Abstract
The usefulness of a test depends on its reproducibility. This determines how closely the test result indicates the actual pathophysiologic state, how well it will predict that state in the future, and if interventions or further pathologic changes are reflected by the test. There is a variation in the parameters of the signal-averaged ECG, more so with spectral than with time domain measurements. These must be accounted for when estimating risk. If one presumes that risk is proportional to the extent of abnormality, then the variation in measurements simply means that only borderline cases can potentially be miscategorized. More important, the lack of reproducibility of measurements made from the signal-averaged ECG indicates that changes noted in an individual after an intervention, such as a surgical intervention, must be viewed with a jaundiced eye. Group changes are perhaps meaningful, and indicate a physiologic effect, but clinical decisions cannot be made unless the changes observed in an individual patient exceed the confidence limits of expected variation. There has been debate as to the usefulness of measurements made from the signal-averaged ECG in predicting antiarrhythmic drug effects (the effect of drugs is discussed elsewhere in this symposium). Here an analogy must be made to the suppression of asymptomatic ventricular ectopy. First, we cannot make a statement that there has been a drug effect unless the parameter measured changes beyond the confidence limits of normal variation or reproducibility. Second, we cannot translate a change in a measurement into a change in risk for arrhythmic events without subjecting that hypothesized relationship to a long-term placebo-controlled clinical trial, albeit acute electrophysiologic trials correlating changes in the signal-averaged ECG to ventricular tachycardia induction provide some insight. And perhaps the relationship must be tested independently for each drug assessed. In the same regard, there is much excitement about the benefits of thrombolytic therapy, but when diagnosing benefit to the individual patient we have to remember the lack of reproducibility of the measurements and also keep in mind that an improved signal-averaged ECG cannot be translated into an improved prognosis without long-term controlled studies. In summarizing the variation and reproducibility of measurements made from the signal-averaged ECG we avoided providing more than a sense of the extent of variation expected because precise confidence intervals depend on the particular techniques used to make the measurements.(ABSTRACT TRUNCATED AT 400 WORDS)
Collapse