A search to the target tissue in which RA-specific inflammation starts: A detailed MRI study to improve identification of RA-specific features in the phase of clinically suspect arthralgia

Abstract

Objective: Based on a unique cohort of clinically suspect arthralgia (CSA) patients, we analysed which combinations of MRI features at onset were predictive for rheumatoid arthritis (RA) development. This was done to increase our comprehension of locations of RA onset and improve the predictive accuracy of MRI in CSA. Methods: In the discovery cohort, 225 CSA patients were followed on clinical arthritis development. Contrast-enhanced 1.5 T MRIs were made of unilateral metacarpophalangeal (MCP) (2-5), wrist, and metatarsophalangeal (1-5) joints at baseline and scored for synovitis, tenosynovitis, and bone marrow edema. Severity, number, and combinations of locations (joint/tendon/bone) with subclinical inflammation were determined, with symptom-free controls of similar age category as reference. Cox regression was used for predictor selection. Predictive values were determined at 1 year follow-up. Results were validated in 209 CSA patients. Results: In both cohorts, 15% developed arthritis < 1 year. The multivariable Cox model selected presence of MCP-extensor peritendinitis (HR 4.38 (2.07-9.25)) and the number of locations with subclinical inflammation (1-2 locations HR 2.54 (1.11-5.82); ≥ 3 locations HR 3.75 (1.49-9.48)) as predictors. Severity and combinations of inflammatory lesions were not selected. Based on these variables, five risk categories were defined: no subclinical inflammation, 1-2 locations, or ≥ 3 locations, with or without MCP-extensor peritendinitis. Positive predictive values (PPVs) ranged 5% (lowest category; NPV 95%) to 67% (highest category). Similar findings were obtained in the validation cohort; PPVs ranged 4% (lowest category; NPV 96%) to 63% (highest category). Conclusion: Tenosynovitis, particularly MCP-extensor peritendinitis, is among the first tissues affected by RA. Incorporating this feature and number of locations with subclinical inflammation improved prediction making with PPVs up to 63-67%.