IntroductionDiscontinuation of tyrosine kinase inhibitor (TKI) treatment is emerging as the main therapy goal for Chronic Myeloid Leukemia (CML) patients. The DESTINY trial showed that TKI dose reduction prior to cessation can lead to an increased number of patients achieving sustained treatment free remission (TFR). However, there has been no systematic investigation to evaluate how dose reduction regimens can further improve the success of TKI stop trials.MethodsHere, we apply an established mathematical model of CML therapy to investigate different TKI dose reduction schemes prior to therapy cessation and evaluate them with respect to the total amount of drug used and the expected TFR success.ResultsOur systematic analysis confirms clinical findings that the overall time of TKI treatment is a major determinant of TFR success, while highlighting that lower dose TKI treatment for the same duration is equally sufficient for many patients. Our results further suggest that a stepwise dose reduction prior to TKI cessation can increase the success rate of TFR, while substantially reducing the amount of administered TKI.DiscussionOur findings illustrate the potential of dose reduction schemes prior to treatment cessation and suggest corresponding and clinically testable strategies that are applicable to many CML patients.

The transfer of new insights from basic or clinical research into clinical routine is usually a lengthy and time-consuming process. Conversely, there are still many barriers to directly provide and use routine data in the context of basic and clinical research. In particular, no coherent software solution is available that allows a convenient and immediate bidirectional transfer of data between concrete treatment contexts and research settings. Here, we present a generic framework that integrates health data (e.g., clinical, molecular) and computational analytics (e.g., model predictions, statistical evaluations, visualizations) into a clinical software solution which simultaneously supports both patient-specific healthcare decisions and research efforts, while also adhering to the requirements for data protection and data quality. Specifically, our work is based on a recently established generic data management concept, for which we designed and implemented a web-based software framework that integrates data analysis, visualization as well as computer simulation and model prediction with audit trail functionality and a regulation-compliant pseudonymization service. Within the front-end application, we established two tailored views: a clinical (i.e., treatment context) perspective focusing on patient-specific data visualization, analysis and outcome prediction and a research perspective focusing on the exploration of pseudonymized data. We illustrate the application of our generic framework by two use-cases from the field of haematology/oncology. Our implementation demonstrates the feasibility of an integrated generation and backward propagation of data analysis results and model predictions at an individual patient level into clinical decision-making processes while enabling seamless integration into a clinical information system or an electronic health record.

Risk stratification and treatment decisions for leukemia patients are regularly based on clinical markers determined at diagnosis, while measurements on system dynamics are often neglected. However, there is increasing evidence that linking quantitative time-course information to disease outcomes can improve the predictions for patient-specific treatment responses. We designed a synthetic experiment simulating response kinetics of 5,000 patients to compare different computational methods with respect to their ability to accurately predict relapse for chronic and acute myeloid leukemia treatment. Technically, we used clinical reference data to first fit a model and then generate de novo model simulations of individual patients’ time courses for which we can systematically tune data quality (i.e. measurement error) and quantity (i.e. number of measurements). Based hereon, we compared the prediction accuracy of three different computational methods, namely mechanistic models, generalized linear models, and deep neural networks that have been fitted to the reference data. Reaching prediction accuracies between 60 and close to 100%, our results indicate that data quality has a higher impact on prediction accuracy than the specific choice of the particular method. We further show that adapted treatment and measurement schemes can considerably improve the prediction accuracy by 10 to 20%. Our proof-of-principle study highlights how computational methods and optimized data acquisition strategies can improve risk assessment and treatment of leukemia patients.