Background/Aim. Traditional methods for detection of mycobacteria, such as microscopic examination for the presence of acid-fast bacilli and isolation of the organism by culture, have either a low sensitivity and/or specificity, or take weeks before a definite result is available. Molecular methods, especially those based on nucleic acid amplification, are rapid diagnostic methods which combine high sensitivity and high specificity. The aim of this study was to determine the usefulness of the Cobas Amplicor Mycobacterium tuberculosis polymerase chain reaction (CAPCR) assay in detecting the tuberculosis cause in respiratory and nonrespiratory specimens (compared to culture). Methods. Specimens were decontaminated by the N-acetyl-L-cystein- NaOH method. A 500 μL aliquot of the processed specimen were used for inoculation of Löwenstein-Jensen (L-J) slants, a drop for acid-fast staining, and 100 μL for PCR. The Cobas Amplicor PCR was performed according to the manufacturer's instructions. Results. A total of 110 respiratory and 355 nonrespiratory specimens were investigated. After resolving discrepancies by reviewing medical history, overall sensitivity, specificity, and positive and negative predictive values for CA-PCR assay compared to culture, were 83%, 100%, 100%, and 96.8%, respectively. In comparison, they were 50%, 99.7%, 87.5%, and 98%, respectively, for the nonrespiratory specimens. The inhibition rate was 2.8% for respiratory, and 7.6% for nonrespiratory specimens. Conclusion. CA-PCR is a reliable assay that enables specialists to start treatment promptly on a positive test result. Lower value for specificity in a group of nonrespiratory specimens is a consequence of an extremely small number of mycobacteria in some of them.