Browsing Theses and Dissertations School of Pharmacy by Author "Knebel, William"
Factors that are important in determining the statistical significance of covariates during population pharmacokinetic analysesKnebel, William; Young, David G. (2000)Population pharmacokinetic analysis has become an integral part of the drug development process. Patient demographic and pathphysiologic characteristics, expressed as either dichotomous (yes/no) or continuous (weight, age, creatinine clearance) covariates, aid in population pharmacokinetic model development by helping to explain some of the variability that exists in the study population. Their inclusion is often used to determine groups in the population that may be more susceptible to adverse drug effects and to determine the proper dose and dosing regimen. There are number of graphical and statistical methods to aid in the detection of dichotomous and continuous covariates during population pharmacokinetic modeling. However, the effect of the range of a continuous covariate and the percentage of patients in a population analysis who are positive for a dichotomous covariate on the ability to detect the covariate during the population analysis have not been determined. This research was designed to investigate these issues. For dichotomous covariates, it was clear based upon the results of this research that a dichotomous covariate can be present in as little as 5% of the population and still be statistically significant in a population pharmacokinetic analysis as long as the effect of the covariate on the pharmacokinetic parameter is 20% or more and the interindividual variability was 10% or less. In addition, the ability to detect the covariate was highly dependent on the interindividual variability and the effect of the covariate on the pharmacokinetic parameter. In order to increase the ability to detect a dichotomous covariate, either the interindividual variability of the pharmacokinetic parameter being affected by the covariate must decrease or the number of subjects with the covariate must increase. For continuous covariates, the results of this research indicated that the magnitude of effect of a continuous covariate in population pharmacokinetic modeling should be 23% or more and the interindividual variability on the pharmacokinetic parameter affected by the covariate should be 10% or less for the covariate to be statistically significant. The ability to detect the continuous covariate can be increased by increasing the range of the covariate or by combining studies to increase the number of subjects in a population pharmacokinetic analysis. However, the latter approach of increasing the number of subjects may prove to be more beneficial with respect to increasing the ability to detect the continuous covariate. In general, the results of this research should be used to plan additional population pharmacokinetic studies in the event of the failure of the initial study or to investigate why a covariate was not detected in a population pharmacokinetic analysis. Guidelines of how to use these results for these purposes are given.