# Efficacy and Safety of Tofacitinib in Patients With Rheumatoid Arthritis and Inadequate Response to Methotrexate: A Real-World Study

## TL;DR

Tofacitinib helps rheumatoid arthritis patients who don't respond well to methotrexate, with good results and manageable side effects in real-world settings.

## Contribution

This study evaluates tofacitinib's real-world efficacy and safety in RA patients with methotrexate failure and explores machine learning for predicting treatment response.

## Key findings

- Tofacitinib significantly reduced disease activity and improved physical function in RA patients.
- Adverse events were mostly mild, with infections being the most common.
- Machine learning models identified CRP, disease duration, and baseline DAS28 as key predictors, but predictive accuracy was limited.

## Abstract

Background: Despite advances in pharmacologic management, achieving sustained remission in rheumatoid arthritis (RA) remains challenging, particularly among patients who exhibit suboptimal responses to conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) such as methotrexate (MTX). The emergence of targeted synthetic disease-modifying antirheumatic drugs (tsDMARDs), including Janus kinase (JAK) inhibitors, has expanded therapeutic options by directly modulating intracellular signaling pathways central to inflammation and immune activation. Among these, tofacitinib has demonstrated efficacy in clinical trials, yet its real-world performance, safety profile, and predictors of treatment response are less clearly defined. Real-world data, reflecting diverse patient populations and routine clinical practice, are essential for complementing randomized controlled trials (RCTs) and for guiding evidence-based, individualized treatment strategies in RA management.

Methods: This retrospective real-world study included 450 RA patients treated with tofacitinib following MTX failure. Demographics, clinical characteristics, laboratory results, treatment history, and comorbidities were assessed. Outcomes included changes in disease activity score 28 (DAS28), Health Assessment Questionnaire (HAQ), pain, and stiffness at six months, along with adverse event monitoring. Exploratory data analysis and machine learning models (logistic regression, random forest, XGBoost, LightGBM, and support vector machine {SVM}) were applied to predict treatment response.

Results: The mean age was 51.2±12.4 years, with 236 females (52.4%). At six months, DAS28 decreased significantly (5.4±1.1-3.2±1.0; p<0.001), with 286 patients (63.6%) achieving low disease activity and 142 patients (31.6%) reaching remission. HAQ improved from 1.5±0.6 to 0.9±0.5 (p<0.001). Adverse events occurred in 132 patients (29.3%), mostly mild infections. Machine learning models identified CRP, disease duration, and baseline DAS28 as key predictors; however, the predictive performance for treatment response was generally limited, with most models showing area under the curve (AUC) values below 0.65.

Conclusion: Tofacitinib demonstrated significant clinical benefit in RA patients with inadequate MTX response, with acceptable safety. While machine learning highlighted key predictors, future work with larger datasets is needed to optimize predictive accuracy and personalize therapy.