Intermediate Markers Underlying Electrocardiographic Predictors of Incident Atrial Fibrillation: the MESA.

HomeCirculation: Arrhythmia and ElectrophysiologyVol. 14, No. 12Intermediate Markers Underlying Electrocardiographic Predictors of Incident Atrial Fibrillation: The MESA Free AccessReview ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissionsDownload Articles + Supplements ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toSupplementary MaterialsFree AccessReview ArticlePDF/EPUBIntermediate Markers Underlying Electrocardiographic Predictors of Incident Atrial Fibrillation: The MESA Eric Xie, MD, Colin Wu, PhD, Mohammad Ostovaneh, MD, Wendy S. Post, MD, Shelby Kutty, MD, PhD, Elsayed Z. Soliman, MD, MSc, MS, David A. Bluemke, MD, PhD, Susan R. Heckbert, MD, PhD, Joao Lima, MD and Bharath Ambale-Venkatesh, PhD Eric XieEric Xie Cardiology Division, Department of Medicine (E.X., M.O., W.S.P., S.K., J.L.), Johns Hopkins Hospital, Baltimore. , Colin WuColin Wu Office of Biostatistics Research, National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, MD (C.W.). , Mohammad OstovanehMohammad Ostovaneh Cardiology Division, Department of Medicine (E.X., M.O., W.S.P., S.K., J.L.), Johns Hopkins Hospital, Baltimore. , Wendy S. PostWendy S. Post https://orcid.org/0000-0002-8655-5204 Cardiology Division, Department of Medicine (E.X., M.O., W.S.P., S.K., J.L.), Johns Hopkins Hospital, Baltimore. , Shelby KuttyShelby Kutty https://orcid.org/0000-0001-9428-0979 Cardiology Division, Department of Medicine (E.X., M.O., W.S.P., S.K., J.L.), Johns Hopkins Hospital, Baltimore. , Elsayed Z. SolimanElsayed Z. Soliman https://orcid.org/0000-0001-5632-8150 Epidemiological Cardiology Research Center, Department of Epidemiology and Prevention (E.Z.S.), Wake Forest School of Medicine, Winston-Salem, NC. Department of Medicine, Section of Cardiology (E.Z.S.), Wake Forest School of Medicine, Winston-Salem, NC. , David A. BluemkeDavid A. Bluemke https://orcid.org/0000-0002-8323-8086 Department of Radiology, University of Wisconsin School of Medicine and Public Health, Madison (D.A.B.). , Susan R. HeckbertSusan R. Heckbert https://orcid.org/0000-0002-7100-512X Department of Epidemiology, School of Public Health, University of Washington, Seattle (S.R.H.). , Joao LimaJoao Lima https://orcid.org/0000-0001-8756-6995 Cardiology Division, Department of Medicine (E.X., M.O., W.S.P., S.K., J.L.), Johns Hopkins Hospital, Baltimore. and Bharath Ambale-VenkateshBharath Ambale-Venkatesh Correspondence to: Bharath Ambale-Venkatesh, PhD, Division of Radiology, Johns Hopkins University, 600 N Wolfe St, Baltimore, MD 21287. Email E-mail Address: [email protected] https://orcid.org/0000-0002-2330-2373 Department of Radiology (B.A.-V.), Johns Hopkins Hospital, Baltimore. Originally published30 Nov 2021https://doi.org/10.1161/CIRCEP.121.009805Circulation: Arrhythmia and Electrophysiology. 2021;14Other version(s) of this articleYou are viewing the most recent version of this article. Previous versions: November 30, 2021: Ahead of Print It remains unclear whether the role of the ECG in prediction of atrial fibrillation (AF)1 is primarily as surrogate for cardiac structure and function, or if it captures unique myocardial prognostic electrophysiological information unobtainable from imaging. We sought to investigate ECG predictors of incident AF, with cardiac structure/function as intermediate markers of AF. We hypothesized (1) a subset of ECG measures might predict AF independent of imaging and (2) such ECG measures may reflect a pathophysiologic process distinct from or preceding structural/functional changes. Our objectives were 3-fold. First, we identified ECG predictors of AF using machine learning. Second, we observed the robustness of predictors after adjustment for clinical factors and imaging data. Finally, we explored the correlation of top predictors with additional prognostic factors of AF.MethodsThe MESA (Multi-Ethnic Study of Atherosclerosis) is a multicenter prospective cohort study with 6814 participants. Subjects gave informed consent and the study was approved by institution review boards at all participating institutions. Inclusion and exclusion criteria are summarized (Figure). Requests to access the dataset from qualified, trained researchers may be sent to [email protected]The protocols