Radiographic assessment of knee osteoarthritis: reproducibility and sensitivity to change

To determine the cross sectional and longitudinal reproducibility of various measures used for assessing radiographic knee osteoarthritis (OA) and to compare the sensitivity to change over a one year period of these measures.

We studied 55 patients referred to hospital with knee OA (clinical and radiographic ACR criteria). Anteroposterior radiographs at baseline and after 12 months were read by methods both qualitative (Kellgren and Lawrence grading scale, joint space narrowing scale) and quantitative (joint space width measurement at 3 different points). All qualitative methods used standardized atlases. The intraclass correlation coefficient and the graphical method of Bland and Altman were used to assess cross sectional and longitudinal reproducibility. Reproducibility was tested using 2 readers (interreader) and 2 readings for one of the readers (intrareader). Sensitivity to change was assessed using standardized response mean (SRM).

All methods tested were shown to be reproducible both for cross sectional and longitudinal data. Intrareader was higher than interreader reproducibility for most radiographic features. Significant changes were observed after one year for methods measuring joint space narrowing. The sensitivity to change of the methods assessing joint space narrowing (joint space narrowing scale and joint space width measurement) was higher, with SRM varying from 0.37 to 0.57, than for the Kellgren and Lawrence grading system (SRM 0.19 and 0.23).

These data suggest that methods measuring narrowing should be preferred as outcome measures in clinical trials or longitudinal epidemiologic studies; and show that in a particular subset of patients with very active disease, significant radiographic changes in knee OA can be detected after a one year period.