Abstract 2391: Towards better prediction of treatment resistance: Directly modeling progression from multiple lesions

Abstract

Abstract Late-stage cancers in the clinic present as a set of dispersed tumors, each with distinct growth kinetics. As most patients enrolled in trials of experimental anticancer therapeutics have late-stage malignancies, it is important to understand the impact of data analysis methodologies on the ability to detect treatment effect. In particular, we are interested in the potential risk posed by aggregating individual tumor measurements from multiple-lesion clinical data into a single tumor burden measurement, as this may cause a loss of information. In this work, we use quantitative modeling to assess how analyzing multiple lesions separately versus using an aggregated readout affects tumor response and detection of progression free survival and relapse. To assess the impact of data aggregation, we built a two-population clonal dynamics model consisting of sensitive and resistant tumor cells undergoing growth in the presence or absence of treatment. We performed parameter sweeps to assess the impact of individual model parameters on progression free survival and relapse. We then extended these sweeps to the multi-lesion situation, using nonlinear regression to compare parameter values of aggregated models and the parameter values of the individual models that comprised it. From our simulations, the most influential parameters in determining rebound are the growth rate of the resistant cells and the fraction of these cells, while growth rate of sensitive cells and growth rate inhibition affect effectiveness of treatment. In analyzing the multiple lesion model in different situations, we found that estimates for progression free survival were significantly more sensitive when the lesions were analyzed separately, rather than being aggregated. Our work suggests that aggregating the lesions into a single readout loses impact of individual lesion growth rate inhibitions, impairing early detection as the ability to detect fast growing, small, resistant lesions (despite overall tumor burden decreasing) is lost. Taken together, our findings suggest that data aggregation causes a loss of information. Relapse can be predicted at earlier times when using the multiple lesion model compared to an aggregated model. Citation Format: Samuel Protich, Madison Stoddard, Lin Yuan, Douglas White, Dean Bottino, Arijit Chakravarty. Towards better prediction of treatment resistance: Directly modeling progression from multiple lesions [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2391.