Background ACR scores are widely accepted binary endpoints in Rheumatoid Arthritis (RA) clinical trials to characterize response to treatment. The Disease Activity Score calculated from twenty-eight tender and swollen joints (DAS28) is often assessed as a secondary endpoint in RA trials and is believed to be a more sensitive measure of dose-response due to the continuous nature of the scale.
Objectives To characterize the quantitative relationship between the change from baseline (CFB) in DAS28 score and the proportion of responders based on ACR20/50/70 criteria using summary level data available in the literature from RA trials reporting both measures of response.
Methods Summary level data from RA trials reporting DAS28-CRP, DAS28-ESR or unspecified DAS28 and ACR20/50/70 variables at the primary time point (with a minimum of 20 subjects per arm) were included in the analyses. The dataset included 82 unique trials with information from more than 12,000 patients (trial durations from 4 to 104 weeks). Three separate mixed-effects logistic regression analyses of the relationships between CFB in each DAS28 score and ACR20/50/70 responses with random inter-trial variability (in the intercept) were conducted and the sample size for each trial was used for weighting
Results The observed and model-estimated relationships between DAS28 and ACR20/50/70 are presented in Figure 1. For a CFB in DAS28 CRP of -1 point, the point estimate [90% prediction interval (PI)] for ACR20, ACR50 and ACR70 responders is 33% [25% to 43%], 13% [8% to 22%] and 4% [2% to 7%], respectively. For a treatment that increases the CFB in DAS28 CRP from -1.0 to -2.0 points, the increase in ACR20, ACR50 and ACR70 responders is estimated to be 26% [24% to 26%], 19% [13% to 24%] and 9% [6% to 14%], respectively. For a treatment that increases the CFB in DAS28 CRP from -2.0 to -3.0 points, the increase in ACR20, ACR50 and ACR70 responders is estimated to be 22% [18% to 25%], 27% [24% to 28%] and 22% [16% to 28%], respectively. The relationship between the ACR scores and DAS28 was estimated to be slightly shallower and more variable for DAS28 ESR than DAS28 CRP.
Figure 1: The relationship between DAS28 continuous scores (DAS28 CRP, DAS28 ESR and unspecified DAS28) and the proportion of responders based on the ACR criteria. The solid lines and the shaded areas depict the median and 90% prediction interval for the proportion of responders at the end of treatment (EOT) for ACR20 (red), ACR50 (green) and ACR70 (blue) responders. The symbols represent the observed relationships for the analyzed trials with the symbol size weighted by the arm sample size.
Conclusions A quantitative relationship between the continuous DAS28 score and proportion of subjects achieving ACR response has been established. The relationship between the DAS28 scores and ACR responders is estimated to be non-linear, most prominently for the deeper ACR responses. DAS28 CRP was better correlated with the ACR scores than DAS28 ESR based on the lower heterogeneity estimated in the relationship
Disclosure of Interest M. Minocha Shareholder of: AbbVie, Employee of: AbbVie, J. Mandema Consultant for: AbbVie, A. Othman Shareholder of: AbbVie, Employee of: AbbVie