Zinnia P Parra-Guillen1, Björn Frendéus1, Ingrid Karlson1, Mindaugas Meller1, Ingrid Teige, Andres McAllister1, Johan E Wallin1
1BioInvent International AB
Objectives Tumor necrosis factor receptor 2 (TNFR2) is a type I transmembrane protein highly expressed by myeloid cells and specific T cells subsets [1]. TNFR2 is involved in both anti- and pro-inflammatory. Accordingly, TNFR2 has been suggested as a promising novel target for anticancer treatment, with both agonists and antagonists demonstrating potent anti-tumor activity in preclinical settings [2]. We are developing BI-1910, an agonistic human IgG2 monoclonal antibody targeting TNFR2 that does not block the engagement of TNFR2 with its natural ligand, TNF-a. BI-1910 is currently being evaluated in an ongoing Phase 1/2a, open-label, dose-escalation, multicenter, first-in human, consecutive-cohort, clinical trial, as a single agent and in combination with pembrolizumab, in subjects with advanced/metastatic solid tumors whose disease has progressed after standard therapy (NCT06205706). The purpose of this work is to develop a population model to characterize pharmacokinetics (PK), receptor occupancy (RO), and the response of the target engagement biomarker soluble TNFR2 (sTNFR2) at different BI-1910 doses, to support early clinical development and guide dose recommendations. Methods: PK, RO and sTNFR2 data from 30 patients receiving BI-1910 as a single agent every 3 weeks in doses ranging from 4 to 900 mg were available for the analysis. Model building was performed in a sequential and integrative manner, first developing the PK model, then adding data for the RO data, to finally extend the model to include sTNFR2 information. Interindividual variability (IIV) was modeled exponentially, except for bounded parameters where additive models in the logit domain were used. Additive and proportional error models were explored for the different analytes. Goodness of fit plots, parameter precision in terms of relative standard errors (RSE), and visual predictive checks were used for model selection and evaluation. Analyses were performed in NONMEM 7.5, using FOCEI algorithm with the support of PsN, R and RStudio for data preparation and postprocessing of results. Results: A two-compartment model with linear clearance (0.0113 L/h, IIV = 49 %) and restricted total volume of distribution (~ 6 L, IIV = 20.5 %) well described BI-1910 concentration time profiles across all dose ranges and with a low residual unexplained variability (RUV= 13.2 %). Although some trends towards non-linear PK were identified at highest dose level, associated parameters were not estimated with sufficient precision (relative standard errors, RSE, an > 75 %), thus this feature was not included at this stage. Regarding RO, a direct and saturable model driven by BI-1910 drug levels was selected, estimating an almost complete maximum receptor occupancy (EMAX > 90 %) attained at already low drug levels (concentration triggering 50 % of maximum effect, EC50 = 1630 ng/mL) and sustained over the whole dosing interval at doses 300 mg and above. Finally, an indirect response model was selected to characterize the time course of sTNFR2, with BI-1910 levels being capable of completely inhibiting biomarker degradation rate constant (kout = 0.0677 1/h). Remarkably, the estimated EC50 for sTNFR2 effect was one order of magnitude higher than the one estimated for RO, suggesting a potential disconnection between RO and pharmacodynamic effects. The final joint model provided a satisfactory description of all sources of data, at both individual and population level, with adequate parameter precision (RSE < 35 %). Stochastic simulations were performed to explore model performance at different dosing regimens and guide dose selection. Conclusions: A PK/RO/sTNFR2 joint model has been successfully developed to characterize BI-1910 pharmacokinetics and pharmacodynamics across a broad range of doses using early clinical stage information, and to support dose selection for upcoming studies.
[1] Ye et al. Frontiers in Immunology, 9:583 (2018) [2] Medler et al. Cancers, 14: 2603 (2022)
Reference: PAGE 33 (2025) Abstr 11618 [www.page-meeting.org/?abstract=11618]
Poster: Drug/Disease Modelling - Oncology