RESUMEN
The aim of the study was to assess agreement between three-dimensional volumetric ultrasound (3D US) performed by inexperienced staff and real-time conventional ultrasound (2D US) performed by experienced rheumatologists in detecting and scoring rheumatoid arthritis (RA) lesions. Thirty-one RA patients underwent examination of seven joints by 2D and 3D US for synovitis and tenosynovitis in B and PD modes and erosions in B mode. A global score for synovitis and global counts for synovitis, tenosynovitis and erosions were also calculated for every patient. Agreement between 2D and 3D US was analysed for counts and scores at the patient level with the intraclass correlation coefficient (ICC) and for counts at the joint level with Cohen's kappa coefficient. B-mode synovitis was detected at a median of five joints in each patient, frequently in wrists and hand joints but less frequently in foot joints. PD-mode synovitis, tenosynovitis and erosions were detected less frequently. All ICCs for agreement between 2D and 3D US findings were significant. All kappa coefficients were significant for B- and PD-mode synovitis and for erosions (except PIP3), while those for tenosynovitis were only significant for MCP2 (B and PD modes) and PIP2 (B mode). Although the 3D US volumes were acquired by inexperienced operators, agreement between 2D and 3D US was acceptable in detecting and scoring synovitis. A higher level of agreement was attained for patient-level global scores and counts than for individual joints.