Evidence-Based Comparison of Commercial Interferon- g Release Assays for Detecting Active TB

Transcription

1 CHEST Special Feature Evidence-Based Comparison of Commercial Interferon- g Release Assays for Detecting Active TB A Metaanalysis Roland Diel, MD, MPH ; Robert Loddenkemper, MD, FCCP ; and Albert Nienhaus, MD, MPH Test accuracy of interferon- g release assays (IGRAs) for diagnosing TB differs when using older or precommercial tools and inconsistent diagnostic criteria. This metaanalysis critically appraises studies investigating sensitivity and specificity of the commercial T-Spot. TB and the QuantiFERON-TB Gold In-Tube Assay (QFT-IT) among definitely confirmed TB cases. We searched Medline, EMBASE, and Cochrane bibliographies of relevant articles. Sensitivities, specificities, and indeterminate rates were pooled using a fixed effect model. Sensitivity of the tuberculin skin test (TST) was evaluated in the context of IGRA studies. In addition, the rates of indeterminates of both IGRAs were assessed. The pooled sensitivity of TST was 70% (95% CI, ) compared with 81% (95% CI, ) for the QFT-IT and 88% (95% CI, ) for the T-Spot. TB. Sensitivity increased to 84% (95%CI, ) and 89% (95% CI, ) for the QFT-IT and T-Spot. TB, respectively, when restricted to performance in developed countries. In contrast, specificity of the QFT-IT was 99% (95% CI, ) vs 86% for the T-Spot. TB (95% CI, ). The pooled rate of indeterminate results was low, 2.1% (95% CI, ) for the QFT-IT and 3.8% (95% CI, ) for the T-Spot. TB, increasing to 4.4% (95% CI, ) and 6.1% (95% CI, ), respectively, among immunosuppressed hosts. The newest commercial IGRAs are superior, in comparison with the TST, for detecting confirmed active TB disease, especially when performed in developed countries. CHEST 2010; 137(4): Abbreviations: IGRA 5 interferon- g release assay; LTBI 5 latent TB infection; QFT-IT 5 QuantiFERON-TB Gold In-Tube assay; TST 5 tuberculin skin test More than 1 year has passed since the release of a metaanalysis on the sensitivity and specificity of interferon- g release assays (IGRAs) and the tuberculin skin test (TST) for TB. Pai and colleagues 1 summarized Manuscript received October 3, 2009; revision accepted November 10, Affiliations: From the Department of Pulmonary Medicine (Dr Diel), Medical School (MHH) Hannover; German Central Committee against Tuberculosis (Dr Loddenkemper), Lungenklinik Heckeshorn, HELIOS Klinikum Emil von Behring, Berlin; and Institution for Statutory Accident Insurance and Prevention in the Health and Welfare Services (Dr Nienhaus), Hamburg, Germany. Correspondence to: Roland Diel, MD, MPH, Assistant Professor, Department of Pulmonary Medicine, Hannover Medical School (MHH), Carl-Neuberg-Straße 1, Hannover, Germany; Diel.Roland@mh-hannover.de. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians ( site/misc/reprints.xhtml ). DOI: /chest a body of studies that explored the emer ging technology predominately using the second-generation Quanti- FERON Gold (Cellestis Limited; Carnegie, Australia) whole blood enzyme-linked immunosorbent assay test on the one hand and a wide selection of ELISpot assays with different cutoffs and preparations on the other. Since then, a constant flood of publications on IGRAs has appeared in journals, most covering work done with the standardized, commercially available T-Spot. TB ELISpot (T-Spot; Oxford Immunotec; Abingdon, UK) and the new QuantiFERON-Gold In-Tube ([QFT-IT] Cellestis Limited) assay, the latter incorporating a third RD-11 antigen, TB 7.7. The large body of publications on these two IGRA tools, which meanwhile have both received market approval by the US Food and Drug Administration, offers the opportunity to reassess their performance with respect to sensitivity and specificity. It was Special Feature

2 our intention to establish performance benchmarks that the routine laboratorian can expect to achieve using these products. To this end, our review takes into account only work based on accepted gold standards of diagnostic confirmation of active TB disease (see criteria discussed in the next section), also enabling us to verify TST sensitivity where it has been assessed within the scope of IGRA studies. One aspect of the new IGRA tests that has yet to be systematically evaluated is that of invalid res ults (indeterminates). Both commercially available IGRAs in contrast to the TST offer a mitogen stimulation of cells performed in parallel to the specific-antigen stimulation in order to measure the ability of the harvested cells to produce interferon- g. If an individual does not respond sufficiently to either specific antigens or the mitogen control, they are deemed indeterminate. The absence of a mitogen response may suggest T-cell anergy, but may also be indicative of error in performance of the assay. Indeterminate results may be meaningful, but may also reduce the value of a procedure if they occur too frequently. For this reason, an overview on the frequency of such invalid results has been included in this analysis. Methods of the Review Inclusion and Exclusion Criteria for establishing the diagnosis of TB. If the number of unconfirmed and confirmed patients in mixed studies was not presented separately, those studies were also rejected to avoid selection bias. (5) Studies assessing sensitivity of TB among patients screened after being treated for more than 2 weeks were excluded because of the risk of inhibited IGRA reactivity. 