatment before performing an on-treatment PSG study after ⱖ 2 weeks’ familiarization. A treatment PSG study within 180 days of diagnostic PSG was used as a conservative surrogate measure of treatment. A standardized protocol existed for data entry, and checks were made for all data outliers.
946 Original Research
Figure 1 – Study flow diagram. AF 5 atrial fibrillation; AHI 5 apnea/ hypopnea index; PSG 5 polysomnography.
[
148#4 CHEST OCTOBER 2015
]
study cases and to identify incident AF during follow-up, as described previously.23 A history of various conditions at baseline was defined as having any hospital admission with a primary or secondary discharge diagnosis for that condition during the 15 years before the date of the sleep study. (See Table 1 for baseline variables and Table 2 for International Classification of Diseases codes.) A history of hypertensive disease or diabetes was based on hospital admission records or patient-reported treatment with antihypertensive or diabetic medications. Incident AF events during follow-up to mid-2009 following PSG study were defined as a hospital admission with a primary or secondary diagnosis of AF/flutter (International Classification of Diseases, Ninth Revision 427.31, 427.32, 427.3; International Classification of Diseases, Tenth Revision I48).23 However, AF hospitalization that occurred in the setting of a coronary artery bypass graft procedure was censored.23 The use of patient file and linked data for this protocol was approved by the Human Research Ethics Committees of the Sir Charles Gairdner Hospital (Application 2001-083), who granted a waiver of consent, and by the Department of Health of Western Australia (Project 2011/48). Data contained no personal identifiers prior to analysis. Statistical Analysis: Study population characteristics were described by counts and percentages, means and SDs, or medians and interquartile
Results Table 1 shows the baseline clinical and PSG characteristics of the 6,841 sleep clinic patients in the study cohort, stratified by incident AF. The cohort was predominantly middle aged (48.3 ⫾ 12.5 years old) and male sex (77%), with a wide range of AHI (median [lower quartile, upper quartile] 8.90 [2.50, 23.50] events/h) (Table 1). Patients had a history of cardiovascular risk factors and comorbidities commonly found among patients attending sleep clinics, but a relatively low baseline rate of cardiovascular disease, such as coronary artery disease (7.6%) and heart failure (1.9%) (Table 1). Over a median follow-up of 11.9 years (IQR, 4.7 years), there were 455 incident cases of AF. The individuals who developed AF were approximately 10 years older and more often men and had a slightly higher BMI than those without AF at baseline (Table 1). Univariate predictors of incident AF hospitalization are shown in Table 1. OSA diagnosis and most OSA severity measures were univariately predictive of incident AF (P , .001) (Table 1). The cumulative probability of incident AF rose with increasing OSA frequency categories (Fig 2). Multivariate predictors of AF were age, male sex, BMI, hypertension, diabetes, valvular heart disease, coronary or peripheral artery disease, and heart failure (data not shown). After adjustment for these other multivariate predictors, OSA diagnosis (AHI ⱖ 5), OSA severity (using AHI clinical cut points), log (AHI 1 1), and log
journal.publications.chestnet.org
ranges (IQRs). Variables with (positively) skewed distributions were natural log transformed for use in regression models. Cox regression models for time from PSG study to first AF event were used to obtain adjusted hazard ratios (HRs) and 95% CIs for a range of risk factors. Each baseline variable was first assessed for its association with incident AF using a Cox model that included only that risk factor (Table 1). AHI was assessed in three ways: as a categorical variable with two levels (nil OSA: AHI , 5, and OSA: AHI ⱖ 5 events/h), as a categorical variable with four levels (nil: , 5 events/h; mild: 5-14.99 events/h; moderate: 15-29.99 events/h; and severe: ⱖ 30 events/h), and with log (AHI 1 1) as a continuous variable. The unadjusted cumulative risk of AF over time for OSA groups is presented as a cumulative frequency curve graph (Fig 2). Other indexes of OSA severity (ie, cortical arousal index and oxygenation indexes) were analyzed using the same approach. The effect of each OSA variable on incident AF was assessed after adjustment for age, sex, height, and BMI (model 1, Table 3) and again after further adjustment for comorbid risk factors (model 2, Table 3). We also tested interactions between OSA variables and AF separately for age, sex, and BMI. To assess the effect, if any, of OSA treatment on the relationship between AHI and AF, we performed a sensitivity analysis by refitting the multivariate models after excluding individuals who underwent treatment PSG within 180 days of diagnosis. We also performed a sensitivity analysis excluding cases with predominant central sleep apnea defined as AHI ⱖ 5 events/h where . 50% were central events. P values , .05 were considered statistically significant. All statistical analyses were performed in R.24
(Sao2t , 90% 1 1) were all independently predictive of developing AF (Table 3). The fully adjusted HR for AHI ⱖ 5 vs AHI , 5 was 1.55 (95% CI, 1.21-2.00; P , .001) and showed evidence of increasing hazard with increasing AHI categories. The fully adjusted multivariate model with log (AHI 1 1) gave an HR of 1.15 (95% CI, 1.06-1.26; P 5 .001) for each increment of 1 in log (AHI 1 1), which corresponds to increments of 0 to 6.5, 6.6 to 19, and 19.1 to 54 in AHI. The effect of AHI on the risk of AF was not modified by age, sex, or BMI (interaction P . .05). Additional multivariate models were constructed without AHI but using other measures of OSA severity (Table 1), of which only the Sao2t , 90% (log [Sao2t , 90% 1 1]) was an independent predictor of AF with an adjusted HR of 1.12 (95% CI,1.06-1.19; P , .001). AHI as a categorical variable and log (Sao2t , 90% 1 1) were both independently predictive of AF when included in the same multivariable model. Only log (Sao2t , 90% 1 1) was independently predictive when included with AHI as a continuous covariate because both OSA severity measures were strongly correlated (r 5 0.68). The effect of OSA severity on AF was negligibly changed after removing patients treated for OSA from the analysis (n 5 832) or excluding patients with predominant central sleep apnea (n 5 58)
Discussion We found that a PSG-confirmed diagnosis of OSA (AHI ⱖ 5 events/h) is associated with increased rates of 947
948 Original Research
[
148#4 CHEST OCTOBER 2015
]
4.3 (273)
14.7 (67)
2.84 (1.47)
Log (AHI 1 1), median (IQR)
37.7 (2,408)
31.9 (145)
Severe (% ⱖ 30)
18.6 (1,187)
15.8 (1,010)
27.9 (1,781)
29.2 (133) 21.1 (96)
Mild (% 5-14.9)
37.7 (2,408)
62.3 (3,978)
