Development of a schedule-dependent population pharmacodynamic model for rhizoxin without quantitation of plasma concentrations

In previous phase I reports of short bolus infusion of rhizoxin, problems i n assay sensitivity prevented the description of pharmacokinetic-pharmacody namic relationships, and a pharmacologically guided approach to dose escala tion was deemed not feasible. In this report, we describe a mathematical mo del, which explains the schedule-dependent interpatient pharmacodynamic var iability of rhizoxin administered on a continuous infusion schedule. Using patient demographic and toxicity data from 45 patients treated in a phase I dose and duration escalation study of rhizoxin, we sought to model the nad ir neutrophil count. We hypothesized that a surrogate derived variable base d on dose and duration would reflect a pharmacokinetic parameter that would be a significant covariate. Multiple linear regression analysis was carrie d out to determine the other significant covariates, Dose/m(2) x Log-DUR/AL B was significantly correlated with the LogANC(nadir) (Log(10) neutrophil n adir; I = 0.56, P < 0.001). Other significant covariates included baseline performance status (PS), baseline serum bilirubin (BIL), and Log(10) baseli ne neutrophil count (LogANC(baseline)). Model bias and precision were asses sed using the mean prediction error (MPE) and the root mean square error (R MSE) of the ANC(nadir), respectively. We constructed 1-4 covariate models. The variability of ANC(nadir) was modeled with good precision and accuracy with a 4-covariate model (MPE and RMSE 0.113 +/- 0.182 x 10(3) cells/mu l a nd 1.22 x 10(3) cells/mu l, respectively). This model should be validated a nd improved on with further clinical data. We believe that such pharmacodyn amic modeling should be explored further to determine its performance and c linical relevance compared with modeling using pharmacokinetic parameters.