Developing patient specific population pharmacokinetic models and comparing compartmental vs. non-compartmental analysis for the prediction of individual pharmacokinetic parameters

One of the most challenging aspects facing clinical pharmacology is the optimization of patient dosage so that the desired therapeutic effect is maximized and toxic effects related to miss-dosage are minimized. The field of pharmacokinetics (PK) has generated several approaches towards overcoming this challenge such as the use of compartmental and non-compartmental analyses for estimating individual PK parameters that are used for determining dosage guidelines. Between these analyses the challenge of optimizing patient dosage is then translated towards a challenge in parameter estimation and model specification. Patient specific characteristics, for example, age, weight, or gender, are thought to be significant factors that may contribute to between subject PK variability. Several approaches have been developed in the field of population pharmacokinetics (PPK) to account for the relationship between such patient specific characteristics and PK parameters in hopes of explaining the observed variability among the target population. The two overall goals of this research study were: (1) Compare the techniques of compartmental versus non-compartmental approaches towards estimating individual PK parameters prior to their use in a population analysis. (2) Develop a PPK modeling methodology that will reduce the observed variability among PK parameter estimates by accounting for differences in patient factors which in turn will enable the optimization of patient dosage on an individual basis.;The approach that was undertaken for this study is related to that of the two-stage approach. In the first stage individual drug concentration time profiles are analyzed with the compartmental and non-compartmental analyses for the estimation of individual PK parameters. In the second stage we test two approaches as our population analyses that investigate a functional relationship between the estimated PK parameters and available patient specific characteristics. The first population analysis utilizes a principal component analysis of an expanded matrix of nonlinear patient specific characteristic terms and subsequent multiple linear regression. The second approach utilizes a multi-linear regression of logarithmic data.;Drug plasma concentration vs. time data sets collected for 61 individuals who were orally administered a 0.25 mg dose of the common hypnotic agent, Triazolam, were provided by our collaborator at the Tufts School of Medicine (Dr. David Greenblatt) in order to test our methodology. The compartmental and Non-compartmental analyses were observed to yield very similar results. No statistically significant reduction in the percent error of prediction was found in either of the population analyses. The results of the second population were simulated and gave no observable decrease in variability when compared with raw data.