Machine learning and artificial intelligence for predicting anti-TNF treatment response in rheumatoid arthritis: a systematic review

Authors: Boumechache, Malika; Dilekh, Tahar; Malki, Issam; Slimani, Samy

Source journal: Clinical rheumatology (2026-11-19)

DOI: 10.1007/s10067-026-08525-w

Abstract

Rheumatoid arthritis causes joint inflammation, progressive structural damage, and marked variability in treatment response. Anti-tumour necrosis factor (anti-TNF) therapies have transformed management, yet 30-40% of patients do not respond adequately, so predicting response before treatment would be clinically valuable. This systematic review evaluates artificial intelligence and machine learning models developed to predict anti-TNF response in rheumatoid arthritis. PubMed, IEEE Xplore, arXiv, and Google Scholar were searched for studies published between 2019 and 2026 following PRISMA 2020, with an initial search in June 2025 updated in May 2026. Selection, extraction, and risk-of-bias assessment using the Prediction model Risk of Bias Assessment Tool were performed independently by two reviewers. Thirty studies met the inclusion criteria, of which 14 developed or validated at least one model with a treatment-specific analysis of response to a TNF inhibitor, using random forests, gradient boosting, support vector machines, and neural networks applied to clinical, registry, genetic, and transcriptomic data. Reported discrimination was highest in small transcriptomic studies and lowest in large registry studies, but this ordering tracked validation design rather than modelling approach: internally validated models reported the highest apparent performance, whereas externally validated clinical models plateaued at an area under the curve of approximately 0.70-0.72. Applying the risk-of-bias tool according to its published categories and combination rules, most models were at high risk of bias, driven by the analysis domain. No model has been evaluated for clinical impact or is available in a deployable form. The evidence is hypothesis-generating and does not yet support use of these models in routine rheumatology practice.

Read the full text

Related topics

Related open access articles

Abstract and bibliographic data from PubMed (U.S. National Library of Medicine). Copyright of the abstract belongs to the original publisher or authors.