OpiniãoGiving TPT to someone who has TB disease can delay resolution of disease and favour the emergence of drug resistance. Excluding TB disease before initiating preventive treatment is one of the critical steps in the TBI care pathway. This section proposes approaches for ruling out TB disease and diagnosing TBI in people at risk of TB according to HIV status, symptoms, household contact, other risk factors, age, TBI test results and abnormality on CXR (Fig.1). The evidence and the recommendations for these steps are briefly discussed, as are tools for TB screening, first recommended in 2021 (13).
People with HIV
- Infants and children living with HIV who have poor weight gain,4 fever or current cough or who have a history of contact with a person with TB should be evaluated for TB and other diseases that cause such symptoms. If TB disease is excluded after an appropriate clinical evaluation or according to national guidelines, these children should be offered TB preventive treatment, regardless of their age. (Strong recommendation, low certainty of the estimates of effect)
- Adults and adolescents living with HIV should be screened for TB according to a clinical algorithm. Those who do not report any of the symptoms of current cough, fever, weight loss or night sweats are unlikely to have TB disease. Those who report any of these symptoms may have TB, should be evaluated for TB and other diseases and should be offered TB preventive treatment if TB disease is excluded, regardless of their antiretroviral treatment status. (Strong recommendation, moderate certainty of the estimates of effect)
- Among adults and adolescents living with HIV, chest X-ray may be used to screen for TB disease. (Conditional recommendation, low certainty of the estimates of effect)
- Among adults and adolescents living with HIV, C-reactive protein at a cut-off of > 5 mg/L may be used to screen for TB disease (Conditional recommendation, low certainty in the estimates of effect)
- Among adults and adolescents living with HIV, molecular WHO-recommended rapid diagnostic tests may be used to screen for TB disease. (Conditional recommendation, moderate certainty in the estimates of effect)
Justification and evidence
A systematic review of studies on infants and children, conducted for the 2011 guidelines provided limited evidence on the best approach to screening (26). Using these few studies and expert opinion, the previous GDG recommended a screening rule of poor weight gain, fever, current cough and a history of contact with a person with TB (recommendation 9). A systematic review was undertaken for the 2018 update to assess the performance of the screening rule; however, the only publication found was of a study of 168 children aged ≤ 12 years hospitalized with HIV in Kenya (51). In this study, the sensitivity was 100% (95% CI, 94 ; 100.0) and the specificity was 5% (95% CI, 1 ; 11). The systematic review conducted for the 2021 TB screening guidelines comprised two studies conducted in outpatient settings, with a total of 20 926 participants (13). In this review, the combined symptom screen (in which the presence of any symptom constituted a positive screen) had a pooled sensitivity of 61% (95% CI 58% ; 64%) and a pooled specificity of 94% (95% CI 86% ; 98%) (Table 3). Despite the lack of high-certainty evidence, the GDG considered that a strong recommendation for symptom screening was warranted for children < 10 years who were living with HIV, given the high risk of disease and of mortality when the diagnosis is missed and TB is left untreated.
Infants and children with HIV should be screened for TB as part of standard, routine clinical care, regardless of whether they are receiving TPT or ART. Symptom-based screening is generally acceptable to caregivers and people and is feasible even in resource-limited settings. Therefore, the GDG decided to make a strong recommendation for use of symptom screening in children with HIV. TB disease should be ruled out for those who have one or more symptoms. The GDG also noted that clinicians should broaden the differential diagnosis to include other diseases that may cause current cough, fever and poor weight gain in children with HIV. If the evaluation shows no signs of TB disease and the clinician decides not to treat for TB disease, children with HIV should be offered TPT, regardless of their age. Infants < 12 months of age should, however, be given TPT only if they have a history of household contact with a person with TB and TB disease has been excluded according to national guidelines. Guidance on further testing for TB in people with HIV who have suggestive clinical features is available elsewhere (44).
The text of recommendation 10 is a combination of two related recommendations in the 2015 guidelines that were updated in 2018 (17,37). In 2011, WHO conducted a systematic review and a meta-analysis of data for individual patients and recommended a symptom-screening rule of a combination of current cough, weight loss, night sweats and fever to exclude TB disease in adults and adolescents (52). The review showed that the rule had a sensitivity of 79%, a specificity of 50% and a negative predictive value of 97.7% at a TB prevalence of 5%. Most of the people with HIV in the studies included in the systematic review were not receiving ART.