17.8 (81)
Moderate (% 15-29.9)
Nil OSA (% , 5)
AHI clinical cut points
17.8 (81) 82.2 (374)
No OSA (% , 5)
OSA (% ⱖ 5)
2.24 (1.93)
8.4 (20.0)
1.6 (99)
0.16 (10)
7.3 (33) 16.1 (28.5)
AHI binary
1.1 (71) 8.4 (533)
1.6 (105)
AHI, median (IQR)
Heart failure
6.6 (423) 0.8 (53)
1.8 (8)
0.66 (3)
Thyroid disease
Chronic renal disease
5.5 (25) 15.4 (70)
Peripheral artery disease
COPD
1.3 (6)
21.3 (97)
Coronary artery disease
Stroke or TIA
3.3 (15)
0.9 (58)
22.9 (1,463)
18.5 (84)
Valvular heart disease
Current
31.1 (1,986)
27.9 (127) 39.3 (179)
Never
30.9 (1,972)
11.7 (744)
30.7 ⫾ 6.1
29.7 (135)
31.9 ⫾ 6.29
171.8 ⫾ 9.2
Former
Smokinga
Diabetes
Hypertension
History
BMI, kg/m2
172.0 ⫾ 9.2
47.6 ⫾ 12.2
59.2 ⫾ 10.6
Height, cm
Age, y
76.5 (4,888)
No AF (n 5 6,386)
AF (n 5 455) 82.2 (374)
Characteristic
(N 5 6,841)
19.5 (1,332)
16.2 (1,106)
28.0 (1,914)
36.4 (2,489)
63.6 (4,352)
36.4 (2,489)
2.29 (1.95)
8.9 (21.0)
1.9 (132)
1.7 (113)
0.19 (13)
8.8 (603)
1.4 (96)
0.9 (59)
7.6 (520)
1.1 (73)
22.6 (1,547)
31.7 (2,165)
30.68 (2,099)
5.0 (340)
12.9 (879)
30.7 ⫾ 6.1
171.8 ⫾ 9.2
48.3 ⫾ 12.5
76.9 (5,262)
All (N 5 6,841)
3.31 (2.53-4.35)
2.66 (1.98-3.57)
2.12 (1.12-2.80)
1.00
2.62 (2.06-3.34)
1.00
1.39 (1.29-1.50)
1.008 (1.006-1.01)
4.8 (3.37-6.85)
1.16 (0.57-2.33)
5.42 (1.74-16.86)
2.11 (1.63-2.72)
4.71 (3.15-7.05)
1.59 (0.71-3.57)
3.59 (2.87-4.50)
3.79 (2.26-6.34)
0.89 (0.68-1.18)
1.35 (1.08-1.70)
1.00
3.99 (3.07-5.17)
3.25 (2.66-3.98)
1.04 (1.02-1.05)
1.00 (0.99-1.01)
1.08 (1.07-109)
1.32 (1.04-1.68)
HR (95% CI)
(Continued)
, .0001
, .0001
, .0001
…
, .0001
…
, .0001
, .0001
, .0001
.68
.004
, .0001
, .0001
.26
, .0001
, .0001
.42
.009
…
, .0001
, .0001
, .0001
.91
, .0001
.022
P Value
] Characteristics of the AF-Free Cohort at Baseline and Univariate Associations (Cox Model Hazard Ratio) With Incident AF During Follow-up
Male sex
TABLE 1
Data are given as % (No.) or mean ⫾ SD unless otherwise indicated. Cortical arousal index was defined as the number of EEG arousals/h of sleep. AF 5 atrial fibrillation; AHI 5 apnea/hypopnea index (events/h); HR 5 hazard ratio; IQR 5 interquartile range; SaO2 5 oxygen saturation; SaO2t , 90% 5 time with oxygen saturation , 90%; TIA 5 transient ischemic attack. aMissing observations: 1,030. bMissing observations: 242. cMissing observations: 315. dMissing observations: 230. eMissing observations: 211.
, .0001 1.003(1.002-1.004) Log (SaO2t , 90% 1 1), median (IQR)
6.35 (39.50)
1.40 (10.30)
1.60 (11.30)
, .0001 0.98 (0.97-0.98) 86.00 (8.50) 83.00 (12.15) Log (lowest SaO2 1 1),d median (IQR)
e
86.00 (8.40)
, .0001 1.004 (1.002-1.005) 21.50 (16.8) Log (cortical arousal index 1 1),c median (IQR)
25.50 (19.3)
21.80 (16.90)
P Value
, .0001
HR (95% CI)
0.98 (0.97-0.98)
All (N 5 6,841)
79.00 (18.2)
78.80 (18.5)
No AF (n 5 6,386) AF (n 5 455)
73.25 (19.71)
Characteristic
Log (sleep efficiency index),b median (IQR)
] (continued) TABLE 1
journal.publications.chestnet.org
incident AF hospitalization over a median 12 years of follow-up, independent of obesity and other established AF risk factors. Additionally, there is a dose-response relationship between OSA severity and rates of incident AF. In particular, the AHI, whether expressed as a continuous variable or stratified by commonly used clinical cut points, and Sao2t , 90% have independent graded associations with incident AF risk. A series of retrospective studies have shown a relationship between OSA and AF, usually in the time-limited setting of a single overnight PSG or in case-control analysis.4-6,25,26 However, there has been a paucity of epidemiologic studies investigating the long-term association between OSA and incident AF. To our knowledge, there has only been one previous longitudinal study of a sleep clinic population with PSG-diagnosed OSA and incident AF.18 That study, by Gami and colleagues,18 reported that a decrease in nocturnal Sao2, but not AHI, was independently predictive of subsequent development of AF and only in those , 65 years old. Although the current study is in broad agreement with Gami and colleagues,18 we found that the independent OSA-related predictors of incident hospitalization with AF were both AHI and Sao2t , 90%, an oxygenation index not assessed in the study by Gami and colleagues.18 The current study examined a larger sample with a longer follow-up duration and more than three times as many incident cases of AF, and the greater power probably explains why we were able to confirm that AHI was a significant AF predictor. Additionally, the current study diagnosed OSA using full-night PSG in all cases, whereas Gami and colleagues18 used a mixture of full-night and split-night studies, which are likely to be less accurate. The mean age of their study cohort was very similar to ours, yet we found no evidence that the effect of OSA on the risk of AF was limited to subjects , 65 years old or that the association was affected by sex or degree of obesity. In our study, Sao2t , 90% was a stronger predictor than was continuous AHI, indicating that oxygen desaturation may be an important mediator in the interaction between OSA and AF. Notably, Sao2t , 90% is a more complete descriptor of the extent and degree of hypoxemia than the lowest Sao2, a solitary data point. Further, neither the current study nor that of Gami and colleagues18 found an association between EEG arousal rates and incident AF, which suggests that the severity of OSA-related intermittent hypoxia is more important than sleep fragmentation in the development of AF. 949
TABLE 2
] Classification of Outcomes and Comorbidities Using Hospital Admission Diagnosis Codes and Patient File Data
Hospital Admission Diagnostic Codes Disease
ICD-9, ICD-9-CM, 1979-1999
ICD-10, 1999-Present
Patient File Data
Heart failure
428.0-428.9
I50.0-I50.9
…
Hypertensive disease
401.0-405.99
I10.0-I15.9
Antihypertensive medication
Valvular heart disease (includes V codes)
394.0-397.9
I05.0-I09.0
…
V42.2, V43.3
T82, Z95
424
I34–I38
History of coronary heart disease (includes procedure codes)
410-414, 429.2, V42.2, V43.3
5-363, 5-361 (ICD-9) 36.01, 36.02, 36.05, 36.06, 36.07, 36.10-36.19
I20-I25, T82, Z95
…
ICD-10 procedure codes 38215-00 to 38218-02 35304-00 to 35305-02 35310-00 to 35310-05
Stroke, TIA, or systemic embolism
431, 432.0-432.9, 434.1, 435.0-435.9, 436
I61.0-I61.9, I62.0-I62.9
…
Peripheral artery disease
440.0-448.9
Thyroid disease
240.0-246.9
E00.0-07.9
…
Chronic renal disease
585.1, 585.9, 586
N18.1-18.9
…
G45.0-G45.9, I67.8 I63.4 I70.0-I78.9
…
COPD
490.0-496.9
J40-J47
…
Type 2 diabetes mellitus
250.0-250.9
E10.0-14.9
Diabetes medication
AF/flutter
427.31, 427.32, 427.3
I48
…
ICD-9 5 International Classification of Diseases, Ninth Revision; ICD-9-CM 5 International Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10 5 International Classification of Diseases, Tenth Revision. See Table 1 legend for expansion of other abbreviations.
