Until recently, the only tool available to detect latent tuberculosis infection (LTBI) was the tuberculin skin test (TST). Although the TST has proven to be useful in clinical practice, it has known limitations . An important advance in recent years has been the development of in vitro T cell—based IFN-γ release assays (IGRAs). These assays use antigens more specific to Mycobacterium tuberculosis than the PPD used in the TST. As reviewed elsewhere [2, 3], there is now considerable evidence that prior bacille Calmette-Guárin vaccination does not affect IGRA results, meaning that IGRAs have higher specificity than does the TST. Besides the high specificity, other potential advantages of IGRAs include logistical convenience, avoidance of subjective measurements (such as skin induration), need for fewer patient visits, and the ability to perform serial testing without inducing the boosting phenomenon.
Although IGRAs are promising for LTBI diagnosis and are being used in place of the TST in some settings, what is their potential role in the evaluation of individuals with suspected active tuberculosis? There are 2 reasons for evaluating IGRAs among patients with active tuberculosis: (1) to use active tuberculosis as a surrogate reference standard for LTBI (most published studies would fit in this category), and (2) to determine whether IGRAs will be helpful in diagnosing active tuberculosis, per se. Both approaches have limitations. The first reason is based on the logic that anyone with active tuberculosis must have tuberculosis infection—albeit not latent infection. Therefore, the assumption is that active tuberculosis is an acceptable surrogate for LTBI, which may or may not be valid. From an immunological perspective, active tuberculosis occurs when the host immune response is unable to contain the latent infection; thus, it is possible that the sensitivity of IGRAs in active disease may not reflect their sensitivity in LTBI. The second reason to perform the test is based on the assumption that evidence of tuberculosis infection (i.e., a positive IGRA result) is useful to diagnose active disease. This is problematic, because IGRAs, like the TST, are based on cellular immune response, and as such, they are incapable of distinguishing between LTBI and active disease. This is an issue because most cases of tuberculosis disease occur in populations with a high prevalence of LTBI.
It has been suggested that IGRAs have the potential to serve as useful tests to rule out active tuberculosis in selected populations (e.g., children and immunocompromised persons) in which microbiologic diagnosis is hard to establish [3, 4]. In other words, although a positive IGRA result may not always indicate active disease, a negative IGRA result may indicate lack of tuberculosis infection and, therefore, of disease. For IGRAs to be useful in excluding active disease, they must be highly sensitive in patients with active tuberculosis. In this issue of Clinical Infectious Diseases, Dewan et al.  suggest that this may not always be the case for the QuantiFERON-TB Gold (QFT-G; Cellestis) assay. Their study estimated the sensitivity of QFT-G among individuals with suspected tuberculosis disease in San Francisco. They found that only 64% (95% CI, 48%–78%) of 36 patients with culture-positive tuberculosis had positive QFT-G results . These data, although limited by a small number of patients (reflected in the imprecision of their sensitivity estimate) and lack of a prospective head-to-head comparison with TST, add to the growing body of literature on the sensitivity of QFT-G in various populations.
Table 1 and figure 1 present the results of 10 published studies of QFT-G among patients with active tuberculosis [5–14]. Most studies recruited patients with microbiologically confirmed tuberculosis disease who were untreated or who received treatment for <4 weeks. A majority of these studies, including the study by Dewan et al. , used the second-generation QFT-G assay, which is currently licensed in the United States and has been approved by the US Food and Drug Administration. This second-generation version has already been replaced in many countries by the newer “In Tube” version of the assay (currently under US Food and Drug Administration review). The In Tube assay is simpler to use and may be more sensitive than the second-generation format .
