Among the 757 patients who underwent both QFT-GIT and TST simultaneously from September 2008 to November 2010, patients with indeterminate QFT-GIT results (n = 21), with active (n = 11) or suspicious (n = 1) findings for tuberculosis on a chest radiograph, were excluded. Finally, 724 patients were recruited for this study: 497 patients with rheumatoid arthritis (RA), 198 with ankylosing spondylitis (AS), and 29 with juvenile rheumatoid arthritis (JRA). The agreement between the 2 tests was estimated by Cohen's ¦Ê and factors influencing discrepancies were identified using multivariate analysis.
The positivity of QFT-GIT was higher in RA than AS or JRA (30.2 % , 16.2 % , and 3.4 % , respectively). In contrast, TST positivity was highest in AS compared to RA and JRA (45.5 % , 28.2 % , and 17.2 % , respectively). The agreement between the 2 tests was low in all patients (¦Ê = 0.285). The only predictor of a discrepancy between the 2 tests was older age. Factors associated with discordant QFT-GIT-negative/TST-positive results were female [odds ratio (OR) = 2.33, confidence interval (CI) 1.11 to 4.89] and AS (OR = 3.12, CI 1.44 to 6.79), whereas a discordant QFT-GIT-positive/TST-negative result was associated with glucocorticoid use (OR = 2.44, CI 1.24 to 4.81).
The agreement between the 2 tests is low; therefore, it would be better to perform both tests than to use any 1 test alone for the detection of LTBI in TB-endemic regions. Female and underlying AS are related to being QFT-GIT-negative/TST-positive, and the use of glucocorticoid is associated with being QFT-GIT-positive/TST-negative.