Machine-Learning Model Predicts Treatment Response in Patients with MDS

Article

Timely identification of non-responders could reduce toxicities and costs, investigators wrote.

ai brain

A new machine-learning model can predict response to hypomethylating agents in patients with myelodysplastic syndromes, according to new research presented at the American Society of Hematology (ASH) 2019 Annual Meeting in Orlando.

Investigators from the Cleveland Clinic developed a clinical artificial intelligence (AI) model to predict response and resistance to hypomethylating agents after 90 days of initiating therapy. The model is based on changes in blood counts using time series analysis technology.

The team trained the AI using absolute values and changes in complete blood count values.

Nathan Radakovich, MD, from the Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, and his team screened 107 patients with myelodysplastic syndromes who received hypomethylating agents at the Cleveland Clinic from 2005—2013. Patients included had regular complete blood count draws during the treatment.

The investigators collected nearly 20,800 unique data points, including complete blood count, clinical, and genomic data.

The median age of patients included was 69 years old, between a range of 37—100. Fewer than 30 patients were female.

From the cohort, 37.4% of patients were very low or low risk; 29.9% were intermediate; 17.8% were high; 14.9% were very high risk, according to the Revised International Prognostic Scoring System for myelodysplastic syndromes Risk Assessment Calculator (IPSS-R).

Of the 107 patients, 57% received only azacitidine, 17.8% received only decitabine, 3.7% received both, and 21.5% received hypomethylating agents with an additional agent.

“While the hypomethylating agents azacitidine and decitabine improve cytopenias and prolong survival in (myelodysplastic syndromes) patients, response is not guaranteed,” the investigators wrote. “Timely identification of non-responders could prevent prolonged exposure to ineffective therapy, thereby reducing toxicities and costs.”

The model’s area under the curve was 0.95 in the training cohort, and 0.83 in the validation cohort. When the investigators added more patients to the cohort, the validation increased to an area under the curve of 0.89.

The model can be used to develop novel trial designs, so that patients who were predicted to not respond after 90 days of hypomethylating agent treatment could be assigned to an investigational agent, Radakovich and colleagues suggested.

The AI can also help decide whether a patient who was predicted to respond should continue hypomethylating agent therapy.

The investigators increased the sample to increase model accuracy and are developing a larger cohort of patients treated at different institutions.

The study, titled "Predicting Response to Hypomethylating Agents in Patients with Myelodysplastic Syndromes Using Artificial Intelligence," was presented at ASH 2019.

Related Videos
Addressing HS Risks at the Genetic Level, with Kai Li, BSc
Building a Psoriasis Knockout Regimen Around Risankizumab, with Andrew Blauvelt, MD, MBA
Joel Gelfand, MD, MSCE: Phototherapy Utility in Psoriasis
Pediatric Hidradenitis Suppurativa Severity not Linked to Obesity
Rizankizumab and the KNOCKOUT Study, with Andrew Blauvelt, MD, MBA
© 2024 MJH Life Sciences

All rights reserved.