Background: Prolonged QT dispersion (QTD) is associated with an increased risk of arrhythmic death but its accuracy varies substantially between otherwise similar studies. This study describes a new type of bias that can explain some of these differences.
Material: One dataset (DiaSet) consisted of 356 subjects: 169 with diabetes, 187 nondiabetic control persons. Another dataset (ArrSet) consisted of 110 subjects with remote myocardial infarction: 55 with no history of arrhythmia and 55 with a recent history of ventricular tachycardia or fibrillation.
Methods: 12-lead surface ECGs were recorded with an amplification of 10 mm/mV at a paper speed of 50 mm/s. The QT interval was measured manually by the tangent-method. The bias depends on the magnitude of the measurement errors and the measurable part of the bias increases with the number of the repeated measurements of QT.
Results: The measurable bias was significant for both datasets and decreased for increasing QTD in the DiaSet (P < 0.001) and in the ArrSet (P = 0.11). The bias was 2.5 ms and 1.9 ms at QTD = 38 ms and 68 ms, respectively, in the ArrSet, and 7.5 ms and 2.8 ms at QTD = 19 ms and 55 ms, respectively, in the DiaSet.
Conclusions: This study shows that random measurement errors of QT introduces a type of bias in QTD that decreases as the dispersion increases, thus reducing the separation between patients with low versus high dispersion. The bias can also explain some of the differences in the mean QTD between studies of healthy populations. Averaging QT over three successive beats reduces the bias efficiently.