Objective: To evaluate the inter- and intrareader variability for interpretation of a modified Larsen's radiographic classification system for juvenile rheumatoid arthritis (JRA) focused on osteochondral lesions and a conventional Larsen's classification system, compared to a reference MR scoring system of corresponding images.
Materials and methods: Seventy-five radiographs of 60 children with JRA, performed within a short interval of time from the MR examinations, were independently evaluated by three experienced radiologists, three diagnostic imaging residents and three rheumatologists, in two separate sessions, according to the two different classification methods, blinded to the corresponding MR images.
Results: The inter- and intrareader concordance rates between the two radiographic classification systems and the MR-related radiographs were respectively poor and poor/moderate. The interobserver range of weighted kappa values for the conventional and the modified Larsen's system respectively was 0.25-0.37 vs 0.19-0.39 for radiologists, 0.25-0.37 vs 0.18-0.30 for residents and 0.19-0.51 vs 0.17-0.29 for rheumatologists. The intrareader rate ranged from 0.17-0.55 for radiologists, 0.2-0.56 for residents, and 0.14-0.59 for rheumatologists.
Conclusion: Although the proposal of a new radiographic classification system for JRA focused on osteochondral abnormalities sounds promising, the low inter- and intrareader concordance rates with an MR-related radiographic system makes the clinical applicability of such a radiographic system less suitable.