Good reliability, questionable validity of 25 different classification criteria of knee osteoarthritis: a systematic appraisal

Objectives: Despite extensive epidemiological and clinical research, there is no consensus on classification criteria to define knee osteoarthritis (OA). No gold standard is available and many different definitions are used. For future research and interpretation of epidemiological studies, we aimed to evaluate reliability and validity of commonly used classification criteria. Study Design and Setting: Systematic searches were performed in Medline/Pubmed and Embase for articles evaluating reliability, construct validity, and content validity of knee OA classification criteria. Results: In 18 articles, 25 classification criteria were found that could be summarized in three categories (radiological clinical and radiological combined classification criteria, and clinical classification criteria). No classification criteria based on magnetic resonance imaging could be included. In general, intra- and interrater reliabilities were good. Construct validity was low when radiological criteria were compared with clinical classification criteria. Associations between classification criteria and symptoms and risk factors like pain and obesity were moderate. Conclusion: More research is needed to investigate the impact of different classification criteria in epidemiological research and to reach consensus about which criteria should be used to define knee OA. Meanwhile, to create uniformity in epidemiological research we recommend separate lesion scoring, overall scoring, and pain registration to define knee OA. (C) 2008 Elsevier Inc. All rights reserved