Abstract Introduction: Heterogeneity between patients and within tumors poses challenges to effective cancer treatment. Drug combinations can overcome heterogeneity by increasing the probability that a patient will have a response to one or more drugs. This approach is successful in numerous disease settings including multiple lymphomas, but technological advances have led to a new approach to addressing heterogeneity: precision approaches that stratify patients into subpopulations with more uniform, stronger responses to specific treatments. These strategies can be implemented together to improve patient outcomes. We have developed computational mechanistic models of cancer treatment in the context of both sources of heterogeneity which explain and predict clinical outcomes of treatment of Large B-cell lymphoma (LBCL) with combination chemoimmunotherapy and CAR T-cell therapy. Combination therapy: Our novel population tumor kinetics (pop-TK) model simulates clinical combination therapy outcomes by implementing multidrug dose response functions in heterogeneous populations of tumor cells, within heterogeneous cohorts of patients. Informed by the outcomes of novel drugs in phase 2 trials in relapsed/refractory (r/r) patients, the pop-TK model accurately predicted of outcomes of seven first-line phase 3 trials of novel combinations. Most notably, informed by treatment of r/r LBCL with polatuzumab-vedotin (pola), the pop-TK predicted the success of modifying the standard treatment regimen (RCHOP) with pola before the clinical trial results were published. The pola combination improves overall survival (OS) and PFS in Activated B-Cell (ABC) LBCL (hazard ratio (HR): OS 0.3, PFS 0.4), but reduces OS (HR: 1.6) and does not improve PFS (HR: 1.0) in Germinal Center B-Cell (GCB) LBCL. The subtype-specific difference in PFS was also predictable based on subtype results of r/r trials (predicted HR: ABC 0.3, GCB 0.9). CAR T-cell therapy: We used a population pharmacokinetic/pharmacodynamic (pop-PK/PD) model of CAR T-cell therapy to reproduce the PFS distribution and the distribution CAR T-cell abundance for CAR T therapy in r/r LBCL patients. The impact of tumor size on patient outcomes in our model and in clinical studies led us to computationally explore the impact of treating smaller tumors with CAR T-cell therapy on treatment outcomes. We found that using detectable circulating tumor DNA (ctDNA) as a biomarker for administering CAR T-cell therapy immediately after RCHOP increased the proportion of simulated patients without a second disease progression event by 30% (HR: 0.43 [0.34, 0.55]). Conclusions: The pop-TK model supports LBCL cell-of-origin subtype as a biomarker for pola performance, and the pop-PK/PD model demonstrates that detectable ctDNA is a potential biomarker for early CAR T-cell administration. These examples of model-driven precision trial design have the potential to improve outcomes using readily available biomarkers and existing therapies. Citation Format: Amy E. Pomeroy, Adam C. Palmer. Mechanistic models of cancer heterogeneity explain and predict clinical outcomes of Large B-Cell Lymphoma (LBCL) treatment [abstract]. In: Proceedings of the Fourth AACR International Meeting on Advances in Malignant Lymphoma: Maximizing the Basic-Translational Interface for Clinical Application; 2024 Jun 19-22; Philadelphia, PA. Philadelphia (PA): AACR; Blood Cancer Discov 2024;5(3_Suppl):Abstract nr PO-005.

Abstract PO-005: Mechanistic models of cancer heterogeneity explain and predict clinical outcomes of Large B-Cell Lymphoma (LBCL) treatment