2 (6) Participants in studies assessing IGRA specificity of latent TB infection had to be healthy, native residents of lowincidence countries without any previously known exposure to TB, irrespective of bacille Calmette-Guérin vaccination status. (7) In studies comparing results of two or more tests, all tests had to be performed simultaneously in the same persons to ensure comparability. We excluded studies that performed sequential testing in which performing the second test was conditional to the positivity of the first test. Comparisons of changing fractions of the study populations were excluded because of possible selection bias. (8) In cases of TST performance, results using the cutoff chosen by the authors as part of the respective study design were taken for our review. Results As shown in Figure 1, a total of 679 English and non-english articles were obtained through database searching, and of these, 124 publications were eligible for inclusion. Of the 124 studies, 40 reported sensitivity of any of the three screening tests, a further seven (of these five also included in the studies reporting sensitivity) investigated specificity of IGRAs, and a total of 116 studies (of these 82 additional studies unrelated Search methods for identification of studies, data collection and analysis, as well as statistical analysis were performed as described in the online supplement. Studies were identified as potentially relevant as outlined in the Search Methods section. From these, original articles or letters to the editor were selected if they met the following selection criteria: 679 potentially relevant citations identified by electronic databases and supplementary sources Excluded before full text screening: Editorials (34) Economic analyses (10) Animal TB infection (3) (1) Studies had to present original data and to have followed a study design allowing data comparison; case reports, editorials, and reviews were excluded a priori. (2) Only studies using the most recent commercially available IGRAs (QFT-IT and T-Spot. TB ) were included; here we aimed to provide the best possible estimate of the diagnostic accuracy of those tests most likely to be considered today for routine work. For all published data for the T-Spot. TB a cutoff of 6 spots for a positive result was used. (3) Because one of the topics of this analysis is the rate of indeterminate results, studies in which a mitogen stimulation was not performed or in which blood processing was performed in a manner deviating from the manufacturers instructions (ie, in vitro incubation. 24 h or freezing of cells) were also excluded. (4) Only studies in which participants had active TB disease confirmed by culture and/or polymerase chain reaction and/or histologic examination were included for the assessment of sensitivity, as false TB diagnosis may lead to an underestimation of true IGRA sensitivity. Studies were excluded that relied solely on clinical evaluation or radiologic features, on improvement of symptoms while on anti-tb therapy, and/or on smear positivity as the criteria 632 full text articles included for further review 341 full text studies screened for further eligibility criteria 124 studies finally included into review Sensitivity: 40 Specificity (IGRAs): 7* Indeterminates: 116** * including 5 studies also providing sensitivity data ** including 82 additional studies Excluded before in-depth examination: Reviews and meta analyses (65) Guidelines (35) Letters without original data (19) Case reports (11) Immunological background studies (99) Laboratory studies (62) Figure 1. Flow diagram for study selection. IGRAs 5 interferon- g release assays. CHEST / 137 / 4 / APRIL,

3 to sensitivity or specificity) provided data on indeterminate results. Seventeen studies reported outcomes in children aged less than 15 years, and 45 studies investigated immunosuppressed individuals. The main outcomes of the studies are described as follows. Sensitivity of Screening Tests Sensitivity of the TST: Strictly following our inclusion criteria as described above, 25 studies could be included into this analysis ( Fig 2 ); 3-27 more than half of those (n 5 13) were published in 2008 or 2009 and only two were conducted in developing countries: India 24 and South Africa. 27 As shown in Figure 2, the overall detection rate of the TST for active TB among the 1,238 patients included was 69.9% (865/1,238) with a small CI of 67% to 72%. Of note, however, is that most studies had only a small proportion of confirmed TB cases, only 10 studies included 50 patients or more, and the degree of heterogeneity between the studies was considerable (I %). Five studies listed in the update of Pai et al 1 were not taken into consideration because numbers of confirmed TB cases were not presented or because of a possible selection bias. 32 Sensitivity of the QFT-IT Assay: For determining sensitivity in confirmed TB cases for the QFT-IT, 19 studies could be included, resulting in a pooled sensitivity of 81% (797/988; 95% CI, ) ( Fig 3 A ) The degree of heterogeneity between the studies was high (I %). In the subgroup of studies coming from developing countries, 9,24,33,34,37,39 ( Fig 3 C ) there was a low sensitivity of only 274 of 369 TB patients, or 74.3% (95% CI, ). In contrast, combined sensitivity from the remaining 13 studies performed in developed countries ( Fig 3 B ) increased to 84.5% (95% CI, ) or 523/619, and heterogeneity between studies was only moderate (I %). The difference in sensitivity for the QFT-IT as shown in developed and developing countries was highly significant (523/619 vs 274/369, P,.001). Sensitivity of T-Spot. TB : For determining pooled sensitivity of the T-Spot, 17 studies were available, detecting 732 of 837 confirmed TB cases, and resulting in a pooled sensitivity of 87.5% (95% CI, ) ( Figure 4) The degree of heterogeneity was comparable to that among the QFT-IT studies (I %). Of note, only two of the studies 27,46 had been performed in developing countries. Thus, focusing on the subgroup of developed countries here resulted only in a slight increase in sensitivity to 88.5% (95% CI, , difference not significant). Sensitivity in Head-to-Head Comparisons of QFT-IT and T-Spot. TB : To date, seven direct comparisons of the sensitivities of T-Spot. TB and QFT-IT have been Figure 2. Forest plot showing sensitivity results from studies investigating detection of active TB disease (here by the tuberculin skin test). Studies are ordered by author and year of publication. The red squares and horizontal lines correspond to the recorded percentage of true positive results among confirmed TB cases and their respective 95% CIs. The area of the red squares reflects the weight each study contributes to the analysis. The diamond represents the pooled value with its 95% CI. 954 Special Feature

