Drug-drug interaction prediction

A Bayesian meta-analysis approach

Lang Li, Menggang Yu, Raymond Chin, Aroonrut Lucksiri, David A. Flockhart, Stephen D. Hall

Research output: Contribution to journalArticle

18 Citations (Scopus)

Abstract

In drug-drug interaction (DDI) research, a two drug interaction is usually predicted by individual drug pharmacokinetics (PK). Although subject-specific drug concentration data from clinical PK studies on inhibitor/inducer or substrate's PK are not usually published, sample mean plasma drug concentrations and their standard deviations have been routinely reported. In this paper, an innovative DDI prediction method based on a three-level hierarchical Bayesian meta-analysis model is developed. The first level model is a study-specific sample mean model; the second level model is a random effect model connecting different PK studies; and all priors of PK parameters are specified in the third level model. A Monte Carlo Markov chain (MCMC) PK parameter estimation procedure is developed, and DDI prediction for a future study is conducted based on the PK models of two drugs and posterior distributions of the PK parameters. The performance of Bayesian meta-analysis in DDI prediction is demonstrated through a ketoconazolemidazolam example. The biases of DDI prediction are evaluated through statistical simulation studies. The DDI marker, ratio of area under the concentration curves, is predicted with little bias (less than 5 per cent), and its 90 per cent credible interval coverage rate is close to the nominal level. Sensitivity analysis is conducted to justify prior distribution selections.

Original languageEnglish
Pages (from-to)3700-3721
Number of pages22
JournalStatistics in Medicine
Volume26
Issue number20
DOIs
StatePublished - Sep 10 2007

Fingerprint

Bayes Theorem
Drug Interactions
Meta-Analysis
Drugs
Pharmacokinetics
Prediction
Interaction
Pharmaceutical Preparations
Sample mean
Markov Chains
Area Under Curve
Monte Carlo Markov Chain
Credible Interval
Statistical Simulation
Model
Random Effects Model
Model Analysis
Prior distribution

Keywords

  • Area under the concentration curve ratio (AUCR)
  • Bayesian model
  • Drug-drug interaction (DDI)
  • Meta-analysis
  • Pharmacokinetics (PK)
  • Prediction

ASJC Scopus subject areas

  • Epidemiology

Cite this

Li, L., Yu, M., Chin, R., Lucksiri, A., Flockhart, D. A., & Hall, S. D. (2007). Drug-drug interaction prediction: A Bayesian meta-analysis approach. Statistics in Medicine, 26(20), 3700-3721. https://doi.org/10.1002/sim.2837

Drug-drug interaction prediction : A Bayesian meta-analysis approach. / Li, Lang; Yu, Menggang; Chin, Raymond; Lucksiri, Aroonrut; Flockhart, David A.; Hall, Stephen D.

In: Statistics in Medicine, Vol. 26, No. 20, 10.09.2007, p. 3700-3721.

Research output: Contribution to journalArticle

Li, L, Yu, M, Chin, R, Lucksiri, A, Flockhart, DA & Hall, SD 2007, 'Drug-drug interaction prediction: A Bayesian meta-analysis approach', Statistics in Medicine, vol. 26, no. 20, pp. 3700-3721. https://doi.org/10.1002/sim.2837
Li L, Yu M, Chin R, Lucksiri A, Flockhart DA, Hall SD. Drug-drug interaction prediction: A Bayesian meta-analysis approach. Statistics in Medicine. 2007 Sep 10;26(20):3700-3721. https://doi.org/10.1002/sim.2837
Li, Lang ; Yu, Menggang ; Chin, Raymond ; Lucksiri, Aroonrut ; Flockhart, David A. ; Hall, Stephen D. / Drug-drug interaction prediction : A Bayesian meta-analysis approach. In: Statistics in Medicine. 2007 ; Vol. 26, No. 20. pp. 3700-3721.
@article{d7c7a0df58564b8abfb203d5f00edf78,
title = "Drug-drug interaction prediction: A Bayesian meta-analysis approach",
abstract = "In drug-drug interaction (DDI) research, a two drug interaction is usually predicted by individual drug pharmacokinetics (PK). Although subject-specific drug concentration data from clinical PK studies on inhibitor/inducer or substrate's PK are not usually published, sample mean plasma drug concentrations and their standard deviations have been routinely reported. In this paper, an innovative DDI prediction method based on a three-level hierarchical Bayesian meta-analysis model is developed. The first level model is a study-specific sample mean model; the second level model is a random effect model connecting different PK studies; and all priors of PK parameters are specified in the third level model. A Monte Carlo Markov chain (MCMC) PK parameter estimation procedure is developed, and DDI prediction for a future study is conducted based on the PK models of two drugs and posterior distributions of the PK parameters. The performance of Bayesian meta-analysis in DDI prediction is demonstrated through a ketoconazolemidazolam example. The biases of DDI prediction are evaluated through statistical simulation studies. The DDI marker, ratio of area under the concentration curves, is predicted with little bias (less than 5 per cent), and its 90 per cent credible interval coverage rate is close to the nominal level. Sensitivity analysis is conducted to justify prior distribution selections.",
keywords = "Area under the concentration curve ratio (AUCR), Bayesian model, Drug-drug interaction (DDI), Meta-analysis, Pharmacokinetics (PK), Prediction",
author = "Lang Li and Menggang Yu and Raymond Chin and Aroonrut Lucksiri and Flockhart, {David A.} and Hall, {Stephen D.}",
year = "2007",
month = "9",
day = "10",
doi = "10.1002/sim.2837",
language = "English",
volume = "26",
pages = "3700--3721",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "20",

