Comparison of radiographic and MRI osteoarthritis definitions and their combination for prediction of tibial cartilage loss, knee symptoms and total knee replacement: a longitudinal study

Objective: To describe the value of radiographic- and magnetic resonance imaging (MRI)-defined tibiofemoral osteoarthritis (ROA and MRI-OA, respectively) and in combination for predicting tibial cartilage loss, knee pain and disability and total knee replacement (TKR) in a population-based cohort. Design: A radiograph and 1.5T MRI of the right knee was performed. ROA and MRI-OA at baseline were defined according to the Osteoarthritis Research Society International atlas and a published Delphi exercise, respectively. Tibial cartilage volume was measured over 2.6 and 10.7 years. Knee pain and disability were assessed at baseline, 2.6, 5.1 and 10.7 years. Right-sided TKRs were assessed over 13.5 years. Results: Of 574 participants (mean 62 years, 49% female), 8% had ROA alone, 15% had MRI-OA alone, 13% had both ROA and MRI-OA. Having ROA (vs. no ROA) and MRI-OA (vs. no MRI-OA) predicted greater tibial cartilage loss over 2.6 years (−75.9 and −86.4 mm³/year) and higher risk of TKR over 13.5 years (Risk Ratio [RR]: 15.0 and 10.9). Only MRI-OA predicted tibial cartilage loss over 10.7 years (−7.1 mm³/year) and only ROA predicted onset and progression of knee symptoms (RR: 1.32–1.88). In participants with both MRI-OA and ROA, tibial cartilage loss was the greatest (over 2.6 years: −116.1 mm³/year; over 10.7 years: −11.2 mm³/year), and the onset and progression of knee symptoms (RR: 1.75–2.89) and risk of TKR (RR: 50.9) were the highest. Conclusions: The Delphi definition of MRI-OA is not superior to ROA for predicting structural or symptomatic OA progression but, combining MRI-OA and ROA has much stronger predictive validity.