Dmitry Shchelokov 1, Oleg Demin Jr 1, Galina Saynukova 1
1 InSysBio CY (Paphos, Cyprus)
Objectives: Tarlatamab is a DLL3/CD3 bispecific T-cell engager (TCE) recently approved by the FDA for the treatment of extensive-stage small cell lung cancer (ES-SCLC). The dosing regimen for the first cycle follows a step-up approach, with 1 mg administered on day 1, and 10 mg on days 8 and 15. Alternative step-up dosing (SUD) regimens for the target dose of 10 mg Q2W were not investigated in clinical studies [1]. Our aim was to test a mechanistic model that links cytokine release syndrome (CRS) directly to the trimer levels formed by tarlatamab (the complex of TCE with target receptors) and to explore alternative SUD regimens to reduce the frequency of CRS.
Methods: The quantitative systems pharmacology (QSP) model integrated the pharmacokinetics of tarlatamab, its distribution into the tumors of patients with ES-SCLC, binding with CD3 expressed by T-cells and DLL3 expressed by cancer cells, formation of trimers in the immunological synapse between T-cells and cancer cells, turnover of free and bound receptors, and trimer internalization. Pharmacokinetics were simulated using parameters from a published population PK model [1]. Variability was also incorporated in tumor volume, the number of T-cells in the tumor, and the percentage of DLL3-positive cancer cells. Data on the number of DLL3 molecules per primary tumor cell were limited, with a rough estimate of 10,000 [2]. CRS grades were directly linked to the peak levels of trimers in the tumor. If the trimer levels exceed the calibrated threshold, the virtual patient is considered to have CRS of a specific grade. These threshold levels were calibrated using data on CRS frequency from different doses in a dose-escalation trial.
Results: CRS incidence (any grade) was 0% in the 0.003, 0.01, 0.03, and 0.1 mg cohorts in both the model and clinical trial. The model successfully described the CRS incidence for other flat doses (0.3 and 1 mg) and two SUD regimens (1/3 mg and 1/30 mg). Model predictions were validated using the SUD regimens tested in the phase 2 trial: 1/10 mg and 1/100 mg. CRS incidence following the approved 1/10 mg regimen was described with adequate precision, whereas it was overpredicted for the 1/100 mg regimen. The model predicted alternative SUD regimens that mitigate CRS compared to the approved regimen. For instance, administering 0.3 mg instead of 1 mg on day 1 and 1 mg instead of 10 mg on day 8 in the approved regimen resulted in a decrease in the frequency of any grade of CRS from 50% to 20-25%, including almost complete avoidance of CRS grade 3 or higher. An additional step-up dose on day 4 further reduces CRS incidence.
Conclusions: The direct link between trimer levels in the tumor and CRS grades allows for the description of clinical data on CRS incidence for all tested doses and SUD regimens, except the 1/100 mg regimen, which was overpredicted by the model. One reason for this discrepancy is the limited data on the number of DLL3 molecules per cell in the primary tumor. The QSP model, with a mechanistic description of turnover and internalization of free and bound target receptors and trimers, enables the prediction of optimal SUD regimens.
References:
[1] U.S. Food and Drug Administration. FDA Multidisciplinary Review: Imdelltra (Tarlatamab). NDA 761344. 2024. https://www.accessdata.fda.gov/drugsatfda_docs/nda/2024/761344Orig1s000MultidisciplineR.pdf
[2] Sharma et al. Cancer Res. 2017 Jul 15;77(14):3931-3941
Reference: PAGE 34 (2026) Abstr 12170 [www.page-meeting.org/?abstract=12170]
Poster: Drug/Disease Modelling - Oncology