Predictive value of MRI features for development of radiographic osteoarthritis in a cohort of participants with pre-radiographic knee osteoarthritis-the CHECK study

Objective. To determine whether MRI features are associated with development of radiographic knee OA and can be used as a predictive tool in early knee OA. Methods. In 148 participants of the Cohort Hip and Cohort Knee study (mean age 56 years, 78% women), with a Kellgren Lawrence (KL) score <= 1, we obtained semi-quantitatively scored knee MRI scans and radiographs at baseline. After 5 years, we determined the development of radiographic knee OA (KL >= 2). We calculated odds ratios (ORs), with 95% CIs adjusted for age, sex and BMI, to identify MRI features associated with OA development. With these MRI features, we constructed an internally validated prediction model, for which we measured the area under the receiver operating characteristics curve, sensitivity and specificity. Results. Radiographic OA developed in 28% of the participants after 5 years. Statistically significant associations were: cartilage defects OR = 1.7 (95% CI: 1.1, 2.6), osteophytes OR = 3.1 (1.7, 5.7), bone marrow lesions OR = 2.0 (1.2, 3.4), effusion OR = 2.1 (1.2, 3.5) and meniscal pathology OR = 2.8 (1.3, 6.3). With the combined MRI features in a prediction model, the sensitivity was 66%, the specificity 67% and the optimism-corrected area under the receiver operating characteristics curve 0.685. Conclusion. In early knee OA, MRI depicts significantly associated pathology in cartilage, bone and menisci, whereas the radiograph fails to detect these changes. Although MRI has potential for identifying patients at risk for developing radiographic knee OA, it cannot be used as an absolute diagnostic tool in early knee OA due to its low discriminative ability.