During the 2018 updating of the guidelines, a systematic review was undertaken to compare the performance of the four-symptom screen in people with HIV who were and were not receiving ART (see PICOs 2 and 3 in Annex 3 and Table 2 in (53)). Data from 17 studies were used in the analysis. The pooled sensitivity of the four-symptom screen for people with HIV on ART was 51.0% (95% CI 28.4 ; 73.2), and the specificity was 70.7% (95% CI 47.7 ; 86.4); in people with HIV who were not receiving ART, the pooled sensitivity was 89.3% (95% CI 82.6 ; 93.6), and the specificity was 27.2% (95% CI 17.3 ; 40.0). In two studies on addition of abnormal CXR findings to the screening rule for people with HIV on ART (54,55), the pooled sensitivity was higher (84.6%, 95% CI 69.7 ; 92.9), but the specificity was lower (29.8%, 95% CI 26.3 ; 33.6) than for the symptom screen alone. In all studies, the median prevalence of TB among people with HIV on ART was 1.5% (interquartile range, 0.6–3.5%). At a 1% prevalence of TB, the negative predictive value of the symptom screening rule was 99.3%; addition of abnormal CXR findings increased the negative predictive value by 0.2%. No studies of the addition of CXR to the symptom rule for pregnant women were found. The GDG agreed that, in adults and adolescents with HIV, the four-symptom screen (current cough, fever, weight loss or night sweats) is useful for ruling out TB disease, regardless of ART use, although confirmation of TBI with IGRA, TST or TBST would be desirable before starting TPT. It noted the potential benefits of adding normal CXR findings to the rule, while recognizing that the improvement in performance was marginal. Moreover, increased use of CXR would add more false-positive results to the screening rule, which would require more investigations for TB and other illnesses. Therefore, the GDG reiterated that CXR may be added as an additional investigation only if it does not pose a barrier to the provision of preventive treatment for people with HIV. It should not be a requirement for initiating TPT. Although no study was found of the effect of adding CXR in testing pregnant women, the GDG noted that pregnant women with HIV could also benefit, as long as good practices are observed to prevent harmful exposure of the fetus to radiation (56).
In 2020, a systematic literature review and meta-analysis of individual data on patients were conducted to assess further the accuracy of the WHO-recommended four-symptom screen (W4SS) of people with HIV and of important subgroups and to identify other screening tools and strategies to increase detection of TB in people with HIV (13). The screening tools and strategies reviewed by the GDG included CRP, CXR and mWRD, as both stand-alone tests and in combination with the W4SS. Culture was the reference standard for assessing the accuracy of the screening strategies (Table 3). The meta analysis of data on individual patients comprised 23 studies of 16 269 participants with HIV, in which the accuracy of the W4SS was reviewed. Most of the studies addressed pulmonary TB disease.
The W4SS has suboptimal accuracy for some subgroups of people with HIV. The specificity is low, 37–38%, among all people with HIV and even lower among people newly enrolled or not on ART. Therefore, people who do not have TB disease are frequently screened as positive and are referred unnecessarily for diagnostic evaluation. This reduces the efficiency of screening programmes (e.g. with higher costs for diagnostic testing) and slows initiation of TPT. The sensitivity of W4SS is also low (53%) among people with HIV on ART; thus, almost half of prevalent TB cases are not identified in routine symptom screening alone. In a setting in which the prevalence of TB is 1%, 742 of 1000 outpatients screened with the W4SS and CRP would be true negatives and eligible for TPT, while only 416 would be found to be eligible with the W4SS alone. Restricted access to CRP or CXR should not be a barrier to initiating TPT.
Table 3. Diagnostic accuracy of screening tests in people with HIV
Recommendation 11 on use of CXR for people with HIV was first made in 2018 to update the position in the 2011 guidelines (26). Since 2021, WHO has also conditionally recommended use of CAD software programmes to interpret digital CXRs for pulmonary TB during screening and triage of people aged ≥ 15 years, regardless of HIV status (13).
