The variability and burden of severe sleep apnea and the relationship with atrial fibrillation occurrence: analysis of pacemaker-detected sleep apnea

Recently, Boston Scientific/Vitalio pacemakers have been reported to have a transthoracic impedance sensor and AP Scan algorithm screen for SA in pacemaker patients [8]. Briefly, transthoracic impedance fluctuated regularly with respiratory movement. The difference of transthoracic impedance in a cycle may be utilized to estimate the variation in tidal volume. According to the AP Scan algorithm, respiratory disturbance event was defined as a decline of tidal volume of at least 26% from the baseline or a breathe suspended for at least 10 s or more [9]. An RDI was the average number of respiratory disturbance events the patients experienced per hour during a programmed sleep time. Sleep time was programmed individually according to patients’ sleep diaries. The RDIs were recorded and plotted once a day to form the AP Scan trend (Fig. 1).

This prospective single-center study was conducted at the Fuwai Hospital (Beijing, China). A series of consecutive patients undergoing implantation of a Vitalio J237 or J274 (Boston Scientific, St. Paul, MN, USA) from August 2016 to December 2017 were enrolled in the study, irrespective of whether or not the patients had symtpoms of SA or a SA diagnosis. All patients had an indication for pacemaker implantation according to current guidelines [10].

The exclusion criteria included clinical diseases affecting the transthoracic impedance sensor function or resulting in sleep alteration, such as chronic obstructive pulmonary disease, interstitial pulmonary diseases, heart failure, parasomnias, alcoholism, use of hypnotics, habits, or occupations leading to sleep deprivation or altered sleep–wake cycle. The respiratory signal amplitude could vary according to the posture or movement of patients. If the respiratory signal amplitude was unstable or signal-to-noise ratio was out of range, the RDI would then be discarded and the night would be considered invalid. Patients with > 10% invalid AP Scan data were excluded. Information including demography, underlying heart and pulmonary disease, pacing indication, atrial arrhythmia, and N-terminal pro brain natriuretic peptide (NT-proBNP) were obtained from medical record. All patients underwent standard two-dimensional and Doppler echocardiographic examination at rest. Measurements of the LV cavity dimension was derived from the parasternal long axis view of two-dimensional imaging. LVEDD was measured at end diastole (the onset of the QRS complex). LVEF was measured by modified Simpson rule from the apical four-chamber plane.

This study was approved by local ethics committees, and all enrolled patients were required to sign an informed consent.

After implantation. Patients underwent an attempt at overnight PSG. The overnight PSG was performed and interpreted by a sleep specialist who was blinded to the PM-RDI data. PSG examination included electroencephalogram (EEG) for sleep analysis, electromyogram for chest and abdominal respiratory movement, oral-nasal airflow, and arterial oxyhemoglobin saturation. A hypopnea event was defined by a decline of tidal volume at least 30% from baseline lasting 10 s or longer. An apnea event was defined by a suspension of oral-nasal airflow lasting 10 s or more. The PSG-AHI was the average number of apnea and hypopnea events per hour during sleep time. The severity of SA was graded by the AHI scores [11], that is, PSG-AHI ≥ 30: severe SA; 15 < or =  PSG-AHI < 30: moderate SA; 5 < or = PSG-AHI < 15: mild SA.

RDI criteria has been described in our previous study [5]. That is, a 26 ≤ RDI < 41 identified patients as moderate SA, corresponding to 15 ≤ AHI < 30; an RDI ≥ 41 identified patients as severe SA, corresponding to AHI ≥ 30 with a sensitivity of 82.1%, a specificity of 88.6% [5]. The RDI data of the night during which time the patient underwent the PSG was deemed as the first night RDI. The mean RDI was equal to the average RDI during the follow-up.

During the follow-up, we found that the RDI, like other physiology parameters, such as heart rate and blood pressure, also fluctuated every day. We raised a concept as RDI variability, which was defined as standard deviation of RDI (RDI-SD). RDI burden was referred to as the percentage of nights with RDI ≥ 26.

The duration of AF each day was recorded in the pacemaker. Boriani’s study demonstrated that daily burden of 6 h was associated with 17% increase in the risk of stroke [12]. In Turakhia’s study, a burden of ≥ 5.5 h raised the short-term risk of stroke 4- to 5-fold [13]. In this study, an AF event was defined as the cumulative AF duration > 6 h in a day.

The patients were routinely arranged follow-up at 1 month after implantation, every 6 months thereafter, and more frequently according to the clinical needs. The interrogations of the store data were performed during each follow-up. The store data included the RDI data and AF events. In this study, AF event was defined as the occurrence of AF lasting more than 6 h. All the data could be recorded in PM as long as 1 year. During the follow-up, the RDI data was restored in special memory card and sent to technical department of Boston Scientific company to be extracted automatically by special software.

