Magnitude of error introduced by application of heart rate correction formulas to the canine QT interval

Abstract

Background: Accurate detection of drug-induced QT interval changes is often confounded by concurrent heart rate changes. Application of heart rate correction formulas has been the traditional approach to account for heart rate-induced QT interval changes, and thereby identify the direct effect of the test article on cardiac repolarization. Despite numerous recent studies identifying the imprecision of these formulas they continue to be applied.

Methods: Using a chronic atrioventricular dissociated His-paced canine model, heart rate correction methods were evaluated for their ability to generate a corrected QT interval independent of original heart rate. Additionally, His bundle pacing at a heart rate of 60 beats/min allowed calculation of the magnitude of error introduced by application of heart rate correction formulas.

Results: Of the fixed parameter heart rate correction formulas, only Van de Water was able to predict corrected QT values independent of the original heart rate. The magnitude of error discovered by application of heart rate correction formulas varied, but in many cases was very large. Bazett's formula was associated with a mean overcorrection of 67.9 ms; Fridericia's 28.7 ms. Van de Water was the best fixed parameter formula with a mean error of 10.8 ms. As expected, group and individual corrections derived from linear regression of the HR-QT data offered improvement over the traditional formulas. Both were able to predict QTc values independent of the heart rate. However, errors of the magnitude of 10 and 6 ms, respectively, were still introduced.

Conclusion: Van de Water and linear regression correction methods were superior to others in this study, but all methods generated QTc errors equal to or much greater than the magnitude of interest for drug safety evaluation.