RetourDiagnostic accuracy studies evaluating sensitivity, specificity and concordance (agreement) of TBSTs were identified. There were no identified studies on the efficacy of TPT based on diagnostic test results, on the predictive value for progression to TB disease or on the proportion started on TPT.
The assessed evidence for Cy-Tb and C-TST has included a manufacturer-recommended induration of at least 5 mm as the cut-off. According to the Diaskintest instructions for use, the presence of induration of any size is considered a positive response. However, the assessed evidence also included some studies for Diaskintest that used an induration of at least 5 mm as a cut-off, specified where applicable.
Sensitivity
A total of 20 studies involving 1627 participants provided data for evaluating the sensitivity of TBSTs in people with microbiologically confirmed TB, which was used as a proxy for sensitivity to diagnose TB infection. Of these, six studies with 539 participants were head-to-head comparisons with the TST or IGRAs (or both); 17 studies included 1276 participants who were HIV-negative or whose HIV status was unknown; five studies included 317 People with HIV; and four studies included 34 participants aged under 18 years. Of the included studies, 14 evaluated Diaskintest, four Cy-Tb and three C-TST, as shown in Figs. 1–4 (all of which are from Web Annex A).
Fig. 1. Sensitivity of TBSTs in head-to-head studies
CI: confidence interval; FN: false negative; IGRA: interferon-gamma release assay; QFT: QIAGEN QuantiFERON; TBST: Mycobacterium tuberculosis antigen-based skin test; TP: true positive; TST: tuberculin skin test.
The pooled sensitivity against the microbiological reference standard for TB disease in six headto-head studies (Fig. 1) was 78.1% (95% confidence interval [CI]: 70.6–84.1%). The evidence was considered to be of high certainty and was not downgraded. Starshinova 2018 (5) and Starshinova 2019 (6) evaluated Diaskintest results with a cut-off of induration of at least 5 mm; the rest of the studies were head-to-head studies evaluating Cy-Tb. The assessed evidence for Cy-Tb included a cut-off of at least 5 mm in all studies. The TST cut-off was 5 mm for People with HIV and 15 mm for people who were HIV-negative in four studies (7–10). Only studies on Diaskintest and Cy-Tb were included in this analysis.
Fig. 2. Sensitivity of TBSTs in all studies in individuals with HIV-negative or unknown status
CI: confidence interval; DST: drug susceptibility testing; FN: false negative; HIV: human immunodeficiency virus; TBST: Mycobacterium tuberculosis antigen-based skin test; TP: true positive.
The pooled sensitivity in 17 studies presented in Fig. 2 among participants who were HIVnegative or HIV status unknown was 76.0% (95% CI: 70.3–80.8%). The sensitivity estimates were lower in the studies using Diaskintest (any induration size). The reason for this is unclear; it may reflect different study populations or study quality. As a result, the evidence certainty was downgraded one level for inconsistency and another level for imprecision. Consequently, the certainty of the evidence was considered very low. Despite the manufacturer’s recommendation to use induration of any size as a positive result, the sensitivity in studies using a Diaskintest result of at least 5 mm as the cut-offwas more closely aligned with the other tests in the class, which all use a cut-off of at least 5 mm.
Risk of bias was considered serious due to the person having knowledge of the reference standards when interpreting the results of index tests. In most Diaskintest studies, the selection of participants and of the reference standard were unclear; hence, the certainty of the evidence was downgraded one level for risk of bias. The sensitivity ranged from 55% to 100% (the reasons for this heterogeneity are unknown); consequently, the certainty of the evidence was downgraded one level for inconsistency. Thus, the overall certainty of the evidence was considered low.
Fig. 3. Sensitivity of TBSTs in People with HIV
CI: confidence interval; DST: drug susceptibility testing; FN: false negative; HIV: human immunodeficiency virus; People with HIV: people living with HIV; TBST: Mycobacterium tuberculosis antigen-based skin test; TP: true positive
Only studies on Diaskintest and Cy-Tb were included in the analysis presented in Fig. 3. The pooled sensitivity among People with HIV in five studies was 63.5% (95% CI: 52.6–73.2%). Risk of bias was considered serious for Diaskintest studies because of the person having knowledge of the reference standards when interpreting the results of index tests; hence, the evidence certainty was downgraded one level for risk of bias. The sensitivity estimates were lowest (39.8%) in the one study that used Diaskintest (any induration size). The reason for low sensitivity for Diaskintest (any induration size) is unclear, and the evidence certainty was downgraded one level for inconsistency. Certainty was also downgraded one level for imprecision. Consequently, the certainty of the evidence was considered to be very low.
