Population Based Pharmacodynamics for In Vitro Drug Sensitivity Assays: Prediction of Model Based Parameters of Drug Activity and Relationship to Clinical Outcome
A. Quartino(1), M.O. Karlsson(1), A. Freijs(1), N. Jonsson(1), P. Nygren (3), J. Kristensen (2), E. Lindhagen (2) and R. Larsson (2)
(1)Department of Pharmaceutical Biosciences, Division of Pharmacokinetics and Drug Therapy, Uppsala University, Sweden; (2)Department of Medical Sciences, Division of Clinical Pharmacology, Uppsala University, Sweden; (3)Department of Oncology, Radiology and Clinical Immunology, Section of Oncology University Hospital, Uppsala, Sweden
Background: Due to limited availability of patient tumor cells, predictive in vitro drug sensitivity assays often measures a variable effect at a few fixed concentrations. The information on the drug concentration-response relationship is not obtained and individual estimate of drug potency is limited.
Purpose: To develop a population-based pharmacodynamic model for the in vitro drug sensitivity of leukemic cell samples from patients with acute myelocytic leukemia (AML) to predict individual pharmacodynamic parameters for cytosine arabinoside (AraC) and daunorubicin (Dnr). Further to relate in vitro parameters to clinical outcome.
Methods: 179 consecutive samples of tumor cells from patients with AML were analyzed, with respect to respons to increasing concentrations of the drugs, using the fluorometric microculture cytotoxicity assay (FMCA). The population pharmacodynamic model was developed using the FOCE method in NONMEM. 124 samples were assigned to the learning data set and the final model was evaluated using another 30 samples. Additional 25 samples was added to the data set for the analysis of clinical outcome. The probability of clinical response was estimated using a logistic regression on in vitro parameter estimates of 46 individuals actually treated with the AraC+Dnr combination.
Results: The data was best described by an Emax model for AraC and a sigmoid Emax model for Dnr. The population mean values for Emax and EC50 were 78% and 0.30 µg/ml, and 90% and 0.11 µg/ml for AraC and Dnr, respectively. The prediction interval (10-90%) of individual EC50 for AraC and Dnr was 0.098-0.82 µg/ml and 0.022-0.57 µg/ml, respectively. The correlation between EC50 for the two drugs was low (0.04) whereas that for Emax was high (0.72). EC50 could be predicted with adequate precision from only one concentration. For patients treated with the AraC+Dnr combination, the probability of complete response was significantly (p<0.05) related to the product of the ratio of Emax to EC50 of the two drugs.
Conclusion: A joint pharmacodynamic model for AraC and Dnr including covariances across drugs, could adequately describe the in vitro sensibility data. Even with sparse sensitivity measurements adequate information on drug potency can be obtained. The model for clinical outcome is mechanistically reasonable and supports the dual therapy.