Background Detailed quantitative evaluation of knee radiographs might enable the identification of osteoarthritis (OA) earlier in the disease.
Objectives To evaluate whether and which separate features of radiographic damage, measured quantitatively in knees with early symptoms related to OA, are associated with the incidence of radiographic OA and the persistence and/or progression of clinical OA during five-year follow-up.
Methods From the CHECK cohort, participants with knee pain at baseline were evaluated (n=829). Radiographic OA incidence was defined on joint level as Kellgren & Lawrence (K&L) grade≥II at T5-year. For analyses of predictors for “poor” radiographic outcome, knees with K&L<II at T0 were selected (n=985 knees). Clinical OA was defined as persistent knee pain, and as progression of WOMAC pain and function score during 5-year follow-up. Persistent pain outcome was defined as knee pain during joint motion at T4-year and T5-year. For these predictive analyses participants with at least a painful knee were included (n=1060 knees). Progression of WOMAC pain and function scores from baseline to follow-up was classified using a quintile approach described by Sharma et al. For these predictive analyses, participants with only painful knees were included (n=286 and n=279 respectively; no painful hip, no hip K&L≥II). At T0 Knee Images Digital Analysis (KIDA), provided quantitative measurements of joint space width (JSW), varus angle, osteophyte area, eminence height, and subchondral bone density. Logistic regression analyses were performed to evaluate which radiographic characteristics, in addition to basic demographic and clinical characteristics, at T0 were predictive of “poor” radiographic or clinical OA outcome.
Results In this early OA cohort 19% had “poor” radiographic outcome, and 48%, 54%, and 56% had “poor” clinical outcome (persistent pain, WOMAC pain, and function) at T5y. The prediction at baseline of “poor” radiographic outcome five years later improved significantly by measuring osteophyte area (odds ratio (OR) =7.0) and minimum JSW (OR=0.7) at T0, in addition to demographic and clinical characteristics (area under curve (AUC)=0.74 vs. 0.64 without radiographic features). By calculating a predictive score of the KIDA model and by determining a cut-off for optimal specificity, individuals could be selected with a chance of incident radiographic OA of 54% instead of the prior probability of 19%. Radiographic characteristics hardly added to prediction of “poor” clinical OA outcome.
Conclusions In individuals with onset knee pain, separate quantitative radiographic characteristics added to the prediction of radiographic OA incidence five years later. Although the prediction models can not be used for individual decision making (yet), quantitative measurement of radiographic features in individuals suspected for OA might be valuable in identifying individuals at high risk of developing radiographic osteoarthritis, which is worthwhile when designing clinical trials.
Disclosure of Interest None Declared