Vasiliki Kostiou1, Eric Jurgens2, Vijayalakshmi Chelliah1, Piet H. van der Graaf1,3,4, Andrzej M. Kierzek1, Ross S. Firestone2, Kevin Miller2, Bruno Almeida Costa2, Sridevi Rajeeve2, Alexander M. Lesokhin2, Neha Korde2, Carlyn R. Tan2, Hamza Hashmi2, Hani Hassoun2, Kylee Maclachlan2, Urvi A. Shah2, Malin Hultcrantz2, Issam Hamadeh2, Sergio A. Giralt2, David J. Chung2, Heather J. Landau2, Michael Scordo2, Gunjan Shah2, Saad Z. Usmani2, Sham Mailankody2
1Certara, 2Memorial Sloan Kettering Cancer Center, 3Systems Pharmacology and Pharmacy, LACDR, Leiden University, 4Cincinnati Children’s Hospital Medical Center
Introduction: Despite promising outcomes in CAR T-cell therapy for relapsed/refractory multiple myeloma (RRMM), nearly all patients eventually relapse. Resistance and relapse may be driven by CAR T-cell and tumour intrinsic factors. Several strategies are currently being investigated to improve response durability, with multi-antigen targeting being a promising area of clinical research. Objective: A mechanistic Quantitative Systems Pharmacology (QSP) model of multiple myeloma growth and CAR T-cell therapy using measurable biomarkers was developed to evaluate anti-BCMA and anti-GPRC5D CAR T-cell therapies, to assess combination treatment approaches and identify patient-specific factors associated with response and relapse. Methods: We developed a QSP model, describing the key biological processes and clinical biomarkers involved in myeloma development and CAR T-cell treatment. Key input parameters included tumour intrinsic factors (i.e. tumour burden, proliferation rate, antigen expression), CAR T-cell intrinsic factors (i.e. biodistribution, antigen binding, activation and expansion, kill rate), and serum biomarkers (i.e. M protein, serum free light chains [sFLC]). The model was calibrated to published patient pharmacokinetic and biomarker response data from anti-BCMA (Ciltacabtagene autoleucel [cilta-cel], Idecabtagene vicleucel [ide-cel]) and anti-GPRC5D (MCARH109) CAR T-cell therapy studies [1] [2] [3]. We then utilized clinical biomarker data from two real world relapsed/refractory multiple myeloma (RRMM) patient cohorts treated with commercial cilta-cel and ide-cel respectively at Memorial Sloan Kettering Cancer Center (MSKCC), to validate our model. Global sensitivity analysis via Morris Method was used to identify model parameters associated with the greatest impact on response and virtual trial simulations with varied input parameters were performed to test response variability. Results: The QSP model reproduced observed pharmacokinetic and response biomarker data (serum FLCs, M protein) from publicly available ide-cel, cilta-cel, and MCARH109 CAR T-cell therapy studies. The calibrated model recapitulated biomarker data from 41 RRMM patients treated with commercial cilta-cel or ide-cel (n=19 ide-cel, n= 22 cilta-cel). Virtual trial simulations exploring the impact of variable baseline disease and CAR T-cell characteristics on response predicted that worse outcome is associated with rapidly proliferating tumors, low antigen expression, and low CAR T-induced killing rate. Model predictions comparing anti-BCMA and anti-GPRC5D CAR T-cell monotherapies suggested that GPRC5D-targeted treatment is more sensitive to GPRC5D antigen downregulation. Additionally, CAR T-cell combination treatment simulations predicted a better outcome in scenarios considering high BCMA and GPRC5D antigen expression. Conclusion: We developed a QSP model of CAR T-cell therapy for RRMM that accounts for both BCMA and GRPC5D targeting CAR T-cell products, allowing the exploration of both mono- and combination therapies. We calibrated/validated our model using published clinical trial and real-world clinical response data from patients treated at a single institution. Our model provides useful insights on factors influencing clinical outcome by associating tumour related properties (high proliferation, low target antigen expression) and CAR T-cell related properties (low CAR T-induced killing rate) with disease relapse. Simulations also suggested that anti-GPRC5D treatment is more sensitive to antigen downregulation compared to anti-BCMA treatment. Our model is well-positioned as a representative model and can serve as a framework to investigate additional mechanisms and multi-antigen targeting. The platform is suitable for exploratory studies and can be used to conduct virtual patient population simulation experiments for comparative analysis, patient selection and clinical decision-making.
[1] Raje N, Berdeja J, Lin Y, et al. N Engl J Med, 18:1726-1737, 2019. [2] Xu J, Chen LJ, Yang SS, et al. Proc Natl Acad Sci U S A, 19:9543-9551, 2019. [3] Mailankody S, Devlin SM, Landa J, et al. N Engl J Med, 13:1196-1206, 2022.
Reference: PAGE 33 (2025) Abstr 11513 [www.page-meeting.org/?abstract=11513]
Poster: Drug/Disease Modelling - Oncology