Reproducibility and sensitivity to change of 5 methods for scoring hand radiographic damage in patients with...

To compare intrarater and interrater reproducibility and sensitivity to change of 5 scoring methods for radiographic damage on hand radiographs in patients with rheumatoid arthritis (RA).

Radiographs of 22 patients from Norway and France with average 2 years' disease duration at baseline and mean 30 months' followup were assessed by 2 readers according to Larsen, Larsen/Rau, Sharp, Sharp/van der Heijde, and Simple Erosion Narrowing Score (SENS) methods. Reproducibility at baseline and on progression was assessed using intraclass correlation coefficients (ICC) and Bland-Altman graphs. Sensitivity to change was compared across methods by computing the country-adjusted standardized response means (SRM) ratio.

Intrarater reproducibility varied with the reader (ICC ranging from 0.90 to 0.97), with Larsen and Larsen/Rau ranking highest. Interrater reproducibility was highest with Sharp and Sharp/van der Heijde (ICC 0.76 to 0.93). Bland-Altman graphs showed a decrease of concordance in cases of more severe damage. Sensitivity to change was higher with Sharp and Sharp/van der Heijde modified for erosions (SRM ratio 1.44 and 1.70), than with Larsen/Rau and SENS. The differences between Sharp, Sharp/van der Heijde, and Larsen were less for joint space narrowing. There was a significant reader effect (p < 0.05) in all but the Sharp method. Expressed as percentage of the maximum score, the smallest detectable difference varied between 3.5% (Sharp/van der Heijde) and 14.2% (SENS erosion).

All methods have high intraobserver and interobserver reliability. The interrater reproducibility decreases with disease severity. Recent modified methods perform best to detect changes, but the advantages of SENS seemed to be lost when applied on hand radiographs alone. Training the readers appears to be essential.