Background International guidelines for treatment-to-target in rheumatoid arthritis (RA) recommend methotrexate (MTX) as first line therapy because of its favorable benefit/risk ratio and long-term safety. However, about 50% of patients fail to achieve remission and require other treatment. Identification and application of predictors of inadequate response to MTX may improve clinical outcome by enabling timely step-up treatment in those who need it.
Objectives To evaluate the added predictive value of protein biomarkers compared to use of clinical variables alone, to identify early RA patients failing a MTX step-up strategy, including hydroxychloroquine (HCQ).
Methods In the treat-to-target U-ACT-EARLY trial, 108 DMARD-naïve RA patients initiated MTX step-up therapy (10 mg to 30 mg or maximum tolerable dose) until remission (DAS28 <2.6 with SJC ≤4) was achieved. If no remission after 20 weeks, HCQ was added, which after 12 weeks was replaced by tocilizumab (TCZ) if remission still was not achieved. Clinical assessments at baseline were analyzed and missing values were imputed using multiple imputations (n=20). Additionally, serum was collected before treatment and analyzed for 85 inflammatory proteins using Luminex® xMAP technology. Predictors for failing the MTX step-up strategy (MTX+HCQ) were selected using univariate preselection (p≤0.15) followed by multivariate backward selection (p≤0.10). A model including only clinical predictors was developed and subsequently, a second model in which protein biomarkers was added. The Area Under the Receiver Operator Curve (AUC-ROC), Net Reclassification Index (NRI, quantifies the improvement in prediction between two models) and Likelihood-ratio (LLR) test were used to assess both models.
Results Within 1 year, 56 (52%) of the 108 patients failed the strategy for various reasons: adverse events (n=4), added TCZ according to treatment protocol (n=42), withdrawal from study because of inefficacy (n=10). Multiple logistic regression identified baseline DAS28, current smoking and alcohol consumption as clinical predictors (model 1, AUC-ROC 0.75, 95%>CI 0.66–0.84) and in addition MCP1, IL6, IL33, sCD14 and PD1 (model 2, AUC-ROC 0.87, 95%>CI 0.80–0.94) as significant protein biomarkers (Table 1). The LLR test showed a better fit of model 2 (59.83, p<0.001). When using a risk of non-response >80% as cut off, model 1 correctly predicted 24/51 patients not failing the strategy and model 2 predicted 15 (29% improvement) extra patients not failing the strategy (Table 2). In those failing the strategy, 50/56 patients were correctly predicted by model 1 and 4 (-7%) patients less was correctly predicted by model 2.
Conclusions Adding protein biomarkers to clinical predictors increases the predictive accuracy of identifying patients at baseline who will need other treatment after initiating a MTX step-up strategy. These findings could contribute too more personalized treatment in early RA patients to optimize long-term outcomes.
Disclosure of Interest X. Teitsma: None declared, J. Jacobs: None declared, P. Welsing: None declared, T. Woodworth Employee of: Roche, A. Pethö-Schramm Employee of: Roche, M. Borm Employee of: Roche, C. Bernasconi Employee of: Roche, J. van Laar Consultant for: received fees from MSD, Pfizer, Roche, Eli Lilly and BMS, J. Bijlsma Grant/research support from: received research grants (to his department) from AbbVie, BMS, Crescendo, MSD, Mundipharma, Pfizer, Roche, Sun and UCB, Consultant for: received consulting fees from AbbVie, BMS, Crescendo, MSD, Mundipharma, Pfizer, Roche, Sun and UCB, F. Lafeber: None declared
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.