Fig. 4. Sensitivity of TBSTs in children and adolescents
AI: any induration size; CI: confidence interval; DST: drug susceptibility testing; FN: false negative; TBST: Mycobacterium tuberculosis antigen-based skin test; TP: true positive.
Sensitivity of TBSTs among children and adolescents is shown in Fig. 4. The pooled sensitivity in four studies for this class of tests was 97.1% (95% CI: 81.9–99.6%). The number of participants included in this analysis was small – only 34 participants in four studies; hence the studies were downgraded two levels for imprecision. Therefore, the evidence certainty was considered low. Only studies on Diaskintest were available for this analysis. Aggerbeck (7) estimated the sensitivity of Cy-Tb in 12 children and adolescents with TB, of whom only two were bacteriologically confirmed and were not included in the figure.
Specificity
A total of 14 studies involving 3792 participants provided data for evaluating specificity of TBSTs (including difference in specificity compared with the reference test); three of the studies included 1104 children and adolescents and three included 587 BCG-vaccinated individuals. Specificity was measured in healthy individuals with negative IGRA results. Difference in specificity was used as an alternative specificity measure, and was calculated as the difference in the proportion of negative results between TBSTs and the TST or IGRAs in healthy populations.
Fig. 5. Specificity in healthy individuals with negative IGRA results
CI: confidence interval; DST: drug susceptibility testing; FN: false negative; IGRA: interferon-gamma release assay; QFT: QIAGEN QuantiFERON; TN: true negative; TST: tuberculin skin test.
The specificity assessed in the five studies presented in Fig. 5 was high for all three tests in the TBST class. For Diaskintest it was 99.1% (95% CI: 93.6–99.9%), as compared with QFT; for Cy-Tb it was 98.0% (95% CI: 92.6–99.5%), as compared with QFT; and for C-TST it was 95.5% (95% CI: 92.6–97.3%), as compared with T-Spot. During the GDG meeting, participants noted that – considering the totality of evidence (which included studies of very low quality) – the overall certainty of the evidence on tests’ effects for specificity was very low.
Specificity in children and adolescents (2 studies, 176 patients), as determined in individuals with negative IGRA results, was high. For Diaskintest with a cut-off of at least 5 mm it was 99.1% (95% CI: 94.9–99.9%), as compared with QFT, and for Cy-Tb it was 91.4% (95% CI: 82.2–96.1%), as compared with QFT. Specificity in BCG-vaccinated individuals (3 studies, 292 patients), as determined in healthy individuals with negative IGRA results, was also high, being 97–99% (depending on the test), with a pooled value of 99.0% (95% CI: 96.9–99.7%). More details can be found in Web Annex A.
Fig. 6. Difference in specificity – TBSTs versus the TST
AI: any induration size; CI: confidence interval; DST: drug susceptibility testing; TBST: Mycobacterium tuberculosis antigen-based skin test; TST: tuberculin skin test.
Values above “0” indicate higher specificity for TBSTs, while values below “0” indicate higher specificity for the TST.
The overall pooled difference in specificity in 14 studies (Fig. 6) comparing TBSTs and the TST was 33.5% (95% CI: 18.2–48.8%) higher for TBSTs. In studies of Diaskintest and C-TST done in high TB incidence settings, the differences in specificity were higher for Diaskintest versus the TST (with both tests having a cut-off of at least 5 mm) (57.3%, 95% CI: 40.2–74.3%), than with Diaskintest (any induration size) versus the TST with a cut-off of at least 5 mm (29.9%, 95% CI: –3.66–63.5%). For C-TST versus the TST with a cut-off of at least 5 mm, the difference in specificity was 39.9% (95% CI: 34.0–45.8%). In contrast, in studies of Cy-Tb undertaken in low TB incidence settings, the difference in specificity between Cy-Tb and the TST was less prominent, but was greater with the TST with a cut-off of at least 15 mm (4.61%, 95% CI: –28.6–37.9%) than with the TST with a cut-off of 5 or 15 mm (–2.0%, 95% CI: –12.3–8.3%). The difference may be explained by the background level of BCG in the study populations or by the cut-offs that were used. Fig.7 has more details on the specificity of TBSTs versus the TST in BCG-vaccinated people. Overall risk of bias was considered serious because test allocation by arm was not blinded in any of the studies except those for Cy-Tb. In most Diaskintest studies, the selection of participants and the diagnosis of the reference standard were unclear. The certainty of the evidence was therefore downgraded one level for risk of bias. The difference in specificity ranged from –2% to 72%; hence, the certainty of the evidence was downgraded one more level for inconsistency. Consequently, the certainty of the evidence for difference in specificity between TBSTs and the TST was low.
