Victor Poon, MS: No financial relationships to disclose
Objectives: Efbalropendekin alfa (EBA) (IL-15/IL-15Rα-Fc) is a Fc-fusion protein engineered to induce proliferation in NK and CD8+ T cells to boost cancer immunotherapy1. In this drug’s ongoing first-in-human dose escalation study, EBA was given as monotherapy in Phase Ia and combined with 840 mg of atezolizumab (ATZ) in Phase 1b. EBA was dosed from 0.01–0.12 mg/kg every 2 weeks (Q2W) or 0.06-0.18 mg/kg every 4 weeks (Q4W). In preclinical and clinical studies, EBA exhibited complex dose- and time-dependent nonlinear pharmacokinetics (PK), which were attributed to dynamically changing target-mediated drug deposition (TMDD). EBA increases target expansion and, consequently, its own clearance. This complex PK-PD behavior was captured well in a previously developed QSP model2. However, the QSP model is not well suited for population PK (popPK) analysis and covariate assessment in late-stage clinical development. Therefore we aim to develop plausible semi-mechanistic popPK models, identify covariates of clinical interest and apply the models for simulating PK.
Methods: Based on Q2W data, an Auto-Induction1 model was developed with central and peripheral drug compartments and a third compartment representing the amount of EBA binding receptors (IL15R). Further, a more mechanistic PD-enhanced TMDD model was developed with central and peripheral drug compartments, a third compartment of free IL15R, and a fourth compartment of EBA-IL15R complex. This model used in vitro measured drug-receptor binding constants. The model performance was compared between the two models.
After developing the base models, effects of baseline covariates and ATZ combination on PK parameters were evaluated. Using these models, various EBA PK dosing regimens were simulated. Model predicted exposure metrics were then calculated and compared between the two models. Lastly, both models were applied to pooled Q2W and Q4W data to assess model performance.
Results: For the model development using Q2W data, both the Auto-Induction and TMDD models performed better than a two-compartment popPK model and both fit observed data equivalently in terms of RMSE and R2. In both models, a statistically non-significant (P>0.1) ATZ effect on EMax and nonspecific drug clearance (CL) was observed, and a significant (P < 0.05) body weight effect on CL and the central volume of distribution was identified.
Using the developed models, less than dose-proportional increase in exposures across the simulated doses were predicted for both Q2W and Q4W regimens. The model-simulated exposures diverged slightly between the two models when extrapolating to doses higher than those used for modeling (>0.12mg/kg). The simulated average weekly AUC was overall comparable between the monotherapy and ATZ combo arm (5% lower in patients in the ATZ combo arm). Lastly, both model structures identified based on Q2W data were also able to describe the Q4W data.
Conclusions: Two types of popPK models were developed which can both describe EBA PK data. The model suggests a low risk for ATZ to affect EBA PK. These models are important to characterize EBA PK, assess the impact of key baseline covariates, and generate exposure metrics for exposure-response analyses in support of future development trials of EBA.
Citations: [1] Tran B. et. al., Phase Ia/Ib dose-escalation study of efbalropendekin alfa (XmAb24306) as single-agent and with atezolizumab in solid tumors. AACR 2025
[2] Lu D. et. al., Complex PK-PD of an engineered IL-15/IL-15Rα-Fc fusion protein in cynomolgus monkeys: QSP modeling of lymphocyte dynamics. Eur J Pharm Sci. 2023
Keywords: Population PK, Cancer Therapy, Clinical Trials