}

TY - JOUR

T1 - Drug-drug interaction prediction

T2 - A Bayesian meta-analysis approach

AU - Li, Lang

AU - Yu, Menggang

AU - Chin, Raymond

AU - Lucksiri, Aroonrut

AU - Flockhart, David A.

AU - Hall, Stephen D.

PY - 2007/9/10

Y1 - 2007/9/10

N2 - In drug-drug interaction (DDI) research, a two drug interaction is usually predicted by individual drug pharmacokinetics (PK). Although subject-specific drug concentration data from clinical PK studies on inhibitor/inducer or substrate's PK are not usually published, sample mean plasma drug concentrations and their standard deviations have been routinely reported. In this paper, an innovative DDI prediction method based on a three-level hierarchical Bayesian meta-analysis model is developed. The first level model is a study-specific sample mean model; the second level model is a random effect model connecting different PK studies; and all priors of PK parameters are specified in the third level model. A Monte Carlo Markov chain (MCMC) PK parameter estimation procedure is developed, and DDI prediction for a future study is conducted based on the PK models of two drugs and posterior distributions of the PK parameters. The performance of Bayesian meta-analysis in DDI prediction is demonstrated through a ketoconazolemidazolam example. The biases of DDI prediction are evaluated through statistical simulation studies. The DDI marker, ratio of area under the concentration curves, is predicted with little bias (less than 5 per cent), and its 90 per cent credible interval coverage rate is close to the nominal level. Sensitivity analysis is conducted to justify prior distribution selections.

AB - In drug-drug interaction (DDI) research, a two drug interaction is usually predicted by individual drug pharmacokinetics (PK). Although subject-specific drug concentration data from clinical PK studies on inhibitor/inducer or substrate's PK are not usually published, sample mean plasma drug concentrations and their standard deviations have been routinely reported. In this paper, an innovative DDI prediction method based on a three-level hierarchical Bayesian meta-analysis model is developed. The first level model is a study-specific sample mean model; the second level model is a random effect model connecting different PK studies; and all priors of PK parameters are specified in the third level model. A Monte Carlo Markov chain (MCMC) PK parameter estimation procedure is developed, and DDI prediction for a future study is conducted based on the PK models of two drugs and posterior distributions of the PK parameters. The performance of Bayesian meta-analysis in DDI prediction is demonstrated through a ketoconazolemidazolam example. The biases of DDI prediction are evaluated through statistical simulation studies. The DDI marker, ratio of area under the concentration curves, is predicted with little bias (less than 5 per cent), and its 90 per cent credible interval coverage rate is close to the nominal level. Sensitivity analysis is conducted to justify prior distribution selections.

KW - Area under the concentration curve ratio (AUCR)

KW - Bayesian model

KW - Drug-drug interaction (DDI)

KW - Meta-analysis

KW - Pharmacokinetics (PK)

KW - Prediction

UR - http://www.scopus.com/inward/record.url?scp=34548149708&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34548149708&partnerID=8YFLogxK

U2 - 10.1002/sim.2837

DO - 10.1002/sim.2837

M3 - Article

VL - 26

SP - 3700

EP - 3721

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 20

ER -