Prediction of Cardiac Death in Patients with Bundle Branch Block After Myocardial Infarction zy Gilberto Sierra, Ph.D.*, Patrick Morel, M.ScA.*, Jean-Lucien Rouleau, M.D.?, John Ferguson, M.D.$, Richard F. Davies, M.D., Ph.D.9, Duncan J. Stewart, M.D.11, Mario Talajic, M.D.?, Martin Gardner, M.D.#; Robert Dupuis, M,D.**, Claude Lauzon, M.D.**, Bruce Sussex, M.D.??, Wayne Warnica, M.D.43, Pierre Le Guyader, M.ScA.*, Reginald Nadeau, M.D.*, and Pierre Savard, Ph.D.* From the *Centre zyxwvutsrq de Recherche, H6pital du Sacre'-Coeur, Montrkal, Que'bec, Canada; TDe'partement zyxwvuts de Mkdecine, lnstitut de Cardiologie de Montrial, Montriaal, Canada; #Memphis Vascular Foundation, Memphis zyxwvuts , Tennessee , USA; §Department of Medicine, Ottawa Heart Institute, Ottawa, Ontario, Canada; IlDepartment of Medicine, Royal Victoria Hospital, Montreal, Canada; #Department of Medicine, Victoria General Hospital, Nova-Scotia, Canada; **Dipartement de Midecine , Centre Hospitalier Rigionul de L'Amiunte, Thetford Mines, Que'bec, Canada; ??Department of Medicine , Health Science Center, St-John's, Newfoundland, Canada; ##Department of Medicine, Foothills Hospital, Calgary, Alberta, Canada zyxw Background: Left (LBBB) and right (RBBB) bundle branch block (BBB) patients have an increased incidence of cardiac death after myocardial infarction (MI). The purpose of this study was to assess the value of the signal-averaged electrocardiogram (SAECG) and other clinical variables for the prediction of cardiac death after MI in BBB patients. Methods: SAECGs were recorded 5-15 days after MI in 76 LBBB and 79 RBBB patients. The SAECG was analyzed in the time domain and the frequency domain (wavelet analysis in 7 frequency bands ranging from 0.05 to 250 Hz). Results: During follow-up (1 7 2 8 months), cardiac death occurred in 22 LBBB (28.9%) and 12 RBBB patients (1 5.2%). None of the SAECG time-domain variables were significantly different between patients with and without cardiac death. In LBBB patients, univariate analysis showed that one wavelet parameter in the 3.9- to 7.8- Hz frequency band (P = 0.008), inhospital recurrent MI (P = 0.002), left ventricular ejection fraction (LVEF) < 30% (P = 0.004), lack of percutaneous transluminal coronary angioplasty zyxwvut (P = 0.02), and history of angina (P = 0.029) were significantly different in cardiac death patients. In RBBB patients, only recurrent angina was significantly different (P = 0.025). In LBBB patients, the combination of recurrent MI zyxwv or LVEF < 30% displayed the best predictive values: sensitivity (85.7%), specificity (76.6%), positive (52.2%), negative (94.7%), and total (78.7%) predictive accuracies and risk ratio of 9.9. The effect of recurrent MI and LVEF < 30% remained after statistical adjustment by means of regression using Cox proportional hazards. Conclusions: High-risk MI patients can be identified by recurrent MI and LVEF < 30% in LBBB patients, and by recurrent angina in RBBB patients. SAECG did not demonstrate incremental bundle branch block; prediction of cardiac death; signal-averaged electrocardiogram; wavelet transform information for the purpose of risk stratification in BBB patients. A.N.E. 1999;4(2):184-194 zyx This study was financially supported through a University-Industry grant fiom the Medical Research Council of Canada and Bristol-Myers Squibb of Canada. The present study is an ancillary trial of CARE. Address for reprints: Gilberto Sierra, Ph. D., Centre de Recherche, H6pital du Sacre'-Coeur, 5400 Ouest, B o d . Gouin, Montre'al, Que'bec, Canada H4J zyxwvutsrq 1 C.5. Far: + +1-514-338-2694. E-mail: sierrag@crhsc.umontreal.ca 184