Use of CXR for screening in parallel with symptom screening improves the sensitivity over that with the W4SS alone in all subgroups of people with HIV. In particular, screening with CXR significantly improves the sensitivity in people with HIV who are on ART and is the most sensitive screening strategy for this group. When available, CXR is recommended for use in parallel with the W4SS to rule out TB disease before initiating TPT in people with HIV who are on ART (13). The evidence on the performance of CXR and the W4SS for all people with HIV reviewed before making the 2021 TB screening guidelines was from eight studies, conducted in Benin, Botswana, Brazil, Guinea, India, Kenya, Malawi, Myanmar, Peru, South Africa and Zimbabwe, with a total of 6238 participants (Table 3). CXR alone was found to have similar sensitivity to and similar or higher specificity than the W4SS in all subpopulations. When CXR was conducted after a positive W4SS, CXR was less or similarly sensitive and more or similarly specific. When CXR was used in parallel with the W4SS, the sensitivity was higher or similar and the specificity was similar.
Recommendation 12 relates to the use of CRP for screening adults and adolescents with HIV for TB disease. CRP is an indicator of general inflammation that can be measured with point-of-care tests in capillary blood collected by finger prick. The evidence reviewed comprised six studies conducted in Kenya, South Africa and Uganda with a total of 3971 participants (13). The accuracy of CRP based on cut-off values of > 5 mg/L and > 10 mg/L as indicators of TB disease was reviewed, and the accuracy of two was considered to be similar or superior to that of the W4SS. The cut-off value of > 5 mg/L was recommended, as it is the lowest threshold for abnormality in many clinical settings and is more sensitive than the value of > 10 mg/L. The meta-analysis of data on individual patients on CRP with a cut-off of > 5 mg/L reported similar sensitivity and higher or similar specificity to the W4SS in all the subpopulations assessed (Table 3). When CRP was combined with the W4SS and used in parallel, it had similar or greater sensitivity and specificity than the W4SS alone in all populations, depending on the cut-off threshold used and the subpopulation assessed, while a positive screen with either tool led to a diagnostic test. CRP was found to be most accurate for outpatients who were not on ART as compared with the W4SS alone, which had a sensitivity of 84% (95% CI 75% ; 90%) and a specificity of 37% (95% CI 25% ; 50%) in this subpopulation. When performed sequentially after a positive W4SS in people with HIV who were not on ART, CRP with a cut-off of > 5 mg/L was as sensitive (84%; 95% CI 73% ; 90%) as the W4SS alone but was significantly more specific (64%; 95% CI 55 ; 72%). Like the W4SS, the specificity of CRP for TB screening in inpatients with HIV was extremely low, probably due to competing comorbidities that would also result in raised CRP levels and symptoms.
Recommendation 13 relates to use of mWRD for screening adults and adolescents with HIV for TB disease. A systematic review of the performance of mWRD in screening for TB among people with HIV comprised 14 studies with a total of 9209 participants. The Xpert MTB/RIF assay was the mWRD used in most of the studies. Use of an mWRD alone had a sensitivity of 69% (95% CI 60% ; 76%) and a specificity of 98% (95% CI 97% ; 99%) as compared with use of the W4SS followed by an mWRD, which had sensitivity of 62% (95% CI 56% ; 69%) and a specificity of 99% (95% CI 97% ; 99%) (Table 3). The accuracy of the mWRD was not significantly different from that of the W4SS followed by the mWRD in various subpopulations.
Household contacts of a person with TB and other risk groups
Infants and children < 5 years of age⁵
Justification and evidence
Symptom-based screening has been reported to be a safe, feasible contact management strategy in children, even in resource-limited settings (58,59). Modelling of the parameters for a high TB burden setting suggested that provision of TPT without TBI testing is cost-effective for child contacts < 5 years (60). See section 1.1 for the background of the recommendation for TBI testing and treatment in this risk group.
