The evaluation of diagnostic accuracy of new in vitro diagnostic assays for tuberculosis infection has been hampered by the lack of a standard reference test. The aim of this study was to compare sensitivity and specificity of interferon gamma assays for latent tuberculosis infection by assessing the association of test results with tuberculosis occupational exposure and by using latent class analysis. We analysed data from 115 healthcare workers on whom tuberculin skin test (TST) and the following in vitro tests were performed: in-house ELISPOT for RD1 proteins, T.SPOT-TB and Quantiferon-TB Gold. Results of all tests were associated with increased occupational risk of exposure to Mycobacterium tuberculosis, but only TST was associated with Bacillus Calmette-Guerin (BCG) vaccination. Sensitivity/specificity (95% confidence intervals) estimated by a latent class model were: 99.9%/64.2% (53.0-74.1) for TST, 95.3% (61.8-99.6)/87.5% (78.0-93.2) for in-house ELISPOT, 96.7% (69.3-99.7)/85.6%(75.3-92.0) for T.SPOT-TB, and 76.3% (55.9-89.1)/93.6% (85.4-97.3) for Quantiferon. The estimated specificity of in vitro assays was higher than that of TST also among individuals who were not BCG-vaccinated. In conclusion, when used in healthcare workers, in vitro assays may provide a significant increase of specificity for tuberculosis infection compared to TST, even among non vaccinated individuals, at the cost of some sensitivity.