The appearance of drug-resistance has followed in the wake of extensive programmatic use of anti-tuberculosis (TB) drugs.1 The increasing problem of drug resistance poses different levels of diagnostic challenge: to identify the presence of drug resistance, define the extent of resistance present; and support the design of an effective anti-TB regimen.2

Defining the drug resistance profile and its severity. The appearance of strains of Mycobacterium tuberculosis resistant to both isoniazid and rifampicin, the two core anti-TB drugs, has defined multidrug-resistant tuberculosis (MDR-TB) from the outset. Later in 2006, when more resistant forms of TB appeared, the definition of extensively drug-resistant tuberculosis (XDR-TB) was agreed, referring to MDR-TB strains of M. tuberculosis with additional resistance to any fluoroquinolone and at least one of the three injectable anti-TB drugs (capreomycin, kanamycin and amikacin). The XDR-TB definition was made on the assumption that these classes of drugs were essential to successfully treat a patient with MDR-TB.3–5

More recent evidence demonstrated that XDR-TB patients had worse outcomes than MDR-TB patients with susceptibility to these drugs (those with resistance to fluoroquinolones faring worse than those resistant to injectables).6 The scientific community also considered codification of patterns of resistance beyond XDR7 and the possibility of defining so-called “total drug-resistance”, although no agreement was found on this.8,9

Although not formally recognized by the World Health Organization (WHO), clinicians widely adopted the term pre-XDR to indicate MDR-TB patients with additional resistance to either fluoroquinolones or injectables.9,10

In 2021 WHO introduced the definition pre-XDR-TB (e.g., MDR-TB plus resistance to fluoroquinolones) and modified the definition of XDR-TB, now being MDR-TB plus additional resistance to any fluoroquinolone and at least one WHO Group A drug (bedaquiline, linezolid).5,11,12

Diagnosing and treating pre- and XDR-TB. Considering the difficulty of treating XDR-TB, typically with costly and often toxic second-line anti-TB drugs, the importance of achieving rapid and effective diagnosis to design an effective regimen is obvious.

Of the 4.8 million people diagnosed with pulmonary TB worldwide in 2020, only 59% were bacteriologically confirmed and 71% of them were tested for rifampicin resistance. Notably, a WHO-recommended rapid molecular test was used as the initial diagnostic test for only 33% of new TB diagnoses, contributing to reasons why only about one third of the 157,903 MDR/rifampicin-resistant TB cases detected in 2020 had access to treatment.12

In a recent rapid communication,13,14 WHO recommended the use of shorter regimens on a programmatic basis. The pivotal initial tests which are used globally to classify patients as drug-susceptible or drug-resistant are the Xpert MTB/RIF and Xpert MTB/RIF Ultra, which allow detection of resistance to rifampicin only, and other new low-complexity molecular assays (Truenat MTB-RIF Dx) or moderate complexity nucleic acid amplification tests (NAATs) systems (Abbott Molecular RealTime MTB RIF/INH; Becton Dickinson BD MAX MDR-TB; Bruker-Hain Diagnostics FluoroType MTBDR; Roche Diagnostics Cobas MTB-RIF/INH) have been recently conditionally recommended as well.15 As the shorter MDR/RR-TB regimens are based on WHO Group A drugs, and no rapid WHO-approved tests are available for bedaquiline and linezolid, it is crucial that at least resistance to fluoroquinolones is rapidly detected, together with resistance to isoniazid.16 In case longer regimens are used, rapid detection of resistance to other drugs could be important (e.g., to injectables and ethionamide). Xpert MTB/XDR (isoniazid, ethionamide, fluoroquinolones, amikacin, capreomycin and kanamycin), first-line Line Probe Assays (LPAs; rifampicin and isoniazid) and second-line LPAs (fluoroquinolones, amikacin, capreomycin and kanamycin), plus other moderate complexity automated NAATs and high complexity reverse hybridization-based NAATs (Nipro Genoscholar PZA-TB) are the follow on rapid diagnostic tests currently recommended for the detection of additional drug-resistance beyond rifampicin.15,17

Despite advances, treatment success of MDR-TB remains low (globally at 59%); being lower for pre-XDR and XDR-TB patients and higher in settings where all oral regimens including new drugs are used.12,18,19 Improving treatment success will require access to rapid diagnostic tests able to reliably detect and classify drug resistance, and inform clinicians designing individual regimens.20

