Population Pharmacokinetic Models: Measures of Central Tendency

Warren F. Dodge, Roger W. Jelliffe, C. Joan Richardson, Renee A. Bellanger, James A. Hokanson, Wayne R. Snodgrass

Research output: Contribution to journalArticlepeer-review

17 Scopus citations


The availability of personal computer programs for individualising drug dosage regimens has stimulated interest in modelling population pharmacokinetics. Appropriate use of population models requires knowledge of the distribution of their pharmacokinetic parameter values. If non-Gaussian, it is imperative not to assume that the arithmetic mean is the best measure of central tendency. This has important consequences for therapeutic drug monitoring and for individualising drug dosage regimens. Utilising a new nonparametric expectation maximisation (NPEM) population modelling program and retrospective data from 129 preterm infants who received gentamicin in our own clinical setting, we developed 2 population models. The NPEM algorithm showed that both had significant non-Gaussian distributions of their parameter values (e.g. the mean value for the volume of distribution exceeded the 75th percentile). Although the median parameter values demonstrated only slightly better predictive accuracy than the mean, its choice as the measure of central tendency was associated with excellent agreement between the goal and the subsequently observed average initial peak serum gentamicin concentrations (i.e. 6.0 vs 6.2 mg/L). In contrast, simulation using the mean parameter values predicted an average concentration 30% greater than the goal (i.e. 8.0 vs 6.0 mg/L). Thus, for this group of patients, the median proved a better measure of central tendency.

Original languageEnglish (US)
Pages (from-to)206-211
Number of pages6
JournalDrug Investigation
Issue number4
StatePublished - Apr 1993

ASJC Scopus subject areas

  • Pharmacology (medical)


Dive into the research topics of 'Population Pharmacokinetic Models: Measures of Central Tendency'. Together they form a unique fingerprint.

Cite this