| Gaia: | <p>Tuberculosis (TB) remains a major cause for morbidity and mortality in children, and the very young and immunocompromised are especially at high risk. A major hurdle is a timely diagnosis, as microbiological confirmation – regarded as the reference standard – remains uncommon and the available tools are neither child-friendly, user-friendly, nor sufficiently accurate.</p>
<p>This thesis presents results from a large TB diagnostic cohort of children younger than 15 years recruited in five countries (South Africa, Tanzania, Mozambique, Malawi, India), aiming to evaluate the diagnostic accuracy of three novel tests for childhood TB. Participants suspected of having TB underwent thorough clinical and microbiological testing, and reference standard investigations (culture and GeneXpert® MTB/RIF Ultra, "Ultra®") were conducted alongside the novel tests. The latter included a novel immunoassay using whole-blood, two urinary tests detecting mycobacterial antigen, and a stool processing kit allowing for detection of mycobacterial DNA. Clinical case definitions for TB applied followed published consensus statements.</p>
<p>Of 974 children recruited over 30 months, 732 were eligible for assignment to clinical case definitions, being confirmed TB (32.0%, 234/732), unconfirmed TB (36.2%, 265/732), and unlikely TB (31.8%, 233/732). Of the whole cohort, 46.4% (340/732) were younger than five years, with median ages of 6.1 years (IQR 1.8-11.3) in children with confirmed TB, 4.6 years (IQR 1.7-8.2) in children with unconfirmed TB, and 5.5 years (IQR 2.5-8.0) in children with unlikely TB. HIV-coinfection was reported in 16.2% (117/732) of the overall cohort, with most cases being classified as unconfirmed TB (26.0%, 68/117). Tuberculin skin test positivity, chest radiography findings, and number of symptoms suggestive of TB were strongly associated with the diagnosis of TB when conducting logistic regression.</p>
<p>Microbiological confirmation was achieved via culture and Ultra® in 46.6% (109/234), via culture alone in 17.1% (40/234), and via Ultra® alone in 36.3% (85/234). Of the latter, 73.0% (62/85) were only confirmed by the lowest semi-quantitative readout Trace call.</p>
<p>Diagnostic accuracy of the T-cell activation marker for TB (TAM-TB) was assessed, using phenotypic characterisation of MTB-specific CD4+-cells. In the overall cohort, the sensitivity was determined at 51.3% (95% CI 41.8-60.7) and specificity at 88.2% (95% CI 81.3-93.2), comparing children with confirmed TB to those with unlikely TB. When using a microbiological reference standard requiring TB-culture positivity, the performance indicators were pronouncedly higher, with increases in sensitivity of more than 25% for certain subgroups.</p>
<p>The urinary tests evaluated here were AlereLAM and FujiLAM, both being lateral flow assays targeting the mycobacterial cell-wall protein lipoarabinomannan (LAM). Comparing children with confirmed TB to those with unlikely TB, sensitivity was determined at 31.6% (95% CI 25.2-38.6) for FujiLAM and 13.2% (95% CI 8.9-18.6) for AlereLAM, with determined specificities of 89.1% (95% CI 83.8-93.1) and 94.1% (95% CI 90.0-96.9). Applying regression analyses, test accuracy for both tests was strongly associated with clinical study centre and age of the participant, more pronouncedly so for FujiLAM.</p>
<p>Stool was processed using a kit and subsequently analysed using Ultra®. Determined sensitivity across the cohort was 34.2% (95% CI 27.5-41.4) with a specificity of 85.4% (95% CI 79.6-90.1), when comparing children with confirmed TB to those with unlikely TB. The sensitivity was significantly increased in female children (47.3%, 95% CI 36.7-58.0), compared to male (22.5%, 95% CI 14.9-31.9). Test accuracy was associated with proxies of TB bacillary load and extensive pathological findings on chest radiography.</p>
<p>In conclusion, all tests evaluated here showed modest accuracy to diagnose children with TB. Although neither of the novel tests evaluated did achieve a level of accuracy usually aimed for a diagnostic, it was sufficient to be of value in a setting where easy-to-obtain specimens can be used, circumventing the need for elaborate sampling infrastructure. If implemented in settings with limited infrastructure, their impact might still be considerable. The findings of this thesis suggest that an approach with multiple tests might be more suitable to diagnose children with TB and highlights the need for further evaluations of diagnostic multi-step algorithms based on both clinical data and novel tests.</p>
|
|---|