PREDICTION OF EARLY RESPONSE TO ANTI-TNF THERAPY AND DISEASE ACTIVITY STATE USING THE ANKYLOSING SPONDYLITIS DISEASE ACTIVITY SCORE

Abstract

Background: We previously reported that in the ASSERT and GO-RAISE studies, BASDAI50 response can be accurately predicted with a combination of baseline characteristics. This analysis compares the accuracy of predicted response rates and disease activity as measured by different assessments including the ankylosing spondylitis disease activity score (ASDAS). Methods: The ASSERT and GO-RAISE trial data were combined and baseline characteristics were used to predict the probability for achieving week 12 ASAS20, BASDAI50 and ASDAS major response and BASDAI and ASDAS scores at week 12. Univariate analyses identified baseline predictors for response. Variable associations were explored using Spearman correlation analysis. A stepwise selection procedure using logistic regression, ROC and correlation analyses was used to select a final set of predictors. Logistic regression was used to calculate the predicted week 12 BASDAI and ASDAS scores and the probability of week 12 response to anti-TNF or placebo respective to combined selected predictors at baseline. The accuracy of prediction between the different assessments was compared using ROC analyses and R-square. Results: 479 AS patients (NY modified criteria) treated with anti-TNF and 156 patients treated with placebo with continued conventional, with BASDAI and spinal pain assessment 4 were included. Age (mean 39.5; SD 11.3 yrs), BASFI, (mean 5.4; SD 2.2 cm), Berlin enthesitis-score (mean 2.4; SD 2.9), therapy (anti-TNF or placebo), CRP (mean 2.1; SD 2.4 mg/dL) and HLA-B27 genotype [(þ) or (-)] were retained as final predictors of response. The area under the ROC curve of the model which included the six selected predictors (log transfer of CRP values) and interactions between predictors was 0.85, 0.79, and 0.75 and the R-square was 0.43, 0.33, and 0.24 for week 12 ASDAS major response, BASDAI50 response and ASAS20 response, respectively. This indicates good accuracy for prediction of BASDAI50 and ASDAS major response and fair accuracy for ASAS20 response. The R-squares of 0.49 and 0.27 for ASDAS and BASDAI scores respectively, indicate that the week 12 disease activity measured by ASDAS could be more reliably predicted. Conclusions: Elevated CRP, lower age, lower enthesitis score, lower BASFI, and presence of HLA-B27 genotype are associated with a better treatment response at week 12 in ASSERT and GO-RAISE. The values of these characteristics at baseline allow calculating the predicted response rates and future disease activity of an individual patient. The accuracy of predicting ASDAS and ASDAS response is good and higher than that of other assessments.