Arthritis & Rheumatism, Volume 62,
November 2010 Abstract Supplement

Abstracts of the American College of
Rheumatology/Association of Rheumatology Health Professionals
Annual Scientific Meeting
Atlanta, Georgia November 6-11, 2010.

The ACR/EULAR 2010 Criteria as Well as Other Predictive Algorithms for Rheumatoid Arthritis Show Good Diagnostic Performance.

Alves2,  Celina, Luime1,  Jolanda, van Zeben4,  Jendé, Huisman4,  Margriet, Weel3,  Angelique, Barendregt3,  Pieternella, Hazes1,  Johanna

Erasmus MC
Erasmus MC, Rotterdam, The Netherlands
Maasstad Ziekenhuis
Sint Franciscus Gasthuis


The ACR/EULAR devised new diagnostic criteria to improve early diagnosis of rheumatoid arthritis. The criteria were developed as a diagnostic algorithm for persistent and/or erosive disease, what was previously considered to be rheumatoid arthritis. Several other diagnostic algorithms already exist [1,2].


To evaluate the diagnostic performance of the ACR/EULAR 2010 criteria, van der Helm algorithm [1] and Visser algorithm [2] to identify persistent arthritis at 1 year using different patient sets and 2 outcomes.


Eligible patients in the Rotterdam Early Arthritis CoHort presented with inflammatory arthritis or >= 2 inflammatory tender joints and symptom duration less than 12 months. Diagnostic algorithms were tested on discrimination using ROC-curves and calibration by calibration plots. Robustness was tested using different (sub) sets of patients: (i) all patients (n=736) (ii) classified and unclassified arthritis (n=532), and (iii) undifferentiated arthritis (n=301). Two outcomes evaluated were: (a) persistence, defined as arthritis or the need for DMARDs and (b) the 1987 ACR criteria for rheumatoid arthritis.


Patients (n=736) had mean age 51 yr (sd 14), median ESR 15 (0–121), RF+ 21%, Anti-CCP+ 20%, median SJC 2 (0–44), median TJC 6 (0–38), erosions 5% at baseline. Table 1 shows discrimination performance for each algorithm per patient set (i;ii;iii) and for outcome (a) and (b). The algorithms had comparable performance on discrimination and calibration, with better performance the ACR/EULAR 2010 and van der Helm algorithm on discrimination. The ACR/EULAR 2010 criteria and van der Helm algorithm performed best in sample (i). Figure 1 shows ROC-curves and calibration plots for the algorithms in patient set (i).

[1]van der Helm-van Mil, AH, et al A&R 2007

[2]Visser, H, Et al A&R 2002

Table 1. AUCs (95% CI) for all algorithms in 3 patient subsets with either the outcome persistent arthritis or the 1987 ACR criteria (in italic).

 ACR/EULAR 2010Van der HelmVisser
(i)0.83 (0.79–0.86)0.81 (0.77–0.85)0.74 (0.69–0.78)
(ii)0.78 (0.73–0.83)0.76 (0.71–0.82)0.79 (0.74–0.84)
(iii)0.73 (0.66–0.80)0.74 (0.66–0.82)0.73 (0.65–0.80)
(i)0.83 (0.79–0.87)0.83 (0.79–0.87)0.78 (0.73–0.83)
(ii)0.77 (0.72–0.82)0.77 (0.71–0.82)0.76 (0.70–0.82)
(iii)0.61 (0.51–0.70)0.60 (0.50–0.71)0.59 (0.50–0.69)
(i) all patients; (ii) (un-) and classified arthritis; (iii) undifferentiated arthritis

Figure 1. ROC-curves and calibration plots for patient set (i) and outcome persistence.


The algorithms had comparable performance. However, on discrimination the ACR/EULAR and the van der Helm algorithm performed best in the full patient set. The well performing algorithms enable use in different settings (e.g. primary and specialized care) and suggests that any of these models will improve early diagnosis of patients at risk for rheumatoid arthritis.

To cite this abstract, please use the following information:
Alves, Celina, Luime, Jolanda, van Zeben, Jendé, Huisman, Margriet, Weel, Angelique, Barendregt, Pieternella, et al; The ACR/EULAR 2010 Criteria as Well as Other Predictive Algorithms for Rheumatoid Arthritis Show Good Diagnostic Performance. [abstract]. Arthritis Rheum 2010;62 Suppl 10 :1775
DOI: 10.1002/art.29540

Abstract Supplement

Meeting Menu