As expected, we found that traditional (AF) risk factors were also predictive of AF in the cohort. The finding that OSA is an independent predictor of incident AF supports the likelihood that it is a true risk factor for AF. This is important information because there are effective and safe treatments for OSA, such as CPAP. If treatment of OSA is shown to reduce the incidence of subsequent AF, then the potential exists to significantly reduce critical AF-related comorbidities such as stroke and heart failure27 through systematic identification and treatment. In the current study, we found no evidence that OSA treatment affected the incidence of AF. However, the retrospective nature of this study meant that ascertainment of treatment effects among cases was indirect and relied on surrogate measures, with no long-term objective measure of treatment compliance. However, studies to date suggest that treatment with CPAP may reduce the recurrence rate of AF after cardioversion28 and catheter 950 Original Research
ablation,29 although randomized controlled data are needed. This study has a number of potential strengths and limitations. The use of high-quality person-linked administrative data allowed for the assessment of comorbidities that may confound the association of OSA with incident AF. Although the study is retrospective, all PSG data were collected prospectively and with careful standardized protocols in place. Furthermore, hospital medical record linkage was obtained in a high proportion of sleep study cases (88.4%) and the missing cases are likely to result from random factors rather than any systematic biases. PSG equipment changed during the study period, as techniques improved over time, and it is likely that the accuracy of OSA diagnosis would have improved as a result. However, the potentially lower accuracy of the earlier technology is likely to bias toward underestimating the strength of the associations found. The use of linked hospital
[
148#4 CHEST OCTOBER 2015
]
admission data to detect incident AF cases may have overestimated the true time to AF onset for identified incident cases and missed AF cases who did not have any hospital admission where AF was recognized over the follow-up period. However, the possibility of missed asymptomatic AF cases in the cohort would probably be nondifferential and would be likely to bias toward the null hypothesis; therefore, our HRs may be conservative.30
Conclusions
Figure 2 – Incidence of AF. Cumulative frequency curve for incident AF for all subjects (N 5 6,841) during a median follow-up of approximately 12 y, stratified by OSA severity. See Figure 1 legend for expansion of abbreviation.
TABLE 3
This large clinic-based cohort of patients assessed for OSA using laboratory-based PSG found an independent association between the presence and severity of OSA and incident AF over a median 12-year follow-up period. Our data support a growing evidence base that OSA is an important novel and potentially modifiable risk factor for the development of AF. Randomized interventional studies are required to ascertain whether treatment of OSA can reduce the incident or recurrent burden of AF.
] Hazard Ratios of OSA Measures for Incident AF: Multivariate Models
Measure Nil OSA (AHI: , 5) OSA (AHI: ⱖ 5)
HR (95% CI)a
P Value
HR (95% CI)b
P Value
1.00
…
1.00
…
1.56 (1.21-2.00)
, .001
1.55 (1.21-2.00)
, .001
1.00
…
1.00
…
AHI, clinical cut points Nil OSA (AHI: , 5) Mild (AHI: 5-14.9)
1.48 (1.12-1.96)
.006
1.48 (1.12-1.96)
.006
Moderate (AHI: 15-29.9)
1.49 (1.10-2.02)
.100
1.51 (1.11-2.05)
.008
Severe (AHI: ⱖ 30)
1.78 (1.32-2.39)
, .001
1.73 (1.29-2.33)
, .001
Log (AHI 1 1)
1.16 (1.06-1.27)
, .001
1.15 (1.06-1.26)
.001
Log (SaO2t , 90% 1 1)
1.14 (1.08-1.21)
, .001
1.12 (1.06-1.19)
, .001
See Table 1 legend for expansion of abbreviations. Adjusted for age, sex, height, and BMI. bAdjusted for age, sex, height, BMI, hypertension, valvular disease, stroke/TIA, coronary or peripheral artery disease, COPD, chronic renal disease, heart failure, and diabetes. a
journal.publications.chestnet.org
951
Acknowledgments Author contributions: G. C. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis, including and especially any adverse effects. G. C., N. M., T. B., D. R. H., L. S., M. K., and J. H. contributed substantially to the study design, data analysis, and interpretation and the writing of the manuscript.
8. Somers VK, Dyken ME, Clary MP, Abboud FM. Sympathetic neural mechanisms in obstructive sleep apnea. J Clin Invest. 1995;96(4):1897-1904.
10. Shamsuzzaman AS, Winnicki M, Lanfranchi P, et al. Elevated C-reactive protein in patients with obstructive sleep apnea. Circulation. 2002;105(21):2462-2464.
Role of sponsors: The sponsors had no direct role in the study design; the collection, analysis, or interpretation of the data; or the writing of the manuscript.
11. Dimitri H, Ng M, Brooks AG, et al. Atrial remodeling in obstructive sleep apnea: implications for atrial fibrillation. Heart Rhythm. 2012;9(3):321-327.
References 1. Rahman F, Kwan GF, Benjamin EJ. Global epidemiology of atrial fibrillation. Nat Rev Cardiol. 2014;11(11):639-654. 2. Greenlee RT, Vidaillet H. Recent progress in the epidemiology of atrial fibrillation. Curr Opin Cardiol. 2005;20(1): 7-14. 3. Wang TJ, Parise H, Levy D, et al. Obesity and the risk of new-onset atrial fibrillation. JAMA. 2004;292(20):2471-2477. 4. Gami AS, Pressman G, Caples SM, et al. Association of atrial fibrillation and obstructive sleep apnea. Circulation. 2004;110(4):364-367. 5. Mehra R, Benjamin EJ, Shahar E, et al; Sleep Heart Health Study. Association of nocturnal arrhythmias with sleepdisordered breathing: The Sleep Heart Health Study. Am J Respir Crit Care Med. 2006;173(8):910-916. 6. Stevenson IH, Teichtahl H, Cunnington D, Ciavarella S, Gordon I, Kalman JM. Prevalence of sleep disordered breathing in paroxysmal and persistent atrial fibrillation patients with normal left ventricular function. Eur Heart J. 2008; 29(13):1662-1669.
952 Original Research
22. Holman CD, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage of health records in Western Australia: development of a health services research linked database. Aust N Z J Public Health. 1999;23(5):453-459.
9. Dempsey JA, Veasey SC, Morgan BJ, O’Donnell CP. Pathophysiology of sleep apnea. Physiol Rev. 2010;90(1):47-112.
Conflict of interest: N. M. and D. R. H. have received research support from ResMed. N. M. has received an honorarium for participating at a ResMed breakfast symposium. None declared (G. C., T. B., L. S., M. K., J. H.).
Other contributions: The authors thank the staff at the Western Australian Data Linkage Branch, the Western Australia Department of Health Inpatient Data Collection and Epidemiology Branch, and the Registrar General of the Western Australia Department of the Attorney General for the provision of data. We acknowledge the contributors to the establishment of the database from which these data were extracted, including Sutapa Mukherjee, PhD; Bhajan Singh, PhD; Annette C. Fedson, PhD; and Lyle Palmer, PhD.
21. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. The report of an American Academy of Sleep Medicine Task Force. Sleep. 1999;22(5):667-689.
7. Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S. The occurrence of sleep-disordered breathing among middleaged adults. N Engl J Med. 1993;328(17): 1230-1235.
23. Knuiman M, Briffa T, Divitini M, et al. A cohort study examination of established and emerging risk factors for atrial fibrillation: the Busselton Health Study. Eur J Epidemiol. 2014;29(3):181-190. 24. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2012.
12. Shah RV, Abbasi SA, Heydari B, et al. Obesity and sleep apnea are independently associated with adverse left ventricular remodeling and clinical outcome in patients with atrial fibrillation and preserved ventricular function. Am Heart J. 2014;167(4):620-626.
25. Monahan K, Storfer-Isser A, Mehra R, et al. Triggering of nocturnal arrhythmias by sleep-disordered breathing events. J Am Coll Cardiol. 2009;54(19):1797-1804. 26. Tanigawa T, Yamagishi K, Sakurai S, et al. Arterial oxygen desaturation during sleep and atrial fibrillation. Heart. 2006;92(12): 1854-1855.
13. Todd K, McIntyre WF, Baranchuk A. Obstructive sleep apnea and atrial fibrillation. Nat Sci Sleep. 2010;2:39-45.
27. Fuster V, Rydén LE, Cannom DS, et al; American College of Cardiology/ American Heart Association Task Force on Practice Guidelines; European Society of Cardiology Committee for Practice Guidelines; European Heart Rhythm Association; Heart Rhythm Society. ACC/AHA/ESC 2006 guidelines for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology Committee for Practice Guidelines (Writing Committee to Revise the 2001 Guidelines for the Management of Patients With Atrial Fibrillation): developed in collaboration with the European Heart Rhythm Association and the Heart Rhythm Society. Circulation. 2006;114(7):e257-e354.
14. Peppard PE, Young T, Palta M, Skatrud J. Prospective study of the association between sleep-disordered breathing and hypertension. N Engl J Med. 2000; 342(19):1378-1384. 15. Young T, Peppard P, Palta M, et al. Population-based study of sleepdisordered breathing as a risk factor for hypertension. Arch Intern Med. 1997;157(15):1746-1752. 16. Hung J, Whitford EG, Parsons RW, Hillman DR. Association of sleep apnoea with myocardial infarction in men. Lancet. 1990;336(8710):261-264. 17. 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.
28. Kanagala R, Murali NS, Friedman PA, et al. Obstructive sleep apnea and the recurrence of atrial fibrillation. Circulation. 2003;107(20):2589-2594.
18. Gami AS, Hodge DO, Herges RM, et al. Obstructive sleep apnea, obesity, and the risk of incident atrial fibrillation. J Am Coll Cardiol. 2007;49(5):565-571. 19. Mukherjee S, Hillman D, Lee J, et al. Cohort profile: the Western Australian Sleep Health Study. Sleep Breath. 2012;16(1):205-215.
29. Fein AS, Shvilkin A, Shah D, et al. Treatment of obstructive sleep apnea reduces the risk of atrial fibrillation recurrence after catheter ablation. J Am Coll Cardiol. 2013;62(4):300-305.
20. American Thoracic Society. Medical Section of the American Lung Association. Indications and standards for cardiopulmonary sleep studies. Am Rev Respir Dis. 1989;139(2):559-568.
30. Selmer C, Olesen JB, Hansen ML, et al. The spectrum of thyroid disease and risk of new onset atrial fibrillation: a large population cohort study. BMJ. 2012;345:e7895.
[
148#4 CHEST OCTOBER 2015
]
Original Research Sleep Disorders
]
Severity of OSA Is an Independent Predictor of Incident Atrial Fibrillation Hospitalization in a Large Sleep-Clinic Cohort Gemma Cadby, PhD; Nigel McArdle, MD; Tom Briffa, PhD; David R. Hillman, MBBS; Laila Simpson, PhD; Matthew Knuiman, PhD; and Joseph Hung, MBBS (Hons)
BACKGROUND: OSA is a common condition that has been associated with atrial fibrillation (AF), but there is a paucity of data from large longitudinal cohorts to establish whether OSA is a risk factor for AF independent of obesity and other established risk factors. METHODS: We studied patients attending a sleep clinic referred for in-laboratory polysomnography for possible OSA between 1989 and 2001. Whole-population hospital data in Western Australia for 1970 to 2009 were linked to sleep study cases to determine incident AF hospitalization to 2009. Cox regression analyses were used to assess the independent association of OSA with incident AF.
Study case subjects (6,841) were predominantly middle aged (48.3 ⫾ 12.5 years old) and men (77%), and 455 developed AF during a median 11.9 years of follow-up. Univariate predictors of AF included age, BMI, hypertension, diabetes, valvular heart disease, coronary or peripheral artery disease, heart failure, and COPD (all P , .001). After multivariable adjustment, independent predictors of incident AF were an apnea/hypopnea index (AHI) . 5/h (hazard ratio [HR], 1.55; 95% CI, 1.21-2.00), log (AHI 1 1) (HR, 1.15; 95% CI, 1.06-1.26), and log (time with oxygen saturation , 90% 1 1) (HR, 1.12; 95% CI, 1.06-1.19). There were no interactions between age, sex, or BMI and AHI for AF development. RESULTS:
CONCLUSIONS: OSA diagnosis and severity are independently associated with incident AF. Clinical trials are required to determine if treatment of OSA will reduce the burden of AF.
CHEST 2015; 148(4):945-952
Manuscript received February 5, 2015; revision accepted April 1, 2015; originally published Online First April 30, 2015. ABBREVIATIONS: AF 5 atrial fibrillation; AHI 5 apnea/hypopnea index; HR 5 hazard ratio; IQR 5 interquartile range; PSG 5 polysomnography; Sao2 5 oxygen saturation; Sao2t , 90% 5 time spent with oxygen saturation , 90% AFFILIATIONS: From the Centre for Genetic Origins of Health and Disease (Drs Cadby and Simpson), the School of Anatomy, Physiology and Human Biology (Drs McArdle, Hillman, and Simpson), the School of Population Health (Drs Briffa and Knuiman), and the School of Medicine and Pharmacology (Dr Hung), Sir Charles Gairdner Hospital Unit, University of Western Australia, Crawley; and the West Australian Sleep Disorders Research Institute (Drs McArdle, Hillman, and Simpson), Department of Pulmonary Physiology and Sleep Medicine, Sir Charles Gairdner Hospital, Nedlands, WA, Australia.
