A prospective study was undertaken to assess the diagnostic value of various rapid diagnostic tests for tuberculosis in pleural fluid, and to assess the sensitivity and speed of a radiometric mycobacterial culture system (BACTEC, Johnson Laboratories). Patients presenting to the Department of Medicine at Groote Schuur Hospital with pleural effusions for diagnostic pleural aspiration and biopsy over a 6 month period were entered into the study. Because the incidence of tuberculous effusions was observed to be high in this population (65% of 94 patients), patients from the Department of Radiotherapy with proven malignant disease and the development of new pleural effusions requiring diagnostic or therapeutic aspiration were included in the study in order to increase the number of control patients without tuberculosis. The 111 patients (17 of whom were recruited from the Department of Radiotherapy) were divided into 4 diagnostic categories: tuberculosis - 62 patients, malignant - 28 patients, miscellaneous conditions - 10 patients, and undiagnosed - 11 patients (3 of whom probably had tuberculosis). There were 59 male patients. The racial distribution was 11 whites, 51 of mixed race, and 49 blacks. Exudative pleural effusions were present in 109 patients. Closed pleural biopsies with the Abrams needle were performed on 100 patients using a modified version of the standard technique whereby larger specimens were obtained by stripping pleura off the chest wall. Seven pleural biopsies were reported as inadequate by the pathologist and the diagnostic yield of the procedure was 63%. Tuberculosis was confirmed histologically or by culture in 62 patients. The age distribution of these patients was bimodal, with most cases occuring in the third decade. The presentation was usually acute, with 60% of patients being symptomatic for less than 4 weeks. Granulomata were found on initial pleural biopsy in 52 cases (84%). Pleural biopsy culture was positive in 44 cases (71%). The radiometric culture system tested (12B BACTEC) yielded the same number (14) of positive cultures as conventional mycobacterial culture media in pleural fluid, but was almost twice as fast. Bedside inoculation of pleural fluid into 13A BACTEC bottles more than doubled the yield in the 24 patients tested (11 positive cultures compared with 4 each for conventional and 12B BACTEC media, p=0.046). The rapid diagnostic tests assessed on pleural fluid were adenosine deaminase (ADA), an antigen (BCG) capture enzymelinked immunosorbent assay (ELISA), and a specific DNA probe after amplification with the polymerase chain reaction. ADA was found to have a sensitivity of 0.77 and a specificity of 0.83 in the 109 patients tested, and values were significantly higher in tuberculosis patients compared with the other three diagnostic categories (p< 0.001 ). The ELISA test was performed on 103 patients and showed a sensitivity of only 0.26 and a specificity of 0.72. The DNA probe was performed on 43 patients, and had a sensitivity of 0.93 with a specificity of 0.43. Contamination of samples or latent tuberculous infection may have been responsible for the poor specificity of the DNA probe.