Evidence reviewed for the 2021 TB screening guidelines on the performance of symptom screening in children and adolescents < 15 years who are close contacts of a person with TB comprised four studies with a total of 2695 participants (13). A comparison of a screen of symptoms including any one of cough, fever or poor weight gain, in which the presence of any symptom constitutes a positive screen, with a composite reference standard indicated a pooled sensitivity of 89% (95% CI 52% ; 98%) and a pooled specificity of 69% (95% CI 51% ; 83%) (13). The evidence on the performance of CXR in close contacts < 15 years who were exposed to people with TB comprised four studies with a total of 2550 participants. In comparison with a composite reference standard, screening for abnormalities on CXR suggestive of TB had a pooled sensitivity of 84% (95% CI 70% ; 92%) and a pooled specificity of 91% (95% CI 90% ; 92%).
Household contacts aged ≥ 5 years and other risk groups
- Among HIV-negative household contacts aged ≥ 5 years and other risk groups, the absence of any symptoms of TB and the absence of abnormal chest radiographic findings may be used to rule out TB disease before TB preventive treatment. (Conditional recommendation, very low certainty of the estimates of effect)
- Among individuals aged ≥ 15 years in populations in which TB screening is recommended, systematic screening for TB disease may be conducted with a symptom screen, chest X-ray or molecular WHO-recommended rapid diagnostic tests, alone or in combination. (Conditional recommendation, very low certainty of the estimates of effect)
- Among individuals < 15 years who are close contacts of a person with TB, systematic screening for TB disease should be conducted with a symptom screen of any one of cough, fever or poor weight gain; or chest radiography; or both. (Strong recommendation, moderate to low certainty of the estimates of effect).
Justification and evidence
Recommendation 14 for ruling out TB disease in contacts aged ≥ 5 years and other HIV-negative risk groups is conditional, due to the very low certainty of the evidence, which is from a study originally included in the 2018 guidelines (17). The systematic review determined the sensitivity and specificity of screening based on symptoms and/or CXR for ruling out TB disease in HIV-negative people and people of unknown HIV status for the 2015 guidelines (see PICO 3 in Annex 3) (61). To illustrate how the various screening and diagnostic algorithms are expected to rule out TB disease, a simple model was constructed to compare the following six screening criteria: (i) any TB symptom, (ii) any cough, (iii) cough for 2–3 weeks, (iv) CXR abnormality suggestive of TB, (v) any CXR abnormality and (vi) a combination of any CXR abnormality or any TB symptom. The model indicated that the combination of any CXR abnormality and the presence of any symptoms suggestive of TB (i.e. cough of any duration, haemoptysis, fever, night sweats, weight loss, chest pain, shortness of breath and fatigue) would have the highest sensitivity (100%) and negative predictive value (100%) for ruling out TB.
Before the 2018 guidelines update, the review was updated to include household contacts aged ≥ 5 years of people with pulmonary TB in high TB burden countries (62). Seven studies of the accuracy of “any CXR abnormality” had a pooled sensitivity of 94.1% (95% CI 85.8 ; 97.7) and a pooled specificity 86.8% (95% CI 79.7 ; 91.7). In a hypothetical population of 10 000 HIV-negative individuals in a country with a TB prevalence of 2%, use only of any TB symptoms for screening would wrongly classify 54 people with TB as not having TB disease and being offered TPT. In contrast, use of any abnormal CXR finding would result in 12 people with TB being offered preventive treatment. Use of the combination of any TB symptoms plus any CXR abnormality would result in no people with TB disease being incorrectly offered preventive treatment. At a TB prevalence of 2%, use of any TB symptoms alone as the screening criterion would require investigations of 16 extra non-TB patients for every individual with TB identified, whereas use of any abnormal CXR finding would require TB investigation of 7 extra non-TB patients for every individual with TB identified. Use of the combination of any TB symptoms plus any CXR abnormal finding would increase the number of individuals requiring TB investigation to 15 extra non-TB patients for every individual with TB identified.
Recommendations 15 and 16 are related to use of symptom screen, CXR or mWRD, alone or in combination, to screen adults and adolescents for TB disease. A systematic review of the diagnostic accuracy of using symptoms and CXR to detect TB disease among individuals aged ≥ 15 years with negative or unknown HIV status was undertaken for the 2021 TB screening guidelines (13). Table 4 shows that, overall, screening for cough has low sensitivity but higher specificity, while screening for any TB symptom improves the sensitivity but reduces the specificity. CXR is both highly sensitive and specific. mWRDs are less sensitive when used for screening than when they are used in diagnostic use but are very specific.
