RESUMO
OBJECTIVE: To determine the predictive value of tibiofemoral (TF), patellofemoral (PF) MRI osteoarthritis (OA), and Kellgren and Lawrence grade 1 (KLG1) for the incidence of knee OA (clinical and/or radiological signs) at 2.5 and 6.5 years follow-up in a high-risk cohort. METHODS: Data of the PROOF study were used, consisting of middle-aged obese women without clinical (ACR-criteria) knee OA and free of radiographic signs (KLG < 2) at baseline. To determine the relation between MRI defined knee OA and KLG1 at baseline with clinical (ACR criteria) or radiographic OA (KLG ≥ 2) at the follow-up points, sensitivity, specificity, likelihood ratios and pre-test and post-test probabilities were calculated. RESULTS: The baseline prevalence of KLG1 (42.9%) was higher than TF MRI OA (14.6%) and PF MRI OA (10.0%). All diagnostic performance statistics indicated better prediction for radiographic OA than for clinical OA. For both outcomes and time points, the absolute difference between pre-test and post-test probabilities was the highest for TF MRI OA. The number needed to screen to obtain a certain number of cases with definite knee OA after a given follow-up period was higher (16-524%) for MRI OA, than for KLG1. CONCLUSIONS: When comparing the associations and post-test probabilities, TF MRI OA was more strongly related to the development of radiographic knee OA than KLG1. However, for the selection of people at high risk for developing knee OA, for instance for preventive trials, radiography seems to be sufficient, due to the high baseline prevalence.