On Statistical Properties of QT Correction Methods

• Activity Number: 29
• Type: Contributed
• Date/Time: Sunday, August 8, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yibin Wang*+ and Guohua Pan and Alfred Balch
• Companies: Novartis Pharmaceuticals and Johnson & Johnson Pharmaceutical R&D, LLC and Novartis Pharmaceuticals
• Keywords: QT prolongation ; correction for HR ; statistical property ; QT-RR correlation

Abstract

There is an increasing regulatory emphasis on assessing drug-induced QT interval prolongation. Since QT interval is correlated with heart rate (HR), assessment of QT prolongation should be made at a standard HR, resulting in the need to correct QT interval (QTc) for HR. Numerous methods using fixed correction factors have been proposed; however, none can fully remove the QT-HR relationship when there is substantial variability among subjects or subject populations. This study investigates the statistical properties of QT correction methods that use individual-based (QTcI), population-based (QTcP), or fixed (QTcF) correction factors. It is found that, under both the linear and log-linear models for the QT-HR relationship, QTcP and QTcF are biased with VAR(QTcF) < VAR(QTcP) < VAR(QTcI). Furthermore, QTcI is unbiased under the linear model, but biased under the log-linear model. The Type I error may be inflated in an analysis using QTcP or QTcF (or QTc under the log-linear model) as the response. Therefore, HR should be included in such an analysis as a covariate to adjust for the remaining correlation of QTc with HR. This approach is equivalent to one-step analysis that corrects for HR by using the uncorrected QT interval as the response in a model while including HR as a covariate.