Validation of a transthoracic impendance sensor for screening of sleep apnea in CIED patients: a pooled analysis of AIRLESS and UPGRADE

FUNDING ACKNOWLEDGEMENTS: Type of funding sources: Private company. Main funding source(s): Scientific grant support from the Boston Scientific Investigator Sponsored Research (ISR) Committee, Boston Scientific, St. Paul, MN, USA. BACKGROUND/INTRODUCTION: Cardiac implantable electronic devices and their integrated thoracic impedance sensors, meant to detect sleep apnea, are available for over a decade now. Despite its usage in daily clinical practice, there is only limited data on its diagnostic accuracy. PURPOSE: To further validate an apnea sensor by using polysomnography (PSG) and investigate subgroups with potential differences in diagnostic accuracy. METHODS: AIRLESS and UPGRADE were prospective investigator driven trials meant to validate an apnea sensor. Patients, who either fulfilled the criteria for implantation of an implantable cardioverter-defibrillator (ICD), cardiac resynchronization therapy (CRT) or upgrading to CRT according to current guidelines were eligible for enrolment. Sleep apnea and its severity, measured by apnea-hypopnea index (AHI), were exclusively assessed by PSG. PSG was conducted in a time interval of 1 to 3 months post implantation of the device. For direct comparison, the apnea sensor derived respiratory disturbance index (RDI) was used from the identical night. Bivariate correlations (spearman coefficient) as well as the Bland-Altmann method were applied to assess diagnostic accuracy of the apnea sensor. RESULTS: Overall, 80 patients were analysed, out of whom the majority (n = 35, 43.8%) were diagnosed with central sleep apnea. Sixteen (20.0%) patients received a conventional ICD system, whereas 21 (26.3%) were implanted a CRT-P and 43 (53.8%) a CRT-D. Atrial fibrillation was known in 50.0% (n=40) of patients, 45 patients suffered from periodic leg movement syndrome (PLMS, 56.3%). In the overall cohort, the sensor derived RDI correlated significantly with the AHI (r = 0.61, p < 0.001) with a mean difference (MD) of -12.6 (95% CI -38.2 – 13.0). The sensor did specifically perform well in patients without atrial fibrillation (r = 0.80, p < 0.001; MD -9.0, 95% CI -30.4 – 12.5), while the presence of PLMS did not affect the accuracy of measurements (PLMS: r = 0.60, p = 0.007, MD -16.1, 95% CI -43.9 – 11.8 vs. no PLMS: r = 0.62, p = 0.001, MD -9.9, 95% CI -32.8 – 12.9). Furthermore, the RDI was found to correlate well in patients with central sleep apnea (r = 0.61, p = 0.013), although similar overestimation was observed (MD -11.6, 95% CI -35.4 – 12.2) as in other subgroups. CONCLUSION: In an exclusive heart failure cohort, the apnea sensor did correlate well with the PSG derived AHI. Although similar correlation is found in most subgroups, there remains to be a decent overestimation of events. [Figure: see text] [Figure: see text]