The aim of this study was to compare the results of a commercial assay based on the ligase chain reaction [(LCR) LCx Probe System MTB; Abbott, USA] with those of culture in liquid medium (Septi-Chek AFB; Becton-Dickinson, USA) and culture on the egg-based Löwenstein-Jensen solid medium for the direct detection of Mycobacterium tuberculosis complex in nonrespiratory specimens. The results were analyzed according to the standard definition of a true-positive result. Two hundred thirty-five nonrespiratory samples routinely submitted to rule out tuberculosis were analyzed. All samples were smear-negative. Mycobacterial growth in either culture medium was detected in 18 (7.6%) specimens: Mycobacterium tuberculosis was recovered from seven (38.9%) specimens cultured on Löwenstein-Jensen medium and from 18 (100%) specimens cultured in Septi-Chek AFB. The LCR protocol was positive in 22 specimens. None of the LCR-negative controls showed positive results. All samples positive by culture on Löwenstein-Jensen medium were positive by culture in liquid medium and by the LCR assay. However, Mycobacterium tuberculosis was detected by culture in liquid medium in two specimens that were negative by the LCR assay, whereas six specimens negative by culture in liquid medium were positive by the LCR protocol; three of these were identified as true-positive results of the LCR assay. The sensitivity, specificity, and positive and negative predictive values were 33.3%, 100%, 100%, and 93.8%, respectively, for Löwenstein-Jensen medium; 85.7%, 100%, 100%, and 98.6% for the liquid medium; and 90.4%, 98.5%, 86.3%, and 99% for the LCR assay. These findings indicate that the LCR assay may be a valid method of high diagnostic yield for direct detection of Mycobacterium tuberculosis complex in nonrespiratory specimens.