Hierarchical Nonlinear Models in Pharmacology of HIV

Harvard University (Sch Of Public Hlth), Boston MA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): In the treatment of HIV infection, as the focus shifts toward personalized treatment planning, there is an increased interest in using pharmacokinetic data for assessing optimal treatment options and drug dosage. Improved clinical pharmacology data is needed for the safe and effective use of drugs in children. However, important questions regarding the design and analysis of pharmacokinetic data are still wide open. Use of sparse sampling and population pharmacokinetics analysis using hierarchical models has been advocated. This research proposal focuses on investigating nonlinear hierarchical statistical models for pediatric pharmacokinetics data in HIV, with an emphasis on dealing with values below the limit of quantitation, measures of curvature and explained variability, model parameterization, objective prior distributions, and sparse sampling. A new, fast algorithm for estimation in the nonlinear mixed model in the presence of censored data will be investigated and dedicated PK software implemented. The algorithm is based on missing data and EM algorithm ideas. Definitions of model curvature and relative explained variability will be extended and studied in the nonlinear mixed model. The influence of model parameterization of non-linear PK models will be studied. The influence of specification of prior distribution of the parameters on Bayesian inference will be investigated via a sensitivity analysis, and practical guidelines for non-informative prior specification in pediatric PK will be provided.