Low-complexity manual nucleic acid amplification tests for pulmonary tuberculosis in children

Background

Accurate and prompt diagnosis of tuberculosis in children is challenging due to non‐specific clinical presentation and the low bacillary load of samples. Low‐complexity manual nucleic acid amplification tests (LC‐mNAATs) such as loop‐mediated isothermal amplification (TB‐LAMP) are World Health Organization (WHO)‐recommended rapid molecular diagnostic tests. Even in resource‐limited settings, they have good diagnostic accuracy in adults.

Objectives

To determine the diagnostic accuracy of LC‐mNAATs for the detection of pulmonary tuberculosis in children (< 10 years) with presumptive pulmonary tuberculosis.

Secondary objectives

1. To compare the diagnostic accuracy of LC‐mNAATs and Xpert MTB/RIF Ultra for the detection of pulmonary tuberculosis in children with presumptive pulmonary tuberculosis.

2. To compare the diagnostic accuracy of LC‐mNAATs and smear microscopy for detecting pulmonary tuberculosis in children when TB‐LAMP is considered as a replacement test for smear microscopy.

3. To determine the diagnostic accuracy of LC‐mNAATs for the detection of pulmonary tuberculosis if used as an add‐on test amongst sputum smear‐negative children.

4. To investigate potential sources of heterogeneity in the diagnostic accuracy of LC‐mNAATs due to factors such as smear status, age, HIV status, setting, and tuberculosis burden.

Search methods

We searched CENTRAL, MEDLINE, Embase, Science Citation Index, Biosis Previews, Global Index Medicus, SCOPUS, WHO ICTRP, and ClinicalTrials.gov on 2 October 2023 for published articles and trials in progress without language or time limits. We screened the reference lists of included articles, conference abstracts, tuberculosis reviews, and guidelines. We searched ProQuest Dissertations & Theses A&I for dissertations. We approached the Stop TB Partnership, FIND, and other experts on tuberculosis for ongoing and unpublished studies. A WHO public call was made between 30 November 2023 and 15 February 2024 for ongoing and unpublished studies from manufacturers and researchers.

Selection criteria

We included cross‐sectional and cohort studies that evaluated LC‐mNAATs in children (< 10 years) against microbiological or composite reference standards. Our index test was TB‐LAMP, and comparator index tests were Xpert MTB/RIF Ultra and smear microscopy. The microbiological reference standard included automated liquid culture, solid culture, or a combination of both methods. We considered only design‐locked, marketed technologies.

Data collection and analysis

Four review authors, in pairs, independently screened titles and abstracts and assessed the full texts of potentially eligible articles. A fifth review author resolved any disagreements. We tailored and applied the QUADAS‐2 and QUADAS‐C tools to assess the risk of bias and applicability. Six review authors, in three pairs, extracted data and performed methodological quality assessment. A seventh review author resolved any disagreements. We contacted the primary study authors for missing data. We assessed the certainty of evidence using the GRADEpro GDT online tool.

Main results

We included four eligible studies (303 participants). Three studies took place in low‐ and middle‐income countries, with two studies from countries with a high tuberculosis burden. All four studies assessed different respiratory and non‐respiratory specimen types and evaluated TB‐LAMP against the microbiological reference standard.

We judged one study to have an unclear risk of bias in two domains of QUADAS‐2. The risk of bias was low for most of the studies. One study recruited inpatients from tertiary hospitals, causing high applicability concerns.

Three studies (67 children, including eight with pulmonary tuberculosis) evaluated respiratory samples (sputum, broncho‐alveolar lavage, and tracheal aspirate). The sensitivities were between 60% and 100%, and the specificities were between 95% and 100% (very low‐certainty (sensitivity) and low‐certainty (specificity) evidence). Three studies (176 participants, including 14 children with pulmonary tuberculosis) used gastric aspirate; the sensitivity was not estimable in two studies, and was 64% in the third study. The specificities were between 93% and 100%. The sensitivity was 100% (95% confidence interval (CI) 29 to 100), and the specificity was 96% (95% CI 88 to 100) in gastric lavage from one study. One study (144 participants, 12 children with pulmonary tuberculosis) assessed diagnostic accuracy using nasopharyngeal aspirate. The sensitivity was 58% (95% CI 28 to 85), and the specificity was 94% (95% CI 88 to 97). The same study (seven children with pulmonary tuberculosis) also evaluated stool specimens, and the sensitivity and specificity were 100% (95% CI 59 to 100) and 92% (95% CI 86 to 96), respectively. We did not perform a meta‐analysis due to limited data.

Interpretation of the results

Respiratory samples

For every 1000 children tested, if 100 had tuberculosis according to culture, 60 to 100 with tuberculosis would be identified as positive by the TB‐LAMP. Of the 900 children without tuberculosis, 855 to 900 would be identified as negative by the test.

Gastric aspirate

For every 1000 children tested, if 100 had tuberculosis according to culture, 64 with tuberculosis would be identified as positive by the TB‐LAMP. Of the 900 children without tuberculosis, 837 to 900 would be identified as negative by the test.

Gastric lavage

For every 1000 children tested, if 100 had tuberculosis according to culture, 135 would be TB‐LAMP positive, of which 100 would have tuberculosis (true positives), and 35 would not have tuberculosis (false positives); 865 would be TB‐LAMP negative, of which 864 would not have tuberculosis (true negatives), and one would have tuberculosis (false negatives).

Nasopharyngeal aspirate

For every 1000 children tested, if 100 had tuberculosis according to culture, 112 would be TB‐LAMP positives, of which 58 would have tuberculosis (true positives), and 54 would not have tuberculosis (false positives); 888 would test negative, of which 846 would not have tuberculosis (true negatives), and 42 would have tuberculosis (false negatives).

Stool

For every 1000 children tested, if 100 had tuberculosis according to culture, 171 would be TB‐LAMP positive, of which 99 would have tuberculosis (true positives), and 72 would not have tuberculosis (false positives); 829 would test negative, of which 828 would not have tuberculosis (true negatives) and one child would have tuberculosis (false negative).

Authors' conclusions

Evidence on the diagnostic accuracy of LC‐mNAATs for the detection of pulmonary tuberculosis in children is limited due to few and small studies. Adequately powered studies evaluating LC‐mNAATs in children are needed.