Improving the reproducibility of QT dispersion measures

The low reproducibility of the QT dispersion (QTD) method is a major reason why it is not used in clinics. The purpose of this study was to develop QT dispersion parameters with better reproducibility and identification of patients with a high risk of ventricular arrhythmia or death.

Three institutions using different methods for measuring QT intervals provided QT databases, which included more than 3500 twelve-lead surface ECGs. The data represented low and high risk subjects from the following groups: the normal population EpiSet (survivors vs dead from cardiovascular causes), acute myocardial infarction patients AmiSet (survivors vs dead) and remote myocardial infarction patients ArrSet (with vs without a history of ventricular arrhythmia). The EpiSet, AmiSet, and the ArrSet contributed with N = 122, 0, and 110 ECGs for reproducibility analysis, and 3244, 446, and 100 ECGs for the analysis of prognostic accuracy. The prognostic accuracy was measured as the area under the Receiver Operator Curve. The QT intervals were divided into six QT pairs; the longest pair consisted of the longest and the shortest QT intervals etc. The QT dispersion trend (QTDT) was defined as the slope of the linear regression of the N longest QT pairs after estimation of missing QT intervals by interpolation of measured QT intervals. The QTMAD and the QTSTD methods were defined as twice the mean absolute deviation and the standard deviation of the N longest QT pairs. The reproducibility was improved by 27% and 19% in the EpiSet and the ArrSet relative to the reproducibility of QTD. The accuracy improved for the EpiSet and the ArrSet and was maintained for the AmiSet.

By using the three longest and the three shortest QT intervals in QTDT, QTMAD, or QTSTD, the reproducibility improved significantly while maintaining or improving the prognostic accuracy compared to QTD.