Background Quantitative and semi-quantitative assessments of structural changes on magnetic resonance images (MRI) have been widely adopted in knee osteoarthritis (KOA) research. However, there are no long-term studies that describe the independent association of these structural factors and total knee replacement (TKR) in older adults.

Objectives This study aimed to describe the predictability of MRI-based measures (cartilage defects, bone marrow lesions (BMLs), meniscal tears, meniscal extrusions, bone area increase and cartilage volume loss) for TKR over 10.7 years.

Methods 983 participants (mean age 62.8 years, 50% female) were randomly recruited from the local community of southern Tasmania and followed up at 2.6 years (range 0.9–4.8) and 10.7 years (range 9.2–12.5) later. A 1.5T MRI scan of the right knee was acquired at baseline (n=930) and 2.6 years (n=399). Tibiofemoral and patellar cartilage volume and tibial bone area at baseline and 2.6 years were measured using a manual segmentation method on the T1-weighted fat saturated 3D GRE images. Cartilage defects were assessed on T1-weighted MRI using a modified Outerbridge scoring system (grade 0–4). BMLs were measured on T2-weigthed fat saturated FSE images using a modified Whole-Organ MRI scoring system (grade 0–3). Meniscal extrusion (grade 0–2) and meniscal tears (grade 0–3) were measured using the T1- and T2-weighted MRI. TKR data were recorded at baseline, 2.6 years, 5.1 years and 10.7 years later. Analyses were performed using cox proportional hazards models after adjustment for age, sex, body mass index (BMI), baseline WOMAC knee pain and radiographic KOA status. Other MRI structural pathologies were added to the model in order to explore the independent association of each of these predictors.

Results 863 subjects (88%) responded for 2.6 years follow-up, 758 subjects (77%) for 5.1 years follow-up and 560 subjects (57%) for 10.7 years follow-up. After adjustment for age, sex, BMI, baseline WOMAC knee pain and radiographic KOA status, baseline MR imaging markers such as cartilage defects per grade (HR=2.28, p<0.01), BMLs per grade (HR=2.10, p<0.01), tibial bone area per cm² (HR=1.17, p<0.01), and percentage annual cartilage volume loss at medial tibiofemoral compartment (HR=1.19, p<0.05) were significantly associated with TKR over 10.7 years. These associations were independent of each other and there was a dose-response relationship between cartilage defects and BMLs at baseline and increased risk of TKR (Fig. a,b). Those who had TKR all had grade 3 meniscal tears at baseline, but meniscal extrusion was not associated with TKR. The risk of TKR increased significantly when more MRI pathologies were simultaneously present at baseline (HR=1.84/unit, p<0.01). There was also a number-response relationship between the number of MRI pathologies at baseline and the risk of TKR (Fig. c).

• Download figure
• Open in new tab
• Download powerpoint

Conclusions Baseline MRI structural pathologies such as cartilage defects, BMLs, meniscal tears, bone area increase and cartilage volume loss are independent predictors of TKR over 10.7 years. Presence of multiple pathologies increased the risk of TKR, suggesting that MRI structural markers are good predictors of rapid knee OA progression in the general population.

Acknowledgement National Health and Medical Research Council of Australia funded this study

Disclosure of Interest None declared