Accuracy of magnetic resonance imaging in grading knee chondral defects.

Abstract

To determine the accuracy of routine magnetic resonance imaging (MRI) in the grading of knee cartilage lesions through a meta-analysis. A search of English-language literature published before February 2012 was carried out in PubMed. Articles using arthroscopy as a gold standard, a 6-knee region dividing method, and a 5-level grading system were included in our meta-analysis. After data extraction, a bivariate mixed-effects model and hierarchical weighted symmetric summary receiver operating curve were used to pool the results of diagnostic tests. A sensitivity analysis was conducted to explore the potential sources of heterogeneity. Overall, 8 studies were included in the meta-analysis. The overall sensitivity, specificity, diagnostic odds ratio, positive likelihood ratio, and negative likelihood ratio were 75% (95% confidence interval [CI], 62% to 84%), 94% (95% CI, 89% to 97%), 47 (95% CI, 18 to 122), 12.5 (95% CI, 6.5 to 24.2), and 0.27 (95% CI, 0.17 to 0.42), respectively. There was substantial heterogeneity among the results. Sensitivity analysis showed the inconsistency of 2 studies. However, eliminating the 2 studies had no significant impact on the overall results. Our results showed that MRI was effective in discriminating normal morphologic cartilage from disease but was less sensitive in detecting knee chondral lesions (higher than grade 1). The negative results of MRI should not prevent a diagnostic arthroscopy. Level II, meta-analysis of Level I and II studies.