Abstract
The world health organization estimates that in 2017, close to 1 million children below 15 years developed tuberculosis but only half of them were notified. Difficulty to obtain sputum in children and the paucibacillary nature of intrathoracic childhood tuberculosis challenge the diagnosis of tuberculosis in children. This leads to the common use of empirical treatment with a high risk of over or under diagnosis. Besides that, few facilities in low resource settings have adequate laboratory capacity to diagnose tuberculosis. Samples must be transported to a reference laboratory, which can effect performance of the tests, especially in the absence of cold chain.Three studies were conducted in Mbarara (Uganda) to evaluate non-respiratory samples and specimen preservation methods to improve diagnosis of pediatric tuberculosis. In the first study, we assessed the performance of XpertMTB/RIF on sputum and stool in children with presumptive tuberculosis and documented outcomes of children according to the tuberculosis treatment decision. In the second study, we assessed the performance of stool XpertMTB/RIF and urine lipoarabinomanann (LAM) among children admitted with severe illness. In the 3rd study, we determined XpertMTB/RIF and MGIT culture recovery rates of smear positive sputum specimen kept untreated at room temperature and treated with either Omnigene or ethanol over different time periods.Of 392 children (median age 3.9 years, 45.4% female and 31% HIV infected) enrolled in the 1st study, 4.3% (17/392) were microbiologically confirmed tuberculosis. Using a microbiological reference standard, sputum XpertMTB/RIF had a 90.9% sensitivity and specificity of 99.1%. The sensitivity and specificity of stool XpertMTB/RIF was 55.6% and 98.2%. The study reported mortality of 6.9% within three months with a higher proportion (10.7%) among children treated for tuberculosis compared to the non-treated children (4.5%). None of treated children with bacteriologically confirmed tuberculosis died compared to 12.3% of those treated empirically.Of 234 patients (median age 16.5 months, 48.3% female, 31.6% HIV infected, 58.5% severely malnourished) enrolled in the 2nd study, 5.1% were microbiologically confirmed tuberculosis. Stool XpertMTB/RIF had a sensitivity of 50% and specificity of 99.1%. For the urine LAM test, it was 50% and 74.1%, respectively. False positive LAM results were more common among low grade positive LAM results and occurred more frequently when urine samples had bacterial contamination.The 3rd study documented that by 15th day, there was no difference of XpertMTB/RIF recovery rate between samples treated with Omnigene or ethanol and untreated samples, meaning that in the study conditions there was no benefit of adding any preservative for samples stored at room temperature up to 15 days. We observed a substantial loss of viability of Mycobacterium tuberculosis on samples treated with Omnigene, which does not support the use of Omnigene for sample transportation before MGIT testing.In conclusion, XpertMTB/RIF on stool gave promising results for the use in children unable to provide sputum and could be an interesting alternative to more complex methods such as sputum induction and gastric aspirate for primary health care centers of limited resource countries. The low specificity of the urine LAM requires further investigation before its use for diagnosis of tuberculosis in children. Despite the encouraging XpertMTB/RIF results from specimen preserved either with Omnigene or ethanol further evaluation under routine field conditions is necessary.