ABSTRACT
Objective: To assess intercentre variability in the ACR core set measures, DAS28 based on three variables (DAS28v3) and Routine Assessment of Patient Index Data 3 in a multinational study. Methods: Seven thousand and twenty-three patients were recruited (84 centres; 30 countries) using a standard protocol in the Quantitative Standard Monitoring of Patients with RA study. Analysis of variance (ANOVA) and mixed-effect analysis of covariance models were used to model the relationship between study centre and different patient-reported and physician-reported RA activity measures. These models were built to adjust for the remaining ACR core set measure (for each ACR core set measure or each composite index), socio-demographics and medical characteristics. ANOVA and analysis of covariance models yielded similar results, and ANOVA tables were used to present variance attributable to recruiting centre. Results: The proportion of variances attributable to recruiting centre was lower for patient reported outcomes (PROs: pain, HAQ, patient global) compared with objective measures (joint counts, ESR, physician global) in all models. In the full model, variance in PROs attributable to recruiting centre ranged from 1.53% for patient global to 3.71% for HAQ compared with objective measures that ranged from 5.92% for physician global to 9.25% for ESR; and was lower for Routine Assessment of Patient Index Data 3 (2.6%) compared with DAS28v3 (11.75%). Conclusion: Intercentre variability in PROs is lower than objective measures of RA activity demonstrating that PROs may be more comparable across centres, and the need for standardization of objective measures.