We conducted a meta-analysis of randomized clinical trials to evaluate the effect of thiazolidinediones on serum C-reactive protein levels. Compared with placebo, treatment with thiazolidinediones significantly decreased the serum C-reactive protein levels (mean -0.82 mg/L, 95% confidence interval -1.15 to -0.49 mg/L, p <0.0001). In a subgroup analysis, the effect of thiazolidinediones on the serum C-reactive protein levels was more pronounced in diabetic patients (mean -1.24 mg/L, 95% confidence interval -2.15 to -0.32 mg/L, p = 0.008) compared with nondiabetic patients (mean -0.27 mg/L, 95% confidence interval -0.41 to -0.14 mg/L, p <0.0001).