Background: Disseminated tuberculosis (TB) is a common cause of death among human immunodeficiency virus (HIV)-infected patients in developing countries. Blood culture offers a potential means to diagnose disseminated TB, but optimal blood culture methods have not been studied.
Methods: Two hundred and fifty-eight HIV-infected patients hospitalized in Tanzania with ≥2 weeks fever or cough had diagnostic studies for TB: 3 sputum samples for acid-fast bacilli smear and culture; 40 ml of blood for culture, randomized 1:1 to 40 ml × 1, or 20 ml × 2 collected 12-24 h apart. Blood was processed using automated MB BacT(®) broth and manual Isolator(®) lysis-centrifugation agar. Mortality was assessed at 2 months.
Results: TB was confirmed in 83 (32%) of 258 patients: by sputum only in 42 (51%, median CD4 = 72 cells/μl), blood only in 15 (18%, median CD4 = 44 cells/μl), and in sputum and blood in 26 (31%, median CD4 = 12 cells/μl). Blood was positive in 21 (16%) for 40 ml × 1 vs 20 (15%) for 20 ml × 1 (p = 0.83) vs 20 (16%) for 20 ml × 2 (p = 0.97). MB BacT was positive in 31 (76%) and Isolator was positive in 20 (49%) of 41 samples (p = 0.01). The mean colony-forming units/ml was 8 (range 3-14). Twenty-one (51%) patients with disseminated TB died; median survival was 6 days (range 0-58).
Conclusions: Disseminated TB in HIV is characterized by persistent bacteraemia, delayed microbiological detection, and high mortality. Twenty millilitres of blood processed by automated broth is the optimal culture method to detect disseminated TB. Empiric TB therapy is warranted for HIV-infected patients from TB-endemic countries with prolonged cough or fever.