Model-Based Prediction Of Patient-Specific Residual Disease Levels For Imatinib-Treated Chronic Myeloid Leukemia.

6604 Background: Molecular response to imatinib (IM) therapy in chronic myeloid leukemia (CML) patients is associated with a biphasic decline of BCR-ABL transcript levels exhibiting a substantial heterogeneity. It is of clinical interest to predict the time courses over several years to decide about treatment discontinuation. Methods: We here present a mathematical model of hematopoietic and leukemic stem cell organization, which can explain CML dynamics and the biphasic decline in response to IM treatment accounting for inter-patient heterogeneity. Model parameters are estimated using 7-year follow-up data from a prospective randomized clinical trial (German IRIS cohort). Results: Based on data on the initial decline of BCR-ABL transcript levels, our model predicts individual long-term response to IM treatment. Furthermore, the model provides estimates on the remaining number of residual leukemic stem cells (LSCs), which is important to decide about treatment termination. In order to test the validity of the model, we provide predictions for an independent data set (German CML IV trial). Whereas we find that the majority of patients (66%) will require IM therapy for more than 30 years to achieve a complete eradication of all LSCs, a subset of patients (10%) is predicted to achieve the complete LSC eradication within less than 15 years of continuous IM administration. Conclusions: We conclude that our mathematical model is able to provide individual long-term predictions on IM-treated CML. Although developed and validated for IM treatment, the proposed model will be applicable to forthcoming molecular response data from first-line second-generation tyrosine kinase inhibitor therapy.