The prediCt consortium project, funded under the German CoSysMed2 programme (grant number 031L0136A) and carried out by the Technical University of Dresden from 1 July 2018 to 31 December 2021, aimed to create a patient‑specific decision‑support tool for chronic myeloid leukaemia (CML). The consortium combined clinical data from several trials with mechanistic mathematical modelling and software engineering to predict the risk of molecular relapse after tyrosine‑kinase inhibitor (TKI) dose reduction or cessation.
In the first work package (WP 1) the team redesigned the database schema and performed extensive statistical analyses of anonymised patient records. These analyses produced three peer‑reviewed publications, including a 2020 Blood article that showed that molecular monitoring during dose reduction can predict recurrence after TKI cessation, and a 2020 Leukemia paper that introduced a traffic‑light stratification model for deep molecular response. A 2021 Experimental Hematology paper discussed how treatment alterations may forecast patient‑specific therapy response. The data handling also involved the implementation of a DataVault‑based data warehouse, ensuring traceability, security, and scalability for the project’s clinical datasets.
WP 2 focused on developing and validating a mechanistic model of the leukemic–immune interaction in CML. The model incorporates the dynamics of BCR‑ABL‑positive leukemic cells, the effects of TKI therapy, and the patient‑specific immune response. Key results were published in the Bulletin of Mathematical Biology (2019), Cancer Research (2020), and a 2022 submission to a high‑impact journal. The 2020 Cancer Research paper demonstrated that model‑based inference can classify immunological control mechanisms from clinical data on TKI cessation and dose reduction. The 2022 manuscript suggested that immune‑mediated control could allow for treatment reduction before complete cessation, potentially reducing relapse rates.
In WP 3 the validated model was integrated into a user‑friendly software demonstrator, now available online at https://predict.imb.medizin.tu-dresden.de. The demonstrator accepts patient‑specific input data, runs the mathematical model, and outputs a risk score for molecular relapse, thereby providing clinicians with a quantitative tool to guide TKI tapering decisions. A 2022 manuscript, currently near submission, describes the integration of in silico predictions into routine clinical workflows and highlights the demonstrator’s potential to improve personalized CML care.
The project’s timeline required adjustments due to COVID‑19‑related delays in data acquisition and patient recruitment. These changes were documented in annual progress reports, and a neutral extension request was approved on 18 May 2021. The extension did not affect the achievement of the project’s core objectives.
Collaboration involved clinical partners supplying longitudinal patient data, mathematical biologists developing the immune‑leukemia model, and software engineers creating the demonstrator. The consortium’s interdisciplinary structure enabled a seamless transition from data collection to model validation and finally to a deployable clinical decision‑support system. The resulting tool offers a scientifically grounded, patient‑specific prediction of relapse risk, supporting clinicians in optimizing TKI therapy duration and potentially improving long‑term outcomes for CML patients.