Table 4. Accuracy of tests in HIV-negative people aged ≥ 15 years in high-risk groups
In conclusion, a parallel screening algorithm based on any symptom of TB and any abnormal CXR finding is likely to be highly sensitive. Therefore, the absence of any TB symptoms and any CXR abnormality can be used to exclude pulmonary TB disease before initiating TPT among HIV-negative household contacts aged ≥ 5 years and in other risk groups. mWRDs may be useful when higher specificity is desirable, such as in situations of limited capacity for confirmatory testing after a positive screen.
The GDG reiterated that national guidelines should specify the investigations that are necessary to rule out TB disease. It noted that screening of child contacts could include testing for TBI (see section 1.3) and CXR, although, lack of those investigations should not be a barrier for either diagnosis of TB disease or provision of preventive treatment. In the absence of those tests, clinical assessment alone is sufficient to decide on initiation of TPT, particularly for household contacts aged < 5 years of a person with bacteriologically confirmed pulmonary TB.
The GDG concluded that symptom screening with or without the addition of CXR should be acceptable for individuals and programme managers. CXR could increase the confidence of health-care providers that TB disease has been ruled out and reduce concern that TPT is being administered inappropriately. The GDG for the 2021 WHO guideline on TB screening reviewed the evaluations of three CAD products used with digital CXR and concluded that CAD can be considered accurate when compared with human readers. The GDG therefore conditionally recommended its use for TB screening and triage in individuals aged ≥ 15 years (13).
Implementation considerations
Fig. 1 presents an algorithm for testing for TBI and TPT, with separate entry points for people with HIV, household contacts and other people at risk for TBI. More detailed algorithms for screening and testing for TBI are available in the two handbooks (63,64).
The W4SS is recommended for testing all people with HIV at every visit to a health facility or contact with a health worker to ensure early detection of TB disease. Other clinical features may also be helpful (e.g. poor weight gain in pregnant women and lymphadenopathy). People who have exclusively extrapulmonary TB may have clinical manifestations that are not necessarily pulmonary and may therefore require further evaluation before TB is definitively excluded. Other diseases that cause any of the four symptoms should be investigated in accordance with national guidelines and sound clinical practice. Individuals found not to have TB disease should then be assessed for TPT.
Where CXR or interpretation of radiography is not available, the absence of any TB symptoms alone can be considered sufficient before starting TPT. This would be the most sensitive of all the symptom-based screening rules, and its negative predictive value is high in most settings. Addition of abnormal CXR findings to the symptom screening rule would improve its sensitivity but also increase the logistics and infrastructure required, the cost to individuals and health services, and the requirement for qualified staff or the availability of CXR with CAD. The optimal frequency of CXR in regular TB screening of people with HIV is uncertain. Adding CXR to symptom screening at every visit would represent a significant burden on individuals and health systems. Local authorities should define its application and frequency according to their local epidemiology, health infrastructure and resources. Either CXR with CAD or radiologists or other trained health-care workers must be available to interpret CXR. mWRDs may be useful when greater specificity is desirable, such as when there is limited capacity for confirmatory testing after a positive screen.
The GDG noted that screening with CXR or mWRD should not be a prerequisite or a barrier to initiating TPT in people with HIV because additional resources are required, in view of the marginal gain in negative predictive value. Conversely, in people with HIV and a low CD4 count, TB disease may be present despite a normal CXR. People with HIV who have any of the four symptoms or abnormal CXR findings may have TB disease and should be investigated for TB and other diseases. Xpert® MTB/RIF should be used as the initial diagnostic test.
TPT should not be withheld from an asymptomatic individual at risk of infection if TBI testing and/or CXR is unavailable, as some people may have both risks (e.g. people with HIV who are also contacts of people with TB), in which case the triage shown in the figure would have to be adapted.
It is critical to ensure proper follow-up and evaluation for TB and other diseases in household contacts with abnormal CXR findings or TB symptoms. The investigations should be performed in accordance with national guidelines and sound clinical practice. Contacts found not to have TB disease should be assessed for TPT. Although TBI testing is not a requirement for initiating TPT, it may be done as a part of eligibility screening where feasible (see section 1.3).
