HealthDay News — Machine learning (ML) integrating clinical and genomic data can be used to predict methotrexate treatment response in early rheumatoid arthritis (RA), according to a study published online Dec. 13 in Arthritis Care & Research.
Elena Myasoedova, M.D., Ph.D., from the Mayo Clinic in Rochester, Minnesota, and colleagues examined the ability of ML approaches with clinical and genomic biomarkers to predict treatment response to methotrexate for patients with early RA. Demographic, clinical, and genomic data were obtained from 643 patients with early RA subdivided into training and validation cohorts (336 and 307, respectively). At three-month follow-up, a response to methotrexate monotherapy was defined as good or moderate by the European League Against Rheumatism (EULAR) response criteria.
The researchers found that EULAR response at three months could be predicted by supervised ML methods combining age, sex, smoking, rheumatoid factor, baseline Disease Activity Score with 28-joint count (DAS28), and 160 single nucleotide polymorphisms previously associated with RA or methotrexate metabolism, with an area under the receiver operating characteristic curve of 0.84 in the training cohort. Prediction accuracy was 76 percent in the validation cohort (sensitivity of 72 percent; specificity of 77 percent). The intergenic rs12446816, rs13385025, rs113798271, and ATIC (rs2372536) had variable importance above 60.0 and were among the top predictors of methotrexate response, along with baseline DAS28.
“Predicting a response to rheumatoid arthritis medication can be challenging, but this approach is very promising and is an exciting development in treating the disease,” Myasoedova said in a statement.