4 A B C Figure 3. Forest plot of QuantiFERON-TB In-Tube Gold assay (QFT-IT) sensitivity. ( A ) QFT-IT sensitivity. ( B ) QFT-IT sensitivity in developed countries. ( C ) QFT-IT sensitivity in developing countries. CHEST / 137 / 4 / APRIL,

5 Figure 4. Forest plot of T-SPOT. TB sensitivity. In US Food and Drug Administration-approved criteria a person is T-Spot. TB negative if they have 4 spots above the nil control, positive if 8 spots, and borderline (or equivocal) for five, six, or seven spots. published. Unfortunately, three of them did not meet the inclusion criteria: Chee et al 35 used divergent denominators of patients tested with T-Spot. TB and QFT-IT (274 patients tested with T-Spot. TB and 280 patients tested by the QFT-IT out of a total of 283 culture-confirmed patients), Connell et al 48 did not report the results of culture-confirmed cases, and Soysal and colleagues 23 compared only an equalized number of TB cases after subtraction of indeterminate results for each IGRA. In the remaining four small comparison studies ( Table 1 ) with a total of 108 patients, there was no statistical difference in sensitivity between the two assays (84.3%, 91/108 for QFT-IT; and 77.8%, 84/108 for the T-Spot. TB ). Sensitivity in Head-to-Head Comparisons of QFT- IT or T-Spot. TB and TST: Seven studies simultaneously comparing sensitivity of T-Spot. TB and the TST ( Table 2 ), with a total of 243 patients with TB and five studies comparing QFT-IT and the TST ( Table 3 ) with a total of 130 patients, could be included. For both IGRAs, pooled sensitivity (T-Spot. TB vs TST: 219/243 [90.1%] vs 166/243 [68.3%] and QFT-IT vs TST (108/130 [83.1%] vs 81/130 [62.3%]) was significantly higher than the corresponding rate for the TST ( P,.0001 and P, 0.001, respectively). Specificity of QFT-IT and T-Spot. TB There were only five published studies targeted at evaluating QFT-IT specificity. 10,22,36,49,50 Only four of the 513 tested persons with a low risk of Mycobacterium tuberculosis infection were positive, resulting in a specificity of 99.2% (95% CI, ) ( Fig 5 ). For T-Spot there were only three published studies: Detjen, 10 Higuchi, 45 and Lee. 6 Of the 255 tested subjects, 35 were classified as being false positive, resulting in a specificity of 86.3% (95% CI, ) ( Fig 6 ). Of note, despite of the low number of low-risk individuals included in this analysis, specificity results between the two IGRAs differed significantly ( P,.0001). Table 1 Head-to-Head Comparison of Sensitivity of QuantiFERON-TB In-Tube Gold Assay vs T- Spot. TB in Patients With TB Study Year Sample Country Confirmed Patients, No. Positive QFT-IT Results, No. (%) Positive T-Spot. TB Results, No. (%) Sensitivity Difference Between QFT-IT and T-Spot. TB, Percentage Points Detjen et al Children Germany (93) 26 (93) 0 Domínguez et al Adults and Spain (79) 36 (86) 27 children Kampmann et al Children United Kingdom (80) 14 (56) 24 Markova et al Adults Bulgaria (93) 8 (62) 31 Totals (84.3) 84 (77.8) n.s. n.s. 5 not significant; QFT-IT 5 QuantiFERON-TB Gold In-Tube assay. 956 Special Feature

6 Table 2 Head-to-Head Comparison of Sensitivity of T-Spot. TB vs Tuberculin Skin Test in Patients With TB Study Year Sample Country Confirmed Patients, No. Positive T-Spot. TB Results, No. (%) Positive TST Results, No. (%) Sensitivity Difference Between T-Spot. TB and TST, Percentage Points Lee et al Predominantly adults Korea (93) 39 (71) 22 Detjen et al Children Germany (93) 28 (100) 27 Kang et al Adults Korea (93) 46 (69) 24 Kim et al Adults Korea (100) 10 (46) 12 Vincenti et al Adults Italy (85) 6 (46) 39 Kobashi et al Adults Japan (88) 29 (60) 28 Nicol et al Children South Africa 10 5 (50) 8 (80) 230 Totals (90.1) 166 (68.3) P,.0001 TST 5 tuberculin skin test. See Table 1 for expansion of other abbreviations. Invalid IGRA Results Indeterminates Among Individuals Tested With QFT-IT: For QFT-IT, a total of 72 studies were identified ( Table 4 ), with the number of study participants varying between and 3,678, 90 covering a total of 21,922 subjects. Of these, 469 individuals had indeterminate results (pooled 2.14% indeterminates, 95% CI, ). Indeterminates Among Subjects Tested With T-Spot. TB : For T-Spot, 59 studies were included ( Table 5 ), covering 12,165 patients with 462 indeterminate results (pooled 3.80% indeterminates, 95% CI, ), ranging from to 1,442 study participants.116 Adding the number of 80 reported subjects from whom there were an insufficient number of cells available for performing the T-Spot, the number of invalid results increases to 4.46% (95% CI, ). The difference in the rate of indeterminate results between the two IGRAs was highly statistically significant ( P,.0001) in favor of QFT-IT. Indeterminates in Immunosuppressed Hosts: Limiting evaluation to subgroups of immunosuppressed subjects increased the rate of indeterminate responses in QFT-IT tested to 4.42% from the pooled figure of 2.14% ( Table 6 ) and in T-Spot to 6.12% from 3.80% ( Table 7 ). The difference in indeterminate rates between the two IGRAs in immunosuppressed subjects was highly significant in favor of QFT-IT ( P 5.001). Indeterminates in Head-to-Head Comparisons of QFT-IT and T-Spot. TB : For the 11 studies directly comparing the number of indeterminate responses in individuals tested simultaneously with QFT-IT and T-Spot, a