The outcome of myelodysplastic syndrome (MDS) varies. The disease can lie dormant for decades, or it can become fatal in a matter of months. According to a recent article in ESMO Oncology News, the new model, which was created by Dr. Aziz Nazha in the Cleveland Center for Clinical Artificial Intelligence (AI), has removed some of the guesswork by its ability to predict a patient’s overall survival.

The model, which is based on genomic and clinical data, can be used separately or in combination with established models to predict MDS patients’ outcomes.

The model predicts the survival and probabilities of leukemia transformation at various time points. It has the ability to stratify patients into more accurate risk categories. The model is also a useful tool for enrolling patients in clinical trials. The study was published in the August 18th issue of the Journal of Clinical Oncology.

A More Personalized Prognosis

MDS, a type of cancer, may occur in the bone marrow when its cells become abnormal.

In analyzing MDS, patients are currently divided into several risk categories by testing samples of bone marrow and peripheral blood. The models used in most clinical practices are the International Prognostic Scoring System (IPSS) and a revised version (IPSS-R). Molecular data were added to the scoring systems, resulting in improved accuracy.

Further, in the new study, the researchers relied on an algorithm using clinical, molecular, and pathologic variables including interactions among each of the variables with the resulting prediction model providing a personalized prognosis specific to each patient.

A total of 1,471 MDS patients with well-annotated molecular and clinical data were enrolled in a group that was analyzed using machine learning techniques.

A prediction model was created and validated against external groups. After extensive comparison studies, the authors were confident that they had created and validated a model that can provide probable survival data throughout an MDS disease course.

Note that prognostic systems incorporating analytics of genomic, clinical, and pathologic data are able to give a more accurate prediction of survival in MDS patients who may receive a variety of therapies throughout the course of their disease.

According to the authors, their model outperformed existing models currently used for patient eligibility in clinical studies and will provide more accurate timing for transplants.