Closing the paediatric tuberculosis diagnostic gap – steps to accelerate new TB tools for children

Of the 1.3 million children less than 15 years old estimated to have tuberculosis (TB) worldwide, about half are not reported to public health programs.^1,2 India has the largest estimated number of childhood TB cases, representing 28% of the global burden, and 13% of all TB cases in India.^3,4 However, children represented only 5% of the reported TB cases in 2022 in India,¹ highlighting the large current diagnostic gap. This further contributes to delays in treatment, as we have achieved only 71% of the United Nations high level meeting target for TB treatment in children.¹  There are several important initiatives to increase TB notifications,⁵ including in India that has a large private sector,⁶ and ongoing efforts to improve provider education and integration of TB screening and care with routine child services.^7,8

Despite these programs, an ongoing major challenge is the lack of adequate TB diagnostics for children.² Sputum-based testing to find the TB bacteria is the standard of care for diagnosing TB, but children are often unable to expectorate sputum on their own and frequently require collection with induced sputum or gastric/nasopharyngeal aspiration which may not be available or feasible in lower level facilities.⁹ At the same time, children have a low amount of bacteria in their sputum (paucibacillary), and so the currently available molecular testing is unable to identify a large proportion of TB cases (poor sensitivity). Although there has been greater funding in research and development for paediatric TB, it is far below the goal target.¹⁰ As outlined by the World Health Organization (WHO), a simple, non-invasive, non-sputum-based test that can be implemented at the point-of-care is urgently needed to support TB diagnosis for children and guide the rapid initiation of treatment.¹¹ 

To achieve this goal, we need to reassess and invest in each step of the diagnostic development pipeline to ensure they are serving the unique needs of children, including (1) Discovery; (2) Design; (3) Evaluation; and (4) Implementation.

There are key differences in how TB presents in children compared to adults.¹² Children are more likely to develop TB symptoms after exposure (primary disease), while adults are more likely to reactivate TB disease after previously controlling the infection. Children are also more likely than adults to have disease manifestations outside of the lung, including the lymph nodes and brain. These differences reflect that children have a developing immune system,¹³ which alters how they respond to TB as they get older.¹⁴ Consequently, it is not surprising that biomarkers and biosignatures that have been discovered and used for adults frequently do not translate well for children. For example, C-reactive protein (CRP) and the three-gene Sweeney biosignature are less accurate for TB triage in children than when used for adults.^15,16 On the pathogen marker side, the cell wall marker lipoarabinomannan (LAM) can be detected in urine, but it has low sensitivity in children given their paucibacillary disease.¹⁷ There are a number of biomarkers in plasma, urine and breath that have been found to distinguish TB disease in children,^18–21 but these need greater validation in larger, diverse, cohorts. We also need a better understanding of the fundamental pathogenesis and host response of TB in children to guide key pathways and candidate biomarkers.²²

The same concept also applies to novel artificial intelligence (AI) models and digital biomarkers. There are a number of groups developing algorithms for TB screening to automatically read chest x-rays (Computer-aided detection or CAD),²³ and analyze cough sounds,^24,25 but these predominantly use adult data. These algorithms will need to be fine-tuned or developed specifically with paediatric data to more accurately reflect both their normal physiology and their unique presentation of TB disease.