C. J. Peer, A. Goey, T. M. Sissung, S. Ehrlich, C. Bryla, S. E. Bates, W. D. Figg; National Cancer Institute, Bethesda, MD

BACKGROUND: Belinostat (BEL) is mainly cleared via the polymorphic UGT1A1. Carriers of >= one variant allele are at risk for increased exposure relative to WT. This study sought to develop a population PK model to simulate optimal doses to produce similar exposure metrics based on UGT1A1 genotype.
METHODS: The dataset used was a phase I trial of BEL with cisplatin and etoposide, where BEL was given as a 48 hr CIVI over the dose range 400-800 mg/m2/24hr. PK levels were measured by UHPLC-MS/MS. UGT1A1 genotype status for the *28, *6 and *60 alleles were determined by fragment analysis or direct sequencing. A population PK model was built using Phoenix NLME1.2 to identify relevant covariates (eg. genotype) that may influence parameter estimates and explain inter-individual variability. The model that best fit the data was determined by graphical plots and statistical analyses based on significant improvement in the model. The final model was internally validated through a visual predictive check and a bootstrap simulation (n=1000 replicates). A survival analysis was performed with PFS data using R to identify any predictor variables that may significantly affect time to progression.
RESULTS: A three compartment model with mixed RV best fit the data. No covariates significantly improved the model, however, clearance was lower for carriers of a single copy of either *28 or *60 (n=20) vs WT in both (n=5). Model-based AUCs were similar to NCA. Simulated doses of 400 mg/m2/24hr for carriers (n=50) and 600 mg/m2/24hr for WT (n=200) demonstrated remarkably similar AUC. No PFS differences were observed between WT vs carriers (p=0.837).
CONCLUSION: Dose titration based on UGT1A1 genotype provides a means to better predict BEL exposure metrics but needs to be explored clinically to assess its utility in combination studies.