The sensitivities of the Xpert MTB/RIF test and an in-house IS6110-based real-time PCR using TaqMan probes (IS6110-TaqMan assay) for the detection of Mycobacterium tuberculosis complex (MTBC) DNA were compared by use of 117 clinical specimens (97 culture positive and 20 culture negative for MTBC) that were frozen in sediment. The 97 clinical specimens included 60 respiratory and 37 nonrespiratory specimens distributed into 36 smear-positive and 61 smear-negative specimens. Among the 97 culture-positive specimens, 4 had rifampin-resistant isolates. Both methods were highly specific and exhibited excellent sensitivity (100%) with smear-positive specimens. The sensitivity of the Xpert MTB/RIF test with the whole smear-negative specimens was more reduced than that of the IS6110-TaqMan assay (48 versus 69%, P = 0.005). Both methods exhibited similar sensitivities with smear-negative respiratory specimens, but the Xpert MTB/RIF test had lower sensitivity with smear-negative nonrespiratory specimens than the IS6110-TaqMan assay (37 versus 71%, P = 0.013). Finally, the sensitivities of the Xpert MTB/RIF test and the IS6110-TaqMan assay were 79% and 84%, respectively, with respiratory specimens and 53% and 78%, respectively (P = 0.013), with nonrespiratory specimens. The Xpert MTB/RIF test correctly detected the rifampin resistance in smear-positive specimens but not in the one smear-negative specimen. The Xpert MTB/RIF test is a simple rapid method well adapted to a routine laboratory that appeared to be as sensitive as the IS6110-TaqMan assay with respiratory specimens but less sensitive with paucibacillary specimens, such as smear-negative nonrespiratory specimens.