Background: The objectives were to investigate the factors influencing signal-averaged ECGs (SAECGs) recorded in patients after myocardial infarction (MI) and to develop criteria for predicting arrhythmic events (AEs) that account for these factors.
Methods and results: SAECGs were recorded 5 to 15 days after MI in 2461 patients without bundle-branch block. The duration (QRSd), terminal potential (VRMS), and terminal duration (LAS) of the filtered QRS were measured. During follow-up (17 +/- 8 months), AEs (arrhythmic death; ventricular tachycardia, VT; ventricular fibrillation, VF) occurred in 80 patients (3.3%). Receiver operating characteristic curves showed that QRSd discriminated patients with all types of AEs, but VRMS and LAS discriminated only VT patients; QRSd minus LAS also discriminated AE patients. Sex, age, and MI location significantly affected the SAECG; survivors without VT or VF were divided into subgroups (2 sex x 4 age x 2 MI), and QRSd values exceeding the 70th percentile in each subgroup predicted AEs with a sensitivity of 65.4%. An unadjusted QRSd criterion showed the same overall sensitivity and specificity but with less uniform values for each subgroup. A Cox model was constructed by use of multiple prognostic indicators, and in rank order, QRSd, previous MI, and Killip class were predictive of AEs.
Conclusions: SAECG adjustments for sex, age, and MI location did not improve sensitivity and specificity but produced a more uniform predictive performance. The proposed criteria are based only on QRSd, because late potentials (VRMS and LAS) did not discriminate patients with sudden death. Duration of high-level activity during QRS (QRSd-LAS) can predict AEs, suggesting that the arrhythmogenic substate involves a large mass of myocardium.