Analysis of animal pharmacokinetic data: Performance of the one point per animal design

Abstract

A simulation study was carried out to determine the impact of various design factors on the accuracy and precision with which population pharmacokinetic parameters are estimated in preclinical pharmacokinetic studies. A drug given by intravenous bolus injection and having mono-exponential disposition characteristics was assumed. The factors investigated were (i) number of animals sampled at specified times with one observation taken per animal, (ii) error in observed concentration measurements, and (iii) doubling the number of observations per animal while varying the number of animals. Data were analyzed with the NONMEM program, and the least number of animals per time point (where each animal supplied one concentration-time point) required for accurate and precise parameter estimation was determined. The one observation per animal design yielded biased and imprecise estimates of variability, and residual variability could not be estimated. Increasing the error in the concentration measurement led to a significant deterioration in the accuracy and precision with which variability was estimated. Obtaining a second sample from each animal practically eliminated bias and facilitated the partitioning of interanimal variability and residual intraanimal variability, by introducing information about the latter. Doubling the total number of observations per animal required using half (i.e., 50) the total number of animals required for accurate and precise parameter estimation with the one sample per animal design.

Pharmacokinetic-Pharmacodynamic Modeling of Methylxanthine Derivatives in Mice Challenged with High-Dose Lipopolysaccharide

“…However, the application of NONMEM V to the destructive sampling data and relatively complex models utilized in this study led to inaccurate estimates of the fixed effect pharmacodynamic parameters and highly imprecise estimates of interindividual variability in both pharmacokinetic and pharmacodynamic parameters. The latter is in agreement with the results of the simulation study using a simple 1-compartment pharmacokinetic model with intravenous bolus administration [33] .…”

Effect of P-Glycoprotein on the Pharmacokinetics and Tissue Distribution of Enaminone Anticonvulsants: Analysis by Population and Physiological Approaches

“…The destructive sampling approach of the study allowed for the analysis of enaminone concentrations in a variety of tissues but was limiting in that it only provided one sample on the concentration-time profile for each animal. Therefore, a population pharmacokinetic analysis approach utilizing WinNonMix was used for statistical comparison of the pharmacokinetic parameters between both groups of mice (Ette et al, 1994(Ette et al, , 1995Carapetis et al, 2001). Table 1 displays the chronology of the modeling procedure used to characterize the population pharmacokinetics for DM5 in both wild-type and knockout mice.…”

“…To evaluate variability and statistically determine whether the pharmacokinetics of the enaminone analogs were significantly altered by Pgp, population pharmacokinetic analysis using WinNonMix version 1.0 (Pharsight Corp.) was used to determine the inter-animal variability of the pharmacokinetic parameters. (Ette et al, 1994(Ette et al, , 1995 The first-order estimation method was used to obtain population parameter estimates. A two-stage analysis (naive pooled data method) with the individual pharmacokinetic data was performed initially, and it was determined that a one-compartment model best described the disposition of both DM5 and DM44 after a single i.v.…”

A survey of population analysis methods and software for complex pharmacokinetic and pharmacodynamic models with examples

“…where: D = intravenous bolus dose (500 units); V c , k 10 , k 12 , and k 21 are PK parameters to a 2-compartment model; k eo is the link effect parameter; E max , C e50 , and g are PD parameters to the Emax model; C(t) is the concentration of the drug in plasma; C e (t) is the concentration of drug in the effect compartment; E(t) is the effect as it changes with time. For each of 500 subjects, data were simulated at 3 or 4 PK data points, and 3 or 4 PD data points, with 1 of 3 combinations of time points: (0.03, 1, 20, 65), (0.1, 3, 35), or (0.3, 10, 50) time units, and the same time points were used for both PK and PD sampling in each subject.…”