A multi-parameter response prediction model for rituximab in rheumatoid arthritis

Area of Science:

• Rheumatology
• Immunology
• Genetics

Background:

• Rituximab is a biologic therapy used for rheumatoid arthritis (RA).
• Predicting rituximab non-response is crucial for optimizing patient treatment.
• Interferon response genes (IRGs) have shown potential in predicting treatment outcomes.

Purpose of the Study:

• To validate an IFN response gene (IRG) set for predicting rituximab non-response in RA.
• To assess the predictive performance of the IRG set combined with clinical parameters.
• To develop a multivariate model for predicting rituximab non-response.

Main Methods:

• Two independent cohorts of RA patients (n=93 and n=133) starting rituximab were analyzed.
• Baseline peripheral blood expression of eight IRGs was measured and averaged into an IFN score.
• Logistic regression and stepwise selection were used to develop a multivariate prediction model.

Main Results:

• Higher mean IFN scores were observed in rituximab non-responders compared to responders.
• Univariate analysis identified baseline DAS28, IFN score, DMARD use, and serological markers as associated with non-response.
• A multivariate model (DAS28, IFN score, DMARD use) achieved an AUC of 0.82 in cohort I and 0.78 in prednisone-negative patients of cohort II, but was less effective in prednisone-positive patients (AUC=0.63).

Conclusions:

• The combination of predictive parameters offers a promising model for predicting rituximab non-response in RA.
• Prednisone use may modify the predictive performance of the IFN score, suggesting a need for further investigation.
• Optimization of the prediction model could be achieved by clarifying the interfering effect of prednisone.