ECG-derived respiration: A promising tool for sleep-disordered breathing diagnosis in chronic heart failure patients.

Theprevalenceofsleep-disorderedbreathing(SDB)inchronicheartfailure (CHF) patients exceeds 60%, including both central sleep apnea(CSA)andobstructivesleepapnea(OSA)[1].Asobservedinthegeneralpopulation, SDB in CHF patients is associated with a poor prognosis [2]and a higher risk of nocturnal arrhythmia [3]. Continuous positiveairway pressure or adaptive servo-ventilation treatments might im-provebothLVEFandmortalityandreducetheriskofarrhythmicevents,but this remains debatable [4]. Nevertheless, at least 75% of severe SDBcasesremainundiagnosedanduntreatedduetotheabsenceofdaytimesymptoms, waiting lists in sleep laboratories, and diagnostic costs. Thehigh SDB prevalence in patients referred for cardiologic assessmentcalls for the development of automated screening tools based on ECG-derived parameters for use by cardiologists in routine practice.MethodsbasedonheartratevariabilityanalysisallowdetectingSDBpatients whopresentsinusal rhythm[5] butare not suitable for cardiacfailureordysautonomicpopulations,whofrequentlyexhibitprematureatrial or ventricular beat, are frequently implanted with permanentpacemakers, or present recurrent atrial fibrillation, bundle branchblocks, or flat heart rate variability [6]. ECG-derived respiration (EDR),a method based on the reconstruction of thoracic signals by analysingchangesinECGmorphologythatareinducedbyrespiratorymovements[7], is a promisingalternative,but has not yet been evaluatedincardiacpopulations [8].Werecruited105patientswithLVEFb45%andNYHAclass≥2fromroutine medical follow-ups of stable CHF patients (Saint-Etienne andGrenoble University Hospitals) to test an automated EDR method withreference to visually scored nocturnal in-home ventilatory polygraphy.This study (clinical trial NCT02116686) complies with the DeclarationofHelsinkiandwasapprovedbythelocalethicscommittee.Allpatientsgave their written informed consent to participate.Standard nocturnal in-home ventilatory polygraphy was performedusinganEmbladevice(Embla®,Broomfield,USA)andscoredaccordingto AASM recommendations using RemLogic® software. Based onpreviously published methods [7,9,10] and developed with Matlab®(Mathworks,Natick,USA),anEDRmethodwasusedasfollows:thoracicmovements derived from ECG signals were analysed in order toreconstruct respiratory movements. The two mechanisms involved,changesintheelectricalheartaxisandchangesintransthoracicimped-ance,areeasilyobservedintheQRSheightorarea,allowingreconstruc-tion of the thoracic signal, which is then used to score apnea andhypopnea events [7]. Analyses were performed on the portion of thepolygraphic ECG signal recorded when the light was turned off. TheECG baseline was removed, each R-peak was detected, and R-peakoutliers were excluded. The EDR signal was then calculated as thesuccessive surfaces of each QRS complex under the R-peak. An SDBevent was then scored when the algorithm detected either a 50%decrease in thepeak-to-peak EDR amplitudefor 10 s [9] or an invertedU-shape of the EDR trace corresponding to a baseline drift due torespiratory effort during obstruction [10].ECG results were subjected to automated analysis only, and poly-graph recordings were scored manually and blinded to EDR results. Uni-variate regression analysis for continuous variables and Bland–Altmanplot were used to assess the relationship between the polygraphicapnea/hypopnea index (AHI