Clinical predictors of response to methotrexate in patients with rheumatoid arthritis: a machine learning approach using clinical trial data

Abstract

Background Methotrexate is the preferred initial disease-modifying antirheumatic drug (DMARD) for rheumatoid arthritis (RA). However, clinically useful tools for individualized prediction of response to methotrexate treatment in patients with RA are lacking. We aimed to identify clinical predictors of response to methotrexate in patients with rheumatoid arthritis (RA) using machine learning methods. Methods Randomized clinical trials (RCT) of patients with RA who were DMARD-naïve and randomized to placebo plus methotrexate were identified and accessed through the Clinical Study Data Request Consortium and Vivli Center for Global Clinical Research Data. Studies with available Disease Activity Score with 28-joint count and erythrocyte sedimentation rate (DAS28-ESR) at baseline and 12 and 24 weeks were included. Latent class modeling of methotrexate response was performed. The least absolute shrinkage and selection operator (LASSO) and random forests methods were used to identify predictors of response. Results A total of 775 patients from 4 RCTs were included (mean age 50 years, 80% female). Two distinct classes of patients were identified based on DAS28-ESR change over 24 weeks: “good responders” and “poor responders.” Baseline DAS28-ESR, anti-citrullinated protein antibody (ACPA), and Health Assessment Questionnaire (HAQ) score were the top predictors of good response using LASSO (area under the curve [AUC] 0.79) and random forests (AUC 0.68) in the external validation set. DAS28-ESR ≤ 7.4, ACPA positive, and HAQ ≤ 2 provided the highest likelihood of response. Among patients with 12-week DAS28-ESR > 3.2, ≥ 1 point improvement in DAS28-ESR baseline-to-12-week was predictive of achieving DAS28-ESR ≤ 3.2 at 24 weeks. Conclusions We have developed and externally validated a prediction model for response to methotrexate within 24 weeks in DMARD-naïve patients with RA, providing variably weighted clinical features and defined cutoffs for clinical decision-making.

Machine learning identifies a profile of inadequate responder to methotrexate in rheumatoid arthritis

“…In rheumatology, recent initiatives used genetic data and machine learning to predict response to MTX [10][11][12]. However, implementing these models in clinical practice remains a challenge because genetic data are unavailable in usual practice.…”

Efficacy and gastrointestinal tolerability of methotrexate in late-onset rheumatoid arthritis patients: a prospective cohort study

“…Yet, it still seems not possible to predict the outcomes of patients on methotrexate based on biomarkers or genetic investigations. A retrospective analysis of clinical predictors of methotrexate response conducted by Duong et al showed that baseline disease activity score-28-eryhtrocyte sedimentation rate (DAS28-ESR), positive ACPA, and health assessment questionnare (HAQ) score were the predictors for drug response [36]. Our study demonstrated that only higher DAS 28 score is the predictor for remission in the methotrexate therapy both in LORA and YORA patients.…”

Specific recommendations to improve the design and conduct of clinical trials

“…Thus, they do not seem to improve prediction of therapeutic response [ 41 ]. However, machine learning methods that utilize approaches such as the least absolute shrinkage and selection operator and random forests methods may identify new previously unrecognized predictors of disease severity and response to a therapy [ 42 ]. Clinicians and clinical trial experts in each specialty should contemplate using these “big data” to create expanded risk models that include more than the obvious risk factors.…”