Background To validate the ACR/EULAR criteria by comparison with the 1987 ACR criteria for RA, in an independent community-based very early arthritis cohort (VErA).
Methods VErA cohort was community-based recruited and conservatively treated during the 2 first years. It comprised 310 patients included between 10-1988 and 01-2002: ≥18 year old, ≥2 swollen joints for ≥6 weeks and ≤6 months, naïve of DMARD and steroids. This cohort was not used to develop 2010 criteria. We compared the performances between the 2010 and the 1987 criteria: number of patients classified as RA; sensitivity and specificity. About 2010 ACR/EULAR criteria, we tested (1): in applying sensu stricto the 4 steps (≥1 “clinical synovitis''/''another disease''/≥1 erosion/``score ≥6/10''; (2) in applying the same 4 steps but with ≥3 erosions; (3) excluding “erosion” step. We used 2 “gold standards'': the expert diagnosis at 6 years of follow-up and ≥3 non equivocal erosions at 2 years (total Sharp-van der Heijde score (SHS).
Results Baseline characteristics of the overall population were: female 68%; median age 52 years [19-84]; swollen joint count 7 [2-37]; tender joint count 6 [0-58]; DAS 28 2.95 [0.45-7.53]; mean HAQ 0.75 [0-2.9] ESR 18mm/h [1-110]; CRP 7mg/l [5-206]; IgM RF+ 22.6% and anti-CCP+ 23.2%; ≥1 erosion 16.8%. 41 patients had alternative diagnoses. Among the 269 other patients at baseline, 67.7% and 59.5% fulfilled the 2010 ACR/EULAR and 1987 ACR criteria respectively. When using the expert diagnosis as gold standard, the ACR/EULAR 2010 criteria sensitivity was significantly higher: 85.9% vs 77.9% (Mc Nemar’s test 0.035); the specificity was similar (59.0% vs 64.1%). When we modified the cut off point for typical RA erosions from ≥1 to ≥3 the sensitivity was similar but the specificity was significantly higher: 69.2% vs 59% (Mc Nemar’s test 0.008). These results were comparable when excluding erosion step (Sp: 70.5%). When using ≥3 erosions at 2 years as gold standard, the performances were similar for the two sets of criteria. The sensitivity and the specificity of the 2010 ACR/EULAR and the 1987 ACR criteria were respectively 95.2% vs 95.2% and 40.2% vs 33.5%. The discriminative ability of 2010 ACR/EULAR and 1987 ACR criteria was comparable (p<0.49) with areas under the curve (AUC) 0.76-0.79 respectively. The AUC of the ACR/EULAR criteria without considering the erosion status was significantly higher than the AUC of the criteria applied sensu stricto: 0.832 vs 0.761 (p<0.02). Moreover the ROC curves showed that the score ≥6, proposed by the 2010 criteria for classification as definite RA, was relevant.
Conclusions Using a very early community-based cohort the 2010 ACR/EULAR criteria classified slightly more patients with RA than the 1987 ACR criteria but otherwise they performed similarly. Interestingly, the application of score, after exclusion of “another rheumatism” diagnoses, without considering the item “erosions” of algorithm, gives better performance than the strict application of 2010 ACR/EULAR criteria. Moreover the score ≥6 seems relevant.
Disclosure of Interest None Declared