MRI Diagnosis of Meniscus Tears in the Knee: An Updated Systematic Review and Meta-analysis.
简介
Background Updated benchmarks are needed on the diagnostic performance of MRI for detecting meniscus tears compared with arthroscopy. Purpose To conduct an updated systematic revie…
英文摘要
Background Updated benchmarks are needed on the diagnostic performance of MRI for detecting meniscus tears compared with arthroscopy. Purpose To conduct an updated systematic review and meta-analysis on the diagnostic performance of MRI in the detection of tears of the native menisci compared with arthroscopy and with subgroup analyses to identify factors that impact accuracy. Materials and Methods A literature search was conducted using PubMed, Scopus, and Embase databases to identify peer-reviewed publications on MRI diagnosis of meniscus tears using knee arthroscopy as the reference standard. Random-effects models, pooled weighted sensitivity and specificity, and summary receiver operating characteristic curve analyses were used to determine diagnostic performance, changes according to publication year, and differences based on study design, patient characteristics, imaging parameters, and diagnostic criteria for tears. A meta-regression model was also used. Results Seventy-five studies (published 1986-2023) from 28 countries included 8507 patients (8517 knees). Pooled weighted sensitivity was higher for medial tears (91.0% [95% CI: 89.3, 92.4]) than for lateral tears (78.5% [95% CI: 74.5, 82.0]). In contrast, specificity was higher for lateral tears (94.0% [95% CI: 92.5, 95.3]) than for medial tears (87.7% [95% CI: 85.2, 89.8]). Specificity for diagnosing lateral tears decreased over the years (P = .01). The highest pooled sensitivity for lateral tears was found in studies using both surfacing linear signal intensity and meniscus distortion for diagnosis (81.1%) compared with studies using only signal intensity (79.2%) or not reporting criteria (60.8%) (P = .02); likewise in meta-regression, using both surfacing signal intensity and meniscus distortion was a predictor of higher performance compared with not reporting criteria (adjusted odds ratio, 3.74 [95% CI: 1.37, 10.18]; P = .01). No differences in sensitivity or specificity (P value range, .29-.59) were found between studies using one or more versus two or more images for diagnosing tears in either meniscus. Conclusion The reported accuracy of knee MRI for meniscus tears was consistently high, regardless of individual study designs, with sensitivity higher for the medial meniscus and specificity higher for the lateral meniscus. © RSNA, 2026 Supplemental material is available for this article.