Comparative pharmacokinetic analysis by standard two-stage method versus nonparametric population modeling.

Study Objective: To compare the two-stage method, a widely used analytical method in pharmacokinetic studies, with nonparametric population modeling by using the same data set for determining the oral bioavailability of ribavirin.

Design: Pharmacokinetic analysis. Clinical research center.

Material: Oral bioavailability data of ribavirin determined previously in six healthy adults.

Intervention: After 13C3-ribavirin 150 mg intravenously and unlabeled ribavirin 400 mg orally had been given 1 hour apart, serial serum and urine samples were obtained for up to 169 hours. Concentrations of 13C3-ribavirin and unlabeled ribavirin in serum and urine were determined by a high-performance liquid chromatography tandem mass spectrometric method.

Measurements And Main Results: Serum and urine concentration-time profiles were comodeled with a three-compartment model. The analysis was performed again by using the nonparametric population analysis technique. Serum ribavirin concentrations underwent Monte Carlo simulation for 1000 subjects receiving a single 600-mg oral dose. Both methods were similar in determining the mean +/- SD bioavailability (51.8 +/- 21.8% by the two-stage method vs 54.8 +/- 16.4% by nonparametric modeling, p=0.79). However, the estimates of dispersion of model parameters and simulated drug exposures were substantially reduced by the population-modeling technique, as it takes into account covariance among model parameters and intersubject variability.

Conclusion: Although the study sample was small, our parallel analyses of the same data set clearly demonstrated that more precise parameter estimates are likely to result with the population-modeling technique. Having accurate and precise estimation of population pharmacokinetic parameters and their true variances is crucial, as, at any dose, there'will be a lower probability of encountering a concentration-driven toxicity because of fewer outliers as the variance associated with the parameters decreases.