University of Göttingen Researchers Use Machine Learning to Predict Side Effects in Multiple Myeloma Treatment, Enabling Safer Outpatient Stem Cell Mobilization

University of Göttingen researchers use machine learning to identify low-risk windows for multiple myeloma patients, enabling safer outpatient stem cell therapy.

By: AXL Media

Published: Apr 1, 2026, 4:37 AM EDT

Source: Information for this report was sourced from University of Göttingen

The Challenge of Stem Cell Mobilization in Multiple Myeloma

Multiple myeloma is a complex blood cancer characterized by the uncontrolled growth of plasma cells in the bone marrow. While there is currently no cure, "autologous stem cell transplantation" remains a cornerstone therapy for stabilizing the disease. This intensive process involves a "mobilization phase," where chemotherapy is used to flush stem cells from the bone marrow into the bloodstream for collection. Historically, patients have been required to spend two to three weeks in the hospital during this phase to ensure that life-threatening side effects, such as acute kidney failure or severe infections, can be managed immediately.

Questioning the Necessity of Long Hospital Stays

The traditional inpatient approach, while safe, is often a significant burden on a patient's quality of life and hospital resources. Researchers from the Göttingen Campus Institute for Dynamics of Biological Networks (CIDBN) observed that many patients either develop side effects very late in the process or do not experience serious complications at all. This prompted the question: Is a multi-week hospital stay necessary for every individual? By evaluating the data of 109 patients treated at the University Medical Center Göttingen, the team sought to create a more nuanced, individualized risk assessment.

Machine Learning as a Predictive Roadmap

Using advanced machine learning methods, the researchers identified specific time windows where the probability of serious side effects was extremely low for the majority of the study group. They then developed predictive models designed to forecast precisely which side effects might occur in which individuals and at what time. According to the study published in npj Digital Medicine, these models achieved high accuracy for certain categories of adverse events, providing a data-backed "therapy roadmap" for clinicians.