The results of an international study published in the Journal of Clinical Oncology propose a new approach to improving the management of chronic myelomonocytic leukemia (CMML), a rare blood cancer characterised by high variability among patients and often unfavorable outcomes. By integrating molecular information, clinical parameters, and advanced computational models, researchers have developed tools capable of more accurately estimating disease progression and supporting personalised therapeutic decisions.

The large-scale, multicenter study is the result of collaboration among an international team involving research centers in Europe, Taiwan and the United States. The project was conceived and coordinated by Humanitas, which integrates clinical activity, oncohematology research and advanced artificial intelligence approaches, under the guidance of Matteo Giovanni Della Porta – Head of the Leukemia Unit at Humanitas Research Hospital and Professor at Humanitas University. Contributions also came from Saverio D’Amico – biomedical engineer and data scientist at the Humanitas AI Center, engaged in the development of data analysis toolsand from the first author Luca Lanino – hematologist at Yale School of Medicine and former resident at Humanitas University.

A rare and complex disease

CMML is characterised by an increase in monocytes in the blood, a type of white blood cell involved in the immune response, and shows marked clinical heterogeneity. It mainly affects the adult population and, in some cases, can evolve into acute myeloid leukemia, a more aggressive form of the disease.

Allogeneic hematopoietic stem cell transplantation currently represents the only potentially curative treatment; however, many patients are not eligible due to advanced age or the presence of comorbidities. Available drug alternatives, such as hypomethylating agents, offer limited benefits, highlighting the need for more precise prognostic tools and increasingly personalised therapeutic approaches.

A new genetic map of the disease

By analyzing clinical and genetic data from over 3,500 patients using a multimodal approach, researchers identified nine molecular clusters, each associated with specific genetic alterations and different clinical outcomes. A subset of patients (about 15%) also showed overlapping features with other myeloid neoplasms, suggesting that the boundaries between these diseases are less clearly defined than previously thought. “This new genetic map allows for a more precise description of the disease compared to traditional classifications, offering a tool to better understand differences among patients and the clinical variability we observe in practice,” explains Dr. Luca Lanino. “A more accurate classification provides the foundation for developing increasingly targeted therapeutic strategies.”

How risk assessment is changing

The collected data led to the development of the International CMML Prognostic Scoring System (iCPSS), a new prognostic system that integrates genetic mutations, hematological parameters, and cytogenetic alterations, that is abnormalities of the chromosomes that contain DNA.

The model identifies five prognostic groups, each with different probabilities of overall survival and progression to acute leukemia, improving individual-level predictive accuracy compared to previous systems. In fact, about 55% of patients were reclassified into different risk categories. “Integrating genetic and clinical information allows for a more precise assessment of prognosis and enables the personalisation of therapeutic decisions,” notes Prof. Matteo Giovanni Della Porta. “This approach makes it possible to better identify patients who may benefit from more intensive strategies, such as transplantation, and to plan the treatment pathway in a more targeted way.”

The role of Artificial Intelligence

An innovative element of the study is the use of advanced data analysis tools and decision-making models developed through artificial intelligence techniques, with the contribution of the Humanitas AI Center. Among its research lines is the development of Digital Twin, which are virtual representation of patients built by integrating clinical data, genomics, medical imaging, treatments, and outcomes, for understanding and managing diseases, particularly in the field of oncohematology. In this context, researchers implemented a federated learning platform, which allows continuous model updating using data from different centers without directly sharing sensitive patient information. The use of synthetic data also made it possible to simulate realistic clinical scenarios, testing and validating the model to support future applications.

Toward precision medicine

Taken together, these solutions demonstrate how the integration of clinical research, genomics, and artificial intelligence can support the development of dynamic prognostic tools that can be updated over time, with concrete implications for clinical practice.

A particularly relevant aspect concerns therapeutic strategies: adoption of the model changed treatment planning in 31% of cases, with an expected improvement in survival among patients eligible for more intensive therapies, such as stem cell transplantation.

Overall, the study represents an internationally significant achievement and a potential paradigm shift in the management of CMML, showing how precision medicine can become increasingly applicable in everyday clinical practice.