Innovative strategies to improve the diagnosis of intrathoracic tuberculosis in children

[摘要] ENGLISH ABSTRACT: Paediatric tuberculosis (TB) contributes approximately 10% of the global TBburden, with over one million estimated new cases and 253,000 TB-related deaths inchildren during 2016. Paediatric TB is a particular problem in low and middle incomecountries. However, the majority of paediatric cases were not notified to National TBPrograms or the World Health Organization and >96% of deaths were estimated to haveoccurred in children who were not receiving antituberculosis treatment. Young, HIVinfectedand malnourished children progress rapidly from infection with Mycobacteriumtuberculosis (M.tb) to TB, and are at exquisite risk of significant morbidity and mortalityfrom complicated and disseminated forms of TB. The challenges around the diagnosis andmicrobiological confirmation of pulmonary TB (PTB), the most common manifestation ofTB disease in children, contribute to poor access to appropriate treatment and to underreporting.The diagnosis of TB in young children typically relies on the evaluation of clinicalsymptoms and epidemiological factors, and, if available, includes tests of TB infection andchest radiology. All of these tools have considerable limitations and cannot reliablyconfirm or exclude a diagnosis of PTB. However, the bacteriological confirmation of PTB in children requires the collection of respiratory specimens using procedures that areboth relatively invasive and resource-intensive. Furthermore, the current gold standardof diagnosis, mycobacterial culture, has low sensitivity (approximately 30%) and longturnaround time (up to 6 weeks) in children, who typically have paucibacillary TB (lowbacillary load). In resource-limited settings, the capacity for respiratory sampling ofyoung children is typically low. These diagnostic challenges prevent adequate reportingand global surveillance of paediatric TB. Diagnostic uncertainty also compromises the clinical management of paediatric PTB, resulting in over- and under-treatment, and hasresulted in the systematic exclusion of children from much-needed interventionalresearch, including tuberculosis treatment trials. Diagnostic research in paediatric PTBhas also been poorly standardised, making generalizability and comparability of resultsdifficult. In addition, the insensitive reference standard has hindered progress towardsthe development of new diagnostic tests tailored for children.In an effort to develop and investigate more feasible strategies to improve andpromote microbiological testing of children with suspected PTB living in high TB-burdensettings, I enrolled a large well-characterized cohort of children presenting to hospitalwith suspected PTB. Children were thoroughly investigated, using standard approachesand intensive specimen collection for liquid culture and molecular testing by Xpert ®MTB/RIF (Xpert). Chest radiographs were dual read by blinded experts and reportedusing standard forms. All children were followed regardless of their final diagnosis andthe spectrum of TB and non-TB disease was well described. I evaluated a number of noveldiagnostic strategies, including the use of stool specimens for diagnosis of PTB usingculture and Xpert, using different stool processing methods, and pooling respiratoryspecimens to improve the diagnostic yield and reduce the cost of laboratory testing.Importantly, I developed a framework for future evaluation of novel diagnostic tools/biomarkers for the diagnosis of PTB in children.The total cohort included 608 children and was representative of thedemographics and spectrum of disease observed in many high TB-burden settings, whereyoung children bear the highest burden of TB disease. The median age of the cohort was16.2 months, with 11.8% HIV-infected. Infants below 6 months of age constituted almost15% of the total cohort. More than 20% of children had a non-specific clinicalpresentation, with similar prevalence of acute respiratory symptoms across all age groups and diagnostic categories. Radiological features not typically associated with PTBwere common, and indicate a high burden of respiratory pathology as well as potentiallynon-typical radiological manifestations of PTB. Two hundred and eighty-one (46.2%)children were diagnosed with PTB and were prescribed antituberculosis treatment: 117(41.6%) were microbiologically confirmed by Xpert or culture, which represents a highdiagnostic yield, considering that approximately 50% of children with PTB had nonseverepulmonary disease. In addition, 20/327 (6.6%) children initially consideredsymptomatic controls were initiated on antituberculosis treatment within two months ofenrolment, due to poor clinical progress or positive results from baseline and follow-upbacteriological investigations. This emphasizes the importance and utility of carefulspecimen collection, incorporating different specimen types and different diagnostictests and of follow-up of all children in whom there is a clinical suspicion of PTB.An unexpectedly high proportion of young infants <6 months of age had severePTB, including cavities, associated with high bacillary load and smear-positivity. Inaddition, young infants and HIV-infected children were high-risk groups for disseminated TB. This calls for urgent priority to be given towards the development of tailoreddiagnostic tests that can rapidly confirm and quantify M.tb disease in the youngestchildren, and in the early stages of disease, prior to rapid progression to severe TB.Careful consideration should be given to infection control measures when managing andinvestigating children, including young infants with suspected PTB.I showed that stool as a specimen was useful to confirm M.tb using Xpert inchildren with severe pulmonary disease, particularly in children with cavities on chestradiograph, detecting 45% of those who were bacteriologically confirmed on respiratoryspecimens. A novel centrifugation-free processing method for stool specimens (stoolprocessing kit) showed similar results to the more laborious, centrifugation-dependent methods I initially investigated. This new approach could be used with more sensitivemolecular assays in future to improve stool-based diagnosis of PTB in children. Incontrast, stool culture had limited value in the detection of M.tb, primarily due to veryhigh contamination (>41% of stool cultures) using standard N-acetyl-l-cysteine–sodiumhydroxide (NALC-NaOH 1.25%) decontamination protocols.Finally, I showed that pooling up to three respiratory specimens of different types(gastric aspirate, induced sputum and nasopharyngeal aspirate) per child, in childrenwho could not expectorate sputum, had similar diagnostic yield by Xpert and culture asindividually testing the same three single respiratory specimens. In paired analyses,pooled specimens had significantly higher overall yield than induced sputum andnasopharyngeal aspirate alone, but had similar diagnostic yield as a single gastricaspirate (86.5% vs. 74.4% respectively, p=0.46). The overall yield of three individualspecimens tested individually was 86% of all confirmed cases, similar to the overall yieldof pooled specimens. These results support the substantial diagnostic value of a singlegastric aspirate using culture and Xpert, and of 'front-loading specimens of differenttypes on one day to improve the feasibility of specimen collection in young children.Through this cohort study, I collected comprehensive follow-up data documentingresponse to antituberculosis treatment and clinical progress in children not receivingantituberculosis treatment (symptomatic controls), to 6 months. These data will befurther analysed to validate recently proposed clinical case definitions for TB diagnosticresearch in children, including the diagnostic value of clinical and other follow-upmeasures. Symptomatic controls who initially presented with symptoms suggestive ofPTB will be further analysed to better understand the spectrum of non-TB respiratorydisease borne by children from high-TB burden settings. I have also established a bio-repository of well-characterised blood and urine specimens for evaluation of promisingdiagnostic and prognostic biomarkers of TB disease in children.In summary, through this body of research, I have generated novel data on theutility of several feasible diagnostic strategies for the diagnosis of PTB in HIV-infectedand uninfected children from high TB-burden settings. I have analysed these data inrelation to relevant clinical and laboratory characteristics in order to make specificrecommendations on the most appropriate placement of these strategies, consideringboth target populations and different levels of health care. I was able to do this by carryingout a well-designed study, in a well-described cohort and by comprehensively reportingon all aspects of the study, including non-evaluable results and complex clinical scenarios.These aspects should be considered when future diagnostic studies for paediatric PTBare being designed, implemented and reported. I have created a rigorous framework forthe evaluation of future novel diagnostic strategies, and I have identified numerous areaswhich require further research and intervention.

[发布机构] Stellenbosch University