Site Editors
Shaji K. Kumar, MD
Prashant Kapoor, MD, FACP
Shaji K. Kumar, MD
Prashant Kapoor, MD, FACP
Posted: Tuesday, January 16, 2024
Using machine learning to integrate readily available clinical and laboratory data may accurately predict M-spike values in patients who have multiple myeloma more rapidly than can the current “gold standard” of serum protein electrophoresis with immunofixation, according to James J. Driscoll, MD, PhD, of Case Comprehensive Cancer Center, Cleveland, and colleagues, writing in JCO Clinical Cancer Informatics. What’s more, their algorithm seems to recognize changes in clinical and laboratory data related to critical disease fluctuations, even in patients with low disease volume (ie, M-spike values of less than 1 g/dL).
Currently, the team noted, treatment decisions based upon standard methods are delayed because the turnaround time for results is between 3 and 7 days. With time-sensitive machine learning tools, treatment plans may be established more quickly, they propose, which could in turn positively “affect patient outcomes… and [potentially] thwart significant comorbidities.”
Dr. Driscoll and co-investigators performed a retrospective chart review using the de-identified electronic medical records of 171 patients with multiple myeloma. With random forest analysis, they identified the weighted value of each independent variable (n = 43) integrated into the machine learning algorithm. “The machine learning–predicted M-spike values correlated highly with laboratory-measured serum protein electrophoresis values,” they wrote.
Of note, the results showed that machine learning tools may feasibly monitor disease burden longitudinally, but they have another socially equalizing benefit, according to the study authors. Such tools “support seamless, secure exchange of patient information to expedite and personalize clinical decision-making,” they added. According to the investigators’ findings, these tools may “overcome geographic, financial, and social barriers that currently limit the access of underserved populations to cancer care specialists, so the benefits of medical progress are not limited to selected groups.”
Disclosure: The study authors’ disclosure information can be found at ascopubs.org.
JCO Clinical Cancer Informatics
