Prior population pharmacokinetic (PPK) modeling approaches have found limited success in developing dynamic models that accurately describe the variability that is observed in pharmacokinetic (PK) behavior among individual patients within a population. Many patient specific characteristics, for example, age, weight, or gender, are thought to be the significant factors that cause such variability between individuals. The overall goal of this research is to formulate a PK modeling methodology that will enable one to relate the variability of the model parameters to the appropriate patient specific characteristics. These models will benefit the field of pharmaceutical therapy in several ways: 1) the ability to tailor the administration of therapeutic drugs for the characteristics of a specific patient, and 2) gain insight that will advance the knowledge of the physiological processes that influence patient specific PK behavior.

In order to develop such a methodology the modeling of an example case has been undertaken. Drug plasma concentration vs. time data sets collected for 61 individuals who were orally administered a 0.5 mg dose of the common anesthetic, Triazolam, were provided by Dr. David Greenblatt of the Tufts School of Medicine. The individual factors: gender, age, weight, and height, were recorded for each of the patients to investigate the contribution of such factors to the observed inter-individual PK variability for this single occasion study.

Four different compartmental absorption models, quantifying the physiological processes relevant to drug adsorption, distribution, and excretion kinetics, were used to represent the PK data of each individual patient. The set of four models includes one-compartment and two-compartment models with and without a lag. Parameter estimation, using an ordinary or weighted least-squares approach along with constrained nonlinear optimization, resulted in satisfactory representation of the majority of the patient data. The models with a lag showed better representations of the data, and the one–compartment lag model was selected for further examination because of its smaller number of parameters. Partial Least Squares and Stepwise Multi-linear regression of linear and nonlinear terms is used to investigate the functional relationship between the model parameters and patient specific factors.