Performance of urinary diagnostic tests in inpatients treated for tuberculosis in a referral hospital in Rwanda

Background Current diagnostic tests for tuberculosis (TB) among patients hospitalized in resource-poor settings are suboptimal. Sensitivity of the Determine™ TB LAM lateral flow Lipoarabinomannan test (LF-LAM) in urine, a low-cost point-of-care test, is too low in outpatients, but may be adequate in HIV co-infected TB inpatients. Methods Adult inpatients starting TB treatment at Kigali University Teaching Hospital were prospectively enrolled. Diagnostic workup for TB included sputum for auramine phenol microscopy and Xpert MTB/RIF, urine for Xpert MTB/RIF and LF-LAM assays, HIV testing and medical imaging. Results From November 2016 to September 2017, 114 patients with urine LF-LAM results, 57 (50%) HIV co-infected, were included. TB was confirmed by sputum auramine microscopy in 51 (45%), by sputum Xpert MTB/RIF in 46 (40%), by urine LF-LAM in 56 (46%) and by urine Xpert MTB/RIF in 17 (19%). All urine Xpert MTB/RIF positive samples were also LF-LAM positive. In addition to sputum analysis, urine LF-LAM increased diagnostic yield from 61/114 (54%) to 86/114 (75%). A positive urine LF-LAM test was not associated with HIV co-infection. Conclusion In hospitalized TB patients, irrespective of HIV status, urine LF-LAM provides a useful diagnostic test in addition to sputum samples. Sensitivity of urine Xpert MTB/RIF is disappointing and provides no additional sensitivity beyond urine LF-LAM.