journal.publications.chestnet.org
Drs Cadby and McArdle contributed equally to this manuscript. The study was supported by the National Health and Medical Research Council project [Grant 1020373] and a Ray Florence Shaw Award [2013/14-001]. CORRESPONDENCE TO: Nigel McArdle, MD, West Australian Sleep Disorders Research Institute, Department of Pulmonary Physiology and Sleep Medicine, Sir Charles Gairdner Hospital, Nedlands, WA, Australia 6009; e-mail: [email protected] © 2015 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.15-0229 FUNDING/SUPPORT:
945
Atrial fibrillation (AF) is the dominant chronic arrhythmia in adults, affecting 1% to 4% of the general population.1 AF is associated with significant morbidity and increased mortality.1,2 Furthermore, the prevalence and cost of AF is escalating, underpinned by an aging population and an increased prevalence of associated risk factors.1,2 Many of the established risk factors for AF, such as older age, male sex, heart failure, myocardial infarction, hypertension, diabetes, and obesity,1,3 are also associated with OSA. Whether the reported association of OSA with prevalent or incident AF4-6 is independent or is a consequence of comorbid obesity or other risk factors warrants further investigation. OSA is a sleep-related breathing disorder characterized by repetitive upper-airway obstruction during sleep that occurs to a clinically significant degree in at least 2% of women and 4% of men.7 The physiologic consequences of OSA include intermittent arterial hypoxemia, central nervous system arousal, and large swings in intrathoracic
Materials and Methods Study Population We studied a consecutive cohort of adults aged ⱖ 17 years referred for suspected sleep disorder to the Western Australian Sleep Disorders Research Institute sleep clinic between 1989 and 2001.19 All patients were clinically assessed by a sleep physician and, where indicated, were referred for an overnight laboratory-based PSG study. Among those who completed a PSG study (n 5 9,244), 6,841 patients were available in the study cohort with linked Western Australian hospital morbidity data, after appropriate exclusions (Fig 1).
pressure during sleep,8,9 which, in turn, are associated with sympathetic activation,8 endothelial dysfunction, and chronic inflammation.10 These same mechanisms have also been implicated in the pathogenesis of AF, either by triggering its initiation or by atrial remodeling so as to promote or maintain the arrhythmia.11-13 OSA is also a predictor of hypertension,14,15 myocardial infarction,16 and heart failure,17 and, in concert with obesity, these conditions can lead to cardiac remodeling and arrhythmias.3,12 Therefore, there is strong biologic plausibility that OSA may predispose toward the development of AF.13 To our knowledge, only one previous clinic-based study18 has demonstrated that OSA predicts the incidence of AF. Accordingly, the current study uses a large (N 5 6,841) sleep clinic sample and person-linked hospital data to determine whether the presence or severity of OSA independently predicts incident AF over 20 years following diagnostic polysomnography (PSG) study.
Western Australian Hospital Morbidity and Mortality Data: The records of all Western Australians who die or are admitted to a public or private hospital are available for high-quality data linkage. 22 The patient file data in this cohort were linked with mortality and hospital morbidity data from 1970 to 2009 (“linked data”). Linked data were used to establish baseline comorbid conditions among sleep
Measurement and Follow-up During the initial consultation, age, sex, and smoking history were recorded. Current medications were determined and BMI (kg/m2) measured using a stadiometer for standing height (m) and routinely calibrated scales for weight (kg) was tabulated on the night of PSG study. Diagnostic PSG study signals, data acquisition systems, and scoring rules used for apneas and hypopneas have been described previously.19 Notably, the hypopnea rule required . 4% desaturation prior to 200020 and subsequently required a ⱖ 3% desaturation or an associated cortical arousal.21 The apnea/hypopnea index (AHI) was the main index of OSA severity, defined as the number of apneas plus hypopneas per hour of sleep. Those with an AHI of , 5 events/h were considered not to have OSA. Predominant central sleep apnea was identified when the AHI was . 5/h, with . 50% of events determined as central. Cortical arousals were scored when there was an increase in EEG frequency for ⱖ 3 s. Overnight oxygen desaturation was assessed by time spent with oxygen saturation , 90% ([Sao2t , 90%], min) and lowest oxygen saturation ([lowest Sao2], %). Following PSG, patients were reviewed by their physicians, and treatment of OSA was initiated as clinically indicated. Treatment commonly consisted of CPAP or, infrequently, a mandibular advancement splint. Our practice was to initiate outpatient treatment before performing an on-treatment PSG study after ⱖ 2 weeks’ familiarization. A treatment PSG study within 180 days of diagnostic PSG was used as a conservative surrogate measure of treatment. A standardized protocol existed for data entry, and checks were made for all data outliers.
946 Original Research
Figure 1 – Study flow diagram. AF 5 atrial fibrillation; AHI 5 apnea/ hypopnea index; PSG 5 polysomnography.
[
148#4 CHEST OCTOBER 2015
]
study cases and to identify incident AF during follow-up, as described previously.23 A history of various conditions at baseline was defined as having any hospital admission with a primary or secondary discharge diagnosis for that condition during the 15 years before the date of the sleep study. (See Table 1 for baseline variables and Table 2 for International Classification of Diseases codes.) A history of hypertensive disease or diabetes was based on hospital admission records or patient-reported treatment with antihypertensive or diabetic medications. Incident AF events during follow-up to mid-2009 following PSG study were defined as a hospital admission with a primary or secondary diagnosis of AF/flutter (International Classification of Diseases, Ninth Revision 427.31, 427.32, 427.3; International Classification of Diseases, Tenth Revision I48).23 However, AF hospitalization that occurred in the setting of a coronary artery bypass graft procedure was censored.23 The use of patient file and linked data for this protocol was approved by the Human Research Ethics Committees of the Sir Charles Gairdner Hospital (Application 2001-083), who granted a waiver of consent, and by the Department of Health of Western Australia (Project 2011/48). Data contained no personal identifiers prior to analysis. Statistical Analysis: Study population characteristics were described by counts and percentages, means and SDs, or medians and interquartile
