Session Information
Session Type: Abstract Submissions (ACR)
Background/Purpose: In rheumatoid arthritis (RA), erosion detection on radiography (X-ray) compared to magnetic resonance imaging (MRI) is characterized by low sensitivity and high specificity. This supports the hypothesis that MRI has a lower limit of detection for erosion than X-ray. To date, however, no studies have directly assessed measurement reliability. The objective of this study was to determine the relative diagnostic test accuracy of MRI and X-ray for erosion detection while accounting for inter-rater reliability.
Methods: A paired, cross-sectional study of 65 RA patients with a range of symptom duration was conducted. For each participant, MRI scans of the bilateral metacarpophalangeal joints (MCP) 2-5 and X-ray of both hands, wrists and feet were taken. Comparisons were limited to the MCP 2-5 joints. The Outcome Measures in Rheumatology (OMERACT) rheumatoid arthritis magnetic resonance imaging score (RAMRIS) and the van der Heijde-modified Sharp (vdHSS) scores were used to evaluate the MRI and X-ray images, respectively. Data were paired at the smallest level of analysis common to both measures: the joint. A total of 488 paired joints were compared. Odds ratio (OR), sensitivity (Se), specificity (Sp), and accuracy were calculated and the smallest detectable difference (SDD)-adjusted and unadjusted evaluations were compared.
Results: The association between erosion detection on MRI and X-ray of MCP 2-5 had an OR of 1.8 (1.2-2.9), Se of 0.31±0.03, Sp of 0.80±0.03, and accuracy of 0.47. Adjusting for measurement reliability increased the OR, Sp, and accuracy to 3.2 (1.5-6.1), 0.93±0.01, and 0.79, respectively, while decreased the Se to 0.19±0.04. Reliability-adjustment decreased the number of erosions detected per joint from 67.8% to 18.6% on MRI and 27.5% to 9.8% on X-ray. Per MCP joint, 2.6- to 8.0-fold the erosions detected on X-ray were detected on MRI. Compared by affected MCP 2-5 joint set, adjustment resulted in MRI detection of 2.1-fold the erosive disease detected on X-ray. At the patient level of analysis, bilateral MRI of the MCP 2-5 joints resulted in the detection of erosive disease in 1.1-fold the number detected on X-ray of hands, wrists and feet (McNemar’s test, p = 0.83, Cohen’s k=0.17±0.13, p=0.16). The correlation between SDD-adjusted vdHSS erosion score and symptom duration was 0.37 (p<0.0001). Correlation between MRI and symptom duration was non-significant (0.10, p=0.26).
Conclusion: Following reliability-adjustment of evaluations at the unit of measurement, a greater proportion of erosive disease is detected on MRI compared to X-ray per joint imaged. At the patient level of analysis, the relative performance of the two imaging modalities is highly dependent on the anatomy imaged. Despite detecting similar proportions with erosive disease when bilateral MRI of MCP 2-5 and X-ray of the hands, and feet are compared, the non-significant low level of agreement indicates that the proportions detected by each modality are unique. The interaction with symptom duration suggests that MRI may detect a greater proportion of patients with erosions at earlier stages of disease progression.
Disclosure:
R. Tavares,
None;
N. Parasu,
None;
K. Finlay,
None;
E. Jurriaans,
None;
H. Wu,
None;
K. A. Beattie,
None;
M. Larche,
None;
L. E. Hart,
None;
W. G. Bensen,
None;
R. S. Bobba,
None;
A. A. Cividino,
None;
C. E. Webber,
None;
J. E. Tarride,
None;
J. D. Adachi,
None.
« Back to 2012 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/relative-reliability-adjusted-diagnostic-test-accuracy-of-erosion-detection-between-magnetic-resonance-imaging-and-radiography-in-rheumatoid-arthritis/