Intensive tuberculosis screening for HIV-infected patients starting antiretroviral therapy in Durban, South Africa.

BACKGROUND: The World Health Organization (WHO) recommends cough as the trigger for tuberculosis screening in human immunodeficiency virus (HIV)-infected patients, with acid-fast bacillus (AFB) smear as the initial diagnostic test. Our objective was to assess the yield and cost of a more intensive tuberculosis screening in HIV-infected patients starting antiretroviral therapy (ART) in Durban, South Africa.

METHODS: We prospectively enrolled adults, regardless of tuberculosis signs/symptoms, who were undergoing ART training from May 2007 to May 2008. After the symptom screen, patients expectorated sputum for AFB smear, tuberculosis polymerase chain reaction (PCR), and mycobacterial culture. Sensitivity and specificity of different symptoms and tests, alone and in combination, were compared with the reference standard of 6-week tuberculosis culture results. Program costs included personnel, materials, and cultures.

RESULTS: Of 1035 subjects, 487 (59%) were female; median CD4 cell count was 100 cells/microL. A total of 210 subjects (20%) were receiving tuberculosis treatment and were excluded. Of the remaining 825 subjects, 158 (19%) had positive sputum cultures, of whom 14 (9%) had a positive AFB smear and 82 (52%) reported cough. The combination of cough, other symptoms, AFB smear, and chest radiograph had 93% sensitivity (95% confidence interval, 88%-97%) and 15% specificity (95% confidence interval, 13%-18%). The incremental cost of intensive screening including culture was $360 per additional tuberculosis case identified.

CONCLUSIONS: Nearly 20% of patients starting ART in Durban, South Africa, had undiagnosed, culture-positive pulmonary tuberculosis. Despite WHO recommendations, neither cough nor AFB smear were adequately sensitive for screening. Tuberculosis sputum cultures should be performed before ART initiation, regardless of symptoms, in areas with a high prevalence of HIV and tuberculosis.