Advances in TB diagnostics: A critical element for the elimination toolkit

Currently available rapid molecular tests

The WHO first recommended an initial rapid molecular diagnostic test for TB in 2010. The Xpert MTB/RIF assay (Cepheid, USA) is a low-complexity, semi-automated, cartridge-based molecular test that is suitable for implementation in high-burden settings without needing containment level-3 molecular pathology laboratories⁷. This assay was modified to increase sensitivity (through additional targets) in 2017, and the Xpert MTB/RIF Ultra is now widely implemented with a global access price of US $8 per cartridge⁸. However, Xpert is not truly point-of-care, as it requires a reliable power supply, a favourable operating environment and frequent maintenance. The Truenat platform (Molbio, India) is a more recently WHO-recommended, low-complexity, chip-based real-time PCR assay⁹. Due to this platform being battery-operated, it is more suitable for point-of-care settings. Both Truenat and Xpert assays can also detect >90 per cent of mutations associated with rifampicin (RIF) resistance.

Despite the gradual rollout of these and other more centralized molecular tests for TB over the last decade, most (38% in 2021 and 47% in 2022) patients do not have access to these diagnostics¹. There is a real need for more accessible and truly point-of-care molecular assays with high sensitivities. Both near-patient and instrument-free point-of-care TB molecular tests are in development and will require parallel market interventions to ensure quick adoption to meet global End TB targets.

Alternative sampling and testing approaches: Moving away from sputum

Sputum is still the most relied on sample type for TB testing, but it is challenging to collect and process, and therefore, contributes to the large diagnostic gap for TB. Many people with presumed TB are unable to produce adequate sputum samples for testing, especially children, people living with HIV and those with the extra-pulmonary disease. Recent innovations in TB diagnostics have focussed on more accessible and useable, non-sputum-based sample types.

The lipoarabinomannan (LAM) antigen is a major cell wall constituent of MTB and was discovered in the urine of individuals with TB in the 1990s¹⁰. The Determine^TM TB LAM Ag test (Abbott, USA) lateral flow assay uses only 60 µL of urine and takes 20 minutes for TB detection. It was found to be useful for TB diagnosis in people living with HIV, and despite moderate overall sensitivity, it has been recommended for TB diagnosis by WHO in this population. Several developers are working on the next-generation urine antigen tests to improve sensitivity with higher affinity antibodies and different urine concentration methods. These tests offer promise to meet WHO TPP targets in all patient groups (including those not living with HIV), provided the variability in detection around the cut-off, as first described for SILVAMP TB LAM (Fujifilm, Japan), is resolved¹¹.

Oral swabs became a mainstay sample during the COVID-19 pandemic and have also emerged as a potential alternative sample type to sputum for TB. When coupled with rapid molecular tests such as Xpert Ultra and Truenat, the sensitivity of tongue swabs for TB can be as high as 70-75 per cent compared to sputum¹². Performance appears to be linked to swab type, collection, storage, and the pairing with the appropriate back-end molecular test (with tests that are able to detect free-DNA being more preferable), therefore, further optimization may well improve the sensitivity¹³. Furthermore, current studies compare swabs to sputum-based testing, whereas a major benefit is likely to be an increase in the yield of microbiological diagnoses of TB in those unable to produce adequate sputum samples, for example, in community-based TB screening programmes.

Breath is the ultimate non-invasive sample for TB diagnosis with a potential link to transmission in addition to the diagnostic use-case¹⁴. Several volatile organic compounds (VOC)-based tests assays from exhaled breath have been evaluated for TB, from electronic noses to Giant African Pouched Rats, with sensitivities ranging from 60-90 per cent¹⁵. The lack of specificity and variable performance hinder current assays for TB. Nevertheless, the promise of VOC detection from breath is shown by commercialized tests for SARS-CoV-2¹⁶. Direct detection of MTB in breath is also feasible. The respiratory aerosol sampling chamber (RASC) demonstrates the proof-of-principle with high specificity. However, the main challenge remains the low abundance of MTB in breath¹⁷. More realistic point-of-care solutions coupled with MTB detection by molecular assays are being explored, such as facemask sampling¹⁸. Other non-sputum sampling being evaluated includes molecular detection of TB from stool and blood sampling for host-response transcriptional signatures^19,20.

Digital health technologies and artificial intelligence (AI) also offer opportunities for diagnostic solutions for TB. Computer-aided detection (CAD) softwares are being used to automate reading and interpreting digital chest X-rays, although implementation challenges remain²¹. Other AI-based tools to classify lung and cough sounds are being explored for TB screening²².

Drug-resistant TB

The current diagnosis of DR-TB requires several steps and is challenging even in well-resourced settings. WHO recommends that all bacteriologically confirmed TB undergo drug susceptibility testing (DST) for commonly used anti-TB drugs²³. First-line molecular assays provide sensitive detection of RIF resistance, but testing for isoniazid and fluoroquinolone resistance is increasingly important. Traditional culture-based phenotypic DST are highly complex, labour-intensive, have long turnaround time and not accessible for most people at risk of DR-TB.

Increased understanding of the molecular level mechanisms of drug resistance over the past decade is leading to advances in the diagnosis of DR-TB. Several rapid molecular platforms now provide DST results for RIF and isoniazid, and two assays have been recommended by WHO for fluoroquinolone resistance (Xpert MTB/XDR [Cepheid, USA], and Genotype MTBDRsl [Bruker/HainLifescience, Germany])²⁴. However, more comprehensive rapid molecular diagnostics to accurately detect first- and second-line drug resistance as well as better point-of-care solutions are needed to close the diagnostic gap for DR-TB.

Next-generation DNA sequencing (NGS) has also facilitated the management of DR-TB. WHO has collated almost 200 drug-resistance mutations in its catalogue of mutations²⁵. However, current workflows require culture of MTB prior to NGS, and therefore its use has been constrained by high cost and turnaround time. Targeted NGS approaches straight from patients’ specimens hold the most promise, with research ongoing to amplify MTB genomes to improve sensitivity and the possibility of all processing and workflow on a single platform. Recent WHO guidelines recommend targeted NGS and are now supported by an operational handbook²⁶. One potential limitation is the confidence of resistance gene targets for newer DR-TB drugs for example, bedaquiline, pretomanid and linezolid.

Conclusions and the way forward

Ending TB requires accessible, affordable and accurate diagnostics to be coupled with better treatment strategies. Technical innovations present the possibility of a range of diagnostic solutions to allow this. We envisage a future where people with presumed TB can provide a range of samples (sputum, swabs, urine, blood) to be tested for TB with rapid diagnostics at the point of care, including the rapid detection of DR-TB to allow for early initiation of an effective TB treatment regimen. This will not only require technology innovation but also political will and considerable investment to translate promising innovations into implementable diagnostic interventions that reach the ‘missing millions’ of people with undiagnosed TB.