ABSTRACT
RATIONALE: The World Health Organization recently revised its recommendations for tuberculosis (TB) diagnosis in people with HIV. Most studies cited to support these policies involved HIV-uninfected patients and only evaluated sputum specimens. OBJECTIVES: To evaluate the performance of acid-fast bacilli smear and mycobacterial culture on sputum and nonsputum specimens for TB diagnosis in a cross-sectional survey of HIV-infected patients. METHODS: In Thailand and Vietnam, we enrolled people with HIV regardless of signs or symptoms. Enrolled patients provided three sputum, one urine, one stool, one blood, and, for patients with palpable peripheral adenopathy, one lymph node aspirate specimen for acid-fast bacilli microscopy and mycobacterial culture on solid and broth-based media. We classified any patient with at least one specimen culture positive for Mycobacterium tuberculosis as having TB. MEASUREMENTS AND MAIN RESULTS: Of 1,060 patients enrolled, 147 (14%) had TB. Of 126 with pulmonary TB, the incremental yield of performing a third sputum smear over two smears was 2% (95% confidence interval, 0-6), 90 (71%) patients were detected on broth-based culture of the first sputum specimen, and an additional 21 (17%) and 12 (10%) patients were diagnosed with the second and third specimens cultured. Of 82 lymph nodes cultured, 34 (42%) grew M. tuberculosis. In patients with two negative sputum smears, broth-based culture of three sputum specimens had the highest yield of any testing strategy. CONCLUSIONS: In people with HIV living in settings where mycobacterial culture is not routinely available to all patients, a third sputum smear adds little to the diagnosis of TB. Broth-based culture of three sputum specimens diagnoses most TB cases, and lymph node aspiration provides the highest incremental yield of any nonpulmonary specimen test for TB.
