Pos0942 development of Prediction Model for Flare After Tapering of TNF Inhibitors in Patients With Axial Spondyloarthritis

Abstract

BackgroundTumor necrosis factor inhibitors (TNFi) have become a mainstay of management for axial spondyloarthritis (axSpA). However, it remains unclear whether patients with axSpA should continue the standard-dose TNFi after achieving stable disease activity. Although complete discontinuation of TNFi is followed by early relapse in most cases, several studies documented that reduced doses of TNFi in patients with prolonged low disease activity showed similar effects on disease control and drug survival compared to standard dose of TNFi. One of the main problem in the dose-tapering strategies for TNFi is a selection of the appropriate patient. However, there has been a lack of robust evidence regarding clinical factors predicting the flare after tapering of TNFi in patients with axSpA.ObjectivesThis study aims to develop and validate the prediction model to select the patients in whom tapering of TNFi does not lead to flare.MethodsWe used the data from Korean College of Rheumatology Biologics registry, which included a total of 1,730 patients receiving biologic DMARD from 2017 to 2019 in South Korea. In this study, a total of 526 patients who were initially treated with the standard-dose TNFi and tapered the dose after at least 1 year of the treatment were analyzed. Dose quotient (DQ, 0-1) was applied to quantified TNFi used during interval. The main outcome was an occurrence of flare defined as ASDAS-CRP score of ≥2.1 after 1 year of tapering TNFi. To develop the prediction model, clinical factors having relevant association (p < 0.1) with the outcome were first selected as candidate predictors. Logistic regression using a stepwise approach through backward elimination was used for the final model.ResultsPatients’ mean (SD) age was 37.5 (11.9) years, 418 (79.5%) were men, and 474 (90.1%) were HLA-B27 positive. Mean disease duration was 5.0 (6.1) years and 433 (82.3%) were TNF naïve. The mean BASFI and ASDAS-CRP at baseline were 3.4 (2.6) and 3.7 (1.0), respectively. Approximately two-thirds of the patients (65.8%) were initiated TNFi tapering at the first 1 or 2 years from baseline. At the time of TNFi tapering, the mean DQ was 0.67 (0.15) and 381 (72.4%) were prescribed concurrently with NSAIDs, and the mean BASFI and ASDAS-CRP were 1.3 (1.8) and 1.6 (0.9), respectively. During 12 months of follow up starting from the TNFi tapering, 127 (24.1%) experienced the flare. The multivariable analysis revealed that HLA-B27 positivity (OR 0.337; 95% CI 0.161-0.705; p=0.004), inflammatory back pain (OR 2.920; 95% CI 1.283-6.648; p=0.011), ASDAS-CRP at tapering (OR 2.798; 95% CI 2.030-3.856; p<0.001), and BASFI at tapering (OR 1.214; 95% CI 1.051-1.402; p=0.008) were significantly associated with flare. Based on the results of the logistic regression analysis, the predicted probability was calculated by the following formula: P=1/[1+ exp{-(1.088 x HLA-B27 negativity + 1.072 x inflammatory back pain + 1.567 x psoriasis + 0.623 x family history of axSpA + 1.092 x diabetes mellitus + 0.435 x DQ at TNFi tapering + 1.029 x ASDAS-CRP at TNFi tapering + 0.194 x BASFI at TNFi tapering)}]. The best cut-off value of the model to define the flare was 0.2416 (95% CI 0.176, 0.301) with sensitivity 74.0% and with specificity 81.0%. AUC was 0.828 (95% CI 0.786-0.869) indicating a good predication (Figure 1). The internal validation with bootstrapping showed minimal overfitting (estimated AUC 0.794) and good calibration between observed and predicted values (calibration slope 1.110, 95% CI 0.903, 1.317; intercept 0.026, 95% CI -0.091, 0.039).Figure 1.Apparent performance of developed model for prediction of flare after 12 months of tumor necrosis factor inhibitors tapering.ConclusionWe developed the prediction model for the flare after 12 months of TNFi tapering in patients with axSpA. It might be applicable in real world setting, although external validation will be required in the future investigation.References[1]Zavada J, et al. Ann Rheum Dis. 2016;75(1):96-102.AcknowledgementsWe greatly thank to the the Clinical Research Committee of the Korean College of Rheumatology and all participating hospitals.Disclosure of InterestsNone declared