Effective and efficient TB diagnostic algorithms are key components of a diagnostic cascade designed to ensure that patients with TB are diagnosed accurately and rapidly, and are promptly placed on appropriate therapy. In turn, that therapy should improve patient outcomes, reduce transmission and avoid development of drug resistance. This section presents a set of four model algorithms that incorporate the goals of the End TB Strategy and the most recent WHO recommendations for the diagnosis and treatment of TB and DR-TB. The algorithms, which emphasize the use of WRDs, are illustrative and countries must adapt them to their local situation.
When selecting a diagnostic algorithm to implement, it is important to consider the characteristics of the population being served. Thus, the four model algorithms are as follows:
- Algorithm 1 relies on mWRDs as the initial diagnostic tests and is appropriate for all settings, although the choice of which mWRD to use may differ in a setting with high MDR/RR-TB prevalence (e.g. a test that detects MTBC and RIF with or without INH resistance may be needed), with high HIV prevalence (e.g. a more sensitive test may be needed) or with high Hr-TB prevalence (where a test that detects MTBC, RIF and INH resistance simultaneously will be needed).
- Algorithm 2 incorporates the most recent WHO recommendations for the use of the LF-LAM as an aid in the diagnosis of TB in PLHIV and is most relevant to settings with a high HIV prevalence. However, Algorithm 2 is applicable to any PLHIV who meet the testing criteria, regardless of the underlying prevalence of HIV in that setting.
- Algorithm 3 and Algorithm 4 are for follow-up testing, after TB is diagnosed, to detect additional drug resistance:
- Algorithm 3 is used when the purpose is to detect resistance to second-line drugs in patients with RIF resistance; and
- Algorithm 4 is used when the purpose is to detect resistance to INH in patients at risk of Hr-TB and with RIF susceptibility, or when Hr-TB (RIF-susceptible and INH-resistant TB) is detected and follow-on testing is needed (e.g. FQ resistance detection).
Algorithms 3 and 4 are appropriate for all settings; however, the resource requirements for follow-up testing may differ strongly between settings with a high burden of DR-TB and settings with a low burden of DR-TB.
Each algorithm is accompanied by explanatory notes and is followed by a decision pathway that provides a detailed description of the various decisions in the algorithm. In general, the algorithms do not show a specific test; rather, they provide a flow of the expected outcomes of the test that are common across the algorithms (e.g. TB detected) and the follow-up action required.
Although the algorithms are presented separately, they are interlinked and cascade from one to the other. This is illustrated in the overview (Fig. 4.1), which also lists the various diagnostic tests that are currently recommended within each algorithm. Algorithm 2 is complementary to Algorithm 1, and currently it includes only a single product option that is intended as an additional test for PLHIV.
The diagnostic pathway begins with a person being screened positive for TB. WHO has released updated recommendations on TB screening, and readers are encouraged to consult the latest guidance (39). However, presumptive TB patients may not always present with symptoms that match the latest screening guideline recommendations but still have an increased probability for TB disease requiring diagnostic testing. The modalities for screening, beyond the four-symptom screen, now include chest X-ray, an mWRD used as a screening tool or C-reactive protein in PLHIV. The addition of mWRD for screening of selected at-risk populations and settings goes beyond its primary use as an initial line diagnostic tool and the different uses should not be confused. However, priority should be given to ensuring universal access to mWRDs as a diagnostic test for TB and DR-TB before extending its use to screening. Furthermore, the use of mWRDs as a screening test would have large financial and operational implications, and should be carefully considered.
People who are referred for diagnostic evaluation after a positive screen for TB using modalities other than an mWRD should go through a clinical evaluation, including other relevant investigations if available (e.g. chest X-ray) and follow Algorithm 1 to reach a bacteriologically confirmed diagnosis. Among those who are referred for diagnostic evaluation after a positive screen for TB using an mWRD,⁵⁶ i.e. used in a screening rather than a diagnostic context, the pretest probability is an important consideration in addition to the clinical picture when deciding to repeat the mWRD or proceed with treatment. Due to the high specificity even when used as a screening test (99%), a positive test is likely to be a true positive when the pretest probability is high.
In the scenario recommending mWRD for screening all PLHIV who are medical inpatients and where the prevalence of TB is at least 10%, the likelihood of the mWRD being a true positive is high and treatment should be considered if the clinical picture is in keeping with a diagnosis of TB. However, the detection of TB DNA does not necessarily indicate that the person has active TB. This may occur in a patient with a history of prior TB treatment (<5 years) and is particularly true for the more sensitive molecular tests (Xpert Ultra); culture would aid in interpretation. Furthermore, in PLHIV who require hospital admission, more than one infection may be present, and patients should be fully investigated.
