A limited sample model to predict area under the drug concentration curve for 17-(allylamino)-17-demethoxygeldanamycin and its active metabolite 17-(amino)-17-demethoxygeldanomycin.

摘要

PURPOSE: The Hsp90-directed anticancer agent 17-(allylamino)-17-demethoxygeldanamycin (17-AAG) is currently undergoing phase I and phase II clinical investigation. Our goal was to develop a simple limited sampling model (LSM) for AUC of 17-AAG and its active metabolite, 17-(amino)-17-demethoxygeldanomycin (17-AG) using drug concentrations from a few time points. METHODS: Pharmacokinetic data from 34 patients treated at 11 dose levels on a Mayo Clinic Cancer Center phase I clinical trial of 17-AAG was utilized. Blood samples were collected at 11 different time points, spanning 25 h. Graphical methods and correlations were used to assess functional forms and univariate relationships. Multivariate linear regression and bootstrap resampling were used to develop the LSM. RESULTS: Using log-transformed data, the two and three time point 17-AAG LSMs are log-AUC (17-AAG) = 0.869 + 0.653*(C(55min)) +0.469*(C(5h)) and log-AUC (17-AAG) = 2.449 + 0.400*(C(55min)) +0.441*(C(5h)) +0.142*(C(9h)). The two and three time point LSMs for 17-AG are log-AUC (17-AG) = 3.590 + 0.747*(C(5h)) +0.169*(C(17h)), and log-AUC (17-AG) = 3.797 + 0.650*(C(5h)) +0.111*(C(9h)) +0.122*(C(17h)). Ninety-seven percent and 94% of the predicted log-AUC values were within 5% of the observed log-AUC for the two and three time point models for 17-AAG and 17-AG respectively. CONCLUSIONS: The precise calculation of AUC is cumbersome and expensive in terms of patient and clinical resources. The LSM developed using a multivariate regression approach is clinically and statistically meaningful. Prospective validation is underway.