QSP Model Identifies Biomarkers for Tebentafusp Treatment in Uveal Melanoma
Key Insights
A quantitative systems pharmacology (QSP) model was developed to simulate tebentafusp therapy in uveal melanoma (search) (UM), mimicking a Phase 3 clinical trial.
The model identified neo-antigen specific T cell clone as the most significant parameter differentiating responders from non-responders to tebentafusp.
CD8 (search)+ T cell density in the tumor showed the highest response probability and responder inclusion score, suggesting its potential as a predictive biomarker.
A quantitative systems pharmacology (QSP) model has been developed to identify potential biomarkers for treatment response to tebentafusp in patients with uveal melanoma (search) (UM). The model simulates the effects of tebentafusp, a T cell engager, and predicts biomarkers that could help in patient selection for this therapy.
The QSP model incorporates four compartments: central (blood), peripheral (other tissues and organs), tumor, and tumor-draining lymph node. It simulates immune activation, T cell trafficking, cancer cell killing, immune evasion, and antigen release. The model was parameterized using clinical data from a Phase 3 trial (NCT03070392) of tebentafusp in metastatic UM, with a virtual clinical trial conducted using 1500 virtual patients (VPs) generated through Latin-Hypercube Sampling (LHS).
Virtual Clinical Trial Results
The virtual clinical trial reproduced the clinical setting of the tebentafusp trial, using the same dosing schedule: 20 μg on day 1, 30 μg on day 8, and 68 μg weekly thereafter. The model predicted an overall response rate (ORR) consistent with the clinical trial results. Sensitivity analysis identified tumor growth rate as the most positively correlated parameter to tumor size change, while neo-antigen specific T cell clone was the most sensitive negatively correlated parameter.
Identification of Predictive Biomarkers
Thirty biomarker candidates were selected based on relevance to cancer immunotherapy, TCEs, and feasibility of measurement in patients. These included model parameters like initial tumor diameter, gp100 (search) and CD47 (search) expression, CD3 (search) expression, and emergent properties like tumor purity, immune cell counts, and cytokine concentrations.
Statistical comparisons between responders and non-responders revealed that neo-antigen specific T cell clone was the most significant parameter differentiating the two groups. Among model components, CD8 (search)+ T cell density in the tumor and central compartment, CD4+ T cells, helper T cells, and the CD8+/Treg ratio in the tumor were higher in responders. IFN-γ concentration was also higher in responders.
Biomarker Ranking and Combinations
Biomarkers were ranked based on response probability and responder inclusion score (RIS). CD8 (search)+ density in the tumor had the highest response probability (0.43) and RIS (0.55). Combinations of biomarkers, such as naïve CD4+ T cell density in the central compartment with the ratio of CD8+ T cells to Treg in the tumor compartment, achieved a higher response probability of 0.67. The best three and four biomarker combinations had a response probability of 0.8.
Predictive Power of On-Treatment Biomarkers
Early on-treatment biomarkers, measured at day 15 and day 30 after treatment initiation, showed higher predictive power compared to pre-treatment biomarkers. M1 macrophage density showed a >300% increase in response probability when relative change at day 15 was evaluated. The highest RIS attained was 0.82 with a combination of four biomarkers, including relative changes of M1/M2 macrophage ratio in the tumor, CD8 (search)+ T cell density in the tumor, ratio of CD8+ to Treg in the tumor and naïve CD8+ density in the central compartment, at day 15 with respect to the baseline.
These findings suggest that a combination of pre-treatment and on-treatment biomarkers could improve patient selection for tebentafusp therapy in uveal melanoma (search).