Nucleic acid amplification techniques (NAATs) have been demonstrated to make significant improvements in the diagnosis of tuberculosis (TB), particularly in the time to diagnosis and the diagnosis of smear-negative TB. The BD ProbeTec strand displacement amplification (SDA) system for the diagnosis of pulmonary and non-pulmonary tuberculosis was evaluated. A total of 689 samples were analysed from patients with clinically suspected TB. Compared with culture, the sensitivity and specificity for pulmonary samples were 98 and 89 %, and against final clinical diagnosis 93 and 92 %, respectively. This assay has undergone limited evaluation for non-respiratory samples and so 331 non-respiratory samples were tested, identifying those specimens that were likely to yield a useful result. These were CSF (n = 104), fine needle aspirates (n = 64) and pus (n = 41). Pleural fluid (n = 47) was identified as a poor specimen. A concern in using the SDA assay was that low-positive samples were difficult to interpret; 7.8 % of specimens fell into this category. Indeed, 64 % of the discrepant results, when compared to final clinical diagnosis, could be assigned as low-positive samples. Specimen type did not predict likelihood of a sample being in the low-positive zone. Although the manufacturers do not describe the concept of a low-positive zone, we have found that it aids clinical diagnosis.