Director, Clinical PK & PMx Merck Sharp & Dohme, Japan
Disclosure(s):
Chihiro Hasegawa, PhD: No financial relationships to disclose
Objectives: Antibody-drug conjugates (ADCs) are a rapidly emerging class of agents for therapeutic areas such as oncology, immunology, etc. The ADCs consist of mixtures of antibodies with different number of drug molecules attached (DAR, drug to antibody ratio). For some purposes, a fully mechanistic PK model including each component with different DARs will be useful, while measurements of all the components are rarely available. Instead, other types of components such as total antibody (tAB) and unconjugated drug (T) are often measured, but it is not so clear as to what analytes/measurements are required for the analysis using a fully mechanistic approach. We explored this aspect via a simulation study.
Methods: This study was motivated by another simulation study [1]. Concentration data of tAB and T were simulated using the fully mechanistic 9-compartment model consisting of six ADC components with different DARs (central + peripheral), two free antibody components (central + peripheral), and one drug component (central only) for 30 virtual subjects with intravenous infusion and rich sampling. In addition, number of all drug molecules attached to antibodies (antibody conjugated drug, acT) and number of antibodies with at least one drug molecule attached (ADC) were also simulated. These data were used as observations with different scenarios; S1 with all analytes (tAB, T, acT and ADC), S2 with three analytes (tAB, T and acT), S3 with another set of three analytes (tAB, T and ADC), and S4 with two analytes (tAB and T). Parameter values were adopted from Li H, et al. [2]. NONMEM 7.4 with FOCE-I was used for parameter estimation. The accuracy and precision of estimates for individual exposure metrics (i.e., AUC and Cmax for each analyte that can inform subsequent exposure-response analyses) were assessed using the relative estimation error (REE = (Est-True)/True).
Results: In all scenarios, the fully mechanistic model converged successfully. Shrinkages for all ETA parameters related to clearance and volume of antibody/drug were < 20% in S1-S3, but this was not the case in S4. In the scenarios S1-S3 (more than three analytes used as data), the estimates for individual AUC and Cmax were accurately estimated for all four analytes in all subjects (absolute REE of < 20%). However, in the scenario S4 (two analytes used as data), for three analytes (tAB, acT and ADC), in several subjects the estimates for individual posthoc AUC were inaccurately estimated (absolute REE of >20-100%). For the remaining analyte (T), the absolute REE was < 20% for all subjects. The individual Cmax was generally well estimated for all four analytes also in S4.
Conclusions: These findings suggest that when applying the fully mechanistic model, at least three analytes described above are required to accurately estimate individual posthoc AUC for all four analytes. In the case of having difficulty to measure the three analytes, more simplified/reduced models will be required.
Citations: [1] Gibiansky L, et al. ACoP10, Orlando, FL, 2019. [2] Li H, et al. J Clin Pharmacol (2017) 57: 1148-1158.