Fig. 7. Difference in specificity – TBSTs versus the TST in BCG-vaccinated population
BCG: bacille Calmette-Guérin; CI: confidence interval; DST: drug susceptibility testing; TBST: Mycobacterium tuberculosis antigenbased skin test; TST: tuberculin skin test.
Two studies (three analyses) provided data on difference in specificity in BCG-vaccinated populations, which was even higher for this population than in populations where only some people had received BCG vaccination; the pooled difference in specificity was 67.4% (95% CI: 24.0–110.7%). Overall risk of bias was considered serious because test allocation by arm was not blinded; hence, the certainty of the evidence was downgraded one level for risk of bias. The CI was broad, ranging from 24.0% to 110.7%, so the certainty of the evidence was downgraded one more level for imprecision. Consequently, certainty of the evidence for difference in specificity between TBSTs and the TST in BCG-vaccinated populations was low.
The pooled difference in specificity in six studies comparing TBSTs and IGRAs was low, at 2.3% (95% CI: –1.6–6.2%), meaning that TBSTs were similar to IGRAs in terms of specificity. More details can be found in Web Annex A.
Agreement
Overall, 16 studies involving 3198 participants (among which four studies with 1307 participants recruited people aged under 18 years) were included to assess agreement of the index tests with comparator tests (the TST or IGRAs, or both).
In participants without TB disease, agreement was high (≥90%) for Cy-Tb and Diaskintest – (any induration size) and Diaskintest 5 mm induration – compared with QFT (Fig. 8). Agreement was slightly lower at 85.5% (95% CI: 75.7–91.7%) for C-TST compared with T-Spot. In one study, which evaluated Diaskintest with induration of at least 7 mm compared with T-Spot, the agreement was considerably lower, at 60.9% (95% CI: 54.3–67.2%). Risk of bias was considered serious because the allocation of tests was not blinded in five studies; hence, certainty of the evidence was downgraded one level for risk of bias. Agreement ranged widely (from 61% to 97%) for various tests and studies, so the certainty of the evidence was downgraded one level for inconsistency. Consequently, certainty of the evidence for agreement between TBSTs and IGRAs was low.
Fig. 8. Agreement of TBSTs versus IGRAs in all studies including participants without active TB
AI: any induration size; CI: confidence interval; DST: drug susceptibility testing; HIV: human immunodeficiency virus; IGRA: interferongamma release assay; QFT: QIAGEN QuantiFERON; TB: tuberculosis; TBST: Mycobacterium tuberculosis antigen-based skin test; TST: tuberculin skin test.
In participants with TB disease, high agreement between TBSTs and IGRAs as the comparator (85.7%) was observed (Fig. 9). Some variability in agreement was seen between the different tests: 79.6% (95% CI: 76.3–82.6%) for Cy-Tb 5 mm compared with QFT; 97.3% (95% CI: 72.7–99.8%) for Diaskintest (any induration size) compared with QFT; and 97.0% (95% CI: 92.3– 98.9%) for DST 5 mm induration compared with QFT. Agreement was slightly lower at 85.4% (95% CI: 72.4–92.9%) for C-TST compared with T-Spot. Risk of bias was considered serious because, in four studies, the allocation of tests by arm was not blinded; hence, the certainty of the evidence was downgraded one level for risk of bias. The agreement ranged from 75% to 100% for various tests and studies, so certainty of the evidence was downgraded one level for inconsistency. The overall certainty of the evidence for agreement between TBSTs and IGRAs in people with TB disease was considered low.
Fig. 9. Agreement of TBSTs versus IGRAs in all studies including people with active TB
AI: any induration size; CI: confidence interval; DST: drug susceptibility testing; HIV: human immunodeficiency virus; IGRA: interferongamma release assay; QFT: QIAGEN QuantiFERON; TB: tuberculosis; TBST: Mycobacterium tuberculosis antigen-based skin test;
TST: tuberculin skin test