Objective: To evaluate the ability of hand radiographs collected at study inclusion to predict a diagnosis of rheumatoid arthritis (RA) 2 years later, in a cohort of patients with early arthritis.
Methods: We evaluated 270 patients with arthritis of less than one year duration. At the first visit, all patients underwent a standardized evaluation including laboratory tests and radiographs. Followup was 30+/-11.3 mo. The hand radiographs were read by observers blinded to patient data who looked for item 7 of the 1987 ACR criteria for RA and used Sharp's method to score erosions and joint space narrowing.
Results: The kappa coefficient for ACR item 7 was < 0.65 for bony decalcification and > 0.8 for erosions. Intra and interobserver correlation coefficients for Sharp score ranged from 0.90 to 0.95. The "erosion" component of ACR item 7 was more specific than the full item 7 (96% versus 87.5%; p = 0.02). Sharp erosion score was not better than the erosion component of item 7 (sensitivity 17%; specificity 96%).
Conclusion: Regardless of the criterion used, hand radiographs were of limited value to predict which patients would be considered as having RA 2 years later. Diagnostic performance was similar for the "erosions" component of the 1987 ACR item 7 and for Sharp erosion score. The full 1987 ACR item 7 (erosions or bony decalcification) performed less well.