A robust machine learning approach to predicting remission and stratifying risk in rheumatoid arthritis patients treated with bDMARDs

Aug 5, 2026

Upload by Sage

Rheumatoid arthritis is a chronic autoimmune disease where biological disease-modifying antirheumatic drugs (bDMARDs) are effective yet costly, and up to 40% of patients fail to achieve remission within six months. This study develops a robust machine learning framework using baseline routine clinical data to predict six-month remission in patients starting bDMARD therapy. Multiple models are compared, calibrated probability estimates are produced for actionable risk stratification, and SHAP explains key predictors. Results are evaluated with external validation on an independent hospital cohort to support clinical integration.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Introduction

Clinical problem and need for prediction

Challenges for clinical implementation

Methods

Data sources and model training

Calibration and probability reliability

Explainability with SHAP

Results

Model performance and calibration outcomes

Key predictors for remission

Risk stratification categories

Validation and Clinical Impact

External validation and robustness

Implications for routine workflows

[www. nature.com/scientificreports](www. nature.com/scientificreports) OPEN A robust machine learning approach to predicting remission and stratifying risk in rheumatoid arthritis patients treated with bDMARDs Fatemeh Salehi1􀀍, Emmanuelle Salin1, Benjamin Smarr2, Sara Bayat3, Arnd Kleyer4, Georg Schett3, Ruth Fritsch-Stork5,6 & Bjoern M. Eskofier1,6 Rheumatoid arthritis (RA) is a chronic autoimmune disease affecting millions worldwide, leading to inflammation, joint damage, and reduced quality of life. Although biological disease-modifying antirheumatic drugs (bDMARDs) are effective, they are costly, and up to 40% of patients do not achieve remission within six months. Accurate prediction of treatment response is crucial for optimizing care, minimizing side effects, and enhancing cost efficiency. This study proposes a robust machine learning framework for predicting six-month remission in RA patients using baseline routine clinical data. The framework also integrates risk stratification and explainability to enhance its clinical applicability. We evaluated multiple machine learning models, AdaBoost, Random Forest, XGBoost, and Support Vector Machines, using data from Austrian RA patients. We externally validated the results on an independent dataset from the Erlangen Hospital. To improve the reliability of probability estimates for actionable risk stratification, we employed calibration techniques, including Platt scaling, Isotonic regression, Beta calibration, and Spline calibration. We generated calibration curves to assess and visualize the alignment between predicted probabilities and observed outcomes. In addition, we used SHapley Additive exPlanations (SHAP) to analyze the contributions of different patient characteristics to the prediction of RA remission. AdaBoost demonstrated stronger performance than the other models, achieving an accuracy of 85.71% and a Brier score of 0.13 with isotonic regression calibration. SHAP identified DAS28, visual analog scales (VAS), age, and swollen joint count (SJC) as important characteristics for the prediction of RA remission. We also stratified patients into low-, medium-, and high-risk categories based on model predictions to support follow-up scheduling and treatment prioritization. Our framework predicts RA remission before the initiation of bDMARD therapy. It enables personalized care, actionable risk stratification, and optimized resource allocation. Its robustness was validated on two different individual cohort datasets, which highlights its potential for integration into routine clinical workflows. Rheumatoid arthritis (RA) is a chronic inflammatory condition that affects 40 to 80 million people worldwide. It primarily impacts the small joints ofthe hands and feet, causing pain, stiffness, and reduced mobility. Beyond joint damage, RA can exhibit systemic features, causing significant physical disability and a diminished quality of life1–3. Biological disease-modifying antirheumatic drugs (bDMARDs) are highly effective in treating RA, particularly for patients in whom conventional synthetic DMARDs (csDMARDs) fail to provide adequate disease control. During bDMARD therapy, patients are typically monitored every 1 to 3 months, with treatment 1Machine Learning and Data Analytics Lab, Department Artificial Intelligence in Biomedical Engineering, FriedrichAlexander-Universität Erlangen-Nürnberg, 91052 Erlangen, Germany. 2Halıcıoğlu Institute for Data Science, La Jolla, CA, USA. 3Department of Internal Medicine 3, Rheumatology and Immunology, Universitätsklinikum Erlangen, 91054 Erlangen, Germany. 4Department of Rheumatology and Clinical Immunology, Charité – University Medicine Berlin, 10117 Berlin, Germany. 5BIOREG, Health Care Centre Mariahilf, ÖGK and Rheumatology Department, Sigmund Freud Private University, 1060 Vienna, Austria. 6Ruth Fritsch-Stork and Bjoern M. Eskofier These authors contributed equally. 􀀍 email: [fatemeh.salehihafshejani@fau.de](fatemeh.saleh

FAQs

What clinical goal does the framework address for rheumatoid arthritis patients on bDMARDs?

It predicts six-month remission before starting bDMARD therapy and stratifies patients by predicted risk to support follow-up planning and prioritization of treatment.

Which machine learning models are evaluated in the study?

AdaBoost, Random Forest, XGBoost, and Support Vector Machines are trained and compared using Austrian patient data.

How does the approach make risk stratification more actionable?

It applies calibration techniques (including Platt scaling and isotonic regression) to improve the reliability of predicted probabilities, and it uses SHAP to interpret which patient characteristics drive remission predictions.

Aug 5, 2026

Upload by Sage