Introduction

Sleep-disordered breathing (SDB), including obstructive sleep apnea (OSA) and central sleep apnea (CSA), is prevalent in approximately half of patients with atrial fibrillation (AF) [1, 2]. Untreated OSA is associated with a higher risk of the recurrence of AF after medical therapy [3], electrical cardioversion [4], or AF ablation [5–7]. While moderate-to-severe OSA causes structural and functional changes in left ventricular (LV) function compared to that in healthy subjects [8, 9], the severity of CSA is also associated with the incidence of AF among patients with LV dysfunction [10]. Most previous studies of these relationships between SDB and LV dysfunction in AF were evaluated in patients already diagnosed with SDB; however, the impact during the subclinical period of SDB has not been elucidated. We conducted a prospective cross-sectional study of the association between the severity of previously undiagnosed SDB among AF ablation candidates and adverse cardiac structural remodeling, particularly regarding left atrium (LA) dilatation and LV diastolic function parameters which are associated with AF ablation failure.

Materials and methods

Study population

The study was conducted under a prospective, single-center observational design using a prospective registry. We recruited 269 consecutive Japanese non-mitral valvular AF patients hospitalized at our institution as candidates for de novo pulmonary vein isolation (PVI) from December 2011 to May 2014. We excluded 5 patients from analysis due to the use of positive airway pressure (CPAP) for known SDB, and 18 due to inappropriate oxygen desaturation data (n = 5) and concomitant disease affecting ventricular diastolic function, such as post-cardiac surgery, creatinine level >=2.0, moderate aortic valve disease, and hypertrophic cardiomyopathy (n = 13; Fig. 1). The study was conducted in the remaining 246 patients. The study protocol was approved by the ethical committee of Tenri Hospital and all patients provided written informed consent to all procedures associated with the study.

Fig. 1
figure 1

Flowchart of the subjects in the study

Full size image

Pulse oximetry

The evening before the AF ablation procedure, all patients underwent a medical interview, physical examination, and oxyhemoglobin saturation (SaO2) monitoring. This monitoring was conducted with a high-resolution pulse oximeter (PULSOX-Me300; Minolta Co, Osaka, Japan), which has emerged as a simplified screening tool via a finger probe pulse oximeter for SDB [11]. A sleep diary was recorded to minimize any potential overestimation of sleep duration. The sampling frequency of the PULSOX- Me300 oximeter was 1 Hz during the memory interval for an average time of 3 s each. The resolution for SpO2 was 0.1 %. The algorithm for analyzing the PULSOX-Me300 data was compatible with that implemented in commercially available software (DS-Me, Minolta Co, Osaka, Japan). In this study, we used a 3 % oxygen desaturation index (ODI) as the SDB cutoff value, as described in previous studies. This ratio has 80 % sensitivity and 95 % specificity for detecting an apnea–hypopnea index (AHI) of ≥5 events/h by polysomnography (PSG) using a cutoff threshold of a 3 % ODI = 5, and an 85 % sensitivity and 100 % specificity for detecting an AHI of ≥20 using a 3 % ODI = 15 [12, 13]. To classify the severity of the SDB, we subdivided the patients into three groups according to 3 % ODI level, namely ODI < 5 events/h as the normal SDB group, 5 ≤ ODI < 15 as the mild SDB group, and 15 ≤ ODI as the moderate-to-severe SDB group (50 patients) (Fig. 1).

Echocardiographic measurements

All transthoracic echocardiographic data were obtained and analyzed by the same echocardiography specialist before the procedure. Two-dimensional echocardiographic images of the LA and LV were obtained in accordance with the American Society of Echocardiography (ASE) recommendations [14]. The mitral inflow velocity and tissue Doppler imaging (TDI) signals were adjusted to provide a smooth velocity distribution without introducing any noise at a 100-mm/s sweep speed. LV ejection fraction (LVEF) was assessed by the modified Simpson’s method. The early (E) and late (A) diastolic peak velocities from the apical 4-chamber (4C) view, and E-wave deceleration time were also measured. The peak early mitral annular velocities (e’) were derived by tissue Doppler analysis at the septal and lateral margin of the mitral annulus in the apical 4C view. The E/e’ was then calculated as an index of the LV filling pressure at both annular sites. Although the baseline data were managed to be acquired during sinus rhythm, measurements in persistent AF cases were obtained from a single cardiac cycle with an RR interval corresponding to a heart rate of 70–80 beats/min during an AF rhythm [15].

