II-20

Comparison of Analysis Methods for Population Exposure-Response in Thorough QT Studies

Parviz Ghahramani (1), Tatiana Khariton (1), Timothy J Carrothers (1)

(1) Forest Research Institute, Jersey City, NJ, USA

Objectives: To compare two approaches for evaluating the relationship between drug concentrations and QTc interval.

Methods: Data from two dedicated thorough QT (TQT) studies were used to compare two common approaches [1,2] used in evaluating concentration-QTc relationships for both escitalopram (ESC, n=107 subjects) and citalopram (CIT, n=119 subjects). The first approach (delta-delta) specified baseline-and-placebo-adjusted QTc as the dependent variable and considered linear mixed-effects models based on observed drug plasma concentrations, with or without log-transformation. The second approach (covariate model) utilized nonlinear mixed-effects models with QTc as the dependent variable, and with terms for baseline, placebo effect, and drug effect (used an Emax model, with intersubject variability on baseline and Emax as a function of individual predicted concentrations). Comparison of the two approaches was made via simulation of the QTc change at steady state Cmax.

Results: For both studies, the delta-delta analysis approach yielded models with population median intercepts that excluded zero (e.g., -19.6 ms for CIT), which would be physiologically unlikely (i.e., indicate QTc shortening at low concentrations). For the CIT study, both approaches gave similar mean predictions of drug-related prolongation, but with different confidence intervals:

CIT Dose

Mean QTc Prolongation (ms, 90% CI)

Delta-delta

Covariate model

20 mg

7.8 (6.2, 9.3)

8.2 (6.5, 10.4)

40 mg

12.6 (10.9, 14.3)

12.5 (10.1, 15.6)

60 mg

16.0 (14.0, 18.1)

15.5 (12.2, 19.7)

For the ESC study, the two approaches gave different predictions (mean and 90 CI) of drug-related QTc prolongation. At 20 mg dose, the delta-delta approach estimated a mean (90% CI) QTc change of 6.6 ms (5.3, 7.9), but the covariate approach estimated 8.3 ms (7.3, 9.2).

Conclusions: Modeling QTc directly using the covariate approach may provide a more physiologically relevant model for predictions across the full concentration range. On the other hand, the delta-delta approach may be computationally less intensive. While both approaches examined above avoid the multiple comparison bias and inefficiency of the Intersection Union Test commonly used in E14-standard TQT studies [3], they may provide different results. Additional validation work is recommended in comparing the relative merits of each C-QT approach.

References:
[1] Garnett CE et al. J Clin Pharm 2008;48:13-18.
[2] Piotrovsky V. AAPSJ 2005 ;7:E609-E624.
[3] Hutmacher MM et al. J Clin Pharm 2008;48:215-224.

Reference: PAGE 22 () Abstr 2994 [www.page-meeting.org/?abstract=2994]

Poster: Safety (e.g. QT prolongation)

PDF poster / presentation (click to open)