MRI Accuracy for Knee Meniscus Tears: Updated Meta-Analysis

Key Takeaways

• MRI accuracy was high overall, with sensitivity favoring medial tears and specificity favoring lateral tears.
• The evidence base included 75 studies, 8,507 patients, 8,517 knees, 28 countries, and used arthroscopy as the reference standard.
• Subgroup analyses found lower lateral-tear specificity over time and higher lateral-tear sensitivity when combined diagnostic criteria were used.

An updated systematic review and meta-analysis found that MRI identified medial meniscal tears with higher pooled sensitivity than lateral tears when measured against arthroscopy, 91.0% versus 78.5%. Focused on native knee menisci, the analysis updated benchmark estimates from a large international literature. Using arthroscopy as the surgical reference standard, pooled MRI accuracy was high overall, with greater sensitivity for medial tears and greater specificity for lateral tears.

The review examined MRI diagnostic accuracy for native knee meniscal tears compared with arthroscopy as the reference standard. Peer-reviewed studies were identified through PubMed, Scopus, and Embase, spanning publications from 1986 through 2023. Across 28 countries, the evidence base covered 75 studies, 8,507 patients, and 8,517 knees. Random-effects models were used for pooled analyses, and summary receiver operating characteristic analyses were also reported.

Medial and lateral tears were analyzed separately, allowing direct compartment-level comparison. Sensitivity reached 91.0% for medial tears, with a 95% CI of 89.3% to 92.4%, and 78.5% for lateral tears, with a 95% CI of 74.5% to 82.0%. For specificity, lateral tears measured 94.0%, with a 95% CI of 92.5% to 95.3%, while medial tears measured 87.7%, with a 95% CI of 85.2% to 89.8%. The pooled results showed higher sensitivity for medial tears and higher specificity for lateral tears.

Subgroup and meta-regression analyses examined study design, patient characteristics, imaging parameters, and diagnostic criteria, with lateral-tear specificity decreasing over time, with P = .01. For lateral tears, pooled sensitivity was highest at 81.1% when studies used both surfacing linear signal intensity and meniscus distortion, versus 79.2% with signal intensity alone. Studies not reporting criteria had a pooled sensitivity of 60.8% for lateral tears. Meta-regression associated use of both criteria with higher performance versus studies not reporting criteria, adjusted odds ratio 3.74, with a 95% CI of 1.37 to 10.18 and P = .01. Image-count comparisons showed no sensitivity or specificity differences, with P values from .29 to .59, and the overall pattern favored medial sensitivity and lateral specificity.