iCPSS Guides Transplant Timing for CMML Patients

A multidisciplinary team built the international CMML Prognostic Scoring System (iCPSS), a machine-learning-backed clinical decision tool that combines clinical features and genomic data from an international cohort of more than 3,000 patients to inform selection and timing for allogeneic stem cell transplant in chronic myelomonocytic leukemia (CMML).

CMML is genetically heterogeneous and clinically biphasic: a myelodysplastic (low white blood cell count) presentation and a myeloproliferative (high white blood cell count) presentation. Disease phenotype can evolve over time, affecting prognosis and therapeutic windows. Allogeneic stem cell transplant is the only curative option but carries substantial mortality and morbidity risk (infections, graft-versus-host disease), making accurate, individualized risk–benefit estimation essential.

iCPSS was trained and validated on clinical and routinely collected genetic mutation data from an international set of >3,000 patients, primarily from North America and Europe. The model outputs prognostic stratification that helps clinicians identify which patients are likely to derive survival benefit from transplantation and suggests optimal timing for referral and transplant discussion. The study includes the practical observation that genetic profiling used by the tool is already part of standard CMML workups. The team reports that iCPSS can change clinical decision pathways by indicating when earlier transplant discussion is warranted versus when observation or non-transplant therapies are appropriate. The paper’s first author, a Yale postdoctoral associate, summarized: “Our tool can help physicians decide when is the best time to discuss transplantation with their patients.”

For ML practitioners in clinical research, iCPSS is an example of a real-world prognostic model that integrates genomics and longitudinal clinical phenotypes at scale and is aimed at an actionable clinical decision (transplant timing). For hematologists and clinical teams, the tool promises to reduce overtreatment and better allocate transplant resources by stratifying transplant benefit. The reliance on routinely collected genomic data increases feasibility for clinical deployment but raises questions about external validation across underrepresented populations and implementation into workflows.

peer-reviewed publication details (model architecture, feature importance, calibration and decision-curve analysis), external validation in non-Western cohorts, prospective impact studies, and UI/EMR integration for shared decision-making.