Vladimir K. Piotrovsky, Achiel Van Peer
Human Pharmacokinetics, Janssen Research Foundation, Beerse, Belgium
Many drugs prolong QT intervals on electrocardiogram (ECG) that may cause serious problems, and may lead to ‘torsades de points’ and even sudden death. Assessment of cardiovascular safety (CVS) is thus an important step in drug development. There is an intrinsic dependence of measured QT intervals on the heart rate and RR intervals. Due to high inter- and intra-individual variabilities in ECG measurements CVS data analysis is a problematic issue.
The currently used methods of CVS data analysis do not account for these variabilities that may lead to under- or over-estimation of drug effects on QT intervals. An alternative approach is developed based on a mixed-effects model described below. At k-th occasion (study day) j-th measured QT interval in i-th individual, QTijk, is given by
QTijk = QTcbsl,ijk*RRijk^Pik*(1+eijk)
where RRijk is j-th RR interval in i-th individual at k-th occasion; QTcbsl,ijk is an individual baseline QT at RR equal to 1; Pik is an individual power parameter at k-th occasion; eijk is a residual error. RRijk^Pik accounts for the individual correction of QT for RR changes. QTcbsl,ijk is subject to diurnal variations and is affected by a drug:
QTcbsl,ijk = 1+DPij+E(Cijk)
where DPik is an individual occasion-specific diurnal profile; E(Cijk) is a drug effect submodel.
The diurnal profile was modelled using a periodic function which included 3 harmonic terms (cosine functions) with periods of 24, 12 and 6 h, respectively. Each harmonic had individual- and occasion-specific amplitude and phase parameters. Various effect submodels were tested including the sigmoidal Emax model. The approach was applied to model CVS data of two drugs and was demonstrated to work properly. The advantage of the approach is that it allows hypothesis testing (is the drug effect significant of not?) and provides point estimates for parameters reflecting the magnitude of the drug effect on QT.
Reference: PAGE 9 () Abstr 108 [www.page-meeting.org/?abstract=108]
Poster: poster