Background: Breast Cancer (BC) patients exhibit a wide variety of responses to neoadjuvant chemotherapy (NACT). This is driven by factors both intrinsic (e.g., mutations, dysregulation, metabolic reprogramming) and extrinsic (e.g., nutrient/drug perfusion, interactions with surrounding healthy tissues and the tumor microenvironment (TME)) to the cells that make up each tumor. The SimBioSys TumorScope is a platform for making individualized predictions of the response of each patient’s tumor to NACT. It employs 3D biophysical simulations that explicitly model the dynamics of cellular response to the ever-changing chemical milieu of drugs and nutrients that perfuse the TME during treatment, in order to predict when and where different regions of the tumor are growing, dying, and ultimately how a given patient will respond to treatment.
Methods: The SimBioSys TumorScope constructs 3D in silico models of each patient’s tumor directly from pretreatment DCE-MRIs. It combines this spatial model with personalized genome scale models of tumor and tissue metabolism, pharmacokinetics, and pharmacodynamics, and vascular perfusion (based on DCE-MRI timeseries). The combined model is then simulated using a custom high-performance reaction-diffusion- material mechanics simulation engine which produces a spatio-temporal trajectory of tumor size, morphology, intra- and extracellular biochemistry.
We evaluated the ability of the TumorScope software to predict volumetric response to NACT. A validation set comprising the pretreatment records (including MRIs) of 780 BC patients that underwent NACT was used. These patients spanned a wide range of tumor sizes, molecular subtypes, and NCCN-recommended treatment regimens. Simulations were initialized using each patient’s pretreatment MRI and pathology data and run from the start of therapy to the specified surgical date. Simulated tumor volumetric percent response (calculated as the ratio of change in tumor volume to initial volume) at the time of surgery was then compared with actual tumor volumetric percent response extracted from presurgical MRIs. Among patients for which event free survival data was available (n = 480), we performed a Cox proportional hazard analysis.
Results: The SimBioSys TumorScope predicted pre-surgical tumor volumetric response with a median error of 0.03% and median absolute deviation of 8.2%. Among the patients for which EFS data was available, we found a hazard ratio of 1.8 associated with having a final simulated volume greater than 0.01 cc (p = 0.00048).
Conclusions: The SimBioSys TumorScope produces accurate patient specific predictions of response to NACT using only standard-of-care pre-treatment data. Such predictions can aid in decision making, enabling physicians to select less-toxic regimens for patients in which a robust response is predicted, and more aggressive treatments and/or clinical trial enrollment when response is likely to be poor
