The Imprecision in Heart Rate Correction May Lead to Artificial Observations of Drug Induced QT Interval Changes

Because of the known limitations of the Bazett and other heart rate correction formulas, it has been proposed that studies of drug induced QT interval changes should use several different heart rate correction formulas and that the consistency of findings by a majority of such formulas should be considered as valid. The aim of this article was to show that such an approach is inappropriate. Using the database of the EMIAT trial, data of QT and RR intervals were taken from electrocardiograms of the first postrandomization visit of 1,402 patients. Of these, 309 were on amiodarone and beta-blockers, 395 on amiodarone and off beta-blockers, 318 on beta-blockers and off amiodarone, and 380 off amiodarone and off beta-blockers. An investigation of drug induced QT interval changes was modeled by evaluating the corrected QT (QTc) interval differences between patients on and off amiodarone, and on and off beta-blockers. A set of 31 previously published heart rate correction formulas was used. In addition to calculating the QTc difference between on and off drug for each formula, the success of heart rate correction was judged by computing correlation coefficients between QTc and RR intervals (ideally corrected QTc values should be independent of heart rate). The difference between on and off drug QT intervals was also evaluated by logarithmic regression models between uncorrected QT and RR intervals in data taken from patients on and off treatment. The QTc interval prolongation on amiodarone was confirmed by all heart rate correction formulas but the extent of the prolongation differed from formula to formula and ranged from 13.6 to 30.9 ms. Of the 31 formulas, 3 reported QTc interval shortening on beta-blockers (up to -11.8 ms) and 28 reported QTc interval prolongation (up to +16.8 ms). The distribution of the results provided by the different formulas suggested that beta-blocker treatment led to a QTc interval prolongation by approximately 7 ms (e.g., +7.4 ms by the Fridericia formula, P = 0.002). The on treatment QTc changes obtained by different formulas were closely correlated to their correction success. Formulas that provided QTc intervals almost independent of the RR intervals estimated approximately 20 ms QTc prolongation on amiodarone and no QTc change on beta-blockers. QT/RR regression analysis confirmed that while amiodarone led to substantial QT prolongation, there was no change of QT interval on beta-blockers beyond the change in heart rate. The study showed that the concept of "majority voting" by different heart rate correction formulas is inappropriate and may lead to erroneous conclusions.