Looking for prognostic information in the ST-T segment--is it really worth it?

The West of Scotland Coronary Prevention Study (WOSCOPS) provided baseline electrocardiograms (ECGs) on 6,595 men without a previous myocardial infarction who were followed for a mean of 4.9 years during which time all events, cardiovascular or otherwise, were recorded. Half of the study group was treated with a lipid lowering drug while the other half was randomly assigned to placebo. This study cohort afforded the opportunity to look at ECG morphology as a marker of risk. All 12-lead ECGs in the study were processed by the Glasgow Program and automated Minnesota Coding was also undertaken. All computer outputs were reviewed to exclude errors due to technically unsatisfactory recordings. Multiple variables were studied. Univariate and multivariate logistic regression analyses were undertaken to determine those electrocardiographic and clinical parameters of predictive value with respect to the primary endpoint of fatal or non fatal myocardial infarction. Those ST-T variables with additional prognostic value in the multivariate analysis, which included the clinical parameters, were used to develop a risk score. Although many ECG measures were of prognostic value in a univariate analysis, only rate, frontal T axis and T+ amplitude in lead I were of significance in a multivariate analysis which included clinical data. With respect to QT dispersion, while it was shown that there was an increased risk for those with QT dispersion exceeding 44 ms (RR 1.38, CI 1.02 - 1.81 P = 0.034) the receiver operating characteristic curve was virtually a straight line. The risk equation also demonstrated that there was as much prognostic value in the use of age and smoking history alone as there was in ECG plus age combined. The conclusion drawn is that the prognostic value of ECG variables has to be considered carefully in the light of other available data.