TOPLINE:

In patients with rheumatoid arthritis (RA), a novel seven-gene transcriptomic signature effectively predicted responses to TNF inhibitor therapy. This signature was validated across multiple patient cohorts and surpassed traditional biomarkers in discriminative capacity.

METHODOLOGY:

• Researchers conducted a retrospective analysis of transcriptomic profiles of peripheral blood mononuclear cells or whole blood from patients with RA to identify a gene signature predictive of responses to TNF inhibitor therapy prior to treatment initiation.
• The analysis used two datasets from the Gene Expression Omnibus database, including data of patients treated with adalimumab or etanercept (n = 80) or with adalimumab or infliximab (n = 42), and conducted external validation in three cohorts: the Reina Sofía Hospital (n = 94), La Princesa Hospital (n = 26), and COMBINE (n = 37) cohorts.
• A small-scale gene signature was defined as the minimal subset of genes that could maximize cross-validation accuracy for predicting responses to TNF inhibitors. Participants were classified as responders or nonresponders based on the European Alliance of Associations for Rheumatology (EULAR) criteria.
• Differential gene expression analysis was performed, and cutoffs for Fisher ratio were used to define a set of discriminatory genes.
• Logistic regression was used to construct the predictive model, and the discriminative performance was assessed using receiver operating characteristic (ROC) curve analysis and the area under the ROC curve (AUC).

TAKEAWAY:

• A gene signature comprising the 18 most discriminatory genes demonstrated a predictive accuracy of 88.75% using leave-one-out cross-validation; only one gene from the differentially expressed genes showed an overlap.
• Further, a refined seven-gene predictive model (MRPL24, COMTD1, DNTTIP1, GLS2, GTPBP2, IL18R1, and KCNK17) effectively distinguished between responders and nonresponders, achieving an AUC of 0.84. ROC curve analysis indicated strong predictive performance, with an AUC of 0.949.
• The seven-gene signature demonstrated superior discriminative capacity, achieving an AUC of 0.949, compared with anticyclic citrullinated peptide and rheumatoid factor, which had AUCs of 0.519 and 0.555, respectively.
• Validation of the seven-gene signature across the Reina Sofía Hospital, La Princesa Hospital, and COMBINE cohorts yielded AUCs of 0.85, 0.851, and 0.939, respectively, confirming its robustness in differentiating responders from nonresponders to TNF inhibitor therapy.

IN PRACTICE:

“This study provides a useful tool to optimize RA treatment strategies, improving the criteria for treatment selection before the start of RA treatment. The gene signature we describe will help to predict the response to [TNF inhibitor] in patients with RA before the initiation of therapy, reducing the time that a patient may spend on ineffective treatment, as well as preventing the potential development of negative side effects,” the authors wrote.

SOURCE:

The study was led by Lucia Santiago-Lamelas, Instituto de investigación Sanitaria de Santiago de Compostela, Santiago de Compostela, Spain. It was published online on September 3, 2025, in Annals of the Rheumatic Diseases.

LIMITATIONS:

The retrospective design may have led to biases due to reliance on existing datasets and clinical records. The use of EULAR criteria could lead to misclassification of patients because of its subjective components. The absence of longitudinal data restricted the assessment of changes in gene expression over time, and the limited ethnic diversity may affect the applicability of the findings to wider populations.

DISCLOSURES:

This study was supported by grants from the Fondecyt, Instituto de Salud Carlos III (ISCIII) (co-funded by the European Union), and GAIN Proyectos de Excelencia. Three authors reported being supported by contracts from different organizations including ISCIII and co-funded by the European Union, and one of them disclosed having a pending patent.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.