Tuberculosis is still a global public health problem. An estimation of one-third of world population were infected with M. tuberculosis, whereas about 8 millions were reported as new active tuberculosis cases each year with annual death of 3 millions. The unexpected reemergence of tuberculosis in industrialized countries and recent outbreaks of multidrug-resistant M. tuberculosis as well as the dramatically enhanced vulnerability of human immunodeficiency virus-positive individuals to tubercle bacilli underline the urgent need for rapid and accurate detection of M. tuberculosis. The rapid diagnosis of tuberculosis is one of the measures required for tuberculosis control. The current detection methods routinely used in the clinical mycobacteriology laboratory are almost entirely dependent on the growth of the organism which is a time consuming technique. The most recent advances in TB diagnostic have, therefore, been concentrated in the fields of direct detection of acid-fast bacilli in clinical specimens by molecular biological methods such as in vitro DNA amplification by polymerase chain reaction (PCR). This study has been designed to develop and to assess the PCR technique for direct detection of M. tuberculosis in sputum samples. The KS4 DNA fragment, which was previously studied in our laboratory and shown to be specific to M. tuberculosis complex by Southern blot hybridization, was isolated from recombinant plasmid pWR6 and was cloned in the plasmid pGEM4. It was later subcloned in the cloning vector, M13 mpl8 and M13 mpl9, and employed for nucleotide sequencing with dideoxy chain termination method. The KS4 fragment was 848 bp long and consisted of 48 copies of 10-bp consensus sequence, CAACA(T)/(,C)CGGC, alternated with 5-bp variable sequences. Two pairs of primers, the outer primer,TPOL/TPOR (5-CCGGCGCTTGCGGGCGGAC CCACCGCC-3/ 5-CAGGCTGCCCTGCCCCACGCCCCGGATC-3) and the inner primer, TPIL/TPIR (5-CTTACCGGCAACAACCAGA-3/ 5-TTGCCG ACGTCCAACATAC-3) were designed from the sequence of KS4 and used in nested PCR. The sensitivity and specificity of these primers were tested by amplifying 10-fold diluted purified M. tuberculosis DNA and DNA from M. tuberculosis complex, 12 MOTT, 13 bacteria, and 2 yeasts, respectively. Detection of 227 bp amplified product was accomplished by ethidium bromide staining of agarose gel and viewing under a UV-transilluminator. The PCR assay could detect as little as 1 fg (approximately 1/5 organism) of purified DNA. These primers were not specific because they could amplify DNA not only from M. tuberculosis complex but also from other 11 mycobacteria e.g. M. aurum, M. avium, M. chelonei, M. duvalii, M. flavescens, M. fortuitum, M. gordonae, M. kansasii, M. neolactis, M. phlei, and M. smegmatis, and 4 bacteria, such as Diplococcus sp, E. coli, N. asteroides, and P. aeruginosa. Thus, these primers were not appropriate to be used in direct detection of M. tuberculosis in clinical samples. In order to assess the PCR technique in direct detection of M. tuberculosis in sputum samples, the new set of primers, 16SIL/16SIR (5GGATAGGACCACGGGAT-3 / 5-TACCGTCAATCCGAGAG-3) was synthesized. They were derived from 16S rRNA gene sequence of M. tuberculosis which were submitted to GenBank database. For increasing ability to detect very small amount of DNA, reamplification with the same pair of primers was employed. Positive reamplification revealed a 304 bp specific DNA band. The PCR assay using these primers had a detection limit of 100 fg (approximately 20 organisms) for purified M. tuberculosis DNA and could amplify DNA only from M. tuberculosis complex, M. kansasii, and M. gordonae. The specificity of 16SIL/16SIR was from this point of view acceptable for further used with clinical samples because of the very low incidence of M. kansasii and M. gordonae infections in Thai population. There were 196 sputum samples, 62 were from patients with definite diagnosis of tuberculosis, 44 were from patients with other proved pulmonary diseases, and 90 were from patients with suspected as having tuberculosis, included in the assessment of PCR for direct detection of M. tuberculosis comparing with culture results. The sensitivity and specificity of the assay were 91.9% (57/62) and 95.4% (42/44), respectively. The positive predictive value and negative predictive value were 96.6% and 89.4%, respectively. From these results, it was concluded that PCR assay using 16SIL/16SIR could be used in direct detection of M. tuberculosis in sputum samples.