In 2021–2022 a Cochrane Review21,22 investigated the efficacy of Xpert MTB/XDR to rapidly detect resistance to isoniazid, fluoroquinolones, ethionamide, and amikacin (but not to the other injectable drugs kanamycin and capreomycin). However, no previous systematic reviews have comprehensively investigated the features of tests available to diagnose pre-XDR/XDR-TB. The aim of this systematic review is to assess all the tools able to rapidly diagnose pre-XDR/XDR-TB (including LPAs, and the Xpert XDR assay) under a point-of-care perspective and describe their technical characteristics (i.e., sensitivity, specificity, positive and negative predictive values).

The aim of our study was to assess all the tools able to rapidly diagnose pre-XDR/XDR-TB (and describe their technical characteristics), as well as to compare the diagnostic techniques and approaches in terms of sensitivity, specificity, laboratory complexity, time to confirmed diagnosis (Table 3).

The results of our study demonstrated that two rapid WHO-endorsed tests are available for these purposes, namely the Xpert MTB/XDR and GenoType MTBDRsl (VER1.0 and VER 2.0).

Xpert MTB/XDR test is a rapid, automated NAAT test at low complexity.21,22 It does not require any sophisticated laboratory infrastructure. The molecular amplification of small quantity of genetic materials takes place inside the cartridge and require less than 2h to provide results. The main advantages over the LPAs are the higher sensitivity for detecting Mycobacterium tuberculosis (it is also recommended in persons with a sputum smear-negative specimen), faster time to response (90min vs. 48–72h), simple handling, reduced number of invalid results for direct testing, its potential implementation outside a sophisticated laboratory, and its safety, as the cartridge represents ‘the laboratory’ where molecular amplification takes place. The Xpert MTB/XDR test can be done taking advantage of existing GeneXpert platforms equipped with 10-colour modules for Xpert MTB/RIF Ultra testing, or through system upgrade from 6-colour modules, with technical capacity to run the Xpert technology available yet on very large scale as infrastructure and procedures are equal. As technology upgrade may be logistically challenging and somehow expensive, different solutions have been conceived, including procurement of new 10-colour GeneXpert modules, new full systems, satellites (instrument only) to connect to existing GeneXpert system or converting an existing system with 6 colours modules to 10 colour modules. In terms of performance of the tests, they are very similar: in our study, it seems to slightly favour Xpert, although only 2 studies were available (sensitivity 91.4–94; specificity 98.5–99; accuracy 97.2–97.7; PPV 88.9–99.1; NPV 95.8–98.9).

The Xpert MTB/XDR is intended as a follow-on test after Xpert MTB/RIF or Xpert MTB/RIF Ultra testing, using left over sample reagent treated specimen, thus not necessarily requesting a second sample or culture to be collected. This approach is attractive with benefits for access to rapid DST, diagnostic laboratory workflow and turnaround time, or after other WHO-endorsed bacteriological tests confirming the presence of pulmonary TB. However, the performance of any rapid assay used as reflex test for a specimen previously determined to be M. tuberculosis positive and harbouring other resistance (e.g., rifampicin) will need to be carefully assessed in future studies, for if the initial rapid test is more sensitive than its reflex assay it may lead to invalid or non-actionable results and challenges with final interpretation of laboratory results.

