The polymerase chain reaction (PCR) is a technique that can be used to amplify a specific DNA genomic sequence, whereby the presence of an extremely small number of bacteria can be detected. The high sensitivity of PCR is particularly useful in paucibacillary situations such as non-pulmonary tuberculosis (TB). The aims of the present study were to establish a PCR assay for the rapid detection of Mycobacterium tuberculosis (MTb) in urine, to compare the sensitivity of PCR with routine culture technique (Bactec) and to determine the optimal type of urine specimen for PCR detection of MTb. In the first phase of the study, a total of 92 urine specimens were collected from 83 patients with suspected urinary tract TB. Two urine specimens in 2 patients were positive for TB by both PCR and Bactec, while 90 specimens from 81 patients were negative by both methods. Inhibition of PCR was present in nine urine specimens (10%). In the second phase of the study, a further seven patients were selected for intensive investigation to determine the optimal urine sampling for PCR detection of MTb. The conclusions of the study are that PCR can provide much faster confirmation of urinary TB (within 24-48 h) than Bactec urine culture (which may take several weeks). About 10% of urine specimens could not be evaluated by PCR due to the presence of inhibitory substances of unknown nature. MTb organisms were found to be excreted intermittently in the urine of infected patients, and single specimens were more likely to be false negative than a 24-h sample. The best method appeared to be the concentration of a large volume of urine, for instance 11 concentrated to 2 ml.