Results Table 1 shows the baseline clinical and PSG characteristics of the 6,841 sleep clinic patients in the study cohort, stratified by incident AF. The cohort was predominantly middle aged (48.3 ⫾ 12.5 years old) and male sex (77%), with a wide range of AHI (median [lower quartile, upper quartile] 8.90 [2.50, 23.50] events/h) (Table 1). Patients had a history of cardiovascular risk factors and comorbidities commonly found among patients attending sleep clinics, but a relatively low baseline rate of cardiovascular disease, such as coronary artery disease (7.6%) and heart failure (1.9%) (Table 1). Over a median follow-up of 11.9 years (IQR, 4.7 years), there were 455 incident cases of AF. The individuals who developed AF were approximately 10 years older and more often men and had a slightly higher BMI than those without AF at baseline (Table 1). Univariate predictors of incident AF hospitalization are shown in Table 1. OSA diagnosis and most OSA severity measures were univariately predictive of incident AF (P , .001) (Table 1). The cumulative probability of incident AF rose with increasing OSA frequency categories (Fig 2). Multivariate predictors of AF were age, male sex, BMI, hypertension, diabetes, valvular heart disease, coronary or peripheral artery disease, and heart failure (data not shown). After adjustment for these other multivariate predictors, OSA diagnosis (AHI ⱖ 5), OSA severity (using AHI clinical cut points), log (AHI 1 1), and log
journal.publications.chestnet.org
ranges (IQRs). Variables with (positively) skewed distributions were natural log transformed for use in regression models. Cox regression models for time from PSG study to first AF event were used to obtain adjusted hazard ratios (HRs) and 95% CIs for a range of risk factors. Each baseline variable was first assessed for its association with incident AF using a Cox model that included only that risk factor (Table 1). AHI was assessed in three ways: as a categorical variable with two levels (nil OSA: AHI , 5, and OSA: AHI ⱖ 5 events/h), as a categorical variable with four levels (nil: , 5 events/h; mild: 5-14.99 events/h; moderate: 15-29.99 events/h; and severe: ⱖ 30 events/h), and with log (AHI 1 1) as a continuous variable. The unadjusted cumulative risk of AF over time for OSA groups is presented as a cumulative frequency curve graph (Fig 2). Other indexes of OSA severity (ie, cortical arousal index and oxygenation indexes) were analyzed using the same approach. The effect of each OSA variable on incident AF was assessed after adjustment for age, sex, height, and BMI (model 1, Table 3) and again after further adjustment for comorbid risk factors (model 2, Table 3). We also tested interactions between OSA variables and AF separately for age, sex, and BMI. To assess the effect, if any, of OSA treatment on the relationship between AHI and AF, we performed a sensitivity analysis by refitting the multivariate models after excluding individuals who underwent treatment PSG within 180 days of diagnosis. We also performed a sensitivity analysis excluding cases with predominant central sleep apnea defined as AHI ⱖ 5 events/h where . 50% were central events. P values , .05 were considered statistically significant. All statistical analyses were performed in R.24
(Sao2t , 90% 1 1) were all independently predictive of developing AF (Table 3). The fully adjusted HR for AHI ⱖ 5 vs AHI , 5 was 1.55 (95% CI, 1.21-2.00; P , .001) and showed evidence of increasing hazard with increasing AHI categories. The fully adjusted multivariate model with log (AHI 1 1) gave an HR of 1.15 (95% CI, 1.06-1.26; P 5 .001) for each increment of 1 in log (AHI 1 1), which corresponds to increments of 0 to 6.5, 6.6 to 19, and 19.1 to 54 in AHI. The effect of AHI on the risk of AF was not modified by age, sex, or BMI (interaction P . .05). Additional multivariate models were constructed without AHI but using other measures of OSA severity (Table 1), of which only the Sao2t , 90% (log [Sao2t , 90% 1 1]) was an independent predictor of AF with an adjusted HR of 1.12 (95% CI,1.06-1.19; P , .001). AHI as a categorical variable and log (Sao2t , 90% 1 1) were both independently predictive of AF when included in the same multivariable model. Only log (Sao2t , 90% 1 1) was independently predictive when included with AHI as a continuous covariate because both OSA severity measures were strongly correlated (r 5 0.68). The effect of OSA severity on AF was negligibly changed after removing patients treated for OSA from the analysis (n 5 832) or excluding patients with predominant central sleep apnea (n 5 58)
Discussion We found that a PSG-confirmed diagnosis of OSA (AHI ⱖ 5 events/h) is associated with increased rates of 947
948 Original Research
[
148#4 CHEST OCTOBER 2015
]
4.3 (273)
14.7 (67)
2.84 (1.47)
Log (AHI 1 1), median (IQR)
37.7 (2,408)
31.9 (145)
Severe (% ⱖ 30)
18.6 (1,187)
15.8 (1,010)
27.9 (1,781)
29.2 (133) 21.1 (96)
Mild (% 5-14.9)
37.7 (2,408)
62.3 (3,978)
17.8 (81)
Moderate (% 15-29.9)
Nil OSA (% , 5)
AHI clinical cut points
17.8 (81) 82.2 (374)
No OSA (% , 5)
OSA (% ⱖ 5)
2.24 (1.93)
8.4 (20.0)
1.6 (99)
0.16 (10)
7.3 (33) 16.1 (28.5)
AHI binary
1.1 (71) 8.4 (533)
1.6 (105)
AHI, median (IQR)
Heart failure
6.6 (423) 0.8 (53)
1.8 (8)
0.66 (3)
Thyroid disease
Chronic renal disease
5.5 (25) 15.4 (70)
Peripheral artery disease
COPD
1.3 (6)
21.3 (97)
Coronary artery disease
Stroke or TIA
3.3 (15)
0.9 (58)
22.9 (1,463)
18.5 (84)
Valvular heart disease
Current
31.1 (1,986)
27.9 (127) 39.3 (179)
Never
30.9 (1,972)
11.7 (744)
30.7 ⫾ 6.1
29.7 (135)
31.9 ⫾ 6.29
171.8 ⫾ 9.2
Former
Smokinga
Diabetes
Hypertension
History
BMI, kg/m2
172.0 ⫾ 9.2
47.6 ⫾ 12.2
59.2 ⫾ 10.6
Height, cm
Age, y
76.5 (4,888)
No AF (n 5 6,386)
AF (n 5 455) 82.2 (374)
Characteristic
(N 5 6,841)
19.5 (1,332)
16.2 (1,106)
28.0 (1,914)
36.4 (2,489)
63.6 (4,352)
36.4 (2,489)
2.29 (1.95)
8.9 (21.0)
1.9 (132)
1.7 (113)
0.19 (13)
8.8 (603)
1.4 (96)
0.9 (59)
7.6 (520)
1.1 (73)
22.6 (1,547)
31.7 (2,165)
30.68 (2,099)
5.0 (340)
12.9 (879)
30.7 ⫾ 6.1
171.8 ⫾ 9.2
48.3 ⫾ 12.5
76.9 (5,262)
All (N 5 6,841)
3.31 (2.53-4.35)
2.66 (1.98-3.57)
2.12 (1.12-2.80)
1.00
2.62 (2.06-3.34)
1.00
1.39 (1.29-1.50)
1.008 (1.006-1.01)
4.8 (3.37-6.85)
1.16 (0.57-2.33)
5.42 (1.74-16.86)
2.11 (1.63-2.72)
4.71 (3.15-7.05)
1.59 (0.71-3.57)
3.59 (2.87-4.50)
3.79 (2.26-6.34)
0.89 (0.68-1.18)
1.35 (1.08-1.70)
1.00
3.99 (3.07-5.17)
3.25 (2.66-3.98)
1.04 (1.02-1.05)
1.00 (0.99-1.01)
1.08 (1.07-109)
1.32 (1.04-1.68)
HR (95% CI)
(Continued)
, .0001
, .0001
, .0001
…
, .0001
…
, .0001
, .0001
, .0001
.68
.004
, .0001
, .0001
.26
, .0001
, .0001
.42
.009
…
, .0001
, .0001
, .0001
.91
, .0001
.022
P Value
] Characteristics of the AF-Free Cohort at Baseline and Univariate Associations (Cox Model Hazard Ratio) With Incident AF During Follow-up
Male sex
TABLE 1
Data are given as % (No.) or mean ⫾ SD unless otherwise indicated. Cortical arousal index was defined as the number of EEG arousals/h of sleep. AF 5 atrial fibrillation; AHI 5 apnea/hypopnea index (events/h); HR 5 hazard ratio; IQR 5 interquartile range; SaO2 5 oxygen saturation; SaO2t , 90% 5 time with oxygen saturation , 90%; TIA 5 transient ischemic attack. aMissing observations: 1,030. bMissing observations: 242. cMissing observations: 315. dMissing observations: 230. eMissing observations: 211.
