(T-085) Modeling progression and treatment of mCRPC using the Thales QSP software platform

Ashley Markazi, PhD – Scientist I, Simulations Plus; Ray Joe, PhD – Scientist II, Simulations Plus; Conner Sandefur, PhD – Senior Scientist, Simulations Plus; Matt McDaniel, PhD – Senior Scientist, Simulations Plus; Ryan Suderman, PhD – Associate Director / Senior Principal Scientist, Simulations Plus

Objectives: Metastatic, castration resistant prostate cancer (mCRPC) is an aggressive form of prostate cancer in which patients progress despite androgen deprivation therapy. Clinical need exists for therapies which are effective against these aggressive tumors. Here, a QSP platform model of mCRPC was developed, and a virtual population (vpop) of mCRPC patients was created by training and validating to clinically relevant measures from several clinical trials. This model and vpop can be used to evaluate new therapies targeting mCRPC patients.

Methods: A QSP model of mCRPC was created in Thales, a model building platform that automates, streamlines, and enhances the QSP modeling process. The model structure includes representations of immune cells, immune killing of cancer cells, cytokine effects, and prostate specific antigen (PSA) as a biomarker of disease state. Additionally, patients were modeled as having two lesions, where each contained 3 distinct tumor microenvironments which could vary in size, growth, immune penetration, and drug PK/PD. Parameter values were chosen to align with existing published mCRPC models or estimated from publicly available in vitro/vivo data. A vpop of 100 mCRPC patients was generated by fitting and validating to clinical trial data from eight clinical trials spanning ten therapeutic compounds, administered either as monotherapy or in combination. Key clinical measures were derived from model predictions such as tumor diameter and serum PSA and were used in training. These included: RECIST1.1 criteria, Overall Response Rate, Best Overall Response, 50% and 90% PSA decreases from baseline, Time to Response, Duration of Response, and patient survival (PFS, OS).

Results: A vpop replicating the clinical outcomes of mCRPC patients was created in Thales. The model predictions closely matched the training data, with the clinical endpoint data falling within the 90% confidence interval of the model predictions >66% of the time and >55% for the validation data. Response data from placebo scenarios is used to capture untreated dynamics, including calibrating immune mediated killing and baseline PSA levels allowing for accurate predictions of checkpoint inhibitors. Additionally, patients with higher tumoral regulatory T cell infiltration responding more poorly to checkpoint inhibition therapy is recapitulated. Lastly, a reduction in PSA levels following effective treatment from all treatment types was modeled.

Conclusions:
These results provide proof-of-concept that mCRPC progression and treatment is able to be modeled using a QSP platform model. The model and vpop accurately predict clinical endpoint data for several clinical trials and therapeutic regimens. Additional therapies could be added to further train the population, both enhancing model performance and allowing clinically relevant predictions for novel compounds such as optimal first-in-human dose, special populations of interest, or novel therapeutic combinations.

Citations: None