Molecular Signatures Can Predict Response to TNFi Treatment Before Initiation in RA

Clinical trial research.
Clinical trial research.
Researchers evaluated the role of machine learning in predicting response before anti-tumor necrosis factor treatment in rheumatoid arthritis.

Cell-specific profiles in patients with rheumatoid arthritis (RA) can predict response before treatment initiation with tumor necrosis factor (TNF) inhibitors, including adalimumab and etanercept, according to study results published in Arthritis & Rheumatology.

Previous studies have identified several potential predictors of response to TNF inhibitors treatment in patients with RA. Data suggest that transcriptomic and epigenetic profiling may help in predicting response to TNF inhibitors before initiating treatment.

In the current study, researchers performed gene expression and/or DNA methylation profiling on immune cells and peripheral blood mononuclear cells, as well as clinical profiling of patients with RA, to produce cell-specific profiles that may help in predicting response to adalimumab and etanercept before treatment initiation.

Using the Utrecht’s BiOCURA cohort, 80 patients with RA were identified before initiating treatment with either adalimumab or etanercept. Response to treatment was defined according to the European League Against Rheumatism (EULAR) criteria at 6 months and analyses were completed to identify transcriptional and epigenetic signatures associated with response to treatment. Using these signatures, the researchers built machine learning models to predict response to adalimumab or etanercept.

There were significant differences in transcriptional signatures among patients with RA who were classified as responders or nonresponders to the TNF inhibitors. The response to adalimumab and etanercept was found to be defined by distinct gene signatures. There were 549 differentially expressed genes between adalimumab responders and nonresponders and 460 differentially expressed genes between etanercept responders and nonresponders.

Compared with adalimumab nonresponders, among the adalimumab responders, the expression of genes associated with RA and TNF signaling pathways were higher in CD4+ T cells. These genes were not differential in monocytes between the responders and nonresponders. Similarly, there were differences in gene expression in monocytes and CD4+ T cells between etanercept responders and nonresponders.

On an epigenetic level, there was a distinct hypermethylation pattern between patients who responded to adalimumab and etanercept. Among patients who responded to etanercept, differentially methylated CpG positions were hypermethylated in 76.3%; among adalimumab responders, approximately 46% of differentially methylated positions were hypermethylated.

The machine learning models to predict response to TNF inhibitors were highly accurate, with an overall accuracy to predict response to adalimumab and etanercept of 85.9% and 79%, respectively, using differentially expressed genes of peripheral blood mononuclear cells. The overall accuracy of the machine learning model to predict response to therapy using differentially methylated positions of peripheral blood mononuclear cells was 84.7% for adalimumab and 88% for etanercept. A follow-up study validated the high performance of these models.

The study had several limitations, among them were the inclusion of patients who did not fulfill all 6-month treatment and the study’s observational design.

“Machine learning models based on these molecular signatures could accurately predict response before ADA and ETN treatment, paving the path towards personalized anti-TNF treatment,” the researchers concluded.


Tao W, Concepcion AN, Vianen M et al. Multi-omics and machine learning accurately predicts clinical response to adalimumab and etanercept therapy in patients with rheumatoid arthritis. Published online September 10, 2020. Arthritis Rheumatol. doi:10.1002/art.41516