Machine Learning Predicts Methotrexate Response in RA

A machine learning approach that combined clinical and genomic biomarkers predicted response to methotrexate among patients with early rheumatoid arthritis with 76% accuracy, according to study findings published online ahead of print in Arthritis Care & Research.

The study included demographic, clinical, and genomic data for a total 643 patients of European ancestry with early rheumatoid arthritis. The machine learning training cohort consisted of 336 patients, and the validation cohort consisted of 308 patients. Genomic data included 160 single-nucleotide polymorphisms (SNPs) associated with rheumatoid arthritis or methotrexate metabolism.

Machine learning methods combining age, sex, smoking, rheumatoid factor, baseline Disease Activity Score in 28 joints (DAS28) scores, and SNPs predicted good or moderate response using European Alliance of Associations for Rheumatology criteria at 3 months, according to the study. In the training cohort, the approach had an area under the receiver operating curve of 0.84. In the validation cohort, the machine learning algorithm achieved a prediction accuracy of 76%, with 72% sensitivity and 77% specificity.

The variable importance of the intergenic SNPs rs12446816, rs13385025, rs113798271, and ATIC (rs2372536) was above 60.0, researchers reported. They were among the top predictors of methotrexate response, along with baseline DAS28 scores.

“Pharmacogenomic biomarkers combined with baseline DAS28 scores can be useful in predicting response to methotrexate in patients with early rheumatoid arthritis,” researchers concluded. “Applying machine learning to predict treatment response holds promise for guiding effective rheumatoid arthritis treatment choices, including timely escalation of rheumatoid arthritis therapies.”

—Jolynn Tumolo

Reference:
Myasoedova E, Athreya AP, Crowson CS, et al. Toward individualized prediction of response to methotrexate in early rheumatoid arthritis: a pharmacogenomics-driven machine learning approach. Arthritis Care Res. Version of Record published online April 6, 2022. https://doi.org/10.1002/acr.24834