Abstract
Magnetic resonance imaging (MRI) is a tool with unprecedented capabilities. Rheumatoid arthritis (RA) abnormalities that can be measured with MRI include erosions, articular cartilage thickness, synovial membrane volume, and pannus. However, as access to MRI increases, there is a risk that its use will not be evaluated using rigorous scientific measurement principles. We reviewed published MRI measurement methods for RA and investigated whether the methods were systematically evaluated for reliability, validity, and responsiveness to change--components of the OMERACT filter. Medline and Embase databases were searched from 1966 to 1999. Titles and abstracts were scanned to identify publications on MRI methods used to assess either disease activity or damage in RA. A data extraction template was developed and 68 peer reviewed publications from 40 research groups were appraised; 40 addressed RA disease activity, 4 RA damage, and 24 both activity and damage. Joints most frequently assessed were knee (32 publications) and wrist (31 publications). Ninety-one percent of publications evaluated either reliability or validity or responsiveness to change. Thirteen percent evaluated all 3 and only 9% evaluated none of these measurement properties. Validity was evaluated in 85%, responsiveness to change in 37%, and reliability in 35% of publications. Only 12% of publications evaluated both intra and inter-reliability. Few publications of MRI measures of disease activity or damage in RA met the OMERACT filter for all measurement properties. It would be regrettable if MRI measures are developed ad hoc, with little regard to considerations of scaling, reliability, validity, and responsiveness to change, because this will severely limit their ability to confidently assess treatment efficacy and prognostic indicators.