LA volume and LV diastolic function

LA volume was estimated by the biplane area-length method, using measurements in the apical 4- and 2-C views at end-systole. LA volume index was calculated as the LA volume divided by the body surface area [16]. In this study, an enlarged LA was defined as a left atrial volume index (LAVI) of ≥34 mL/m2 in accordance with previous studies [17, 18].

Furthermore, left ventricular diastolic dysfunction (LVDD) was evaluated using combined categories with tissue Doppler measurements of the septal and lateral mitral annulus and LA volume, which was the first screening step before grading the LV diastolic function with the recent ASE recommendations [18]. An LVDD case was defined as a patient who met all three of the following criteria, a septal annular e’ of <8 cm/s, lateral annular e’ of <10 cm/s, and an enlarged LA. All patients who were not classified as LVDD cases were treated as having normal LV diastolic function.

Statistical analysis

Data were collected and entered prospectively in a computer database. Continuous variables were summarized as the mean and standard deviation or median and interquartile ranges (IQR), depending on the normality of the distribution. Comparisons were made with the χ 2 or Fisher’s exact test for categorical variables, as appropriate, and with the Wilcoxon rank-sum test for continuous variables. A value of P ≤ 0.05 was considered statistically significant. To identify independent predictors of an enlarged LA and LVDD, univariate and multivariate logistic regression analyses were performed using 11 relevant clinical factors, namely age; gender; AF evolution time (< or ≥24 months); body mass index (BMI); AF type; history of hypertension (HT), diabetes mellitus (DM) or congestive heart failure (CHF); use of angiotensin-converting enzyme inhibitors (ACEI) and/or angiotensin receptor blockers (ARB); brain natriuretic peptide (BNP) level; and left ventricular ejection fraction (LVEF) (< or ≥50 %), and three groups categorized according to the severity of the SDB. Variables with a P value ≤0.10 in the univariate models were included in the multivariate analyses. In univariate and multivariate logistic regression analyses, the normal SDB group was used as reference. Odds ratios and corresponding 95 % confidence intervals (CIs) were also computed where appropriate. Statistical analyses were performed with JMP version 10.02 software (SAS Institute Inc., Cary, NC, USA).

Results

Baseline patient characteristics

Of the 246 study patients, SDB was detected in 176 patients (71.5 %), grouped as 126 (51.2 %) with mild SDB and 50 (20.3 %) with moderate-to-severe SDB (Fig. 1). The baseline characteristics and echocardiographic parameters of all patients and in three groups categorized by the severity of SDB are shown in Tables 1 and 2. The baseline characteristics, including age, gender, CHA2DS2-VASc score [19], β-blocker utilization, and LVEF, were similar among the three groups. The prevalence of hypertension and use of ACEI/ARB were significantly lower in the normal SDB group than in the mild and moderate-to-severe SDB groups (P < 0.0001 and p = 0.015, respectively). Therefore, baseline blood pressure was well controlled (systolic/diastolic blood pressure; 125/74 mmHg; median). In contrast, in the moderate-to-severe SDB group, left ventricular hypertrophy exhibited a greater progression than in the normal SDB group regarding the septal wall thickness (p = 0.018) and posterior wall thickness (p = 0.048). In addition, SDB was positively associated with body mass index (BMI) (p = 0.0001), and patients were more likely to have non-paroxysmal AF (p = 0.0012). The severity of SDB was linearly associated with the BNP level with each cutoff (p = 0.0005). Further, LAVI, the ratio of the early transmitral filling velocity (E) and early diastolic mitral annular velocity (e’), significantly differed between the normal and moderate-to-severe SDB group for both the septal and lateral velocities (LAVI; p = 0.0008, septal E/e’; p = 0.0004, lateral E/e’; p = 0.0003, respectively).

