Myasoedova et al., 2021

Toward Individualized Prediction of Response to Methotrexate in Early Rheumatoid Arthritis: A Pharmacogenomics-Driven Machine Learning Approach

At a glance

Objective

To test the ability of ML approaches with clinical and genomic biomarkers to predict MTX treatment response in patients with early RA.

Related articles

For more articles on prediction of MTX treatment response, see Duong et al., 2022, Sergeant et al., 2018, Castrejón et al., 2016.

Link

DOI: https://doi.org/10.1002/acr.24834

Background

• MTX may be the only drug necessary to control RA for some, but in 30-40% there is no response
• 50% of patients discontinue MTX within 3-5 years (Aletaha & Smolen, 2002)
• Composite disease activity measures reflect a clinically meaningful target of reaching low disease activity (see Van Gestel et al., 1996, Van Gestel et al., 1998, and Fransen & van Riel, 2005)

Methods

• Demographic, clinical and genomic data of 643 european-ancestry patients with early RA
• Response to MTX was defined as good or moderate by the EULAR response criteria at the 3-month follow-up
• Random forests were trained and prediction performance was evaluated with AUC

Results

• In-sample AUC 0.84, out-of-sample prediction accuracy 76% (with p values for both, lol?)
• Intergenic SNPs had variable importance above 60 (whatever that means), and among with baseline DAS28 were the top predictors of MTX response

Conclusion

• Pharmacogenomic biomarkers combined with baseline DAS28 can be useful in predicting response to MTX in patients with early RA
• ML is promising in this area, potentially able to predict timely escalation of treatment strategies in early RA