In the scenario where an mWRD is used as a screening tool in a community with a prevalence of 0.5%, the positive predictive value would only be 39.5%, despite having a specificity of 99%, meaning that more than half of the positive mWRDs may be false positive. In such a situation, a repeat mWRD is warranted but should follow a clinical evaluation, and other investigations should be considered. Treatment should be based on the totality of evidence for the patient.
Algorithm 1 is the starting point for the diagnostic pathway for most patients. The number of tests recommended for this purpose has increased from five to nine with the latest addition of the moderate complexity automated NAAT class. This is a great improvement compared with the past, when smear microscopy was the only option. Member States can now make choices that best fit their circumstances, with the ultimate objective being to serve patient needs. The initial test options can be split into those that provide a TB diagnosis only (TB-LAMP) and those that also provide at least detection of RIF resistance (simultaneously or as a two-step process): Xpert MTB/RIF, Xpert Ultra, Truenat MTB, Truenat MTB Plus, Truenat MTB-RIF Dx and moderate complexity automated NAATs. RIF resistance detection is recommended as part of the targets to achieve universal DST. Among the tests that detect RIF resistance, a subgroup provides this as a two-step procedure, separating TB detection and detection of RIF resistance. This second step could be used as a follow-on test to initial tests that do not offer RIF resistance detection.
The test options for Algorithm 1 vary in complexity. The Xpert MTB/RIF, Xpert Ultra, Truenat MTB and Truenat MTB Plus all require basic pipetting skills and are easy to decentralize but have limited throughput with the commonly used instruments. In contrast, although some of the moderate complexity automated NAATs have minimal hands-on time, they have large infrastructure requirements; also, most of these tests provide higher throughput and are suited to established laboratories with reliable sample referral networks, and they detect resistance to INH in addition to RIF. In practice, test needs and associated choices are likely to vary, depending on the setting within a country or province. Consideration should be given towards hybrid models using a combination of tests from different manufacturers; this has the added advantage of providing a safety mechanism in the event of an expected problem with a supplier.
Algorithm 2 may be the starting point for some patients, primarily for PLHIV who would benefit from a rapid point-of-care test to diagnose TB. It is recommended that testing using Algorithm 1 and Algorithm 2 are done in parallel.
Algorithms 3 and 4 follow on from Algorithm 1, with the split based on the RIF result. Algorithm 3 is for those with confirmed RR-TB and is aimed at providing additional rapid DST. The options are the FL-LPA and SL-LPA, and the recently recommended low complexity automated NAATs with both test groups providing similar DST results. The major difference is the complexity of the test procedures. The low complexity automated NAATs are easier to perform, can be used irrespective of smear grade and may be better suited for decentralization. Rapid early diagnosis of FQ resistance among MDR/RR-TB patients is important, and the low complexity automated NAATs would be the preferred option to scale up and increase access. A new molecular test for PZA is recommended and it belongs to the class of high complexity reverse hybridization NAATs. Currently, the recommendation is for use on isolates only, with testing at higher levels of the laboratory network for selected patients. Here again an interlinked laboratory network offering a range of tests may best cater for patient needs.
Algorithm 4 is a follow-on algorithm for those with RIF-susceptible TB, and is aimed at detecting INH resistance (where INH testing was not included in Algorithm 1) or detecting FQ resistance when Hr-TB has been identified by moderate complexity automated NAATs in Algorithm 1. Currently, INH detection is not widely implemented at initial diagnosis, but this will change with the adoption of moderate complexity automated NAATs as initial tests that also detect INH resistance. The adoption of the low complexity automated NAAT for follow-on testing of RIF-susceptible TB could also increase the detection of INH resistance. The estimated incidence of Hr-TB is greater than that of RR-TB, and both require modification of the treatment regimen. Hr-TB patients would need an FQ added to the therapy. For patients diagnosed through a moderate complexity automated NAAT that detects both RIF and INH resistance in Algorithm 1, there are two options for FQ resistance detection: low complexity automated NAAT, which is less complex and can be decentralized; or SL-LPA which, although more complex, may already be available in the same laboratory as the moderate complexity automated NAAT. In scenarios where diagnostic tests for only RIF resistance are available (e.g. Xpert MTB/RIF and Truenat MTB), follow-on testing in settings or subpopulations with high Hr-TB burden using the low complexity automated NAAT will be appropriate, being simpler and cheaper than performing both the FL-LPA and SL-LPA.
⁵⁶ See https://www.who.int/publications/i/item/9789240022676
FL-LPA: line-probe assay for first-line drugs; LF-LAM: lateral flow lipoarabinomannan assay; NAAT: nucleic acid amplification test; RIF: rifampicin; SL-LPA: line-probe assay for second-line drugs; TB: tuberculosis.
ᵃ Text with grey background: currently recommended tests, text with orange background: newly recommended tests. Numbers on grey background refer to the model algorithms.