Early diagnosis of the active TB is especially needed in HIV infected cases, since there is an accelerated progression of TB and higher mortality. Meantime, ruling out active TB in HIV infected patients prior to isoniazid preventive therapy (IPT) is also necessary. However, besides M.tuberculosis culture, there are only limited methods to diagnose active TB with poor accuracy in HIV-infected patients, which urge the development of alternative, rapid and accurate method. In this study, we found that IGRAs are neither sensitive enough to rule out active TB nor specify to distinguish latent TB infection and active TB in HIV-infected patients.
The sensitivities of both IGRAs in diagnosis of active TB increased after excluded indeterminate results in HIV-infected patients. However, they were still not sensitive enough to rule out active TB alone as they missed more than 20% patients. A recent meta-analysis by Cattamanchi and colleagues showed the sensitivities of QFT-GIT and T-SPOT in diagnosis of latent TB infection in HIV-infected patients were 61% and 72% in low/middle-income countries (indeterminate results included), respectively when using active TB as a surrogate reference standard 
. Another meta-analysis from Hoffmann and Ravn showed pooled sensitivities of 79% and 80.5% for QFT-GIT and T-SPOT in diagnosis of active TB after excluded indeterminate results respectively 
. The slight differences between our results and theirs may be mainly because they also enrolled studies that active TB cases were not confirmed by culture and results were not interpreted according to manufacturer-recommended cut-off value. The sensitivities of both IGRAs were higher in a recent meta-analysis performed on general patients (a few studies also included children and immunosuppressed patients; 81% and 87.5% in their meta-analysis versus 76.7% and 77.4% in ours for QFT-GIT and T-SPOT, respectively after excluded indeterminate results) 
. Thus, HIV infection may impact the performance of IGRAs, which is consistent with results from others 
. However, Tsiouris and colleagues showed no impact of HIV infection on IGRAs sensitivity which might be explained by high proportion of active TB cases that under treatment 
The pooled specificities of QFT-GIT, T-SPOT in this analysis were as low as 75.9%, and 63.1%, respectively, indicating the poor ability of IGRAs in distinguishing latent TB infection and active TB. Apparently, the low specificity is due to the high latent TB infection rate in our selected studies. This is mainly because the most studies we enrolled were conducted in low/middle-income countries with a high TB burden. The two studies from Italy and Austria using QFT-GIT did showed relatively higher specificities. However, the specificity of T-SPOT was also low in the trail completed in Italy by Vincenti and colleagues. This may be due to that nearly one third of non-active TB patients in this study were from low/middle-income countries (i.e. South American and African). Thus, IGRAs may potentially have an adjunctive role in ruling in active TB in HIV-infected patients in high-income countries, but it still needs further investigation.
The high proportion of indeterminate results of IGRAs in HIV-infected patients further dampened their application. In current meta-analysis, the high proportion of indeterminate results could be mainly because of the high level of immunosuppression. Cattamanchi and colleagues showed that the pooled proportion of indeterminate results was significantly higher when CD4 T cell count was less than 200 cells/mL versus greater than 200 cells/mL for QFT-GIT but not T-SPOT in HIV-infected patients 
. However, technical error might also be a cause. The procedure of T-SPOT test is relatively more technically-demanding than QFT-GIT, partially explaining the relatively higher indeterminate results of T-SPOT in this meta-analysis.
There are some limitations in our meta-analysis. Due to our strict inclusion/exclusion criteria, most studies (or the number of eligible individuals) were small, especially in terms of the number of confirmed active TB cases. In addition, number of studies conducted in high-income countries was limited.
In conclusion, the current evidence brought forward in this systematic review and meta-analysis shows that the IGRAs in their current formulations have limited accuracy in diagnosing active TB in HIV-infected patients, and should not be used alone to rule out or rule in active TB cases in HIV-infected patients. Further modification is needed to improve their accuracy.