for ECG and cardiac magnetic resonance acquisition have been described.2,3 Total 568 ECG variables included individual lead voltage amplitudes, duration, and axis. Collection of demographic and clinical covariates was completed at study initiation using standardized questionnaires. Random survival forests were implemented in R 3.5.3 (R Foundation for Statistical Computing) using publicly available libraries.4 Two-thirds of participants were randomly split to a training cohort and one-third to validation cohort. Identification of candidate variables was accomplished by calculation of variable importance via permutation.ResultsParticipant clinical and demographic data are shown (Table S1). Included participants were on average healthier than those excluded.The random survival forest C-index was 0.70 (95% CI, 0.64–0.76) in the validation cohort, with top selected ECG variables shown (Table S2). We further identified the following variables as unique predictors using backwards stepwise Cox selection on the top random survival forest variables: P-amplitude, aVR; P-duration, aVR; J-point amplitude, aVR and V5; QTc interval.The association of unique predictors with incident AF was confirmed in adjusted Cox models (Table). The inclusion of atrial structure and function attenuated the significance of P-amplitude and -duration whereas J-point amplitude and QTc remained significant. Moreover, they remained significant after incorporating structural and functional imaging measures from both ventricles. Finally, we assessed the effects of NT-proBNP (N-terminal pro-B-type natriuretic peptide) and coronary artery calcium. In the fully adjusted model, only QTc remained significant (Table).Table. Association of Top ECG Predictors With AF After AdjustmentSelected variablesIncident AF (no imaging)Incident AF (imaging)*Incident AF (all imaging)†Incident AF (all measures)‡Per 0.1 mV or per 0.04 sHRP valueHRP valueCoeffP valueCoeffP valueP-amplitude aVR1.560.0311.390.11P-duration aVR1.350.0051.280.17J-point amplitude V50.680.0100.690.0180.690.0520.970.14J-point amplitude aVR2.110.0032.020.0051.780.0481.040.22QTc interval1.380.0021.420.0011.430.0041.380.014All models adjusted for age, race, gender, BMI, smoking status, HDL, lipid-lowering medication, systolic blood pressure, hypertension medication, fasting glucose, and alcohol use. AF indicates atrial fibrillation; BMI, body mass index; CAC, coronary artery calcium score; EF, emptying fraction; HDL, high-density lipoprotein; HR, hazard ratio; LA, left atrium; LV, left ventricle; NT-proBNP, N-terminal pro-B-type natriuretic peptide; RA, right atrium; and RV, right ventricle.* Model additionally adjusted for RA and LA volume.† Additionally adjusted for RA and LA strain, RV and LV end-diastolic volume index, mass index, and EF.‡ Additionally adjusted for NT-proBNP, and CAC.We separately explored the relationship of predictors with clinical and imaging measures (Tables S3 through S5). All variables were significantly associated with right atrial volume. Only P-duration was associated with left atrial volume. P-amplitude and -duration were mostly associated with atrial function. J-point amplitude and QTc were strongly associated with left ventricular end-diastolic volume, mass, and ejection fraction. NT-proBNP but not coronary artery calcium was positively associated with both J-point amplitude and QTc.DiscussionWe confirmed p-wave amplitude and duration as strongly and positively associated with AF. However, with addition of atrial volumes to the model, all p-wave measures lost significance. Our findings suggest the predictive value of p-wave measures for AF might be as surrogates for atrial structure, with left atrium measures being stronger intermediate markers compared with corresponding right atrium measures.In contrast, there is little precedent for the identification of J-point amplitude as a predictor of AF, which remained significant independent of imaging covariates. J-point amplitude only lost significance with the addition of NT-proBNP, a known serological predictor of AF. In vivo, NT-proBNP reflects the combined modulation of L-type calcium and potassium channel currents, altering action potential duration and conduction velocity, thereby increasing the susceptibility of the substrate to arrhythmia.5 Our findings suggest that J-point amplitude