statistically significant difference ( P,.0001) between the pooled number of indeterminates was also found. QFT-IT was indeterminate for 76/3,075 subjects (2.47%) and T-Spot for 133/3,075 individuals (4.33%) ( Table 8 ). Quality of Analyzed Studies Although the number of publications meeting the criteria for our analysis was large, the quality of the trials varied considerably. The majority of studies for estimation of sensitivity or specificity of either test had small sample sizes: Fewer than one-third (13) of the 41 studies had a sample size of confirmed TB cases of 50 patients or more. Despite the pooling effect of the metaanalysis, this small sample size enlarges the CI around the true estimates and thus limits their meaningfulness. Small sample size did not play a role in studies investigating invalid results, but most studies using T-Spot did not reveal how many patients were not included in the final study population because of insufficient number of cells for further processing, possibly leading to selection bias. A potential limitation of a metaanalysis to make generalizations, as well as its statistical power, is loss of study participants. This could be relevant when Table 3 Head-to-Head Comparison of Sensitivity of QFT-IT vs TST in Patients With TB Study Year Sample Country Confirmed Patients, No. Positive QFT-IT Results, No. (%) Positive TST Results, No. (%) Sensitivity Difference Between QFT-IT and TST, Percentage Points Detjen et al Children Germany (93) 28 (100) 27 Bartu et al Adults Czech Republic (86) 10 (45) 41 Gerogianni et al Adults Greece (95) 18 (90) 5 Bianchi et al Children Italy (94) 14 (88) 6 Kabeer et al Adults India (66) 11 (25) 41 Totals (83.1) 81 (62.3) P,.001 See Tables 1 and 2 for expansion of abbreviations. CHEST / 137 / 4 / APRIL,

7 Figure 5. Forest plot of QFT-IT specificity. See Figure 3 legend for expansion of abbreviation. initially confirmed TB cases were included in a study, but no test results were finally available. It did not, however, play an important role in this analysis. Only the studies of Mori et al 3 (for 42 of the 118 evaluable TB patients a TST result was not available), Ferrara et al 5 (4 of 24), Palazzo et al 22 (8 of 20 for the TST and 3 of 20 for the QFT-IT), and Dosanjh et al 17 (35 of 154), had a drop-out rate of more than 10% of the group of confirmed TB cases, and there is no evident reason to suppose that the losses had any systematic direction. There was considerable heterogeneity ( P,.0001, I 2. 75%) with respect to most outcomes among the studies. This does not, however, mean that all the studies cannot be pooled to obtain a central estimate. It reflects, rather, clinical heterogeneity of the various study settings and populations. Through stratification, heterogeneity was clearly reduced (ie, when a subgroup analysis of sensitivity of the IGRAs was made for studies from developed countries). Another exception was the analysis of IGRA specificity, which was remarkably consistent, although this category contained data from only five (QFT-IT) and three (T-Spot.TB ) studies, respectively. Discussion Until new biomarkers are found to establish a gold standard for latent TB, active TB has to be used as a surrogate for latent TB infection (LTBI). Although the superiority of IGRAs over TST for detecting LTBI seems likely, 132 conclusions made to date have not been based on a rigorous comparison of confirmed TB cases. The early metaanalysis of Pai and colleagues1 took in studies from the limited literature available at that time that used different definitions of the presence of active TB disease and in which the proportion of cases with culture verification varied considerably. With the expansion of the literature, and with the acceptance of these tests in daily practice also expanding, a new metaanalysis was required to more accurately assist practitioners in their decisions and choices on TB infection testing. Remarkably, in our updated metaanalysis, pooled sensitivity of TST, as far as it is measured within the framework of IGRA studies, has dropped from the 77% to 70% reported by Pai et al. 1 Although five of the 20 studies included in the work of Pai et al 1 did not meet our inclusion criteria, the number of studies explicitly showing cases with culture or nucleic acid confirmation of M tuberculosis increased to 23. Although these tightened inclusion criteria seem to be the main explanation for this downgrading of TST sensitivity, it should be pointed out that only two of the 23 studies fulfilling our prerequisites came from developing countries, making a subgroup analysis impossible at this time. For IGRAs, sensitivity of T-Spot was highest, reaching a pooled value of 87.5% (95% CI, ). Because most of the T-Spot studies with respect to sensitivity were also performed in developed countries, a further differentiation between developed and less-developed countries resulted only in a small Figure 6. Forest plot of T-Spot. TB specificity. 958 Special Feature

8 Table 4 Indeterminate Results in Studies Using the QFT-IT Study/Year No. Subjects Type of Subjects Country No. Indeterminates % Indeterminates Pai et al 51 / Adults India Brock et al 52 / Adults Denmark Mahomed et al 53 / Adults South Africa Tsiouris et al 54 / Children South Africa Tsiouris et al 9 / Adults South Africa Cobanoglu et al 55 / Adults Turkey Dogra et al 56 / Children India Dominguez et al 28 / Adults Spain Drobniewski et al 57 / Adults Georgia Franken et al 49 / Adults The Netherlands Jones et al 58 / Adults United States Luetkemeyer et al 59 / Adults United States Manuel et al 60 / Adults Canada Matulis et al 61 / Adults Switzerland Nienhaus et al 62 / Adults Germany Pai et al 33 / Adults India Palazzo et al 22 / Adults Italy Silverman et al 63 / Adults Canada Baba et al 34 / Adults South Africa Balcells et al 64 / Adults Chile Bartalesi et al 65 / Adults Italy