Background: Autopsy studies of HIV/AIDS deaths in medical in-patients in sub-Saharan Africa have all reported a high frequency of disseminated tuberculosis (TB), indicating frequent failure of diagnosis. This observational study aimed to identify improved means of rapid TB diagnosis.
Methodology: Unselected HIV-infected medical admissions to a South African district hospital were intensively investigated. Sputum, urine and blood specimens were systematically obtained within the first 24 hours. Multiple additional respiratory and non-respiratory samples were obtained throughout admission as clinically indicated. Sputum samples were tested using fluorescence microscopy, liquid culture and Xpert MTB/RIF (Xpert). Urine samples were tested using Xpert (urine-Xpert of both unconcentrated and concentrated samples) and Determine TB-LAM (urine-LAM). Other non-respiratory samples were cultured. TB diagnoses were defined by detection of Mycobacterium tuberculosis in any sample using culture or Xpert.
Results: HIV-status was ascertained in 1,013 of 1,018 (99.5%) admissions and 585 of 609 (96.1%) HIV-infected patients were enrolled. All those without an existing TB diagnosis (n=427) were included in this analysis. 3,471 TB investigations were done on 1,745 samples from a median of 3 anatomic sites per patient. TB was diagnosed in 139 patients (median CD4 count, 80 cells/μL) and symptoms were very poorly predictive. TB prevalence was 32.6% (95%CI, 28.1-37.2). Disease was extrapulmonary in 83% of cases and pulmonary in just 54% (P<0.001). Using samples obtained in the first 24-hours, the proportions of final diagnoses made by sputum microscopy, sputum-Xpert, urine-LAM and urine-Xpert (30-40 ml concentrated urine) were 19.4%, 26.6%, 38.1% and 59.0%, respectively. Rapid urine tests used together diagnosed 69.1% (96 of 139) of cases. This further increased to 80.6% (112 of 139) of cases when combined with sputum Xpert testing. However, of those with CD4 counts <100 cells/μL, 85.1% (63 of 74) could be diagnosed with urine rapid tests alone.
Conclusions: The prevalence of TB was so high and the presentation so non-specific that routine microbiological investigation for TB should be done in all HIV-infected medical in-patients in high-burden settings. Compared to Xpert testing of one sputum sample alone, the addition of urine-based testing increased the diagnostic yield of the initial TB screen 3.0-fold from 26.6% to 80.6% (P<0.001). Urine-based rapid diagnostics should be considered for routine use in this patient population.