Senior Principal Scientist Genentech, Inc Miami, Florida, United States
Disclosure(s):
Brendan C. Bender, M.Sc., PhD: No relevant disclosure to display
Objectives: Mosunetuzumab (Mosun) is a CD20xCD3 T-cell engaging bispecific antibody (TCE) approved for patients with 3L+ follicular lymphoma. A population mPBPK modeling approach was developed to compare Mosun CD20 receptor occupancy (RO%) in the various tissues with the RO% reported from a two-compartmental (2CMT) model that used equilibrium binding assumptions (1).
Methods: The mPBPK model was fit to the same Mosun concentration–time data from 439 patients as previously published (1). Patients received Mosun as either a fixed dose (n= 32; 0.05–2.8mg every 3 weeks (q3w)) or as a Cycle 1 step-up dose regimen followed by q3w maintenance dosing (n=407, 0.4/1.0/2.8 – 1/2/60/30mg).
The model was modified from Cao et al. (2), adding three new tissue compartments representing lymph node, tumor, and spleen. CD3 receptor (T cells) and CD20 receptor (B cells) expression and turnover were applied. Fifty percent of patients also had levels of anti-CD20 (aCD20) therapies from prior treatment (e.g., rituximab and obinutuzumab); all drug binding to targets were modeled using fixed KD values and receptor levels.
The mPBPK model estimated parameters of CL and V1 were compared to those from the 2CMT model; CLd and V2 parameters were not part of the model, as these were described by the fixed physiological flow rates and tissue volumes garnered from literature. The Mosun CD20 RO% time course for each patient was derived directly in each tissue compartment by the bound Mosun:CD20 complex, divided by the total CD20 present. NONMEM software (v7.5.1) was used for modeling.
Results: The final mPBPK model consisted of seven tissue compartments and 44 ODEs which fully described the system of free and receptor–bound Mosun and aCD20. Model fits and diagnostics plots were comparable to the 2CMT model. PK parameter estimates for CL and V1 were similar between the both methods: CL = 1.40 vs 1.08 L/day and V1= 5.65 vs 5.49 L for the mPBPK and 2CMT, respectively. The 2CMT model took two hours to converge, whereas the mPBPK model converged after two days. Free Mosun concentrations in each tissue equilibrated with plasma concentrations after a few days; Mosun partitioning between plasma and tissues ranged from 2.3–fold (leaky tissue) to 11–fold (tight tissue). The predicted plasma Mosun CD20 RO% time courses were superimposable between both modeling approaches. In the mPBPK model, Mosun CD20 RO% in the tumor tissue (efficacy surrogate) and leaky tissue (safety surrogate) were highly correlated with the plasma, 3–fold and 2–fold less at most time points, respectively.
Conclusions: The mPBPK model described the data well, and results were similar with the 2CMT approach. The high correlation between the plasma and tumor/leaky tissues suggests that plasma CD20 RO%, predicted by a compartmental model with an equilibrium approximation, is suitable for clinical drug development; subsequent exposure–response (ER) analyses based on plasma RO% for efficacy and safety ER (3) assessments would be relevant. The numerous ODEs needed to describe this TCE system of drug binding in multiple compartments led to long run times, which may limit its utility with large clinical datasets, stepwise covariate modeling, and other filing requirements.
Citations: Citations: 1. Bender BC, et al. Clin Transl Sci. 2024 Jun;17(6):e13825. doi: 10.1111/cts.13825. 2. Cao Y et al., J Pharmacokinet Pharmacodyn. 2013 Oct;40(5):597-607. doi: 10.1007/s10928-013-9332-2. 3. Li C-C, et al. Clin Pharmacol Ther. 2025 Feb;117(2):465-474. doi: 10.1002/cpt.3445
Keywords: Mosunetuzumab, population pharmacokinetics, mPBPK model, receptor occupancy, bispecific antibody
