Tool and threshold predicting a successful biological DMARDs tapering in patients with RA remission determination

Background: Tapering trials confirmed the feasibility of TNF inhibitors (TNFi) tapering for a relevant proportion of patients in remission and/or low disease activity. However, there are no consensual predictors of a good response to therapeutic spacing among patients with rheumatoid arthritis (RA) in remission. Objectives: To determine the most predictive tool and threshold of a successful TNFi tapering. Methods: Population: The Spacing of TNF-blocker injections in Rheumatoid Arthritis Study (STRASS) trial included 137 RA patients fulfilled the ACR 1987 criteria with sustained (at least 6 months) DAS28 <2.6. Patients were randomly assigned to one of the two following strategies: in the Maintain arm, patients continued to receive TNFi at the standard full regimen and in the Spacing arm, the strategy applied progressive spacing of ADA or ETN subcutaneous injections up to discontinuation at the forth step in the spacing arm. We used the data of the Spacing arm. Analysis: The performances of several variables (DAS28, SDAI, CDAI, CRP, ACPA status, HAQ, patient/physician global assessment, and booleen remission criteria) were assessed for the prediction of successful TNFi tapering, defined as reaching at least 25% tapering of the full regimen during at least 6 months, using sensitivity and specificity for dichotomous variables, or the area under the ROC curve (AUC) and its 95% confidence interval for continuous variables. A predictive score of successful tapering was constructed using LASSO regression modeling technique to avoid overfitting (R software version 3.2.1). Results: The main characteristic of the 64 patients of the Spacing arm were the following (mean ± SD): age 54.3±10.7 years, disease duration 8.3±5.4 years, and DAS 28 1.9±0.6. The baseline variables were similar between patients who failed or succeeded at TNFi spacing, except for the HAQ score (0.30 in the group success and 0.89 in the failure group, p=0.01) and the CRP (2.35 mg/l versus 3.48 mg/l, respectively, p=0.02). Baseline variables performance in predicting successful TNFi spacing: None of the tested variables was able to predict successful TNFi spacing, except the HAQ score and the CRP. A HAQ threshold ≥1.125 had a specificity (Spe) of 93% and an

AUC:

0.713 (CI95%: 0.540-0.886). A CRP threshold ≥6.8 mg/l had a Spe of 0.97 and an

AUC:

0.689 (CI95%: 0.547-0.831). Composite criteria: A composite criteria able to predict successful TNFi spacing has been elaborated, including ACPA status, Boolean criteria, SDAI, CRP and HAQ. A composite score lower than 0.502 was able to predict a successful TNFi spacing: Spe: 100%; Se: 54%;

AUC:

0.829 (CI95%: 0.671 - 0.986). Conclusions: The remission maintenance in rheumatoid arthritis after TNFi spacing is possible. Our results showed that in a population of RA patients in remission with TNFi, baseline HAQ and CRP are independent predictor factors of successful tapering. We have developed a composite index able to predict successful TNFi spacing, with an AUC of 0.829 and a specificity of 100%. A validation study will be needed to confirm its ability to select patients for treatment decrease.