, .0001 1.003(1.002-1.004) Log (SaO2t , 90% 1 1), median (IQR)
6.35 (39.50)
1.40 (10.30)
1.60 (11.30)
, .0001 0.98 (0.97-0.98) 86.00 (8.50) 83.00 (12.15) Log (lowest SaO2 1 1),d median (IQR)
e
86.00 (8.40)
, .0001 1.004 (1.002-1.005) 21.50 (16.8) Log (cortical arousal index 1 1),c median (IQR)
25.50 (19.3)
21.80 (16.90)
P Value
, .0001
HR (95% CI)
0.98 (0.97-0.98)
All (N 5 6,841)
79.00 (18.2)
78.80 (18.5)
No AF (n 5 6,386) AF (n 5 455)
73.25 (19.71)
Characteristic
Log (sleep efficiency index),b median (IQR)
] (continued) TABLE 1
journal.publications.chestnet.org
incident AF hospitalization over a median 12 years of follow-up, independent of obesity and other established AF risk factors. Additionally, there is a dose-response relationship between OSA severity and rates of incident AF. In particular, the AHI, whether expressed as a continuous variable or stratified by commonly used clinical cut points, and Sao2t , 90% have independent graded associations with incident AF risk. A series of retrospective studies have shown a relationship between OSA and AF, usually in the time-limited setting of a single overnight PSG or in case-control analysis.4-6,25,26 However, there has been a paucity of epidemiologic studies investigating the long-term association between OSA and incident AF. To our knowledge, there has only been one previous longitudinal study of a sleep clinic population with PSG-diagnosed OSA and incident AF.18 That study, by Gami and colleagues,18 reported that a decrease in nocturnal Sao2, but not AHI, was independently predictive of subsequent development of AF and only in those , 65 years old. Although the current study is in broad agreement with Gami and colleagues,18 we found that the independent OSA-related predictors of incident hospitalization with AF were both AHI and Sao2t , 90%, an oxygenation index not assessed in the study by Gami and colleagues.18 The current study examined a larger sample with a longer follow-up duration and more than three times as many incident cases of AF, and the greater power probably explains why we were able to confirm that AHI was a significant AF predictor. Additionally, the current study diagnosed OSA using full-night PSG in all cases, whereas Gami and colleagues18 used a mixture of full-night and split-night studies, which are likely to be less accurate. The mean age of their study cohort was very similar to ours, yet we found no evidence that the effect of OSA on the risk of AF was limited to subjects , 65 years old or that the association was affected by sex or degree of obesity. In our study, Sao2t , 90% was a stronger predictor than was continuous AHI, indicating that oxygen desaturation may be an important mediator in the interaction between OSA and AF. Notably, Sao2t , 90% is a more complete descriptor of the extent and degree of hypoxemia than the lowest Sao2, a solitary data point. Further, neither the current study nor that of Gami and colleagues18 found an association between EEG arousal rates and incident AF, which suggests that the severity of OSA-related intermittent hypoxia is more important than sleep fragmentation in the development of AF. 949
TABLE 2
] Classification of Outcomes and Comorbidities Using Hospital Admission Diagnosis Codes and Patient File Data
Hospital Admission Diagnostic Codes Disease
ICD-9, ICD-9-CM, 1979-1999
ICD-10, 1999-Present
Patient File Data
Heart failure
428.0-428.9
I50.0-I50.9
…
Hypertensive disease
401.0-405.99
I10.0-I15.9
Antihypertensive medication
Valvular heart disease (includes V codes)
394.0-397.9
I05.0-I09.0
…
V42.2, V43.3
T82, Z95
424
I34–I38
History of coronary heart disease (includes procedure codes)
410-414, 429.2, V42.2, V43.3
5-363, 5-361 (ICD-9) 36.01, 36.02, 36.05, 36.06, 36.07, 36.10-36.19
I20-I25, T82, Z95
…
ICD-10 procedure codes 38215-00 to 38218-02 35304-00 to 35305-02 35310-00 to 35310-05
Stroke, TIA, or systemic embolism
431, 432.0-432.9, 434.1, 435.0-435.9, 436
I61.0-I61.9, I62.0-I62.9
…
Peripheral artery disease
440.0-448.9
Thyroid disease
240.0-246.9
E00.0-07.9
…
Chronic renal disease
585.1, 585.9, 586
N18.1-18.9
…
G45.0-G45.9, I67.8 I63.4 I70.0-I78.9
…
COPD
490.0-496.9
J40-J47
…
Type 2 diabetes mellitus
250.0-250.9
E10.0-14.9
Diabetes medication
AF/flutter
427.31, 427.32, 427.3
I48
…
ICD-9 5 International Classification of Diseases, Ninth Revision; ICD-9-CM 5 International Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10 5 International Classification of Diseases, Tenth Revision. See Table 1 legend for expansion of other abbreviations.
As expected, we found that traditional (AF) risk factors were also predictive of AF in the cohort. The finding that OSA is an independent predictor of incident AF supports the likelihood that it is a true risk factor for AF. This is important information because there are effective and safe treatments for OSA, such as CPAP. If treatment of OSA is shown to reduce the incidence of subsequent AF, then the potential exists to significantly reduce critical AF-related comorbidities such as stroke and heart failure27 through systematic identification and treatment. In the current study, we found no evidence that OSA treatment affected the incidence of AF. However, the retrospective nature of this study meant that ascertainment of treatment effects among cases was indirect and relied on surrogate measures, with no long-term objective measure of treatment compliance. However, studies to date suggest that treatment with CPAP may reduce the recurrence rate of AF after cardioversion28 and catheter 950 Original Research
ablation,29 although randomized controlled data are needed. This study has a number of potential strengths and limitations. The use of high-quality person-linked administrative data allowed for the assessment of comorbidities that may confound the association of OSA with incident AF. Although the study is retrospective, all PSG data were collected prospectively and with careful standardized protocols in place. Furthermore, hospital medical record linkage was obtained in a high proportion of sleep study cases (88.4%) and the missing cases are likely to result from random factors rather than any systematic biases. PSG equipment changed during the study period, as techniques improved over time, and it is likely that the accuracy of OSA diagnosis would have improved as a result. However, the potentially lower accuracy of the earlier technology is likely to bias toward underestimating the strength of the associations found. The use of linked hospital
[
148#4 CHEST OCTOBER 2015
]
admission data to detect incident AF cases may have overestimated the true time to AF onset for identified incident cases and missed AF cases who did not have any hospital admission where AF was recognized over the follow-up period. However, the possibility of missed asymptomatic AF cases in the cohort would probably be nondifferential and would be likely to bias toward the null hypothesis; therefore, our HRs may be conservative.30
Conclusions
Figure 2 – Incidence of AF. Cumulative frequency curve for incident AF for all subjects (N 5 6,841) during a median follow-up of approximately 12 y, stratified by OSA severity. See Figure 1 legend for expansion of abbreviation.
TABLE 3
This large clinic-based cohort of patients assessed for OSA using laboratory-based PSG found an independent association between the presence and severity of OSA and incident AF over a median 12-year follow-up period. Our data support a growing evidence base that OSA is an important novel and potentially modifiable risk factor for the development of AF. Randomized interventional studies are required to ascertain whether treatment of OSA can reduce the incident or recurrent burden of AF.
