Javier Sánchez Fernández: Model-based optimization of cancer immunotherapy combinations

Abstract

The use of cancer immunotherapies has transformed the treatment landscape for many cancer types. Unfortunately, not all patients respond to these therapies, and most of those who do eventually relapse. Combining cancer immunotherapies may improve patient outcomes. However, determining which molecules to combine, at which doses, and under which dosing schedules rarely is straightforward.

Preclinical experiments offer the opportunity to test a wide variety of experimental conditions. This data, together with information about disease biology, can be integrated into a mathematical modeling framework, which can be used to simulate different scenarios, allowing researchers to prioritize the most promising drug combinations in the patient populations where the highest probability of success is expected. In a continuous cycle, the model can inform the design of novel biologic drugs with improved pharmacological properties to improve outcomes for a larger percentage of the patient population. This thesis aimed to develop modeling and simulation approaches to guide the development of cancer immunotherapy combinations by contributing to molecule design, preclinical experimental design, and translation of preclinical knowledge into clinical insights. 

The translation of the preclinical tumor growth inhibition model suggested that identifying a clinical effect with CD3 T-cell bispecific antibodies in monotherapy may be challenging. However, combination with anti-PD-L1 is expected to more than double progression-free survival, duration of response and response rate, highlighting that combination approaches with these molecules need to be considered as early as possible. 

Using preclinical data, a target engagement model for bispecific costimulators was developed that can be used to prospectively predict the clinical range of doses with maximum expected effect. Furthermore, the model allowed differentiating the contribution of drug exposure and target expression to drug pharmacology. Leveraging this model, the impact of binding affinity on drug pharmacology was explored in silico for nineteen different oncology indications. This identified a molecule with a 10-fold increase in binding affinity as a promising follow-up molecule that may lead to increased patient benefit, establishing a workflow that can combine preclinical data with clinical target expression to explore in silico optimized molecule designs. 

Lastly, a novel semimechanistic model was developed to describe clinical pharmacokinetics of biologics under anti-drug antibody formation and associated loss of exposure. The model can be used to accurately establish clinical the dose-exposure-response relationship without excluding patients with loss of drug exposure, as well as to explore the relationship of patient covariates and dosing schedule on drug immunogenicity.  

This work highlights how modeling and simulation can leverage preclinical data to answer key clinical questions, such as the expected clinical benefit of a drug combination, the optimal range of doses for molecules with complex exposure-response relationships, and the design of improved molecules. These approaches offer valuable tools for data-driven drug development.