First step in the development of an ultrasound joint inflammation score for rheumatoid arthritis using a data-driven approach

Objectives To develop and validate candidate sets of joints and tendons for assessment of ultrasound (US) joint inflammation in rheumatoid arthritis (RA). Methods Patients were included in one of two cohorts from 2010 to June 2013: disease-modifying antirheumatic drug naive early RA or established RA starting/switching biologics. An extensive US examination was performed by experienced sonographers using a validated grey-scale (GSUS) and power Doppler (PDUS) semiquantitative scoring system with scores 0-3 for both GSUS and PDUS in 36 joints and four tendons. We performed factor analysis in the early RA US data and selected candidate joint/tendon sets based on these results. The proportion of information in the total US scores retained in these candidate sets was assessed by R-2 from linear regression analysis. Finally, the candidate sets and previously proposed joint scores were tested in the established RA cohort, and we also evaluated the sensitivity to change with standardised response means. Results 227 patients with early RA and 212 patients with established RA were included. We identified two candidate sets of joints/tendons: candidate set A consisted of seven joints/two tendons (meatacarpophalangeal 1 (MCP1), MCP2, proximal interphalangeal 3, radiocarpal, elbow, metatarsophalangeal 1 (MTP1), MTP2, tibialis posterior tendon, extensor carpi ulnaris tendon) and set B of nine joints/two tendons (MCP5 and MTP5 added to set A). Unilateral reduced scores retained 78%-85% of the information in total score, while bilateral reduced scores retained 89%-93%, and both sets performed better than previously proposed reduced joint scores, and similar or slightly better regarding sensitivity to change. Conclusions The reduced GSUS and PDUS scores retained most of the information from the total score and performed well in a validation cohort of established RA.