<ns3:p><ns3:bold>Background: </ns3:bold>Amongst HIV-positive adults in South Africa with initial negative Xpert results, we compared the yield from repeating Xpert MTB/RIF (“Xpert”) on sputum to guideline-recommended investigation for tuberculosis (TB).</ns3:p><ns3:p> <ns3:bold>Methods:</ns3:bold>  A systematic sample of adults attending for HIV care were enrolled in a cohort exploring TB investigation pathways. This substudy was restricted to those at highest risk of TB (CD4&lt;200 cells/mm<ns3:sup>3</ns3:sup> or unknown) who had a negative initial Xpert result.</ns3:p><ns3:p> At attendance for the Xpert result, a repeat sputum sample was stored, and further investigations facilitated per national guidelines. Participants were reviewed monthly, with reinvestigation if indicated, for at least three months, when sputum and blood were cultured for mycobacteria, and the stored sputum tested using Xpert. We defined TB as “confirmed” if Xpert, line probe assay or <ns3:italic>Mycobacterium tuberculosis </ns3:italic>culture within six months of enrolment were positive, and “clinical” if TB treatment was started without microbiological confirmation.</ns3:p><ns3:p> <ns3:bold>Results: </ns3:bold>Amongst 227 participants with an initial negative Xpert result (63% female, median age 37 years, median CD4 count 100 cells/mm<ns3:sup>3</ns3:sup>), 28 (12%) participants had TB diagnosed during study follow-up (16 confirmed, 12 clinical); stored sputum tested positive on Xpert in 5/227 (2%). Amongst 27 participants who started TB treatment, the basis was bacteriological confirmation 11/27 (41%); compatible imaging 11/27 (41%); compatible symptoms 2/27 (7%); and unknown 3/27 (11%). </ns3:p><ns3:p> <ns3:bold>Conclusions:  </ns3:bold>Amongst HIV-positive individuals at high risk of active TB with a negative Xpert result, further investigation using appropriate diagnostic modalities is more likely to lead to TB treatment than immediately repeating sputum for Xpert. TB diagnostic tests with improved sensitivity are needed.</ns3:p>