Effects of supratherapeutic doses of ebastine and terfenadine on the QT interval.

Abstract

Sir, Drs Gillen et al. recently reported [1] that at supratherapeutic doses, ebastine caused increase in heart rate and that, dependent on the method used for heart rate correction of QT interval, it led either to small but statistically significant, or minute and nonsig­nificant prolongation of the QTc interval. The use of Bazett and/or Fridericia formulae for QT interval correction is well placed in clinical practice because the relative magnitude of the errors due to under-or over-correction of the QT interval is unlikely to lead to incorrect clinical decision. However, in order to investigate drug related QTc interval changes in the presence of drug-induced heart rate change, the appropriateness and precision of the method used for correcting the QT interval for heart rate is essential. If the QTc interval is under-or over-corrected, the heart rate changes are projected into the QTc interval data and the analysis of the study becomes potentially meaningless with the possibility of both false positive and false negative findings. There are ways of judging the success of heart rate correction. Perhaps the simplest test is to investigate the correlation coefficient between the RR and QTc interval in the electrocardiograms obtained in drug-free stage. While a correlation coefficient of 0 does not guarantee that the QTc and RR interval data are truly independent, a value different from 0 shows that the heart rate correction formula used to obtain the QTc values has not been successful in removing the dependency of QT interval on heart rate. Drs Gillen et al. rightly say that a large number of heart rate correction formulae have previously been proposed. However, as their multiplicity implies, none of these formulae has found truly universal acceptance. The reasons for such a lack of a universally acceptable heart rate correction have become understood only recently. It has been observed that the QT/RR interval relationship is both different in different subjects and stable in the same individual over time [2, 3]. The intersubject variability of the QT/RR relationship means that a heart rate correction formula that correctly works in one individual, that is, provides QTc values that are independent of heart rate, will not necessarily be the optimum in another individual. Thus, to obtain truly heart rate-independent QTc values, individual characteristics of the QT/RR relationship need to be taken into account. This effectively means that the individual QT/RR pattern needs to be translated into an individual heart rate correction formula [4]. It is not for the first time that the study reported by Drs Gillen et al. has been analysed and published. I had the possibility of analysing the very same data set of this study and I have previously published the results obtained when using the technology of individual heart rate corrections [5]. Briefly, in the drug-free data of the study published by Dr Gillen et al., the heart rate correction formula QTc = QT/RRα has been optimized for each study participant, obtaining different values of the coefficient α for each individual. In these analyses, I have observed that while the coefficient of Fridericia's formula α = 0.33 was within the range of individually optimized coefficients (it was still significantly overcorrecting for some and under-correcting for other subjects), the coefficient of Bazett's formula α = 0.5 was well outside this range (Figure 1). Figure 1 The drug-free RR/QT interval data of each subject of the study reported by Drs Gillen et al.[1] were used to calculate heart rate corrected QTc intervals using the formula QTc = QT/RRα, ranging the values of α from ... The analyses based on the individually optimized heart rate corrections showed that in the study reported by Dr Gillen et al. ebastine did not in fact cause any QTc interval prolongation. The changes of the QTc interval on placebo, ebastine 60 mg daily ebastine 100 mg daily, and terfenadine 360 mg daily were −2. 76 ± 5. 51 ms, −3. 15 ± 9. 17 ms, −2. 61 ± 9. 55 ms, and 12. 43 ± 15. 25 ms, respectively [5].