Abstract
BACKGROUND: Bone metastasis is the most frequent complication in prostate cancer patients and associated outcome remains fatal. Radium223 (Rad223), a bone targeting radioisotope improves overall survival in patients (3.6 months vs. placebo). However, clinical response is often followed by relapse and disease progression, and associated mechanisms of efficacy and resistance are poorly understood. Research efforts to overcome this gap require a substantial investment of time and resources. Computational models, integrated with experimental data, can overcome this limitation and drive research in a more effective fashion. METHODS: Accordingly, we developed a predictive agent-based model of prostate cancer bone metastasis progression and response to Rad223 as an agile platform to maximize its efficacy. The driving coefficients were calibrated on ad hoc experimental observations retrieved from intravital microscopy and the outcome further validated, in vivo. RESULTS: In this work we offered a detailed description of our data-integrated computational infrastructure, tested its accuracy and robustness, quantified the uncertainty of its driving coefficients, and showed the role of tumor size and distance from bone on Rad223 efficacy. In silico tumor growth, which is strongly driven by its mitotic character as identified by sensitivity analysis, matched in vivo trend with 98.3% confidence. Tumor size determined efficacy of Rad223, with larger lesions insensitive to therapy, while medium- and micro-sized tumors displayed up to 5.02 and 152.28-fold size decrease compared to control-treated tumors, respectively. Eradication events occurred in 65 ± 2% of cases in micro-tumors only. In addition, Rad223 lost any therapeutic effect, also on micro-tumors, for distances bigger than 400 μm from the bone interface. CONCLUSIONS: This model has the potential to be further developed to test additional bone targeting agents such as other radiopharmaceuticals or bisphosphonates.