A molecular signature response classifier predicts the likelihood of eular nonresponse to TNF inhibitor therapies in RA: results from a retrospective cohort analysis
L. Zhang1 , C. van der Tog2 , A. den Broeder2 , T. Mellors1 , E. Connolly-Strong1 , J. Withers1 , A. Jones1 , V. Akmaev1 . 1 Scipher Medicine Corporation, Science and Technology, Waltham, United States of America; 2 Sint Maartenskliniek, Department of Rheumatology, Ubbergen, Netherlands
hold
Background: Following RA treatment recommendations, most people with rheumatoid arthritis (RA) begin targeted therapy with TNF inhibitors (TNFi), even though inadequate response to TNFi therapies is widespread. Treatment changes from one medication to the next are currently fueled by disease-activity measures and eventually result in disease control for most patients; however, this “trial-and-error” approach wastes precious time on ineffective treatments. A delay in reaching treat-to-target goals has a negative effect on patient burden and, possibly, disease progression.1 Useful predictors for TNFi response have been challenging to identify but a specific molecular signature response classifier (MSRC) test was shown to be predictive for inadequate response to TNFi therapies.2 The impact of such identification has the potential to result in improved patient outcomes, but further validation would be welcome, especially for response criteria other than ACR50, and in a stringent treat-to-target setting with lower baseline disease activity.
Objectives: To validate the predictive value of the MSRC test in identifying those patients who do not meet EULAR good response criteria after 6 months of TNFi treatment.
Methods: Data from a prospective cohort study conducted in the Sint Maartenskliniek (Nijmegen, the Netherlands) of RA patients who started adalimumab or etanercept TNFi as their first biologic were included.3 Baseline RNA samples and clinical assessments were used to identify patients who had a molecular signature1 of non-response to TNFi therapy. Outcomes were calculated at six months using DAS28-CRP-based EULAR good response, and high and low confidence responders and non-responders were identified using Monte Carlo simulation with 2,000 repeats and 70% precision cut off. Outcome measurements were blinded for test results. Treatment switch before 6 months was imputed as non-response. Odds ratios and area under the ROC curve (AUC) assessments were used to evaluate the ability of the MSRC test to predict inadequate response at 6 months against EULAR good response criteria.
Results: A total of 68 out of 88 RA patients were identified to have a high-confidence response status and were included in analyses (Table 1). EULAR good response was observed in 45.5% (31/68) of patients. Patients were stratified according to detection of a molecular signature of non-response with an AUC of 0.61. The odds that a patient with the molecular signature of non-response at baseline failed to achieve a EULAR good response at 6 months was four times greater than that of a patient lacking the molecular signature (odds ratio 4.0, 95% confidence interval 1.2-13.3).