BACKGROUND: Synovial inflammation (as defined by hypertrophy and effusion) is common in osteoarthritis (OA) and may be important in both pain and structural progression. OBJECTIVE: To determine if decision rules can be devised from clinical findings and ultrasonography (US) to allow recognition of synovial inflammation in patients with painful knee OA. METHODS: A EULAR-ESCISIT cross sectional, multicentre study enrolled subjects with painful OA knee who had clinical, radiographic, and US evaluations. A classification and regression tree (CART) analysis was performed to find combinations of predictor variables that would provide high sensitivity and specificity for clinically detecting synovitis and effusion in individual subjects. A range of definitions for the two key US variables, synovitis and effusion (using different combinations of synovial thickness, depth, and appearance), were also included in exploratory analyses. RESULTS: 600 patients with knee OA were included in the analysis. For both knee synovitis and joint effusion, the sensitivity and specificity were poor, yielding unsatisfactory likelihood ratios (75% sensitivity, 45% specificity, and positive LR of 1.36 for knee synovitis; 71.6% sensitivity, 43.2% specificity, and positive LR of 1.26 for joint effusion). The exploratory analyses did not improve the sensitivity and specificity (demonstrating positive LRs of between 1.26 and 1.57). CONCLUSION: Although it is possible to determine clinical and radiological predictors of OA inflammation in populations, CART analysis could not be used to devise useful clinical decision rules for an individual subject. Thus sensitive imaging techniques such as US remain the most useful tool for demonstrating synovial inflammation of the knee at the individual level.