HIV-associated tuberculosis is difficult to diagnose and results in high mortality. Frequent extra-pulmonary presentation, inability to obtain sputum, and paucibacillary samples limits the usefulness of nucleic-acid amplification tests and smear microscopy. We therefore assessed a urine-based, lateral flow, point-of-care, lipoarabinomannan assay (LAM) and the effect of a LAM-guided anti-tuberculosis treatment initiation strategy on mortality.

We did a pragmatic, randomised, parallel-group, multicentre trial in ten hospitals in Africa—four in South Africa, two in Tanzania, two in Zambia, and two in Zimbabwe. Eligible patients were HIV-positive adults aged at least 18 years with at least one of the following symptoms of tuberculosis (fever, cough, night sweats, or self-reported weightloss) and illness severity necessitating admission to hospital. Exclusion criteria included receipt of any anti-tuberculosis medicine in the 60 days before enrolment. We randomly assigned patients (1:1) to either LAM plus routine diagnostic tests for tuberculosis (smear microscopy, Xpert-MTB/RIF, and culture; LAM group) or routine diagnostic tests alone (no LAM group) using computer-generated allocation lists in blocks of ten. All patients were asked to provide a urine sample of at least 30 mL at enrolment, and trained research nurses did the LAM test in patients allocated to this group using the Alere Determine tuberculosis LAM Ag lateral flow strip test (Alere, USA) at the bedside on enrolment. On the basis of a positive test result, the nurses made a recommendation for initiating anti-tuberculosis treatment. The attending physician made an independent decision about whether to start treatment or not. Neither patients nor health-care workers were masked to group allocation and test results. The primary endpoint was 8-week all-cause mortality assessed in the modified intention-to-treat population (those who received their allocated intervention). This trial is registered with ClinicalTrials.gov, number NCT01770730.

Between Jan 1, 2013, and Oct 2, 2014, we screened 8728 patients and randomly assigned 2659 to treatment (1336 to LAM, 1323 to no LAM). 108 patients did not receive their allocated treatment, mainly because they did not meet the inclusion criteria, and 23 were excluded from analysis, leaving 2528 in the final modified intention-to-treat analysis (1257 in the LAM group, 1271 in the no LAM group). Overall all-cause 8-week mortality occurred in 578 (23%) patients, 261 (21%) in LAM and 317 (25%) in no LAM, an absolute reduction of 4% (95% CI 1–7). The risk ratio adjusted for country was 0·83 (95% CI 0·73–0·96), p=0·012, with a relative risk reduction of 17% (95% CI 4–28). With the time-to-event analysis, there were 159 deaths per 100 person-years in LAM and 196 per 100 person-years in no LAM (hazard ratio adjusted for country 0·82 [95% CI 0·70–0·96], p=0·015). No adverse events were associated with LAM testing.

Bedside LAM-guided initiation of anti-tuberculosis treatment in HIV-positive hospital inpatients with suspected tuberculosis was associated with reduced 8-week mortality. The implementation of LAM testing is likely to offer the greatest benefit in hospitals where diagnostic resources are most scarce and where patients present with severe illness, advanced immunosuppression, and an inability to self-expectorate sputum.

The European & Developing Countries Clinical Trials Partnership, the South African Medical Research Council, and the South African National Research Foundation.