Session S24.5

A Method for Assessing Significant Changes in Serial ECG Comparison

S Perz*, MF Sinner, R Küfner, S Pfeufer, S Kääb

Helmholtz Zentrum München
Neuherberg, Germany

Conventional computerized ECG interpretation is primarily focused on the analyses of single ECG records.In order to provide criteria on a multivariate scale for intra-individual ECG changes we used the approach that extreme changes of measurements in repeated ECG records are associated with the development of, or recovery from pathophysiologic changes (drug effects, electrolyte changes, cardiac disease).
The criteria for significant changes were derived from standardized twelve lead 10 seconds ECG records based on (1) the KORA S4 study as baseline data, and (2) the KMC study where 860 men and women aged 25-74 years were re-examined two years after the initial S4 study. Each ECG was characterized by 96 quantitative measurements (time intervals, amplitudes, and angles) determined by the Hannover ECG System. Measurement changes exceeding predefined percentiles - 99% as upper limits and 1% as lower limits (P1/P99) of the distributions of measurement differences - were considered significant. Adjustment for sensitivity was performed (a) using additional thresholds (P2.5/P97.5, P5/P95), and (b) different numbers of measurements (1, 2 or 5) exceeding the thresholds. The results of the intra-individual ECG comparison were compared with the disease status derived from a standardized interview.
The application of the P1/P99 thresholds resulted in prevalence estimates of significant ECG changes of 2.8% in the sample if at least five measurements exceeded the thresholds, of 10.7%, if at least two (nm=2), and of 23.1% if at least one measurement exceeded the thresholds. Significant changes were more frequent in males than in females (13.5% vs. 7.3%, p<0.01 if nm=2) and more frequent (12.6% vs. 5.8%, p<0.05 if nm=2) in older individuals (age =60 years). Using P2.5/P97.5 and P5/P95 thresholds increased the prevalence of significant serial ECG changes considerably. The ECG changes derived from the repeated records were significantly (p<0.001) associated with the individuals’ history of cardiac disease.
Beyond the scope of clinically established criteria, our approach to derive criteria from ECG analyses in population-based follow-up studies provides additional information for serial ECG analyses. The criteria can easily be implemented in current ECG analysis systems, if algorithms for precise ECG measurement are available.

(Abstract Control Number: 68)