Diagnostic accuracy and effectiveness of the Xpert MTB/RIF assay for the diagnosis of HIV-associated lymph node tuberculosis

Xpert MTB/RIF (Xpert) is recommended for human immunodeficiency virus (HIV)-associated pulmonary tuberculosis but not extrapulmonary tuberculosis. We assessed the performance of Xpert for HIV-associated lymph node tuberculosis (LNTB), the most common type of extrapulmonary tuberculosis. Among HIV-infected adults suspected of LNTB presenting for fine needle aspirate (FNA) at a South African hospital, we assessed the diagnostic accuracy of Xpert using either FNA culture or a composite of microscopy, culture, and cytology as the reference standard, and evaluated the impact of different diagnostics on patient management. Among 344 adults with valid FNA culture and Xpert results, 84 (24 %) were positive on microscopy, 149 (43 %) on culture, 152 (53 %) on Xpert, and 181 (57 %) had a cytology result suggestive of tuberculosis. Using liquid culture as the reference standard, the specificity of a single Xpert was suboptimal (88.2 %) but the sensitivity was high [93.3 %, 95 % confidence interval (CI) 87.6-96.6] and increased with decreasing CD4 count (from 87.0 % for CD4 > 250 to 98.6 % for CD4 < 100 cells/mm(3)). Using a composite reference standard reduced the sensitivity to 79.2 % but increased the specificity to 98.6 %. All Xpert-positive patients initiated treatment within one day, compared to 70 % of culture-positive but Xpert-negative and 13 % of culture- and Xpert-negative but cytology-positive patients. Xpert is accurate and effective and could be endorsed as the initial diagnostic for HIV-associated LNTB.