A machine learning model of response to hypomethylating agents in myelodysplastic syndromes

What’s Next after Hypomethylating Agents Failure in Myeloid Neoplasms? A Rational Approach

“…In an attempt to simplify the prediction of HMA responses, Nazha et al also developed a separate clinical-only ML-based model built through a multicenter effort by collecting serial complete blood counts (CBC) data from a training cohort of 424 patients who had received at least four cycles of HMA therapy [ 59 ]. The most influential parameter was a change from baseline in hemoglobin, followed by platelets, red cell distribution width (RDW), and white blood cell counts (WBC) [ 59 ].…”

“…In an attempt to simplify the prediction of HMA responses, Nazha et al also developed a separate clinical-only ML-based model built through a multicenter effort by collecting serial complete blood counts (CBC) data from a training cohort of 424 patients who had received at least four cycles of HMA therapy [ 59 ]. The most influential parameter was a change from baseline in hemoglobin, followed by platelets, red cell distribution width (RDW), and white blood cell counts (WBC) [ 59 ]. This externally validated approach accurately differentiated patients into categories with a very low, low, intermediate, high, or very high likelihood of responses solely on variations of CBC parameters, easily trackable during patients’ follow up [ 59 ].…”

“…The most influential parameter was a change from baseline in hemoglobin, followed by platelets, red cell distribution width (RDW), and white blood cell counts (WBC) [ 59 ]. This externally validated approach accurately differentiated patients into categories with a very low, low, intermediate, high, or very high likelihood of responses solely on variations of CBC parameters, easily trackable during patients’ follow up [ 59 ].…”

Hematopoietic Stem Cell Transplantation for Myelodysplastic Syndromes

“…Advances in diagnostic tools, such as machine learning, have implications in decreasing the latency between clinical suspicion of disease and diagnosis, which may translate to more prompt initiation of treatment in patients with high-risk disease and even identification of better conditioning regimens. [23][24][25] Survival rates for MDSs vary greatly and are significantly segregated by age with a median survival of 4.6 years for patients younger than 60 years and significantly lower for patients diagnosed at older than 60 years. 7,26 Patients older than 60 years often have higher degrees of frailty and have historically been less likely to proceed to ASCT, accounting for some of the discrepancy of outcomes after the age of 60 years.…”

“…Previously, Nazha et al 22 also proposed a personalized model to predict leukemia transformation and survival in MDS patients. Advances in diagnostic tools, such as machine learning, have implications in decreasing the latency between clinical suspicion of disease and diagnosis, which may translate to more prompt initiation of treatment in patients with high-risk disease and even identification of better conditioning regimens 23–25 …”