may be valuable for signaling such electrical derangements in the pathogenesis of AF, allowing risk stratification independent of structural or functional abnormalities.The association of QTc interval with AF has been reported in several large cohorts. Our study closely corroborates the direction and magnitude of association of QTc with AF identified in previous studies. This was a remarkably robust association, independent of well-known risk factors including comprehensive imaging measures and serum markers. QTc interval may reflect prognostic electrophysiological properties of the myocardium not captured by imaging or represented among the variables in popular models for AF risk stratification.ConclusionsIn this large, multi-ethnic population, we used machine learning to identify top ECG predictors of AF. We found the predictive value of p-wave amplitude and duration could be attributed to atrial structure as quantified by coincident imaging. However, J-point amplitude and QTc were associated with AF independent of any structural or functional imaging measures, suggesting they may uniquely represent electrophysiological properties of a vulnerable substrate. Further investigation into the biologic correlates of J-point amplitude and QTc interval may improve our understanding of the pathophysiology underlying incident AF.Article InformationSources of FundingThis research was supported by contracts 75N92020D00001, HHSN268201500003I, N01-HC-95159, 75N92020D00005, N01-HC-95160, 75N92020D00002, N01-HC-95161, 75N92020D00003, N01-HC-95162, 75N92020D00006, N01-HC-95163, 75N92020D00004, N01-HC-95164, 75N92020D00007, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168, and N01-HC-95169 from the National Heart, Lung, and Blood Institute, and by grants UL1-TR-000040, UL1-TR-001079, and UL1-TR-001420 from the National Center for Advancing Translational Sciences (NCATS).Supplemental MaterialSupplemental Tables S1–S5Supplemental FigureNonstandard Abbreviations and AcronymsAFatrial fibrillationMESAMulti-Ethnic Study of AtherosclerosisNT-proBNPN-terminal pro-B-type natriuretic peptideDisclosures None.FootnotesFor Sources of Funding and Disclosures, see page 1108.Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/CIRCEP.121.009805.Correspondence to: Bharath Ambale-Venkatesh, PhD, Division of Radiology, Johns Hopkins University, 600 N Wolfe St, Baltimore, MD 21287. Email [email protected]eduReferences1. Attia ZI, Noseworthy PA, Lopez-Jimenez F, Asirvatham SJ, Deshmukh AJ, Gersh BJ, Carter RE, Yao X, Rabinstein AA, Erickson BJ, et al. An artificial intelligence-enabled ECG algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of outcome prediction.Lancet. 2019; 394:861–867. doi: 10.1016/S0140-6736(19)31721-0CrossrefMedlineGoogle Scholar2. Edvardsen T, Rosen BD, Pan L, Jerosch-Herold M, Lai S, Hundley WG, Sinha S, Kronmal RA, Bluemke DA, Lima JA. Regional diastolic dysfunction in individuals with left ventricular hypertrophy measured by tagged magnetic resonance imaging–the Multi-Ethnic Study of Atherosclerosis (MESA).Am Heart J. 2006; 151:109–114. doi: 10.1016/j.ahj.2005.02.018CrossrefMedlineGoogle Scholar3. Jain A, Tandri H, Dalal D, Chahal H, Soliman EZ, Prineas RJ, Folsom AR, Lima JA, Bluemke DA. Diagnostic and prognostic utility of electrocardiography for left ventricular hypertrophy defined by magnetic resonance imaging in relationship to ethnicity: the Multi-Ethnic Study of Atherosclerosis (MESA).Am Heart J. 2010; 159:652–658. doi: 10.1016/j.ahj.2009.12.035CrossrefMedlineGoogle Scholar4. Ishwaran H, Kogalur UB, Blackstone EH, Lauer MS. Random survival forests.Ann Appl Stat. 2008; 2:841–860.CrossrefGoogle Scholar5. Moghtadaei M, Polina I, Rose RA. Electrophysiological effects of natriuretic peptides in the heart are mediated by multiple receptor subtypes.Prog Biophys Mol Biol. 2016; 120:37–49. doi: 10.1016/j.pbiomolbio.2015.12.001CrossrefMedlineGoogle Scholar Previous Back to top Next FiguresReferencesRelatedDetails December 2021Vol 14, Issue 12 Advertisement Article InformationMetrics © 2021 American Heart Association, Inc.https://doi.org/10.1161/CIRCEP.121.009805PMID: 34844442 Originally publishedNovember 30, 2021 Keywordsatherosclerosisethnic groupselectrocardiographyatrial fibrillationhumansPDF download Advertisement SubjectsAtrial FibrillationMagnetic Resonance Imaging (MRI)