BACKGROUND Diagnostic tests are critical to management when the clinical picture is unclear. We analyzed the records of patients in whom the initial diagnosis of cutaneous tuberculosis was doubtful to evaluate the utility of the Mantoux test in this setting. METHODS Of 375 patients with cutaneous tuberculosis seen in our department, the initial clinical diagnosis was doubtful in 90 patients. A Mantoux test was performed with 5 TU of purified protein derivative using standard techniques and read after two days. Patients also underwent skin biopsy and other laboratory tests and received treatment with antitubercular drugs. Patients who had case notes recording satisfactory improvement following antitubercular therapy were classified as having cutaneous tuberculosis; those with evidence of another diagnosis based on the laboratory tests and response to therapy were categorized as non-tuberculosis cases. Patients with inadequate evidence for a diagnosis of either cutaneous tuberculosis or another disorder were classified as unresolved diagnosis. RESULTS Readings were available in 79 patients: 39 were categorized as tuberculosis; 16 as non-tuberculosis; while a final diagnosis could not be reached in 24 cases. Readings ranged from 0 to 40 mm in the cases and from 0 to 30 mm in non-cases. The area under the receiver-operating characteristic curve was 0.66 (95% CI 0.55-0.81). Using a cut-off of 10 mm, the sensitivity and specificity of the tests were 58.97 and 62.50%, respectively. CONCLUSIONS The Mantoux test is of low accuracy in the diagnosis of doubtful cases of cutaneous tuberculosis.