Type
Text
Type
Dissertation
Advisor
zhu, wei | gao, yi | benveniste, helene. | wu, song
Date
2015-12-01
Keywords
Applied mathematics
Department
Department of Applied Mathematics and Statistics.
Language
en_US
Source
This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.
Identifier
http://hdl.handle.net/11401/77599
Publisher
The Graduate School, Stony Brook University: Stony Brook, NY.
Format
application/pdf
Abstract
QT interval is a measure of the time from the beginning of the Q wave to the end of the T wave in the heart's electrical cycle, it is often corrected to lessen its dependence to heart rate, and the corrected value is known as the QTc interval. Prolongation of the QTc interval is a primary biomarker for assessing proarrhythmic risk of drugs, thus it is routinely evaluated in new drug development. Concentration-QTc model with data from early phase single ascending dose (SAD) study is widely used to evaluate the drug-induced QTc prolongation, and the mixed effects model (MEM) is popular for this purpose. But some statistical issues in existing concentration-QTc model are rarely addressed, such as the limitation of the conventional sampling strategy in SAD study design, the baseline adjustment in crossover SAD study, and the violation of covariance matrix structure when applying the maximum likelihood (ML) method to analyzing the MEM. Several statistical issues in concentration-QTc model are addressed in this work. First, multi-oscillator function chosen by some model fitting criteria, such as AIC, is widely used to account for circadian rhythm effect in concentration-QTc model. An evaluation of the limitations of the conventional SAD design sampling strategy to recover the true underlying oscillator function is given. Subsequently, we propose a modified sampling strategy to improve the ability to recover the true underlying function. The superiority of this new sampling strategy is examined and confirmed via simulation studies. Secondly, in a crossover SAD study with period-specific pre-dose QTc baseline, the most efficient way to adjust pre-dose baseline in concentration-QTc model is unclear. We propose a novel conditional ΔQTc model by incorporating the pre-dose baseline and the pre-dose-averaged baseline as covariates in this model. We demonstrate the advantage of the proposed conditional ΔQTc model comparing to existing models, both analytically and through simulation studies. Finally, given that the MEM with ML method is dominant in concentration-QTc model, we proposed several alternative modeling and analysis approaches. We first adopted the Bayesian method and compared which to the Frequentist ML method. Furthermore, since the ML method for MEM requires the correctly specified covariance matrix structure which is usually hard to verify, we examined the generalized estimating equation (GEE) and the generalized method of moments (GMM) as more robust alternatives and comparisons between these and the ML method are made. | QT interval is a measure of the time from the beginning of the Q wave to the end of the T wave in the heart's electrical cycle, it is often corrected to lessen its dependence to heart rate, and the corrected value is known as the QTc interval. Prolongation of the QTc interval is a primary biomarker for assessing proarrhythmic risk of drugs, thus it is routinely evaluated in new drug development. Concentration-QTc model with data from early phase single ascending dose (SAD) study is widely used to evaluate the drug-induced QTc prolongation, and the mixed effects model (MEM) is popular for this purpose. But some statistical issues in existing concentration-QTc model are rarely addressed, such as the limitation of the conventional sampling strategy in SAD study design, the baseline adjustment in crossover SAD study, and the violation of covariance matrix structure when applying the maximum likelihood (ML) method to analyzing the MEM. Several statistical issues in concentration-QTc model are addressed in this work. First, multi-oscillator function chosen by some model fitting criteria, such as AIC, is widely used to account for circadian rhythm effect in concentration-QTc model. An evaluation of the limitations of the conventional SAD design sampling strategy to recover the true underlying oscillator function is given. Subsequently, we propose a modified sampling strategy to improve the ability to recover the true underlying function. The superiority of this new sampling strategy is examined and confirmed via simulation studies. Secondly, in a crossover SAD study with period-specific pre-dose QTc baseline, the most efficient way to adjust pre-dose baseline in concentration-QTc model is unclear. We propose a novel conditional ΔQTc model by incorporating the pre-dose baseline and the pre-dose-averaged baseline as covariates in this model. We demonstrate the advantage of the proposed conditional ΔQTc model comparing to existing models, both analytically and through simulation studies. Finally, given that the MEM with ML method is dominant in concentration-QTc model, we proposed several alternative modeling and analysis approaches. We first adopted the Bayesian method and compared which to the Frequentist ML method. Furthermore, since the ML method for MEM requires the correctly specified covariance matrix structure which is usually hard to verify, we examined the generalized estimating equation (GEE) and the generalized method of moments (GMM) as more robust alternatives and comparisons between these and the ML method are made. | 109 pages
Recommended Citation
Feng, Tian, "An empirical study on concentration-QTc model" (2015). Stony Brook Theses and Dissertations Collection, 2006-2020 (closed to submissions). 3398.
https://commons.library.stonybrook.edu/stony-brook-theses-and-dissertations-collection/3398