Nina Beutling 1,2, Nicolas Frances 1, Miro Eigenmann 1, Gregor Lotz 1, Thomas Stuchly 1, Wouter Driessen 1
1 F. Hoffmann-La Roche Ltd (Basel, Switzerland), 2 University of Bern (Bern, Switzerland)
INTRODUCTION
T cell–redirecting bispecific antibodies (TCBs) targeting CEACAM5 are being developed as a therapeutic in colorectal cancer (CRC) [1]. CEACAM5 is expressed in the majority of CRCs and due to its high prevalence it has been established as both a diagnostic biomarker and a therapeutic target [1–3]. However, CEACAM5 can be cleaved from the cell membrane, generating soluble CEACAM5 (sCEA) in circulation [4]. This circulating antigen may act as a drug sink, reducing systemic exposure and introducing nonlinear pharmacokinetics (PK) beyond classical IgG disposition [5, 6]. The extent of this exposure loss may vary between patients depending on baseline sCEA levels, as observed during clinical evaluation of a CEACAM5–CD3 TCB [7]. At low doses, reduced exposure may appear as good tolerability but may in fact reflect underexposure. At higher doses, variability in exposure may obscure true dose–response relationships. Therefore, interaction between the TCB and soluble target adds considerable complexity to early clinical development, particularly in the interpretation of pharmacokinetics, safety signals, and dose selection.
OBJECTIVES
To address these challenges, we aimed to:
– Develop a semi-mechanistic population pharmacokinetic model to characterize the interaction between sCEA and CEACAM5-targeting TCB in the clinical setting.
– Simulate expected exposure loss as a function of dose and baseline sCEA levels.
– Evaluate whether baseline sCEA stratification is warranted during early pharmacokinetic and safety assessment.
– Assess, at projected efficacious dose levels, whether sCEA-mediated impact remains clinically relevant and estimate the proportion of patients achieving the projected efficacious exposure.
METHODS
Clinical PK data were obtained from a first-in-human, open-label phase I dose-escalation study for a CEACAM5-targeting TCB. Patients received multiple ascending intravenous dose levels administered every three weeks. [7]
Four bioanalytical assays quantified free drug, total sCEA, sCEA-bound drug, and total drug, enabling discrimination between binding states and quantitative characterization of drug–sCEA interaction.
A semi-mechanistic population pharmacokinetic model was developed. Reversible binding described by mass-action kinetics between drug and sCEA was implemented in the plasma compartment, and sCEA turnover was described using a production–elimination model. Free and complexed drug were modeled explicitly with separate first-order elimination processes.
Model development and parameter estimation were performed in Monolix 2024R1 using the SAEM algorithm. Model adequacy was evaluated using goodness-of-fit diagnostics and predictive checks. Sensitivity analyses were conducted in R using rxode2.
Simulations were conducted using Simulx 2024R1, with complementary analyses implemented in R (rxode2). Exposure loss and variability in free drug exposure were evaluated as a function of baseline sCEA. The proportion of patients reaching pharmacologically relevant exposure across dose levels was assessed in a virtual population reflecting clinically observed baseline sCEA variability. Finally, patient stratification was evaluated by applying baseline sCEA thresholds of 20, 100, and 300 ng/mL and quantifying the corresponding simulated changes in free drug exposure and proportion of patients reaching efficacious exposure.
RESULTS
The comprehensive bioanalytical measurements enabled a thorough characterization of sCEA–TCB complex formation in vivo. The sCEA-bound fraction represented a relevant proportion of total circulating TCB, particularly at low doses and higher baseline sCEA levels.
The pharmacokinetic model adequately described the concentration-time profiles of all four analytes. Most population parameters were estimated with relative standard errors below 30%. Consistent with the dispersion observed in the clinical PK data, several parameters, notably Vt and k12, exhibited high inter-individual variability, reflected by large random-effect estimates.
Simulations demonstrated a pronounced baseline sCEA-dependent exposure loss at low doses. Patients with higher baseline sCEA exhibited substantially reduced free drug exposure compared to those with lower sCEA concentrations. At higher dose levels, the magnitude of exposure attenuation decreased, consistent with partial saturation of the soluble antigen sink. The proportion of patients predicted to achieve target exposure increased nonlinearly with dose, approaching saturation at higher dose levels. Baseline sCEA-based stratification increased the probability of achieving efficacious exposure at low and intermediate dose levels. At higher doses, the effect of stratification diminished.