A second set of considerations relates to the capacity of these tests to reliably diagnose XDR-TB, as newly defined by WHO in 2021.11 Xpert MTB/XDR detects M. tuberculosis complex DNA and mutations associated with resistance to isoniazid, fluoroquinolones (ofloxacin, moxifloxacin, levofloxacin, gatifloxacin), second-line injectable drugs (amikacin, kanamycin, capreomycin), and ethionamide in a single test, whereas GenoType MTBDRplus detects resistance to isoniazid and rifampicin, and GenoType MTBDRsl to fluoroquinolones (ofloxacin and moxifloxacin) and aminoglycosides/cyclic peptides (injectable antibiotics such as kanamycin, amikacin, capreomycin, and viomycin) in its 2 versions. Unfortunately, these tests are testing the drugs defining XDR-TB according to the previous 2006 WHO definition,64 which now reflects pre-XDR due to their capacity to detect resistance to fluoroquinolones, and not XDR-TB as neither these tests nor other genotypic DST assays can rapidly diagnose resistance to linezolid and bedaquiline. This is an important gap that will become more important with time; currently bedaquiline resistance is being detected in over 40% of cases failing the all oral regimen in South Africa, prevalence of bedaquiline resistance is currently 4% in South Africa and 15% in Moldova.65,66 Higher MICs to pretomanid have been detected in Lineage 1 strains.67 A 3 drug BPaL (bedaquiline, pretomanid, linezolid) regimen for pre-XDR with no rapid resistance test beforehand may therefore not be without risk and a shorter duration of lower dose linezolid may select for resistance to linezolid and thwart future salvage regimens. Targeted Next Generation Sequencing solutions are available but there is insufficient understanding of the role of resistance-conferring mutations to these drugs and such technologies are expensive and not implemented yet in many settings.68,69 Nonetheless, the two rapid assays provide valuable information at least on resistance to isoniazid and fluoroquinolones with now also the Xpert MTB/XDR technology and not only LPAs enabling to distinguish low- versus high-level resistance to these drugs.28 The detection of specific mutations is not implemented in the GeneXpert software automated interpretation. This information is nevertheless very relevant for clinicians who can rapidly design the best regimen with high precision guided by the knowledge of the association between specific genetic mutations and phenotypic resistance levels.70 For example, in the case where gyrA A90V, S91P, and D94A mutations associated with low-level fluoroquinolone resistance are identified by specific melting temperature window patterns, high dose moxifloxacin could still work. The detection of resistance to ethionamide and second-line injectables seems currently less clinically relevant, although these medicines can still be added to complete MDR regimens when other options are not available.

Considering the TB diagnostic algorithms and diagnostic characteristics, Xpert MTB/XDR could be placed at or near point-of-care settings to be used primarily as a follow-on test for laboratory-confirmed TB, complementing existing rapid tests that detect only rifampicin, or rifampicin/isoniazid resistance. Detection of fluoroquinolone resistance is essential to inform the use of all-oral 6–12 month standardized shorter regimen for MDR/rifampicin-resistant TB. The assay is conditionally recommended for the initial detection of resistance to isoniazid and fluoroquinolones, rather than culture-based phenotypic DST in people with bacteriologically confirmed pulmonary TB, and for the initial detection of resistance to these drugs and ethionamide and amikacin in people with bacteriologically confirmed pulmonary TB and resistance to rifampicin.17 Xpert MTB/RIF or Xpert Ultra and Xpert MTB/XDR could also be used as the initial diagnostic tests to detect TB, rifampicin, isoniazid and fluoroquinolone resistance in order to achieve universal DST; this algorithm may be preferable in settings with a high MDR-TB burden and also in those with high risk of isoniazid resistance (algorithm 117). If not feasible, it may be used for further evaluation of patients with rifampicin-resistant or MDR-TB, for patients with rifampicin-susceptible TB but no results available for isoniazid but at a high risk for isoniazid resistance (algorithms 3–4).

Our review is limited by high heterogeneity of the population targets and limited range of laboratory tools available. We also note limitations in the study designs of published work identified, including little clinical outcome data following treatment based on test findings.

Finally, rapid diagnostic tests are not only helpful in existing algorithms for guiding drug selection but may also guide dose selection (higher dose) in case of low-level resistance when limited alternative treatment options exist.20 Further testing options when a patient is demonstrating slower response to TB treatment than expected, despite proven drug susceptibility would be valuable, such as supplementary rapid tests for drug exposure. We foresee that the introduction of saliva- and urine-based point-of-care testing as well as microsampling techniques could be provided in conjunction with rapid DST as they are becoming more affordable for resource-limited settings, and may support laboratory testing efforts towards better outcomes for people affected by TB around the world.

The systematic review was partially funded via an unrestricted grant by Cepheid Europe SAS to the Public Health Consulting Group, Lugano, Switzerland. The donor had no role in conducting the systematic review, as well as analysing and interpreting the results and writing the manuscript.

The authors declare to have no conflict of interest directly or indirectly related to the manuscript contents.