Performance of knee image digital analysis of radiographs of patients with end-stage knee osteoarthritis

Abstract

Objective: Knee Image Digital Analysis (KIDA) is standardized radiographic analysis software for measuring osteoarthritis (OA) characteristics. It was validated in mild OA, but used for severe OA as well. The current goal was to evaluate the performance of KIDA in severe OA. Design: Of 103 patients, standardized radiographs were performed before and one and 2 years after treatment for severe OA. All radiographs were evaluated on subchondral bone density, joint space width (JSW), osteophytes, eminence height, and joint angle, twice within years by the same observer. Part of the radiographs were randomly selected for reevaluation twice within 1 month and evaluation by another observer. The intraclass correlation coefficient (ICC), smallest detectable difference (SDD) and coefficient of variation (CV) were calculated; the SDD and CV were compared to those in mild OA. The relation of severity with KIDA parameters and with observer differences was calculated with linear regression. Results: Intra-observer ICCs were higher in the 98 severe radiographs reanalyzed within 1 month (all >0.8) than the 293 reanalyzed within years (all >0.5; most >0.8) and than inter-observer ICCs (all >0.7). SDDs and CVs were smaller when reanalyzed within a month and comparable to those in mild OA. Some parameters showed bias between readings. Severity showed significant relation with osteophytes and JSW parameters, and with the observer variation in these parameters (all P < 0.04). Conclusions: KIDA is a well-performing tool also for severe OA. In order to decrease variability and SDDs, images should be analyzed in a limited time frame and randomized order.