Assessment of the Xpert MTB/RIF assay for diagnosis of tuberculosis with gastric lavage aspirates in children in sub-Saharan Africa: a prospective descriptive study.

BACKGROUND Rapid and accurate diagnosis of pulmonary tuberculosis in children remains challenging because of difficulties in obtaining sputum samples and the paucibacillary nature of the disease. The Xpert MTB/RIF assay is useful for rapid diagnosis of childhood tuberculosis with sputum and nasopharyngeal samples. We assessed this assay for the detection of tuberculosis and multidrug resistant (MDR) tuberculosis with gastric lavage aspirate (GLA) samples in children admitted to hospital. METHODS We did a prospective study to assess the sensitivity and specificity of the Xpert MTB/RIF assay with GLA samples for the detection of pulmonary tuberculosis and MDR tuberculosis in new paediatric inpatient admissions at the University Teaching Hospital, Lusaka, Zambia. Children aged 15 years or younger were recruited between June, 2011, and May, 2012. GLA and sputum were analysed by standard smear-microscopy, mycobacterial growth indicator tube (MGIT) culture, MGIT drug-susceptibility testing, and the Xpert MTB/RIF assay. Sensitivity of the Xpert MTB/RIF assay was assessed with the Pearson chi(2) or Fishers exact test. FINDINGS Of 930 children, 142 produced sputum and GLA was obtained from 788 non-sputum producers. Culture-positive tuberculosis was identified in 58 (6.2%) of 930 children: ten from sputum producers and 48 from GLA of non-sputum producers. The sensitivity and specificity of the Xpert MTB/RIF assay were similar: sensitivity was 68.8% (95% CI 53.6-80.9) for GLA versus 90.0% (54.1-99.5; p=0.1649) for sputum samples; specificity was 99.3% (98.3-99.8) for GLA and 98.5% (94.1-99.7; p=0.2871) for sputum samples. The Xpert MTB/RIF assay detected an extra 28 tuberculosis cases compared with smear microscopy and was significantly more sensitive than smear microscopy for both sputum (90.0% [54.1-99.5] vs 30.0% [8.1-64.6], p=0.01) and GLA (68.8% [53.6-80.9] vs 25.0% [14.1-40.0], p<0.0001). The assay load did not differ significantly by sample type (p=0.791). 22 children were infected with HIV and tuberculosis and significant differences in assay performance could not be detected when stratifying by HIV status for either sample type. The Xpert MTB/RIF assay detected rifampicin resistance in three GLA samples: two confirmed as MDR tuberculosis and one false positive. INTERPRETATION Analyses of GLA samples with the Xpert MTB/RIF assay is a sensitive and specific method for rapid diagnosis of pulmonary tuberculosis in children who cannot produce sputum. The single site nature of our study invites caution.