Modified Dixon MRI to detect subclinical inflammation in clinically suspect arthralgia as a risk factor for rheumatoid arthritis development: should we image one or two hands?

OBJECTIVES

MRI-detected subclinical inflammation in clinically suspect arthralgia (CSA) predicts progression to inflammatory arthritis (IA) and rheumatoid arthritis (RA) and is incorporated in EULAR/ACR risk stratification criteria. Conventional MRI is hampered by long scan times, intravenous contrast and high costs, while modified Dixon (mDixon) MRI, with 5 min scan time and no intravenous contrast, is much more feasible. To optimise mDixon MRI utility, we compared bilateral versus unilateral hand analysis for detecting subclinical inflammation. We also studied the distribution of subclinical inflammation in CSA.

METHODS

139 patients of the CSA Leiden cohort were included. mDixon MRIs of bilateral wrist and metacarpophalangeal 2-5 joints were scored for subclinical inflammation (synovitis/tenosynovitis/osteitis) using RA MRI scoring. The hand with the most self-reported painful joints was used for unilateral analysis. Patients were followed for ≥6 months, with IA and RA development assessed at 6 months. We compared prognostic performance of bilateral versus unilateral MRI-detected subclinical inflammation.

RESULTS

Subclinical inflammation detected on bilateral MRI was more strongly associated with IA development than on unilateral MRI; 4.80 (95% CI 1.09 to 21.11) versus 2.34 (95% CI 0.84 to 6.49). Bilateral analysis resulted in a 25% higher sensitivity and 15% lower specificity, with a net 10% increase in correctly classified patients. Results for RA development were similar, with HRs of 8.17 (95% CI 1.06 to 62.86) versus 3.23 (95% CI 0.98 to 10.66), and 20% higher sensitivity. Subclinical inflammation was unilateral in 52% of patients and scanning one hand would miss a quarter of patients.

CONCLUSION

Bilateral MRI of the hands is preferable to unilateral MRI for detecting subclinical inflammation in CSA, because of its asymmetrical distribution.