Table 1 Clinical characteristics of patients with or without sleep-disordered breathing
Full size table
Table 2 Echocardiographic parameters in subjects with or without sleep-disordered breathing
Full size table

LA volume, LV diastolic dysfunction, and SDB

Figure 2 shows the prevalence of patients with an enlarged LA or LADD in the three groups according to SDB severity. An enlarged LA was observed in 126 (51.2 %) of all subjects, and this ratio gradually increased with increasing SDB severity (p < 0.0001). LVDD was observed in 42 (17.1 %) of all subjects, and similarly gradually increased with SDB severity (p < 0.0001). Among the three groups, the incidence of both enlarged LA and LVDD was significantly higher in the moderate-to-severe SDB group (76.0 and 40.0 %, respectively) than in the other two groups.

Fig. 2
figure 2

Unadjusted proportion of an enlarged LA and LVDD conditions by the severity of the ODI

Full size image

Multivariate analysis of predictors of enlarged LA and LVDD

Table 3 demonstrates the univariate- and multivariate-adjusted OR and 95 % CI for the selected cardiovascular risk factors associated with enlarged LA. A multivariate regression analysis after adjusting for the age revealed that moderate-to-severe SDB was the strongest independent predictor of enlarged LA [odds ratios (OR); 2.95, 95 % confidence intervals (95 % CI); 1.08–8.76, p = 0.034], followed by BMI ≥ 25 kg/m2 (OR; 2.35, 95 % CI; 1.13–5.09, p = 0.022). Table 4 shows the univariate- and multivariate-adjusted analyses for the selected cardiovascular risk factors associated with LVDD. The multivariate analysis showed that only the presence of moderate-to-severe SDB was independently associated with LVDD (OR; 5.96, 95 % CI; 2.10–19.00, p = 0.0006).

Table 3 Univariate and multivariate models of clinical factors associated with LAVI ≥ 34 ml/m2
Full size table
Table 4 Univariate and multivariate models of clinical factors associated with LV diastolic dysfunction
Full size table

Discussion

Main findings

In this study, we found that SDB, even in subclinical settings, was commonly present in patients who were candidates for AF ablation. It also showed that the prevalence of an enlarged LA and LVDD increased with increasing severity of SDB, and that moderate-to-severe SDB was an independent predictor of enlarged LA and LVDD. These findings suggest that subclinical moderate-to-severe SDB had adversely affected the LA remodeling as well as LV diastolic function before the procedure.

LA volume, LV diastolic dysfunction, and SDB

Among findings, pulse oximetry revealed that 20.3 % of AF ablation candidates had moderate-to-severe SDB (3 % ODI level of ≥15) even in their subclinical state. Our present study revealed that the coexistence of SDB with a pulse oximeter as well as obesity (BMI ≥ 25 kg/m2) independently associated with LA anatomical remodeling among consecutive AF patients who had not been subjectively recognized as having sleep apnea (not only OSA) before AF ablation. Furthermore, moderate-to-severe SDB was the strongest factor of LVDD among AF patients. Although several studies reported that SDB causes LA remodeling and LV diastolic function, these studies were focused on patients with known SDB strictly diagnosed by OSA history and PSG among a male or obese cohort [20–22]. In contrast, we demonstrated these relationships among AF ablation candidates. It is well known that LVDD is a powerful predictor of LA enlargement and AF, and reflects the duration and severity of an increased LA pressure [23]. The concomitance of LVDD and an enlarged LA is an independent risk factor for the outcome of AF ablation [24]. On the other hand, SDB has been independently recognized to be related to the outcome after PVI [5–7]. Therefore, our findings suggested that this negative link between SDB and a poor prognosis of PVI might be related to the LA remodeling based on preexisting LVDD caused by SDB. In the present study, HT and LV hypertrophy, which are major factors in LVDD, were found to be more likely in moderate-to-severe SDB group. This finding is consistent with previous studies reported that SDB was independently associated with HT and LV hypertrophy, which leads to LVDD [25–27]. Recently, SDB has been recognized not only as a risk factor for LV remodeling, but also for LA remodeling. The pathophysiological interaction between SDB and LV remodeling or between SDB and LA remodeling remains complex, and it might differ depending on the type of SDB [25, 28]. However, awareness of a relationship with subsequent nocturnal hypertension due to excessive sympathetic activation, inflammation, oxidative stress, negative intrathoracic pressure swing, and endothelial dysfunction has recently increased as a new pathophysiological insight between SDB and LV remodeling [29–31]. Additionally, recent works demonstrated that the SDB is independently associated with atrial electrical remodeling as well as anatomical remodeling [32, 33]. They also speculated that intermittent chronic hypoxemia, hypercapnia, and increased plasmatic catecholamine levels may represent direct proarrhythmogenic injury to the atrial tissue. Therefore, not only a pressure overload due to LVDD, but also subsidiary factors related to SDB cause atrium tissue remodeling.

