Open Access Open Badges Research article

Evaluating joint-space narrowing and cartilage loss in rheumatoid arthritis by using MRI

Charles G Peterfy1, Julie C DiCarlo1*, Ewa Olech2, Maire-Agnes Bagnard3, Annarita Gabriele3 and Norman Gaylis4

Author Affiliations

1 Spire Sciences, LLC, 72 Rock Road, San Francisco, CA 94904, USA

2 Department of Rheumatology, University of Nevada School of Medicine, 1707 West Charleston Boulevard, Suite 220, Las Vegas, NV 89102, USA

3 F. Hoffman-La Roche AG, Konzern-Hauptsitz, Grenzacherstrasse 124, CH-4070 Basel, Switzerland

4 Arthritis & Rheumatic Disease Specialties, 21097 NE 27th Court, Aventura, FL 33180, USA

For all author emails, please log on.

Arthritis Research & Therapy 2012, 14:R131  doi:10.1186/ar3861

Published: 30 May 2012



Magnetic resonance imaging (MRI) has been shown to be superior to radiography (XR) for assessing synovitis, osteitis, and bone erosion in rheumatoid arthritis (RA), particularly in clinical trials. However, relatively little has been reported on the ability of MRI to evaluate articular cartilage loss, or joint-space narrowing (JSN), in the hands and wrists. In a previous study, we adapted the nine-point Genant-modified Sharp XR-JSN score for use with MRI (MRI-JSN). In this study, we compare MRI-JSN with XR-JSN by using images from two multicenter clinical trials.


Baseline XR and 1.5-Tesla MR images of one hand and wrist from each of 47 subjects with RA enrolled in one of two multicenter clinical trials were evaluated by using the XR-JSN and MRI-JSN methods by a single radiologist experienced in the two methods. Radiographs and MR images were read independently on different occasions.


In total, 575 of 611 joints were compared (one metacarpophalangeal joint of the thumb and 35 proximal interphalangeal joints were outside the MRI field of view and could not be assessed). The 22 (47%) subjects showed JSN with both XR and MRI, and 25 (53%) subjects showed no JSN with either method. No subject showed JSN with only one or the other method. MRI showed high agreement with XR (intraclass correlation coefficient = 0.83). Sensitivity of MRI for JSN, by using XR as the gold standard, was 0.94; specificity was 0.91; accuracy was 0.91; positive predictive value was 0.64; and negative predictive value was 0.99.


This validation exercise suggests that MRI JSN scoring may offer a viable alternative to XR JSN scoring in multicenter clinical trials of RA. However, the relative longitudinal sensitivity of MRI to change and the ability to discriminate therapeutic effect on JSN were not evaluated in this study.