PRETTY

Funding

The project PRETTY is funded by the Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung) with a total volume of approximately 1.4 million €. Of this, approximately 460.000 € will be made available to PLRI (MHH and TU-BS, funding code: 01KD2416A).

Summary

For many leukemia patients, an allogeneic stem cell transplantation currently represents the only curative option. However, up to 25% of transplanted patients die from transplantation-related causes such as infections, graft-versus-host disease (GvHD), or organ failure, e.g., due to nephrotoxicity as a side effect.

The PRETTY project aims to achieve the following objectives:

• Gain a better understanding of the risk factors for nephrotoxicity in individual leukemia patients with Acute Myeloid Leukemia (AML), Acute Lymphoblastic Leukemia (ALL), or Myelodysplastic Syndrome (MDS) undergoing hematopoietic stem cell transplantation (allogeneic stem cell transplantation, alloHCT)
• Develop a predictive model to support patient-individualized decision-making in therapy selection ("chemotherapy only" vs. "alloHCT")

Current challenges are presented by 1. the lack of a comprehensive dataset providing sufficient longitudinal information on nephrotoxicity (often no data available for the time period during or after alloHCT) and 2. the need to consider large patient cohorts, which are available in the form of decentralized datasets distributed across different centers.

PRETTY will enable continuous and prospective data integration into local Medical Data Integration Centers (MeDICs) in the participating clinical centers. Additionally, local personalized learning models for predicting nephrotoxicity as well as a federated model, where patient data does not leave the local centers, will be developed and compared. As a novelty compared to the "classical" purely data-driven approach, local expertise from treating physicians (physician perspective) and patients (patient perspective) will also be integrated into model learning, among other things through techniques for site-wide collaborative model evaluation and integration of patient-reported outcome data, which capture the subjectively experienced severity of nephrotoxic effects.

The overarching long-term goal of the project is to enable personalized prediction of cancer treatment toxicity (model-based clinical decision support), which supports the selection of an appropriate treatment and thus improves patient outcomes.