Methods: We investigated 12,043 adults with RA and used a discordance cutoff of 3 cm. Large joint involvement was investigated using novel joint indices (x, y, z) and the large joint index (Nishiyama S et al. Rheumatol Int. 2012;32:2569), in which x and y were the indices for upper and lower joints, respectively, and z was for large joint predominance. Predictors of PGA-PhGA discordance and determinants of PGA and PhGA were analyzed by multivariate logistic and linear regression models, respectively. In logistic regression analysis, the odds ratio (OR) for positive discordance (PGA ≥ PhGA by 3 cm) was computed using a concordance (no discordance) group as reference.

Results: Multivariate logistic regression identified age (OR: 1.01 [95% confidence interval: 1.01–1.02]), pain (2.15 [2.08–2.22]) and modified Health Assessment Questionnaire score (1.56 [1.39–1.75]) as significant predictors of positive discordance. On the other hand, TJC (0.93 [0.91–0.94]), SJC (0.92 [0.90–0.94]), CRP (0.88 [0.83–0.93]), class 3-4 (0.63 [0.53–0.76]), Z (0.72 [0.59–0.88]), and the large joint index (0.59 [0.47–0.74]) predicted significantly against positive discordance. Linear regression analysis demonstrated that PGA was mainly determined by pain, whereas PhGA was determined by various factors, including the large joint index (Figure 1) and Z.