BACKGROUND Diagnosis of tuberculosis (TB) in children remains challenging due to the paucibacillary nature of the disease. Detection of TB using urine lipoarabinomannan (LAM) antigen was evaluated in children with presumed TB. MATERIAL and METHODS Children with presumed intrathoracic tuberculosis (ITTB) and lymph node TB (LNTB) were enrolled. Expectorated or induced sputum or gastric aspirates from ITTB patients and fine-needle cytological aspirates from LNTB patients were subjected to Ziehl-Neelsen staining, MGIT™960™ culture and Xpert MTB/RIF testing. Urine samples were tested to detect LAM, and the sensitivity and specificity calculated. RESULTS Of 280 children with presumed ITTB and 101 with presumed LNTB, respectively 71 (25.3%) and 25 (24.7%) were categorised as 'confirmed TB', 70 (25%) and 33 (32.7%) as 'unconfirmed TB', and 139 (49.6%) and 43 (42.5%) as 'unlikely TB'. Respectively 8 (2.8%) children with ITTB and 3 (2.9%) with LNTB were positive on smear, 56 (20.0%) and 23 (22.7%) on Xpert, and 50 (17.8%) and 9 (8.9%) on culture. LAM assay sensitivity was 73.2% in confirmed ITTB cases, and 76% in confirmed LNTB cases; LAM assay specificity in children with ITTB and those with LNTB initiated on anti-tuberculosis treatment was respectively 92% and 93%. Detection of TB using the LAM assay was significantly better than detection using Xpert ( < 0.05 vs. < 0.002). CONCLUSION Urinary LAM testing showed high specificity and sensitivity, was detected in more cases initiated on treatment than reference tests, and improved disease detection by 38.5% in ITTB patients and by 41.6% in LNTB patients. .
