Pharmacodynamic modeling of prolonged administration of etoposide.
PURPOSE: A refined pharmacodynamic model for toxicity is necessary for successful adaptive control of the administration of an anticancer drug to avoid toxicity. We sought to establish a pharmacodynamic model of leukopenia in a 14-day administration of etoposide. METHODS: Pharmacokinetic data of 32 patients treated with etoposide infused over 14 days in a phase I study (20 patients) or in an adaptive control study (12 patients) were used to develop a model for the prediction of a leukocyte nadir count. The concentrations of both estimated unbound and total etoposide at steady state, as well as patient demographic factors, were included in linear and nonlinear models. The unbound fraction of etoposide was estimated using an equation based on serum albumin and total bilirubin. The efficacy of the models was evaluated in terms of correlation coefficient (r), mean predictive error (MPE) and root mean square error (RMSE). RESULTS: For both total and unbound drug concentration, a nonlinear model predicted leukopenia more precisely and with less bias than a linear model, and unbound drug explained more variability of leukopenia than total drug concentration in both linear and nonlinear models. The best model was a nonlinear model with three variables of unbound concentration, pretreatment leukocyte count and prior treatment (r = 0.76, MPE +/- SEM = 0.07 +/- 0.17 x 10(3)/microl, RMSE = 0.95 x 10(3) microl), which was better than the best linear model. CONCLUSIONS: The nonlinear model using unbound etoposide concentration explained the interpatient variability of leukocyte nadir count to a fairly large extent. Although the model provided useful information on the pharmacodynamics of etoposide, it was still imprecise and a more refined model is necessary for application to an adaptive control study.