CONCLUSION
This work demonstrates that comprehensive bioanalytical characterization was crucial for understanding the TCB–soluble target interaction. The availability of four distinct assays was critical for disentangling free and complexed drug fractions and for establishing the presence of a clinically relevant antigen sink.
The population pharmacokinetic model provided a mechanistically coherent framework to quantify sCEA-mediated exposure loss and to describe nonlinear behavior observed in the clinical data. While the dose-dependent attenuation of sCEA impact is consistent with target-mediated drug disposition, the key contribution of this work lies in its quantitative and visual characterization of this behavior. By translating the antigen sink concept into explicit exposure–dose relationships, the model makes variability predictable, thereby guiding team expectations and project decisions.
Baseline sCEA is therefore informative for interpreting early safety and exposure data and may support patient stratification at low and intermediate doses. At higher dose levels, where exposure loss is overcome, the benefit of stratification diminishes, supporting a unified dosing strategy.
Overall, this analysis translates a conceptually well-known TMDD behavior into actionable guidance for early clinical development. By enabling prediction of pharmacologically active exposure for individual patients based on baseline sCEA, this approach supports a more personalized development strategy and lays the groundwork for exposure-guided patient selection. The principles established here may also be applicable to other antibody-based therapies affected by soluble target interference, provided that target-specific kinetic and turnover parameters are appropriately adapted.
References:
[1] Hossain MS, Karuniawati H, Jairoun AA, et al. Colorectal Cancer: A Review of Carcinogenesis, Global Epidemiology, Current Challenges, Risk Factors, Preventive and Treatment Strategies. Cancers (Basel) 2022; 14(7)
[https://doi.org/10.3390/cancers14071732][PMID: 35406504]
[2] Segal NH, Melero I, Moreno V, et al. CEA-CD3 bispecific antibody cibisatamab with or without atezolizumab in patients with CEA-positive solid tumours: results of two multi-institutional Phase 1 trials. Nat Commun 2024; 15(1): 4091
[https://doi.org/10.1038/s41467-024-48479-8][PMID: 38750034]
[3] Thompson J, Zimmermann W. The carcinoembryonic antigen gene family: structure, expression and evolution. Tumour Biol 1988; 9(2-3): 63–83
[https://doi.org/10.1159/000217547][PMID: 3041547]
[4] Yamamoto Y, Hirakawa E, Mori S, Hamada Y, Kawaguchi N, Matsuura N. Cleavage of carcinoembryonic antigen induces metastatic potential in colorectal carcinoma. Biochem Biophys Res Commun 2005; 333(1): 223–9
[https://doi.org/10.1016/j.bbrc.2005.05.084][PMID: 15958210]
[5] Boucher D, Cournoyer D, Stanners CP, Fuks A. Studies on the control of gene expression of the carcinoembryonic antigen family in human tissue. Cancer Research 1989; 49(4): 847–52
[PMID: 2912558]
[6] Iwano J, Shinmi D, Masuda K, Murakami T, Enokizono J. Impact of Different Selectivity between Soluble and Membrane-bound Forms of Carcinoembryonic Antigen (CEA) on the Target-mediated Disposition of Anti-CEA Monoclonal Antibodies. Drug Metab Dispos 2019; 47(11): 1240–6
[https://doi.org/10.1124/dmd.119.088179][PMID: 31533926]
[7] U.S. National Library of Medicine. A First-in-Human, Open-Label, Multicenter, Dose-Escalation Phase I Clinical Study of Single-Agent RO7172508 in Patients With Locally Advanced and/or Metastatic CEA-Positive Solid Tumors; 2018 [cited 2026 December 26] Available from: URL: https://www.clinicaltrials.gov/study/NCT03539484.
Reference: PAGE 34 (2026) Abstr 12151 [www.page-meeting.org/?abstract=12151]
Poster: Oral: Drug/Disease Modelling - Oncology