Footnotes

Quality of evidence was rated as high (no points subtracted), moderate (1 point subtracted), low (2 points subtracted), or very low (>2 points subtracted) based on five criteria: study limitations, indirectness of evidence, inconsistency in results across studies, imprecision in summary estimates, and likelihood of publication bias. For each outcome, the quality of evidence started at high when there were randomized controlled trials or high quality observational studies and at moderate when these types of studies were absent. We then subtracted one point when there was a serious issue identified or two points when there was a very serious issue identified in any of the criteria used to judge the quality of evidence.

3 studies were eligible and thus included in the analysis; 1 published (China) and 2 unpublished (Zambia and South Africa). (N refers to numbers that entered follow-up)

Based on the Newcastle-Ottawa scale, study samples were considered to be representative of specific groups of interest (i.e., silicosis patients (China), case-contacts (Zambia), adolescent school-goers) within the population and IGRA exposure groups were drawn from the same sample and therefore unlikely to introduce any bias. However, studies varied with regard to the comparability (adjustments made to effect measures) and outcome (ascertainment, losses to follow-up, reporting) components of the modified NOS. Lack of proper ascertainment of the TB outcome is considered to be the most serious of limitations. A point is deducted.

The results of the studies could be generalized for the specific country/region and for those specific groups of interest. However, the small number of studies warrants caution; a point is deducted for indirectness.

All 3 studies showed similar results and with very little heterogeneity in the pooled incidence rate ratio (I²=0%, p=0.912). No points were deducted.

The number of incident TB cases was small in all studies and the rates of TB fairly moderate; confidence intervals for relative risk estimates were wide (precision > +/- 20%). This is a very serious limitation. Two points are deducted.

Data included did not allow for formal assessment of publication bias using methods such as funnel plots or regression tests. Therefore, publication bias cannot be ruled out. Although no points were deducted, a degree of publication bias is likely because: 1) literature on IGRAs is rapidly exploding and currently unpublished studies may come out in future (although we made an attempt to include unpublished studies, our attempt was not comprehensive; we are aware of at least one unpublished study that was not assessed for this review); 2) there are anecdotal examples of unpublished negative studies on IGRAs; and 3) because a sizeable proportion of IGRA studies have some level of industry involvement or support, the risk of unpublished negative studies (or delayed publication of negative studies) is not trivial.

All three studies provided incidence rates of TB stratified by IGRA as well as TST status at baseline. (N refers to numbers that entered follow-up)

Serious limitations include lack of proper ascertainment of the TB outcome by smear and/culture, IGRA incorporated in the methods to diagnose TB (South Africa) and lack of adjustment of all confounders. A point is deducted.

The results of the studies could be generalized for the specific country/region and for those specific groups of interest. However, the small number of studies warrants caution; a point is deducted for indirectness.

The two tests perform comparably and any differences are not statistically significant as the 95% confidence intervals for the pooled IRRs overlap and there is no heterogeneity in the pooled estimates for either test (IGRA+: IRR=3.2, I²=0%, p=0.899 and TST+: IRR=2.3, I²=0%, p=0.383). No points deducted.

The confidence intervals of the pooled IRRs are wide (precision > +/- 20%). This is a very serious limitation. Two points are deducted.

Publication bias was not formally assessed, but is deemed likely. See ^B6.

Only the Zambian study examined if there was an exposure-gradient relationship between baseline quantitative IGRA levels and subsequent rates of TB in those levels. (N refers to numbers included in this stratified analysis)

Lack of proper ascertainment of the TB outcome by smear/culture for both studies. The Zambian study is unpublished and only an interim report was available, so quality could not be fully assessed. A point is deducted.

There is only one study. There is serious indirectness. A point is deducted.

There is only one study; inconsistency cannot be assessed. A point is deducted.

The 95% confidence intervals per IGRA stratum were extremely wide (precision > +/- 20%). Two points are deducted.

Publication bias was not formally assessed, but is deemed likely. See ^B6.

The Zambia and South Africa studies further explored rates for TB in paired concordant and discordant TST/IGRA results. (N refers to number included in this stratified analysis)

Serious limitations include lack of proper ascertainment of the TB outcome by smear and/culture, IGRA incorporated in the methods to diagnose TB (South Africa) and lack of adjustment of all confounders. A point is deducted

Although results may be generalizable to similar L/MIC, there are only two studies. A point is deducted.

Rates of TB during follow-up may be higher in those with double positive TST+/IGRA+ results than in those with double negative results. Both studies seem to suggest this. However, contrasting results are seen with regard to discordant pairs. Pooled estimates were not derived. The inconsistency in results is deemed serious; a point is deducted.

Observed 95% confidence intervals around the rates per strata are wide (precision > +/- 20%).

Publication bias was not formally assessed, but is deemed likely. See ^B6.

All 3 studies were included in this evaluation of patient-relevant outcomes. The diagnostic accuracy estimates of sensitivity and specificity etc are surrogates of patient-relevant outcomes important for assessing the frequency and impact of either a false negative or false positive IGRA result at baseline. A falsely positive outcome may result in possible isoniazid preventive therapy (IPT) prescription for a period of 6-9months, depending on country guidelines. IPT, although safe, is not without serious adverse effects, notably, clinical hepatitis and the increased possibility of drug resistance in the future. Whilst a falsely negative result may result in no IPT being provided and the individual exposed to at least a 2-fold risk of developing TB in the future.

Serious limitations include lack of proper ascertainment of the TB outcome by smear and/culture, IGRA incorporated in the methods to diagnose TB and lack of adjustment of all confounders for most studies. A point is deducted

Although results may be generalizable to similar L/MIC, there are only three studies. A point is deducted.

There is heterogeneity in individual studies' test accuracy estimates (e.g. specificity/false positive rates). A point is deducted.

The summary estimates of sensitivity and specificity are moderate and the confidence intervals are wide (precision > +/- 20%). Two points are deducted.

Publication bias was not formally assessed, but is deemed likely. See ^B6.