Objectives: Ponsegromab is a first-in-class, humanized anti-GDF15 monoclonal antibody which is being developed to treat cachexia in patients with solid tumors. Phase 1/2 clinical studies had evaluated total and unbound ponsegramab and total and unbound GDF-15 to fully characterize the PK/GDF-15 relationship. In future phase 3 studies, instead of measuring all 4 analytes, we propose to collect and analyze only total ponsegramab and unbound GDF-15 with sparse sampling. The objective of this analysis is to use simulations to evaluate if the proposed PK and GDF-15 assessments monitoring approach in Phase 3 studies are adequate to characterize ponsegromab PK/PD relationship in the Phase 3 population. Furthermore, we evaluated if the proposed body weight assessments are adequate to characterize the PK/body weight relationship.
Methods: A joint target-mediated drug disposition (TMDD)-indirect response model was developed and found to adequately describe the longitudinal ponsegromab PK/GDF-15/body weight data [1]. Using the existing model from Phase 1/2 data, a virtual population was generated based on the anticipated cancer-induced cachexia population in Phase 3. The virtual population was simulated for a proposed Phase 3 study and then a new data set was generated based on the sampling scheme proposed for this Phase 3 study. Stochastic simulation and estimation (SSE, 250 total runs) was performed to evaluate whether collection of unbound ponsegromab PK and total GDF-15 in a current Phase 3 study design is sufficient for PK/PD characterization. The contribution of Phase 1/2 data to PK/PD characterization was accounted for by incorporating a penalty function during estimation using the $PRIOR subroutine weighted by the parameter uncertainty of the Phase 1/2 model [2]. For each SSE run, the error was assessed as the percentage difference from the virtual participant’s true parameter value. The uncertainty of model parameters from SSE were evaluated using the absolute relative standard error (RSE). All SSE simulations were performed using Perl-speaks-NONMEM (version 5.3.0) [3].
Results: PK parameters of the model were estimated with high accuracy (error < 20%) and low uncertainty (RSE < 30%). Similarly, the PD parameters also showed the same degree of high accuracy and low uncertainty.
Conclusions: Using unbound ponsegromab and total GDF-15 data were sufficient to inform the PK/PD model and adequately characterize PK/PD in that model parameters would be estimated with high accuracy and low uncertainty. By leveraging prior PK/PD data and modeling, we minimize of analytical species which will reduce the overall cost in Phase 3 trials.
Citations: 1. Qiu, R., et al., Simultaneous Modeling of Ponsegromab PK, PD and Clinical Response in Patients with Cancer Cachexia, in Submitted to the American Conference on Pharmacometrics. 2025: Denver, CO. 2. Chan Kwong, A.H.P., et al., Prior information for population pharmacokinetic and pharmacokinetic/pharmacodynamic analysis: overview and guidance with a focus on the NONMEM PRIOR subroutine. J Pharmacokinet Pharmacodyn, 2020. 47(5): p. 431-446. 3. Huang, X.H., et al., Random sparse sampling strategy using stochastic simulation and estimation for a population pharmacokinetic study. Saudi Pharm J, 2014. 22(1): p. 63-9.
