Statistical Considerations in Pharmacokinetic Study Design

Summary

Pharmacokinetic studies may generally be categorised into 3 types: (a) population-based investigations, (b) individual-based compartmental, or (c) individual-based noncompartmental research projects. Each type of study has advantages and limitations.

Population-based investigations pool drug concentrations across more than 1 individual subject. From these data, estimates of pharmacokinetic parameters are calculated. NONMEM is the only computer program available to evaluate this type of information. Recently a method has been proposed which derives individual estimates from the information available from NONMEM. By combining these 2 procedures it is possible for the clinician to review and adjust the dosage regimen if necessary.

Population-based studies require fewer design criteria than other methods and are adaptable to the clinical setting, i.e. subjects can be patients currently being treated with the drug under consideration. One distinct advantage to this type of study is the flexibility of sampling times and the capability of the clinician to use information from the critically ill, the geriatric patient or the very young child. These subjects would not be available for the individual-based type of study because of the relatively large number of samples needed.

Individual-based pharmacokinetic studies can be divided into 2 types with respect to their evaluation: (a) compartmental and (b) noncompartmental investigations. The latter type of study was originally thought to require fewer assumptions than the former but subsequently it has been shown that noncompartmental analyses are more restrictive and are basically compartmental in their approach. These studies estimate parameters which the compartmental investigation does not usually consider. These include area under the moment curve (AUMC) and mean residence time (MRT).

The individual-based compartmental approach to pharmacokinetics is best carried out in the laboratory setting. Estimates of the pharmacokinetic parameters are calculated for each individual in the study. This requires obtaining many samples from each subject and hence is not suitable for the critically ill patient.

Hypothesis testing and interval estimation are predicated on the underlying sampling distributions of the estimates. Without information regarding these distributions the investigator is advised to use the appropriate nonparametric statistical procedures.