Examining a whole-organ magnetic resonance imaging scoring system for osteoarthritis of the knee using Rasch analysis

The ability to reliably quantify all the structural abnormalities in osteoarthritis (OA) of the knee is a long-standing goal of OA research. On December 5 and 6, 2002, Outcome Measures in Rheumatology Clinical Trials and Osteoarthritis Research Society, International held a Workshop for Consensus on Osteoarthritis Imaging in Bethesda, MD, with the aim of providing a state-of-the-art review of imaging outcome measures for OA of the knee. As part of the Workshop, data from previous clinical trials and epidemiological studies of OA were analysed with respect to the metrological properties of the measurement methods used. The following report outlines the results of analyses aimed at evaluating the internal construct validity of a whole-organ, ordinal (semi-quantitative) magnetic resonance imaging score (WORMS) using Rasch analysis. The fit of data to the Rasch model offers a measure of the validity of summing different items into a subscale score and the degree to which this score behaves as a unidimensional, interval level measurement tool.

The Rasch model was applied in two OA studies. The first was a clinical cohort comprising OA knee subjects entering a clinical trial; study entry criteria included patients with at least moderate pain, radiographic osteophytes and a minimum of 1.5mm tibiofemoral joint-space width. The second cohort was from the Boston Osteoarthritis Knee Study, an observational cohort of subjects with symptomatic knee OA with pain on most days and a definite osteophyte in either the tibiofemoral or patellofemoral joints. Baseline WORMS scores from both studies were used for the Rasch analysis, performed with RUMM 2020 software.

There was a substantial proportion of subjects in both study populations with zero scores in several of the subscales of WORMS. Few of the subscales met the requirements of the Rasch measurement model when summated across all sites, and summations of some postulated compartmentally based sites also failed to fit the Rasch model. The existing scoring categories also required rescoring at many sites.

There remain important issues in constructing outcome measurements that summate different features across multiple anatomical sites. The whole-organ scoring system evaluated here is no exception. Resolving these issues will improve the ability of imaging studies to assess complex pathological structural change.