Bartu et al 15 / Adults Czech Republic Bocchino et al 66 / Adults Italy Carvalho et al 67 / Adults Italy Chee et al 35 / Adults Singapore Chegou et al 68 / Adults South Africa Connell et al 48 / Children Australia Diel et al 69 / Predominately adults Germany Diel et al 70 / Predominately adults Germany Harada et al 36 / Adults Japan Herrmann et al 71 / Adults France Hursitoglu et al 72 / Adults Turkey Katiyar et al 37 / Adults India Kipfer et al 73 / Adults Switzerland Mackensen et al 74 / Adults Germany Mirtskhulava et al 75 / Adults Georgia Okamba et al 76 / Adults France Perry et al 77 / Predominately adults United States Petrucci et al 78 / Children Nepal Ponce de Leon et al 79 / Adults Peru Raby et al 80 / Adults Zambia, South Africa Ruhwald et al 50 / Adults Western Europe Schoepfer et al 81 / Adults Switzerland Stebler et al 82 / Adults Switzerland Aichelburg et al 38 / Adults Austria Baker et al 83 / Predominately adults Canada Bergamini et al 84 / Children Italy Bianchi et al 16 / Children Italy Bruzzese et al 85 / Children Italy Chegou et al 39 / Adults Africa Cordero-Coma et al 86 / Adults Spain Dheda et al 87 / Adults South Africa Garfein et al 88 / Adults Mexico Gerogianni et al 18 / Adults Greece Kabeer et al 24 / Adults India Kampmann et al 25 / Children United Kingdom Kariminia et al 89 / Adults Iran Khanna et al 90 / Adults United Kingdom Kunimoto et al 91 / Adults Canada Lighter et al 40 / Children United States Lighter et al 92 / Children United States Machado et al 93 / Predominately adults Brazil (Continued) CHEST / 137 / 4 / APRIL,

9 Table 4 (Continued) Study/Year No. Subjects Type of Subjects Country No. Indeterminates % Indeterminates Markova et al 41 / Adults Bulgaria Miranda et al 94 / Adults United States Richeldi et al 95 / Adults Italy Ringshausen et al 96 / Adults Germany Rivas et al 97 / Adults Spain Sauzullo et al 42 / Adults Italy Schablon et al 98 / Adults Germany Talati et al 99 / Adults United States Triverio et al 100 / Adults Switzerland Vinton et al 101 / Adults Australia Totals 21, Pooled percentage of indeterminate results (95% CI, ). Heterogeneity x (degrees of freedom [ df ]5 71), P Inconsistency (I 2 ) %. No. studies See Table 1 for expansion of abbreviations. increase up to 88.5%. Pooled sensitivity from the 19 QFT-IT studies was 81%. This figure is remarkably different from the pooled sensitivity of 70% based on the six QFT-IT studies referenced by Pai et al. 1 When developing world results are removed from consideration, the pooled sensitivity estimate for QFT-IT increases to 84.5% (and 88.5% for T-Spot). The significantly lower QFT-IT sensitivity observed in resource-poor settings requires careful interpretation. Beside obvious and likely explanations relating to the immunologic status of patients in such settings (ie, HIV coinfection, advanced disease, or malnutrition), logistic requirements of upcoming studies should be examined in detail to allow a better definition of the reasons for this imbalance. Reasons for the apparent higher sensitivity of T-Spot are unclear, but may relate to the choice of cutoffs for the respective IGRAs. As previously pointed out by Dheda et al, 133 differences in IGRA sensitivity for active TB may be explained by the fact that the QFT-IT cutoff is drawn to achieve maximum specificity, whereas the commonly used European T-Spot cutoff of 6 spots appears to maximize sensitivity. The latter fact may have been taken into consideration by the US Food and Drug Administration, which weighed the balance of sensitivity and specificity of the T-Spot assay afresh by imposing a higher secure cutoff of 8 spots and introducing a gray zone of 5 to 7 spots. 134 It should be stressed that the sensitivity values reported in this manuscript do not reflect those that would have been obtained if the US cutoff criteria had been used, in which case a lower estimate would be expected. Comparing specificity of the two IGRAs, it is apparent that the price of the higher sensitivity of T-Spot is its reciprocal lower specificity in comparison with the QFT-IT. In our analysis, a specificity of 99% was found for the QFT-IT, significantly higher than the 86% found for T-Spot. Although it is likely that these values are representative, because of the low numbers of studies that could be included in estimating specificity the estimates should be interpreted with caution. Furthermore, because of the lack of a gold standard the selection of healthy, native-born subjects without any known TB exposure for investigating specificity of tests for M tuberculosis infection is only a surrogate procedure. There is a possibility that some individuals enrolled in the specificity studies cited in this metaanalysis had LTBI, despite their having no identified risk factors. Thus, in our view, more studies should be carried out to further investigate the specificity of the commercially available IGRAs. All in all, the results found in this new metaanalysis demonstrate that sensitivity of both IGRAs by far surpasses that of the TST. Although pooled estimates of all three screening tests are compromised by their high degree of heterogeneity, the superiority of the IGRAs over the TST is clearly evident also in the few studies with head-to-head comparisons between TST and QFT-IT or T-Spot. TB. Estimates of invalid (indeterminate) results in the two IGRAs have not yet been the subject of systematic investigation. Determining the risk of receiving those uninterpretable results, however, is important, especially as the assays may be useful in screening patients suffering from immunosuppressed conditions. In that context, it has frequently been suggested that use of IGRAs in immunosuppressed individuals is commonly associated with a high percentage of invalid results, and referring to the earlier liquid antigen version that especially the QFT-IT is hereby negatively affected. 60, However, the results of our analysis provide pooled evidence that the number of indeterminates among individuals tested with QFT-IT (2.14%) and by T-Spot (3.8%) is quite small, even if the analysis was limited to immunocompromised patients (QFT-IT 4.42% and T-Spot 6.12%). In stark contrast to the suggestions in many of the publications referred to above, the number of indeterminate responses by T-Spot was clearly higher than that found 960 Special Feature