] Hazard Ratios of OSA Measures for Incident AF: Multivariate Models
Measure Nil OSA (AHI: , 5) OSA (AHI: ⱖ 5)
HR (95% CI)a
P Value
HR (95% CI)b
P Value
1.00
…
1.00
…
1.56 (1.21-2.00)
, .001
1.55 (1.21-2.00)
, .001
1.00
…
1.00
…
AHI, clinical cut points Nil OSA (AHI: , 5) Mild (AHI: 5-14.9)
1.48 (1.12-1.96)
.006
1.48 (1.12-1.96)
.006
Moderate (AHI: 15-29.9)
1.49 (1.10-2.02)
.100
1.51 (1.11-2.05)
.008
Severe (AHI: ⱖ 30)
1.78 (1.32-2.39)
, .001
1.73 (1.29-2.33)
, .001
Log (AHI 1 1)
1.16 (1.06-1.27)
, .001
1.15 (1.06-1.26)
.001
Log (SaO2t , 90% 1 1)
1.14 (1.08-1.21)
, .001
1.12 (1.06-1.19)
, .001
See Table 1 legend for expansion of abbreviations. Adjusted for age, sex, height, and BMI. bAdjusted for age, sex, height, BMI, hypertension, valvular disease, stroke/TIA, coronary or peripheral artery disease, COPD, chronic renal disease, heart failure, and diabetes. a
journal.publications.chestnet.org
951
Acknowledgments Author contributions: G. C. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis, including and especially any adverse effects. G. C., N. M., T. B., D. R. H., L. S., M. K., and J. H. contributed substantially to the study design, data analysis, and interpretation and the writing of the manuscript.
8. Somers VK, Dyken ME, Clary MP, Abboud FM. Sympathetic neural mechanisms in obstructive sleep apnea. J Clin Invest. 1995;96(4):1897-1904.
10. Shamsuzzaman AS, Winnicki M, Lanfranchi P, et al. Elevated C-reactive protein in patients with obstructive sleep apnea. Circulation. 2002;105(21):2462-2464.
Role of sponsors: The sponsors had no direct role in the study design; the collection, analysis, or interpretation of the data; or the writing of the manuscript.
11. Dimitri H, Ng M, Brooks AG, et al. Atrial remodeling in obstructive sleep apnea: implications for atrial fibrillation. Heart Rhythm. 2012;9(3):321-327.
References 1. Rahman F, Kwan GF, Benjamin EJ. Global epidemiology of atrial fibrillation. Nat Rev Cardiol. 2014;11(11):639-654. 2. Greenlee RT, Vidaillet H. Recent progress in the epidemiology of atrial fibrillation. Curr Opin Cardiol. 2005;20(1): 7-14. 3. Wang TJ, Parise H, Levy D, et al. Obesity and the risk of new-onset atrial fibrillation. JAMA. 2004;292(20):2471-2477. 4. Gami AS, Pressman G, Caples SM, et al. Association of atrial fibrillation and obstructive sleep apnea. Circulation. 2004;110(4):364-367. 5. Mehra R, Benjamin EJ, Shahar E, et al; Sleep Heart Health Study. Association of nocturnal arrhythmias with sleepdisordered breathing: The Sleep Heart Health Study. Am J Respir Crit Care Med. 2006;173(8):910-916. 6. Stevenson IH, Teichtahl H, Cunnington D, Ciavarella S, Gordon I, Kalman JM. Prevalence of sleep disordered breathing in paroxysmal and persistent atrial fibrillation patients with normal left ventricular function. Eur Heart J. 2008; 29(13):1662-1669.
952 Original Research
22. Holman CD, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage of health records in Western Australia: development of a health services research linked database. Aust N Z J Public Health. 1999;23(5):453-459.
9. Dempsey JA, Veasey SC, Morgan BJ, O’Donnell CP. Pathophysiology of sleep apnea. Physiol Rev. 2010;90(1):47-112.
Conflict of interest: N. M. and D. R. H. have received research support from ResMed. N. M. has received an honorarium for participating at a ResMed breakfast symposium. None declared (G. C., T. B., L. S., M. K., J. H.).
Other contributions: The authors thank the staff at the Western Australian Data Linkage Branch, the Western Australia Department of Health Inpatient Data Collection and Epidemiology Branch, and the Registrar General of the Western Australia Department of the Attorney General for the provision of data. We acknowledge the contributors to the establishment of the database from which these data were extracted, including Sutapa Mukherjee, PhD; Bhajan Singh, PhD; Annette C. Fedson, PhD; and Lyle Palmer, PhD.
21. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. The report of an American Academy of Sleep Medicine Task Force. Sleep. 1999;22(5):667-689.
7. Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S. The occurrence of sleep-disordered breathing among middleaged adults. N Engl J Med. 1993;328(17): 1230-1235.
23. Knuiman M, Briffa T, Divitini M, et al. A cohort study examination of established and emerging risk factors for atrial fibrillation: the Busselton Health Study. Eur J Epidemiol. 2014;29(3):181-190. 24. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2012.
12. Shah RV, Abbasi SA, Heydari B, et al. Obesity and sleep apnea are independently associated with adverse left ventricular remodeling and clinical outcome in patients with atrial fibrillation and preserved ventricular function. Am Heart J. 2014;167(4):620-626.
25. Monahan K, Storfer-Isser A, Mehra R, et al. Triggering of nocturnal arrhythmias by sleep-disordered breathing events. J Am Coll Cardiol. 2009;54(19):1797-1804. 26. Tanigawa T, Yamagishi K, Sakurai S, et al. Arterial oxygen desaturation during sleep and atrial fibrillation. Heart. 2006;92(12): 1854-1855.
13. Todd K, McIntyre WF, Baranchuk A. Obstructive sleep apnea and atrial fibrillation. Nat Sci Sleep. 2010;2:39-45.
27. Fuster V, Rydén LE, Cannom DS, et al; American College of Cardiology/ American Heart Association Task Force on Practice Guidelines; European Society of Cardiology Committee for Practice Guidelines; European Heart Rhythm Association; Heart Rhythm Society. ACC/AHA/ESC 2006 guidelines for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology Committee for Practice Guidelines (Writing Committee to Revise the 2001 Guidelines for the Management of Patients With Atrial Fibrillation): developed in collaboration with the European Heart Rhythm Association and the Heart Rhythm Society. Circulation. 2006;114(7):e257-e354.
14. Peppard PE, Young T, Palta M, Skatrud J. Prospective study of the association between sleep-disordered breathing and hypertension. N Engl J Med. 2000; 342(19):1378-1384. 15. Young T, Peppard P, Palta M, et al. Population-based study of sleepdisordered breathing as a risk factor for hypertension. Arch Intern Med. 1997;157(15):1746-1752. 16. Hung J, Whitford EG, Parsons RW, Hillman DR. Association of sleep apnoea with myocardial infarction in men. Lancet. 1990;336(8710):261-264. 17. 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.
28. Kanagala R, Murali NS, Friedman PA, et al. Obstructive sleep apnea and the recurrence of atrial fibrillation. Circulation. 2003;107(20):2589-2594.
18. Gami AS, Hodge DO, Herges RM, et al. Obstructive sleep apnea, obesity, and the risk of incident atrial fibrillation. J Am Coll Cardiol. 2007;49(5):565-571. 19. Mukherjee S, Hillman D, Lee J, et al. Cohort profile: the Western Australian Sleep Health Study. Sleep Breath. 2012;16(1):205-215.
29. Fein AS, Shvilkin A, Shah D, et al. Treatment of obstructive sleep apnea reduces the risk of atrial fibrillation recurrence after catheter ablation. J Am Coll Cardiol. 2013;62(4):300-305.
20. American Thoracic Society. Medical Section of the American Lung Association. Indications and standards for cardiopulmonary sleep studies. Am Rev Respir Dis. 1989;139(2):559-568.
30. Selmer C, Olesen JB, Hansen ML, et al. The spectrum of thyroid disease and risk of new onset atrial fibrillation: a large population cohort study. BMJ. 2012;345:e7895.
[
148#4 CHEST OCTOBER 2015
]
