OBJECTIVE: To assess the interobserver reproducibility of the modified Ahlbäck, Dejour et al, and Kellgreen et al classifications, with the purpose of identifying points of agreement, points of disagreement, and the possible advantages of one of the classifications over the others, and correlating them according to the level of experience of the physician. METHODS: 50 patients were studied who had the diagnosis of osteoarthrosis of the knee (50 knees), 36 female and 14 male, mean age 62 years (ranging from 50 to 82). Inclusion criteria were older than 50, pain in the knee, and no prior surgery or rheumatic disease in the knee joint. Four physicians who are members of the SBCJ - Brazilian Society of Knee Surgery were invited to evaluate the images, the golden standard being one of the members of the SBCJ who had the highest level of titles. Statistical analysis followed the Kappa method, with significance level α = 1%. RESULTS: The kappa for the sample surveyed ranged from 0.028 to 0.501, with none of the combinations achieving a good concordance (0.61 < kappa < 0.80) or an excellent concordance (kappa > 0.80). The Dejour classification presented the best interobserver reproducibility, with a 0.487 kappa for each pair of observers, and a 0.501 kappa when compared to the golden standard, thus showing a moderate degree of concordance (0.41 < kappa < 0.60). CONCLUSIONS: The interobserver assessment showed that there was no absolute concordance of the researchers for any of the classifications studied. The Dejour classification has the highest concordance (moderate) according to our study, as it defines in a more reproducible and adequate manner the severity of the knee degenerative disease. According to our survey, classifications that assess the reduction in joint space are better tools to analyze the progression of the knee degenerative disease. Reproducibility was higher among physicians who are knee specialists, thus showing that observer experience has a direct impact in the results.
Knee joint; Osteoarthritis, knee; Reproducibility of results; Observer variation