10 Table 5 Indeterminate Results in Studies Using T-Spot. TB Study/Year No. Subjects Type of Subjects Country No. Indeterminates No. With Not Enough Cells % Indeterminates Dheda et al 102 / Adults United Kingdom 1 Unknown 2.08 Meier et al 32 / Adults Germany 1 Unknown 1.11 Zellweger et al 103 / Adults Switzerland 0 Unknown 0.00 Ferrara et al 5 / Predominately adults Italy 15 Unknown 3.82 Goletti et al 43 / Adults Italy 0 Unknown 0.00 Lee et al 6 / Predominately adults Korea 0 Unknown 0.00 Piana et al 104 / Adults Italy Arend et al 105 / Adults The Netherlands 23 Unknown 2.94 Clark et al 106 / Adults United Kingdom 9 Unknown 4.48 Detjen et al 10 / Children Germany 5 Unknown 6.85 Dominguez et al 28 / Predominately adults Spain 7 Unknown 1.12 Grimes et al 107 / Adults United States 0 Unknown 0.00 Hoffmann et al 108 / Adults Switzerland 8 Unknown 9.41 Janssens et al 109 / Adults Switzerland 7 Unknown 1.90 Kang et al 11 / Adults Korea 4 Unknown 2.84 Kim et al 12 / Adults Korea 2 Unknown 2.78 Leung et al 110 / Adults Hong Kong 6 Unknown 4.48 Passalent et al 111 / Adults Canada 14 Unknown 6.90 Piana et al 112 / Predominately adults Italy 21 Unknown 4.03 Porsa et al 113 / Adults United States 22 Unknown 4.92 Rangaka et al 114 / Adults Africa 1 Unknown 0.63 Vincenti et al 14 / Adults Italy 0 Unknown 0.00 Wang et al 44 / Predominately adults Taiwan 0 Unknown 0.00 Adams et al 115 / Adults United States 0 Unknown 0.00 Barsegian et al 116 / Adults Germany 0 Unknown 0.00 Beffa et al 117 / Predominately adults Switzerland Bocchino et al 65 / Adults Italy 4 Unknown 5.80 Brodie et al 118 / Predominately adults United States Chee et al 35 / Adults Singapore 4 Unknown 1.40 Connell et al 48 / Predominantly children Australia 14 Unknown Diel et al 70 / Predominately adults Germany Hesseling et al 119 / Predominately adults Africa 1 Unknown 1.28 Janssens et al 120 / Predominately adults Switzerland 15 Unknown 5.08 Kobashi et al 21 / Adults Japan 3 Unknown 6.25 Lee et al 121 / Adults Taiwan 0 Unknown 0.00 Mandalakas et al 122 / Children and adults Africa 2 Unknown 4.65 Soysal./et al 23 / Predominately adults Turkey 4 Unknown 2.72 Soysal./et al 123 / Children Turkey 5 Unknown 2.39 Stephan et al 124 / Adults Germany 8 Unknown 2.91 Vassilopoulos et al 125 / Adults Greece 0 Unknown 0.00 Behar et al 126 / Adults United States 0 Unknown 0.0 Bergamini et al 84 / Children Italy 3 Unknown 1.95 Bodenmann et al 127 / Adults Switzerland 5 Unknown 3.91 Bruzzese et al 85 / Children Italy 10 Unknown Dheda et al 87 / Adults South Africa 28 Unknown Higuchi et al 45 / Adults Japan Jiang et al 128 / Adults China Kampmann et al 25 / Children United Kingdom Kim et al 47 / Adults Korea 3 Unknown 1.65 Lindemann et al 129 / Adults Germany 0 Unknown 0.00 Markova et al 41 / Adults Bulgaria Martin et al 130 / Adults Ireland Nicol et al 27 / Children Africa Richeldi et al 95 / Adults Italy Rivas et al 97 / Adults Spain 1 Unknown 0.74 Storla et al 131 / Adults Norway Talati et al 99 / Adults United States Triverio et al 100 / Adults Switzerland Warier et al 46 / Children India 23 Unknown Totals 12, Pooled percentage of indeterminate results (95% CI, ). Heterogeneity x ( df 5 58) P Inconsistency (I 2 ) %. No. studies See Table 4 for expansion of abbreviation. CHEST / 137 / 4 / APRIL,

11 Table 6 Indeterminate Results Among Immunosuppressed Hosts Tested With QFT-IT Study/Year No. Subjects Type of Subjects Country No. Indeterminates % Indeterminates Brock et al 52 / HIV infected Denmark Cobanoglu et al 55 / Inflammatory disease Turkey Jones et al 58 / HIV infected United States Luetkemeyer et al 59 / HIV infected United States Manuel et al 60 / Chronic liver disease (end stage) Canada Matulis et al 61 / Inflammatory diseases Switzerland Silverman et al 63 / Bladder cancer Canada Balcells et al 64 / HIV infected Chile Baba et al 34 / HIV infected (plus five controls) South Africa Bartalesi et al 65 / Inflammatory diseases Italy Bocchino et al 66 / Inflammatory diseases Italy Hursitoglu et al 72 / Renal dialysis Turkey Ponce de Leon et al 79 / Rheumatoid arthritis Peru Raby et al 80 / HIV infected Zambia, South Africa Schoepfer et al 81 / Inflammatory bowel disease Switzerland Aichelburg et al 38 / HIV infected Austria Bruzzese et al 85 / Immunosuppressed children Italy Cordero-Coma et al 86 / Chronic uveitis immunosuppressed Spain Garfein et al 88 /2009 1,025 IV drug users Mexico Kabeer et al 24 / HIV infected India Markova et al 41 / HIV infected Bulgaria Richeldi et al 95 / HIV infected/other immunosuppressed Italy Talati et al 99 / HIV infected United States Triverio et al 100 / Renal dialysis Switzerland Totals 5, Pooled percentage of indeterminate results (95% CI, to 0.05). x ; df 5 23 ( P ). Inconsistency (I 2 ) %. See Tables 1 and 4 for expansion of abbreviations. when testing by QFT-IT in the pooled estimate studies, in the immunosuppressed, and in studies directly comparing the two tests ( P,.0001, P 5.001, and P,.0001, respectively). There is no doubt that indeterminate results for any of the IGRAs are largely the result of two different factors: a high level of immunosuppression and/or technical error in blood collection and handling or assay performance. The latter is likely the primary reason for the lower rate of indeterminate results with QFT-IT compared with T-Spot and the old second-generation, liquid antigen QFT test. Differences Table 7 Indeterminate Results Among