The aim of this work was to assess the diagnostic value of anti-CCP-3 and anti-CCP-2 for the diagnosis of rheumatoid arthritis (RA) and determine whether anti-CCP-3 more accurately identifies patients with rheumatoid arthritis than anti-CCP-2. PubMed and CNKI databases were searched for studies published in English and Chinese that examined the use of anti-CCP-3 and anti-CCP-2 in the diagnosis of known or suspected rheumatoid arthritis from January 2006 to July 2013. Seventeen included studies of methodological quality were rated by using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tools A random-effects method was used to summarize sensitivities, specificities, positive and negative likelihood ratio (LR+ and LR−, respectively), and diagnostic odds ratio from 17 studies. The pooled sensitivity, specificity, LR+, LR− and diagnostic odds ratio for anti-CCP-3 were 0.737 (95% CI, 0.717–0.757), 0.933 (95% CI, 0.924–0.942), 11.096 (95% CI, 8.876–13.870), 0.274 (95% CI, 0.231–0.326), and 42.908 (95% CI, 33.828–54.426), respectively. For anti-CCP-2, the values were 0.719 (95% CI, 0.699–0.739), 0.960 (95% CI, 0.953–0.966), 17.485 (95% CI, 11.960–25.562), 0.294 (95% CI, 0.258–0.335) and 63.458 (95% CI, 44.214–91.078), respectively. With high specificity and moderate sensitivity, anti-CCP-2 and anti-CCP-3 played an important role in confirming the diagnosis of RA. Anti-CCP-3 did not have better diagnostic performances than anti-CCP-2, but anti-CCP-2 had evident heterogeneity compared to anti-CCP-3, especially in American patients.