Epidemiological data have demonstrated that SDB has prognostic implications for AF among patients both with and without cardiovascular diseases [1, 2, 34]. On this basis, the development of subclinical SDB might precede several known risk factors of LVDD, and play a crucial role in LA enlargement and LVDD among AF ablation candidates prior to the onset of AF. Previous uncontrolled studies indicate that CPAP induces reverse atrial remodeling in patients with severe SDB [35, 36]. Furthermore, CPAP significantly provides an overall lower arrhythmia recurrence rate and improves outcomes following PVI [5–7, 37]. Pulse oximetry can, therefore, be incorporated into clinical pathways as a simple and efficient initial diagnostic tool for the evaluation for SDB in many patients referred for AF ablation.

SDB and progressing sustained AF

In this study, we found that non-paroxysmal AF was more common in the SDB groups and that paroxysmal AF was more common in the normal SDB group, despite the longer total duration of AF history in the normal SDB patients. While the risk factors associated with incident AF have been reported, few studies have examined whether these AF risk factors differ with regard to the development of non-paroxysmal AF rather than paroxysmal AF. Indeed, OSA is increasingly recognized as a potential risk factor for the development of AF, and SDB is more frequently observed in non-paroxysmal AF patients [38]. However, our findings suggested that the severity of SDB might be preferentially associated with early development toward non-paroxysmal AF.

There is some speculation of the relationship between SDB and the progression to sustained forms of AF. One is that SDB might produce multifocal triggers in arrhythmogenic veins (PVs and/or non-PV) [7]. A second is that SDB is associated with anatomical and/or electrical adverse remodeling. Actually, current studies have shown that inter-atrial conduction or intra-atrial conduction abnormalities, which should contribute to the maintenance of AF, are associated with OSA [22, 34, 35]. However, our candidates for AF ablation were intentionally excluded if they had severe LA enlargement over 150 ml/m2, suggesting that the former speculation might be preferable in our setting. Frequent elevations in end-diastolic pressure due to accumulated stress from the SDB can initiate repetitive arrhythmias from PV [1]. Furthermore, the prevalence of non-PV triggers of AF is higher in SDB patients [7]. This complex interplay between the mechanical–electrical feedback mechanisms might provoke sustained AF [39, 40].

Limitations

Several limitations of our study warrant mention. First, while diastolic dysfunction, a common accompaniment of aging, is associated with HT, obesity, diabetes, and coronary artery disease, the cross-sectional design of this single-center study did not permit us to establish causation between the diastolic dysfunction and SDB. Additionally, the lack of a convenient, inexpensive, and accurate method for determining the diastolic dysfunction affected the reproducibility of echocardiographic data, especially for the composite objects of paroxysmal and non-paroxysmal AF. Secondly, we did not show precise data for the type or severity of SDB using PSG, the standard examination for SDB. A different mechanism of SDB for the occurrence of diastolic dysfunction might affect the results. The existence of central sleep apnea, which should be more common in the patients with persistent AF, might not have been properly diagnosed. Because OSA may be more related to genesis of AF and CSA would be considered more as the result of AF (e.g., a higher LA pressure). Accordingly, a degree of residual confounding may have been present, despite our multivariate adjustment.

Conclusions and clinical implications

The presence of moderate-to-severe SDB in candidates for AF ablation adversely affected LV diastolic function in these patients, even when the SDB was in a subclinical state. We recommend that pulse oximeter testing becomes a routine part of the procedural work-up before AF ablation.