Key messages

• Low-complexity rapid molecular tests can help identify people with extrapulmonary tuberculosis and rifampicin resistance.

• These tests can accurately identify tuberculosis in cerebrospinal fluid, pleural fluid, pleural tissue, synovial, peritoneal and pericardial fluid.

Why is using low-complexity rapid molecular tests for extrapulmonary tuberculosis important?

Tuberculosis is one of the top 10 causes of death worldwide. Tuberculosis mainly affects the lungs but can also occur elsewhere in the body (extrapulmonary). Quick and accurate tests help people start treatment sooner, which saves lives. Low-complexity automated nucleic acid amplification tests (LC-aNAATs) are rapid tests that give results in about two hours, unlike culture that takes weeks. They can also detect resistance to important antibiotics for tuberculosis, like rifampicin.

What is the aim of this review?

To update evidence on how well LC-aNAATs detect extrapulmonary tuberculosis and rifampicin resistance in adults and adolescents.

What did we do?

LC-aNAATs are World Health Organization-recommended rapid molecular diagnostic tests for diagnosing tuberculosis and rifampicin resistance. We combined study results to find out:

• sensitivity for tuberculosis detection: proportion of people with tuberculosis correctly diagnosed as having tuberculosis;

• specificity for tuberculosis detection: proportion of people without tuberculosis correctly identified as not having tuberculosis;

• sensitivity for rifampicin resistance detection: proportion of people with rifampicin resistance correctly diagnosed as being rifampicin resistant;

• specificity for rifampicin resistance detection: proportion of people with rifampicin susceptibility correctly identified as being rifampicin susceptible.

We assessed LC-aNAAT results against a microbiological and a composite reference standard (neither is a perfect reference standard because extrapulmonary tuberculosis has fewer bacteria).

What are the main results of this review?

Thirty-seven studies tested lymph node, pleural, and cerebrospinal fluid, and other specimens from people presumed to have extrapulmonary tuberculosis. We found data for 2 LC-aNAATs (Xpert Ultra and Truenat MTB Plus), but could only pool the data for Xpert Ultra to produce summary estimates which are presented below.

For every 1000 people tested, if 100 had tuberculosis:

cerebrospinal fluid (16 studies)

The sensitivity was 88% with a specificity of 96% against a microbiological reference standard. This means that 124 people would test positive in total, of which 36 would be without tuberculosis (false positive); also, 876 people would test negative in total, of which 12 would have tuberculosis (false negative).

pleural fluid (13 studies)

The sensitivity was 74% with a specificity of 88% against a microbiological reference standard. This means that 181 people would test positive in total, of which 107 would be without tuberculosis (false positive); also, 819 people would test negative in total, of which 26 would have tuberculosis (false negative).

lymph node aspirate (6 studies)

The sensitivity was 71% with a specificity of 97% against a composite reference standard. This means that 94 people would test positive in total, of which 23 would be without tuberculosis (false positive); also, 906 people would test negative in total, of which 29 would have tuberculosis (false positive).

rifampicin resistance (13 studies)

The sensitivity was 100% with a specificity of 99% against a microbiological reference standard. This means 105 people would test positive in total for resistance, of which 5 would be without resistance (false positive); also 895 people would test negative for resistance in total.

Who do the results of this review apply to?

People with presumed extrapulmonary tuberculosis.

How confident are we in our results?

We are fairly confident about LC-aNAAT in cerebrospinal fluid and less confident about lymph node aspirate and pleural fluid for Xpert Ultra because our question was to understand how these tests would perform in routine settings. However, most studies for lymph node aspirate and pleural fluid did not always report the settings. Their results also differed a lot between studies and some of the studies were very small. We are less confident about Truenat MTB plus, as there were few studies and few people tested. Both reference standards are imperfect, which may affect accuracy estimates.

What are the implications of this review?

LC-aNAATs may be helpful in diagnosing extrapulmonary tuberculosis, though sensitivity varies across different extrapulmonary specimens. While for most specimens, specificity is high, the test rarely yields a positive result for people without tuberculosis. LC-aNAATs had high sensitivity for tuberculous meningitis and high sensitivity and specificity for rifampicin resistance. However, there is a need for more data on other tests within the same technology class.

What are the limitations of the evidence?

As culture is not a perfect reference standard for this form of tuberculosis, multiple cultures per specimen could help strengthen the reference standard. However, not all studies reported the number of cultures per specimen, which is a limitation in this review.

How up-to-date is this review?

10 October 2023. A World Health Organization public call for data was made between 30 November 2023 and 15 February 2024 to identify unpublished studies.