Modeling and analysis of curative combination therapy for Diffuse Large B-Cell Lymphoma

Abstract This goal of this project is to develop computer simulations of combination therapy for aggressive lymphomas, and apply them to understand past positive and negative trials, and to enable model-guided design of new regimens. Theoretical models of cancer drug response and resistance evolution have provided many

conceptual insights, but models of a `representative' tumor have not been able to predict response distributions in heterogeneous human populations. We recently developed simulations which use clinically observed distributions of single-drug responses to predict distributions of multi-drug responses. These models have been

validated by accurately predicting many trial results in solid cancers (9 FDA approvals), but our prior models were too simple to describe curative treatments because they lacked intra-tumor heterogeneity and kinetics. Here we propose conceptual and technical advances to model the complexity of curative therapies, using

Diffuse Large B-Cell Lymphoma (DLBCL) as a case study. Our simulations of multi-drug response will consider inter-patient and intra-tumor heterogeneity, tolerability and dosage, treatment schedule and response kinetics, and drug interactions. We adopt the conceptual approach of population-pharmacokinetics, where each

parameter has a distribution describing its variance among patients. We will apply this approach to tumor drug response, considering both intra-tumor and inter-patient variation. Parameter distributions are informed by our experimental data from clone tracing and liquid biopsies to quantify clonal heterogeneity and response kinetics,

and digitization of decades of trial data. Aim 1 will analyze past and current trials of drug combinations in first- line DLBCL to test whether the clinical efficacy of drug combinations is predictable from single drug efficacy. Preliminary data shows the past 20-years of novel combination trials in first-line DLBCL confirm model

accuracy, and we prospectively predicted the first success in 2 decades. This aim will produce predictive models that can help design future drug combinations. Aim 2 will investigate explanations for the negative result of trial that added a targeted therapy, ibrutinib, to standard chemotherapy. We will model the influence of

tolerability and dose reductions, enrollment bias, and treatment schedule, comparing model outputs with real- world analyses of how these factors affect outcome. By understanding causes of trial failures this aim can help future trial designs to overcome these problems. In Aim 3 we will collaborate with the ECOG-ACRIN trial group

to apply model-guided design to a trial of precision combination therapy in first-line DLBCL. The LymphoMatch trial aims to match 5 subtypes of DLBCL to 5 targeted therapies, combined with standard chemotherapy. We will use clinical data on single drug efficacies, combination tolerability, subtypes' prognoses, and accuracy of

subtypes as biomarkers of drug sensitivities, to forecast trial results and so optimize design of regimens and endpoints. This research will deliver innovative multi-scale models of tumor heterogeneity, to solve challenges in the design of novel combination therapies and clinical trials that aim to cure more cases of lymphoma.