Immunosuppressed Hosts Tested With T-Spot. TB Study/Year No. Subjects Type of Subjects Country No. Indeterminates No. With Not Enough Cells % Indeterminates Dheda et al 102 / HIV infected United Kingdom 1 Unknown 3.4 Ferrara et al 5 / Immunosuppressed Italy 15 Unknown 3.8 Piana et al 104 / Hematological diseases Italy Hoffmann et al 108 / HIV infected Switzerland 8 Unknown 9.4 Passalent et al 111 / Renal dialysis Canada 14 Unknown 6.9 Piana et al 112 / Hematological diseases Italy 12 Unknown 6.1 Rangaka et al 114 / HIV infected Africa Bocchino et al 66 / Inflammatory diseases Italy 4 Unknown 5.8 Lee et al 121 / Renal disease Taiwan 0 Unknown 0.0 Mandalakas et al 122 / HIV infected South Africa 2 Unknown 4.7 Vassilopoulos et al 125 / Rheumatology patients Greece 0 Unknown 0.0 Behar et al 126 / Inflammatory diseases United States 0 Unknown 0.0 Bruzzese et al 85 / Immunosuppressed children Italy 10 Unknown 13.5 Markova et al 41 / HIV infected Bulgaria Martin et al 130 / Inflammatory diseases Ireland Richeldi et al 95 / HIV infected/other immunosuppressed Italy Talati et al 99 / HIV infected United States Triverio et al 100 / Renal dialysis Switzerland Totals 2, Pooled percentage of indeterminate results (95% CI, ). x ; df 5 17 ( P ). Inconsistency (I 2 ) %. See Table 4 for expansion of abbreviation. 962 Special Feature

12 Table 8 Head-to-Head Comparison of Indeterminate Results Among Individuals Tested With QFT-IT and T-Spot. TB Study/Year No. Participants Type of Participants Country No. QFT-IT Indeterminates % QFT-IT Indeterminates No. T-Spot. TB Indeterminates % T-Spot. TB Indeterminates Boccchino et al 66 / Rheumatoid arthritis Italy Chee et al 35 / Active TB Singapore Connell et al 48 / Children TB suspects, Australia contacts, immigrants Diel et al 70 / Contacts Germany Dominguez et al 28 / TB suspects, contacts, Spain screening Kampmann et al 25 / Children TB suspects United and contacts Kingdom Markova et al 41 / HIV positive/tb Bulgaria suspects Richeldi et al 95 / HIV positive/other Italy immunosuppressed Rivas et al 97 / Detoxification patients Spain Talati et al 99 / HIV positive United States Triverio et al 100 / Renal dialysis Switzerland Totals 3, See Table 1 for expansion of abbreviations. between the methodologies of two versions of the QFT test probably account for much of the lower rate of indeterminate results for the in-tube format; lymphocytes encounter antigen and mitogen as soon as blood is collected into the QFT tubes, thus negating the possibility of loss of activity over the storage time of blood that is reported for the liquid antigen test. For the T-Spot test, the procedure is especially more demanding; the blood must be transported to the laboratory within 12 h, the lymphocytes isolated by density centrifugation, washed using further centrifugation steps, enumerated by manual counting, and adjusted to a set concentration before cells and antigens are manually added to the 96-well ELISPOT plates. All of these are steps in which technical errors can occur, as is the manual counting of the number of spots in each well at the end of the assay. Our metaanalysis has some limitations. Most studies (or the number of individuals fulfilling our strict inclusion criteria that were eligible out of a broader sample size) were small and the number of studies investigating specificity insufficient, which is partly because only the QFT-IT and T-Spot. TB as IGRAs were in the scope of this analysis. In addition, sensitivity of the TST relied only on the results of studies that also included the performance of any IGRA. However, this was justified because comparison of accuracy between TST and IGRAs is inherent in the scientific discussion and the inclusion of older studies on TST without any reference to IGRA results would be comparing apples with peaches. A further limitation not specifically of this metaanalysis, but of studies evaluating sensitivity of IGRAs and the TST in general, is the lack of a gold standard for LTBI. These tests are mainly used for screening people for LTBI, but sensitivity is derived from results of patients with confirmed active TB, an immunologically different condition. Although we would expect test sensitivity to be higher for individuals with LTBI (generally immunocompetent and controlling their infection) than those with active TB (often immunosuppressed), that assumption cannot currently be proven. We also would have liked to include a meaningful analysis of data from children aged less than 5 years, as they are of particular concern in TB control. Regrettably, there are only very few published studies covering this age group, and a metaanalysis of the limited data would not provide a statistically reliable statement on the true performance of the tests. As was already the case in the 2007 analysis of Pai et al, 1 the main limitation of our metaanalysis was the heterogeneity of the study results limiting the ability to make generalizations based on the resulting pooled estimates. Nevertheless, although the results of our analysis should be interpreted with some caution, the differences among the pooled yields are clear enough to justify serious consideration. Conclusions Our metaanalysis of the existing literature has produced results showing that IGRAs are superior to the TST for detecting confirmed active TB disease, irrespective of the degree of economic resources of the particular setting. This advantage appeared even greater for both QFT-IT and the T-Spot. TB when they were performed in developed countries. Although basic sensitivity of the T-Spot is consistently higher than that of the QFT-IT, its specificity is comparatively very low. Our review demonstrates that the frequency of invalid (indeterminate) estimates is CHEST / 137 / 4 / APRIL,

