Direct detection of Mycobacterium tuberculosis was performed in parallel with the Amplicor M. tuberculosis test (Roche Diagnostic System, USA) and the LCx M. tuberculosis (Abbott Diagnostic Division, USA) on 697 samples, collected from 481 patients, in three different Italian laboratories. Though both systems are licensed only for pulmonary specimens, 113 extrapulmonary specimens (represented mainly by pleural fluids, cerebrospinal fluids and urines) were included in the study. Amplification results were compared with acid-fast microscopy, culture, and identification of isolates. Final clinical diagnosis was used to resolve discrepant results. M. tuberculosis was detected in 105 specimens by both assays, whereas 561 were agreeing negatives; 21 and 6 of the remaining true-positive samples scored positive with LCx only and with Amplicor only, respectively. There were three false-positives with LCx and one false-positive with Amplicor. The diagnostic sensitivity of both methods was significantly better when only respiratory specimens were considered (78% versus 59% in nonrespiratory samples with Amplicor, and 88% versus 65% with LCx). Our data reveal a significantly better sensitivity of the LCx (p = 0.026) and a slight better specificity of the Amplicor assay. It is noteworthy that 16 of the 21 Amplicor-negative specimens in which LCx detected M. tuberculosis were culture negative, thus suggesting that the higher diagnostic sensitivity of the latter assay is attributable to its better analytical sensitivity. However, the majority of such samples originated from patients under antimicrobial treatment, which makes uncertain the clinical significance of such increased sensitivity. Considering true-positive for LCx and true-negative for Amplicor, the 16 culture-negative/LCx-positive/Amplicor-negative specimens resulted true-positives after the resolution of discrepancies, the final overall sensitivity and specificity values of the LCx assay were not significantly different from the ones of the Amplicor assay.