Diagnostic accuracy of magnetic resonance imaging (MRI) for symptomatic knee osteoarthritis: a scoping review

In recent years, significant progress in the diagnosis of knee osteoarthritis (OA) has been made due to advancements in imaging techniques, especially magnetic resonance imaging (MRI). However, the diagnostic accuracy of MRI for knee OA lacks agreement in the literature due to limitations such as heterogeneous reference standards and suboptimal methodological assessment tools. This scoping review systematically collated the literature on the accuracy of MRI in diagnosing symptomatic knee OA and evaluated its methodological quality using a novel tool.

A comprehensive search was conducted across seven electronic databases to identify systematic reviews, with or without meta-analyses, published until May 30, 2024. Selected articles were systematic reviews on test accuracy of MRI for knee OA, whose methodological quality was assessed using a tailored version of the A Measurement Tool to Assess Systematic Reviews (AMSTAR) 2 tool, and the test accuracy performance summarized.

The search yielded 8,586 records of which 73 full-text articles were assessed for eligibility. We included 3 systematic reviews with a total of 99 original studies (7,530 participants). The methodological quality assessment of the reviews revealed common deficiencies in research questions and inclusion criteria components (2/3, 66.7%), protocol registration (2/3, 66.7%), use of statistical methods (1/3, 33.3%), and risk of bias assessment (2/3, 66.7%). The reference standards used were histology (1/3, 33.3%), radiography (1/3, 33.3%), computed tomography (1/3, 33.3%), clinical evaluation (1/3, 33.3%) and direct visual inspection (1/3, 33.3%), open surgery (1/3, 33.3%) and arthroscopy (3/3, 100%). In two reviews with meta-analysis the accuracy of MRI was: sensitivity 61% [95% confidence interval (CI), 53-68%] and specificity 82% (95% CI, 77-87%) (low methodological quality); and sensitivity 74% (95% CI, 71-77%) and specificity 95% (95% CI, 94-95%) (high methodological quality). In the third systematic review, without meta-analysis, the MRI sensitivity and specificity ranged from 26% to 96% and from 50% to 100%, respectively (critically low methodological quality).

This scoping review highlights the heterogeneity and gaps in methodological quality of the literature regarding the accuracy of MRI in diagnosing knee OA. Our adaptation of AMSTAR 2 will help to ensure systematic reviews of diagnostic tests are assessed more effectively and reliably. There is a need for standardization in future studies and reviews to strengthen the evidence base for the use of MRI as a diagnostic tool for knee OA.