Abstract Tumor dormancy, early relapse, and treatment resistance remain poorly understood. After neoadjuvant or adjuvant therapy, the locally advanced and systemically disseminated invisible tumor cells may persist undetected for months or years despite clean MRI or CT scans. Reported detection thresholds and residual tumor sizes vary in the literature, further complicating prognostic accuracy in patient risk stratification. To address this unmet clinical need, we developed mathematical models and generated a computational algorithm to predict the time course of invisible tumor relapse, improve predictive accuracy, and model the invisible-to-visible tumor transition within the detection limits of the MRI/CT tumor imaging tools. We simulated remission-to-relapse trajectories for invisible residual tumors beginning with 1, 10, 102, 103, and 104 cells, and calculated the time for each to progress into a 3 cm3 visible tumor. Exponential growth models with unrestricted doubling times from 2 to 30 days were used to capture rapid, unimpeded tumor growth. A modified logistic model was then used to model the time course of invisible tumor growth in vivo, with doubling times of 5 to 30 days to reflect varying aggressiveness, tumor and tumor microenvironment interactions, and tempo-spatial-context-treatment dependence in response to varied tumor remission and extended tumor latency. To improve mathematical modeling, we refined, verified, and authenticated several variable factors/clinicopathological parameters in experimental biology by fitting our hypothetical tumor growth models to the real-time TNBC tumor growth curves from six patient-derived xenograft (PDX) TNBC models in NSG mice. By varying the clinically invisible residual tumor sizes from 10-6 mm3 to 10-2 mm3, we found that tumors relapsed rapidly once tumor growth resumed, often exceeding the clinical detection thresholds before the regularly scheduled follow-up appointments and allowable MRI/CT-imaging intervals could capture them. Our results highlight the hidden risk of residual diseases, the major challenge of invisible tumor modeling, and the lack of accuracy in relapse prediction in the clinic. Exponential and logistic model fits to the PDX tumor implantation data were complete; the model authentication, relapse prediction augmentation, and correct interpretation of these fitted models will be augmented with the new chemo-resistant PDX models in NSG mice. This work underscores the unmet need and critical importance of modeling invisible tumor growth to improve the accuracy of tumor relapse prediction and patient risk stratification. Future work will focus on transforming this mathematical model into a companion prognostic tool in combination with our selected biomarker panel to predict tumor relapse risk, quantify therapeutic efficacy, and assist and support oncologists’ decision-making in real time in the future. Citation Format: Bryan A. Hawickhorst, Daniel J. McWilliams, Jonathan M. Baker, Amy H. Tang, . Dynamic modeling of invisible tumor growth with simulated residual disease and PDX data abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6850.
Hawickhorst et al. (Fri,) studied this question.