Population Pharmacodynamic Model for Bayesian Prediction of Myelosuppression Profiles Based on Routine Clinical Data after Gemcitabine and Carboplatin Treatment.

BACKGROUND: Hematological toxicity is a serious adverse event and is often a dose-limiting factor in anticancer drugs. The objective of the present study was to develop a modeling and simulation (M&S) procedure for predictions of time course profiles of blood cell counts that reflect myelosuppression profiles.

METHODS: A method for Bayesian prediction of myelosuppression profiles during chemotherapy using a population pharmacodynamic model is proposed, and predictabilities of nadir values and times to nadir (Tnadir) after gemcitabine and carboplatin treatment were evaluated. The model is based on an equation for Erlang distribution, which we previously proposed, and it explains time course profiles of platelet (PLT), red blood cell (RBC) and white blood cell (WBC).

RESULTS AND CONCLUSION: PLT, RBC and WBC counts were retrospectively collected from 61 time courses (a total of 472 points) of 27 cancer patients. Predictive performance by a one-point Bayesian prediction was evaluated using data from day 8 in consideration of applicability to outpatients. Some good predictability was obtained for nadir values with some exceptions for PLT and RBC, whereas the predictability of Tnadir was insufficient. Although the predictability was not acceptable enough, this M&S approach could be used for supportive care during cancer chemotherapy.