Authors: Oleg Demin Jr (1), Galina Kolesova (1), Dmitry Shchelokov (1), Stefan Zajic (2), Lourdes Cucurull-Sanchez (3)
Affiliation: (1) InSysBio CY, Paphos, Cyprus; (2) GlaxoSmithKline, Collegeville, PA 19426, USA; (3) GlaxoSmithKline, Stevenage, Hertfordshire, UK
Objectives: Letetresgene autoleucel (lete-cel) is an autologous T-cell transduced to express a high-affinity T-cell receptor that recognizes NY-ESO-1 antigen in complex with HLA-A*02. NY-ESO-1 is a cancer-testis antigen (CTA) expressed in 70%–80% of synovial sarcoma (SS) tissue. The aim of the work is to develop a QSP model of overall response rate (ORR) and progression-free survival (PFS) in SS patients upon treatment with lete-cel and validate it against clinical data from a pilot clinical study.
Methods: The developed QSP model consists of 5 main modules (1) lymphodepletion including pharmacokinetics (PK) of fludarabine and cyclophosphamide, their effects on lymphocytes and IL-7 and IL-15 levels [1]; (2) tumor growth and microenvironment of SS (M2 macrophages, TGFb, etc.) 3) PK and distribution of lete-cel after intravenous infusion [2]; (4) killing of NY-ESO-1 positive cancer cells by lete-cel and activation of engineered T-cells followed by expansion of T-cells and cytokine secretion; (5) biomarkers in blood (IL-6, IFNg, vector copies per µg DNA). The model was calibrated against various data including, but not limited to, in vitro and lete-cel clinical pharmacodynamics (PD) data*. Biomarkers, ORR and PFS from a pilot study including 4 cohorts were used for validation [3]. ORR and PFS were described in the model using response evaluation criteria in solid tumors (RECIST 1.1) [4]. Ranges for parameters and baselines to generate virtual patients (VPs) were identified based on published baseline data for SS and lete-cel PD data. Two types of VPs were simulated: (1) virtual twins (VTs) with fixed baseline and lete-cel product characteristics as for patients from the pilot study; (2) VPs with baseline and lete-cel product characteristics generated from estimated ranges for SS patients. The number of VPs in virtual clinical trials (VCTs) was the same as in the corresponding cohort of the pilot study. Ten VCTs were simulated for each one of the 4 patient cohorts.
Results: The simulated peak levels of proinflammatory cytokines and expansion of engineered T-cells (vector copies per ug DNA) were increased in responders as compared with non-responders, in agreement with biomarker analysis of the pilot study clinical data [3]. The predicted ORR and PFS were in good agreement with the clinical data: mean ORR from VT simulations was 58.3% (95% CI [50,58.3]) vs 50% (6/12) reported for cohort 1; VP simulations for cohort 3 yielded 20% ORR (95% CI [0,40]) vs 20% (1/5); predicted median PFS was 13 (95% CI [10,14]) vs observed 13 weeks for cohort 2; VCT simulations gave a median duration of response of 16 (95% CI [7,106]) vs observed 16.4 weeks (range [14,94]) for cohort 4 in the pilot study.
Conclusions: Mechanistic modeling of tumor dynamics and response criteria predicted clinical ORR and PFS with adequate precision. Utilization of clinical endpoints data can be considered an essential step of QSP modeling in oncology.
Citations:
[1] Demin Jr et al. ASCPT 2022
[2] Demin Jr et al. ACoP12 2021
[3] Gyurdieva et al. Nat Commun 2022 Sep 8;13(1):5296
[4] Eisenhauer et al. Eur J Cancer 2009 Jan;45(2):228-47
*The human biological samples were sourced ethically and their research use was in accord with the terms of the informed consents under an IRB/EC approved protocol.
Thank you for your message! We will get back to you shortly.