As seen in figure 1, the sensitivity (conservatively estimated without excluding indeterminate results) of QFT-G varied from 55% to 88%, with a weighted, pooled mean of 75% (95% CI, 71%–78%). Because few studies used the In Tube version, this estimate probably reflects the sensitivity of the second-generation QFT-G assay. The 64% sensitivity estimate reported by Dewan et al.  is lower than this average but within the range of previously reported estimates. Several factors may influence the sensitivity of IGRAs, including antigen types and combinations used, test format (ELISA vs. enzyme-linked immunospot assay), specimen used (whole-blood sample vs. PBMCs), severity of tuberculosis disease, treatment status of the patient, and host factors, such as age, nutritional status, immunosupression (e.g., HIV infection), and other comorbid conditions . Incorrect specimen handling and laboratory procedures may also result in suboptimal test performance . Some of the differences seen across studies may also be associated with variations in thresholds (cutoff points) used to define a positive result .
Conversely, factors that may lead to an underestimation of the sensitivity of IGRAs include the accuracy of the tuberculosis diagnosis and whether measures have been implemented to exclude cases with false-positive results of laboratory tests from analysis. Because of the nature of the studies and the large number of potential variables that govern IGRA sensitivity, it is unclear which of these factors were relevant in the study by Dewan et al. . Two recent studies suggest that the enzyme-linked immunospot assay format (which uses PBMCs) may be more sensitive than the ELISA assay (which uses whole-blood specimens) [11, 12], but whether such an increase in sensitivity can make this test clinically useful in the evaluation of active tuberculosis remains to be determined.
Table 1 also presents the TST sensitivities for patients with active tuberculosis, although not all studies performed head-to-head comparisons. The pooled, weighted mean sensitivity of TST is 77% (95% CI, 72%–81%). This is consistent with older studies of TST that reported modest and variable sensitivity for cases of active pulmonary tuberculosis [18–21]. Thus, the sensitivity of the QFT-G assay for active tuberculosis is comparable to the sensitivity of TST, and by extension, both tests have suboptimal sensitivity for patients with active tuberculosis. This reflects the well-known diminished immune response in patients with active tuberculosis at the time of diagnosis, particularly for those with more-advanced disease [18, 19], malnutrition , or older age . Diminished immune response (i.e., anergy) may also explain the indeterminate QFT-G results (which were, on average, seen in 7% of patients with active tuberculosis) ( table 1) reported in various studies . The highest proportion of indeterminate QFT-G results was seen in studies that included individuals with other immune suppressive disorders [7, 13].
These observations raise the question of whether distinct IGRA cutoff points should be set for the evaluation of active tuberculosis vis-à-vis LTBI, especially for persons with severe tuberculosis disease or immunosuppressive conditions. However, because neither test can distinguish latent tuberculosis from active tuberculosis, the specificity for active tuberculosis will always be low, especially in areas with a high incidence of tuberculosis, and lowering the cutoff value for active tuberculosis will only add more subjects with such results.
Together, these findings support the guidelines from the US Centers for Disease Control and Prevention  that state that, for reasons of suboptimal sensitivity, a negative QFT-G or TST result cannot be used alone to exclude the diagnosis of active tuberculosis. The study by Dewan et al.  reinforces the conventional understanding that the diagnosis of active tuberculosis primarily rests on microbiological confirmation of tuberculosis. Clinicians who treat patients with suspected tuberculosis disease should align their practice with the International Standards for TB Care  and use sputum smear microscopic evaluation and culture to investigate patients with suspected active tuberculosis. For special groups for whom microbiological confirmation is difficult (e.g., children and HIV-infected individuals), IGRAs may still have a useful supporting role [25–27], but this requires large-scale evaluation in field settings.
We are grateful to Giovanni Ferrara (University of Modena and Reggio Emilia, Modena, Italy) and Delia Goletti (National Institute for Infectious Diseases, Rome, Italy), for sharing additional information on their studies; and to Marcel Behr (McGill University, Montreal, Canada), for providing feedback on a draft of this manuscript.
Financial support. Both authors receive research support from the Canadian Institutes of Health Research.
Potential conflicts of interest. M.P. and D.M.: no conflicts.