The patients were divided into two groups according to the percentage of night with the mean RDI ≥ 26. Higher burden of severe SA indicated the patients with more than 75% nights RDI ≥ 26. The less burden of severe SA suggested the patients with either mean RDI ≤ 26 or mean RDI ≥ 26 but burden < 75%. Categorical variables were presented as the number and percentage and were compared via chi-square analysis or Fisher’s exact test. Continuous variables were presented as mean ± SD if conform to normal distribution and presented as median (quartile) if not. And independent t test was applied to compare continuous variables. Standard deviation (SD) was used to appraise the variability of RDI. Linear regression was calculated between the individual’s mean RDI and SD. The univariate analysis and multivariable logistic regression logistic regression was used to analyze the relationship between RDI and AF occurrence with the Forward method. The variables included AHI, mean RDI, and first night RDI. Age and BMI were also associated with AF occurrence; these two variables were included in the logistic regression analysis. The Kolmogorov–Smirnov test was applied for testing the fit of variables to a normal distribution, including mean RDI, AHI, and first night RDI. All statistical analyses were performed with IBM SPSS 19.0 package program. A p value of < 0.05 was considered significant difference.

A total of 35 patients underwent Vitalio pacemaker implantation, among them 5 patients (14.3%) were excluded due to least 10% invalid nights, with the rest 30 patients being included. The mean age was 65.1 ± 9.8 years, and 15 (50%) were men. The pacing indication was sick sinus syndrome. No patients had documental SA diagnosis or underwent PSG exam before PM implantation. Fourteen (44.8%) patients had a history of paroxysmal AF. All the patients underwent a dual-chamber pacemaker implantation. Twenty-three (76.6%) patients had an AHI ≥ 5 (8, mild; 9, moderate; 6, severe), and the average AHI was 16.1 ± 12.7. The patients’ characteristics are presented in Table 1.

The patients were followed for a mean period of 356 ± 27 days (range 278–365 days). From 10,442 nights, 10,003 (95.8%) nights’ RDI data were extracted for further analysis. Only 3 patients (10%) never had a single RDI greater than 26, 27 patients (90%) had at least 1 night with RDI ≥ 26, and 17 patients (56.6%) had at least 1 night with RDI ≥ 41. Fifteen patients (50%) presented with more than 50% of nights with RDI ≥ 26 (Fig. 2). The average of the first night RDI was 27.8 ± 13.3. The overall mean RDI of the whole follow-up period was 24.5 ± 8.6. As regards the SA diagnosis, there was no significant difference between the first night RDI and the mean RDI (p = 0.192).

Nine (30%) patients were diagnosed as the high SA burden, as the RDI ≥ 26 in more than 75% nights. Compared with the less burden patients, the high burden patients were found with a higher BMI (26.7 ± 4.8 vs 23.2 ± 3.9, p = 0.036) and a higher prevalence of hypertension (85.7% vs 39.1%, p = 0.031). The left ventricular diastolic diameter was larger in high burden group than that in less burden group (46.7 mm vs 49.2 mm, p = 0.036). Despite the fact that AF occurrence during the follow-up was more common in high burden group (71.4% vs 30.4%, p = 0.05), there was no significant difference in history of paroxysmal AF between the two groups (57.1% vs 43.5%, p = 0.526). The mean RDIs in higher and less burden groups are 35.3 ± 7.3 and 21.2 ± 5.9, respectively (p = 0.001). The RDI-SD among all patients is 6.9 ± 3.3. The RDI-SD in higher burden patients is significantly greater than that in the less burden patients (10.7 ± 4.9 vs 5.7 ± 1.4, p = 0.036) (Table 1).

Linear regression demonstrated a significant linear relationship between the mean RDI and the RDI-SD (R = 0.661; p < 0.001), with a tendency of higher RDI and higher RDI-SD (Fig. 3).

During the follow-up, 12 patients (40%) experienced at least an AF episode lasting more than 6 h. K-S test confirms that the mean of RDI, AHI, and first night RDI comply with the normal distribution. After adjusting for age and BMI in logistic regression model, the mean of RDI (OR = 1.118, 95%CI 1.008–1.244, p = 0.044) had an increased risk of AF occurrence, while the first night RDI (OR = 1.001, 95%CI 0.942–1.063, p = 0.982) or the AHI (OR = 1.015, 95%CI 0.954–1.080, p = 0.637) could not predict the onset of AF.

Limitations of this study included the small number of patients and solo pacing indication. As we wanted to appraise the long-term variability of RDI, it would be more interesting to consider longitudinal observations instead of using an aggregated measure such as mean and SD. Besides a mixed model instead of a linear regression could improve the power of the study. In order to evaluate the relationship between the AF event and RDI, we excluded the patients with persistent AF, and all the patients underwent dual-chamber pacemaker implantation. This resulted in atrioventricular block patients being excluded from this study and might come with some systemic error. In addition, in this study, patients were considered to have AF event if the device detected that AF burden was greater than 6 h, which may contribute to the underestimation of AF events.