A maximum likelihood approach for estimating the QT correction factor using mixed effects model.

Abstract

Assessment of QT interval prolongation is often used for assessing the cardiac safety of a new drug. However, the correction of the QT interval for varying heart rates has potential bias due to various different correction factors. This article proposes a maximum likelihood (ML) approach for calculating the appropriate individual correction factor using the data. The data come from a study with 24 subjects participating in a 10 day multiple dose (NEW RX) placebo-controlled cross-over trial with repeat ECGs obtained at baseline and at day 10. ML techniques were used to fit a random-effects model to observed QT and HR values for estimating the pooled and individual correction factors. QT(c) values using four correction factors (Bazett, Friderecia, pooled and individual) were investigated. The relative performance of the various correction factors are given in terms of variability and graphical techniques. The pooled correction factor was estimated to be 0.292 and the individual correction factors ranged from 0.19 to 0.41. The assessment of the treatment effect on QT(c) yielded inconsistent results. Bazett's factor indicated prolongation (6.55+/-1.20), Friderecia's factor indicated no change, while the pooled (-2.92+/-0.94) and individual (-2.82+/-1.00) factors showed a significant decrease. Graphical examination of individual QT(c) data showed a significant advantage in the use of individual correction factors versus Bazett's factor both in terms of sensitivity as well as reduction in bias. Use of individual correction factors is advocated for the assessment of possible drug-induced QT(c) prolongation.