Reconstructing tumor trajectories during therapy through integration of multiple measurement modalities

Background Accurately determining changes in tumor size during therapy is essential to evaluating response or progression. However, individual imaging methodologies often poorly reflect pathologic response and long-term treatment efficacy in patients with estrogen receptor positive (ER+) early-stage breast cancer. Mathematical models that measure tumor progression over time by integrating diverse imaging and tumor measurement modalities are not currently used but could increase accuracy in measuring response and provide biological insights into cancer evolution. Methods For ER+ breast cancer patients enrolled on a neoadjuvant clinical trial, we reconstructed their tumor size trajectories during therapy by combining all available information on tumor size, including different imaging modalities, physical examinations and pathological assessment data. Tumor trajectories during six months of treatment were generated, using a Gaussian process and the most probable trajectories were evaluated, based on clinical data, using measurement models that account for biases and differences in precision between tumor measurement methods, such as MRI, ultrasound and mammograms. Results Reconstruction of tumor trajectories during treatment identified five distinct patterns of tumor size changes, including rebound growth not evident from any single modality. These results increase specificity to distinguish innate or acquired resistance compared to using any single measurement alone. The speed of therapeutic response and extent of subsequent rebound tumor growth quantify sensitivity or resistance in this patient population. Conclusions Tumor trajectory reconstruction integrating multiple modalities of tumor measurement accurately describes tumor progression on therapy and reveals various patterns of patient responses. Mathematical models can integrate diverse response assessments and account for biases in tumor measurement, thereby providing insights into the timing and rate at which resistance emerges. ### Competing Interest Statement Qamar Khan declares research funding from Novartis. Adam Cohen declares research funding from Novartis. Ruth O Regan declares research funding from Pfizer, Novartis, Seattle Genetics, PUMA. Priyanka Sharma declares research funding from Novartis, Merck, Bristol Myers Squibb. Kari Wisinski declares research funding and clinical trial involvement with Novartis, Eli Lilly, Astra Zeneca, Sanofi and Pfizer. Kevin Kalinsky receives institutional support from Immunomedics, Novartis, Incyte, Genentech/Roche, Eli-Lilly, Pfizer, Calithera Biosciences, Acetylon, Seattle Genetics, Amgen, Zentalis Pharmaceuticals, and CytomX Therapeutics. All other authors have no conflicts of interest to disclose.