We aimed to determine the sensitivity and specificity of QuantiFERON-TB Gold (QFT-G) in Japanese rheumatoid arthritis (RA) patients with a past history of tuberculosis (TB). We assessed whether it is possible to decrease the cutoff using receiver operating characteristic (ROC) analysis. We evaluated chest computed tomography (CT) findings, prior history of treatment, and contact with active TB in 370 RA patients. Forty-nine patients before initiation of treatment with tumor necrosis factor (TNF) inhibitors were divided into two groups: 22 with a past history of TB and 27 without. We estimated the efficacy of QFT-G compared with the tuberculin skin test and antituberculosis (anti-TB) glycolipid antigen antibody. QFT-G was positive (≥0.35 IU/ml) in 13.6% with a past history of TB, increasing to 27.3% at the intermediate range cutoff of 0.1 IU/ml. The sensitivity and specificity of QFT-G was 0.27 and 1.00, respectively, at 0.1 IU/ml. Using ROC analysis, the area under the curve (AUC) of QFT-G but not for the other two tests was significantly large. QFT-G is a useful diagnostic method due to its superior specificity, but the use of a cutoff value of 0.35 IU/ml will likely result in an underestimate. We propose that a lower interferon-γ (IFN-γ) titer of 0.1 IU/ml be adopted when deciding to administer anti-TB drugs before initiation of TNF inhibitors.