A previous history of TB or TPT should not be a contraindication for TPT in cases of re-exposure, after exclusion of reactivated disease. Such individuals, including those with fibrotic radiological lesions, may be at increased risk of progression (65,66). The choice of TPT also depends on the presence of contraindications (e.g. active hepatitis or symptoms of peripheral neuropathy when isoniazid is considered) or the likelihood of drug–drug interactions, particularly when rifamycin regimes are being considered (see section 1.4).
Different symptom screening approaches have different sensitivity and specificity. The facility of symptom screening makes it a much more accessible programme option. Symptom screening is standard in a clinical workup and can be repeated as often as necessary. In contrast, additional resources are necessary for CXR and mWRDs. Scaling up mWRDs for diagnosis should be prioritized (if full access has not yet been achieved) before scaling it up for screening, as it requires significant resources, including increased capacity in and expansion of diagnostic and sample transport networks.
Countries should include the W4SS, CRP, CXR and mWRD in national TB screening algorithms according to their feasibility, the level of the health facility, resources and equity. While all four tools are recommended for people with HIV, CRP is particularly accurate for TB screening of people who are not yet receiving ART, and CXR enhances the sensitivity of the W4SS in people receiving ART, both of which might be considered when choosing algorithms. Consideration should also be given to the added benefit of including CRP for ruling out TB disease before initiating TPT among people with HIV.
CXR has been used to screen for TB for several decades. CXRs are also routinely used to triage people presenting for care who show signs, symptoms or risk factors for TB to determine the most appropriate clinical pathway for proper evaluation. In many settings, however, use of CXR for TB screening and triage for TB disease is limited by the paucity of health personnel trained to interpret radiographic images and by substantial intra- and inter-reader variation in its accuracy to detect abnormalities associated with TB (13). CAD is useful in such situations.
Fig. 1. Combined algorithm for screening and testing individuals at risk before starting TB preventive treatment
CAD, computer aided detection of TB; CRP, C-reactive protein; CXR, chest radiography; IGRA, interferon-γ release assay; mWRD, molecular WHO-recommended rapid diagnostic test; TB, tuberculosis; TBST, Mycobacterium tuberculosis antigen based skin test; TPT, TB preventive treatment; TST, tuberculin skin test.
a Including miners with silicosis, people on dialysis or anti-TNF agent treatment, preparation for transplantation or other risks in national guidelines. TB disease should be ruled out for people in this category.
b For children aged ≥ 10 years, a four-symptom screen is used (current cough or fever or weight loss or night sweats). For children aged < 10 years, consider their history of contact with TB or reported or confirmed weight loss or growth curve flattening or weight for age < –2 Z-scores. Asymptomatic infants aged < 1 year with HIV are given TPT only if they are household contacts of people with TB. For other screening options, see the latest WHO guidance (TB-KSP).
c Any one of cough or fever or night sweats or haemoptysis or weight loss or chest pain or shortness of breath or fatigue. In children, poor weight gain (plateau on growth chart), reduced playfulness or lethargy should also be included in symptom screening; cough may be absent. For other screening options see the latest WHO guidance (TB-KSP).
d Including acute or chronic hepatitis; peripheral neuropathy (if isoniazid is used); regular and heavy alcohol consumption. Pregnancy or a previous history of TB are not contraindications. The person is counselled about the benefits and potential risks of TPT.
e In household contacts aged ≥ 5 years, TST, IGRA or TBST is recommended before consideration of TPT.
f CXR is required only if it was not conducted at a previous step.
g Regimen chosen according to age, strain (drug susceptible or otherwise), risk of toxicity, availability and preference. Adherence supported until completion as prescribed.
4 Poor weight gain here is defined as reported weight loss, very low weight-for-age (< –3 Z-scores), underweight (weight-for-age < –2 Z-scores), confirmed weight loss (> 5%) since the last visit or growth curve flattening
5 For TBI testing and TPT in children < 5 years, see recommendations in section 1.1 and the algorithm in Fig. 1.