Responsiveness and reliability of MRI in knee osteoarthritis: An updated meta-analysis

To systematically evaluate the reliability and responsiveness of magnetic resonance imaging (MRI) for assessing structural changes in knee osteoarthritis (OA), and to update the 2011 meta-analysis with newer scoring systems and expanded literature.

A comprehensive search of five databases was conducted up to December 2024, yielding an initial pool of 57008 studies. Eventually, 145 studies were selected for data extraction. Data on reliability (root-mean-square coefficient of variation [RMS-CV%], CV%, intraclass correlation coefficient [ICC], kappa statistics) and responsiveness (standardized response mean [SRM]) were extracted. Three-level random-effects meta-analysis pooled effect sizes by measurement type, tissue, and cartilage region, and meta-regression assessed the association between follow-up duration and SRM.

Reliability (114 studies): In OA patients, intra-reader quantitative MRI measurements showed low test-retest variability. The pooled RMS-CV% was 2.7% (95% CI: 1.9-3.8%) for total cartilage volume, 3.0% (95% CI: 2.1-4.2%) for total cartilage thickness, and 1.5% (95% CI: 0.9-2.7%) for total cartilage surface area. Responsiveness (45 studies): The pooled annualized SRM of medial tibiofemoral cartilage thickness (SRM=-0.350, 95% CI: -0.405 to -0.294) showed the highest responsiveness. Absolute values of quantitative SRM for cartilage volume and thickness increased with study duration in 2010-2024 (β=-0.016 to -0.003, P < 0.05 across most cartilage regions within the knee).

MRI demonstrates high reliability and good responsiveness for structural monitoring of knee OA. Quantitative medial tibiofemoral cartilage thickness shows larger SRM and is recommended as the primary imaging endpoint for clinical trials; extending follow-up increases the ability to detect structural change.