13 lower among subjects tested with QFT-IT than among those tested with T-Spot. TB, whether immunocompetent or immunocompromised. These findings support the primary use of IGRAs, especially in high-risk groups, and particularly for immunosuppressed individuals with high risk of TB reactivation, as has been stipulated in some current recommendations.139 Considering sensitivity for diagnosing active TB as a surrogate parameter for LTBI, TST-based two-step screening strategies (TST first, IGRA second) for contact tracing should be critically reconsidered because of the poor TST accuracy among patients with TB. Acknowledgments Financial/nonfinancial disclosures: The authors have reported to CHEST the following conflicts of interest: Dr Diel received 1,000 for speaking at a conference sponsored by Cellestis and 750 for speaking at a conference sponsored by Oxford Immunotec. Drs Loddenkemper and Nienhaus have reported that no potential conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article. References 1. Pai M, Zwerling A, Menzies D. Systematic review: T-cellbased assays for the diagnosis of latent tuberculosis infection: an update. Ann Intern Med ;149(3): Lalvani A. Counting antigen-specific T cells: a new approach for monitoring response to tuberculosis treatment? Clin Infect Dis ;38(5): Mori T, Sakatani M, Yamagishi F, et al. Specific detection of tuberculosis infection: an interferon-gamma-based assay using new antigens. Am J Respir Crit Care Med ;170(1): Kang YA, Lee HW, Yoon HI, et al. Discrepancy between the tuberculin skin test and the whole-blood interferon gamma assay for the diagnosis of latent tuberculosis infection in an intermediate tuberculosis-burden country. JAMA ;293(22): Ferrara G, Losi M, D Amico R, et al. Use in routine clinical practice of two commercial blood tests for diagnosis of infection with Mycobacterium tuberculosis: a prospective study. Lancet ;367(9519): Lee JY, Choi HJ, Park IN, et al. Comparison of two commercial interferon-gamma assays for diagnosing Mycobacterium tuberculosis infection. Eur Respir J ;28(1): Jafari C, Ernst M, Kalsdorf B, et al. Rapid diagnosis of smear-negative tuberculosis by bronchoalveolar lavage enzyme-linked immunospot. Am J Respir Crit Care Med ;174(9): Kobashi Y, Obase Y, Fukuda M, Yoshida K, Miyashita N, Oka M. Clinical reevaluation of the QuantiFERON TB-2G test as a diagnostic method for differentiating active tuberculosis from nontuberculous mycobacteriosis. Clin Infect Dis ;43(12): Tsiouris SJ, Coetzee D, Toro PL, Austin J, Stein Z, El-Sadr W. Sensitivity analysis and potential uses of a novel gamma interferon release assay for diagnosis of tuberculosis. J Clin Microbiol ;44(8): Detjen AK, Keil T, Roll S, et al. Interferon-gamma release assays improve the diagnosis of tuberculosis and nontubercu- lous mycobacterial disease in children in a country with a low incidence of tuberculosis. Clin Infect Dis ;45 (3 ): Kang YA, Lee HW, Hwang SS, et al. Usefulness of wholeblood interferon-gamma assay and interferon-gamma enzyme-linked immunospot assay in the diagnosis of active pulmonary tuberculosis. Chest ;132 (3 ): Kim SH, Choi SJ, Kim HB, Kim NJ, Oh MD, Choe KW. Diagnostic usefulness of a T-cell based assay for extrapulmonary tuberculosis. Arch Intern Med ;167 (20 ): Mazurek GH, Weis SE, Moonan PK, et al. Prospective comparison of the tuberculin skin test and 2 whole-blood interferon-gamma release assays in persons with suspected tuberculosis. Clin Infect Dis ;45 (7 ): Vincenti D, Carrara S, Butera O, et al. Response to region of difference 1 (RD1) epitopes in human immunodeficiency virus (HIV)-infected individuals enrolled with suspected active tuberculosis: a pilot study. Clin Exp Immunol ;150 (1 ): Bartu V, Havelkova M, Kopecka E. QuantiFERON-TB Gold in the diagnosis of active tuberculosis. J Int Med Res ;36 (3 ): Bianchi L, Galli L, Moriondo M, et al. Interferon-gamma release assay improves the diagnosis of tuberculosis in children. Pediatr Infect Dis J ;28 (6 ): Dosanjh DP, Hinks TS, Innes JA, et al. Improved diagnostic evaluation of suspected tuberculosis. Ann Intern Med ;148 (5 ): Gerogianni I, Papala M, Klapsa D, Zinzaras E, Petinaki E, Gourgoulianis KI. Whole-blood interferon-gamma assay for the diagnosis of tuberculosis infection in an unselected Greek population. Respirology ;13 (2 ): Jafari C, Thijsen S, Sotgiu G, et al. Bronchoalveolar lavage enzyme-linked immunospot for a rapid diagnosis of tuberculosis: A TBNET Study. Am J Respir Crit Care Med ;180(7): Kobashi Y, Mouri K, Yagi S, et al. Clinical utility of the QuantiFERON TB-2G test for elderly patients with active tuberculosis. Chest ;133 (5 ): Kobashi Y, Mouri K, Yagi S, et al. Clinical evaluation for diagnosing active TB disease and transitional change of two commercial blood tests. Scand J Infect Dis ; 40 ( 8 ): Palazzo R, Spensieri F, Massari M, et al. Use of wholeblood samples in in-house bulk and single-cell antigenspecific gamma interferon assays for surveillance of Mycobacterium tuberculosis infections. Clin Vaccine Immunol ;15 (2 ): Soysal A, Torun T, Efe S, Gencer H, Tahaoglu K, Bakir M. Evaluation of cut-off values of interferon-gamma-based assays in the diagnosis of M. tuberculosis infection. Int J Tuberc Lung Dis ;12 (1 ): Syed Ahamed Kabeer B, Sikhamani R, Swaminathan S, Perumal V, Paramasivam P, Raja A. Role of interferon gamma release assay in active TB diagnosis among HIV infected individuals. PLoS One ;4 (5 ):e Kampmann B, Whittaker E, Williams A, et al. Interferongamma release assays do not identify more children with active tuberculosis than the tuberculin skin test. Eur Respir J ;33 (6): Kobashi Y, Mouri K, Yagi S, Obase Y, Miyashita N, Oka M. Clinical utility of a T cell-based assay in the diagnosis of extrapulmonary tuberculosis. Respirology ;14 (2 ): Nicol MP, Davies MA, Wood K, et al. Comparison of T-SPOT.TB assay and tuberculin skin test for the evaluation of young children at high risk for tuberculosis in a community setting. Pediatrics ;123 (1 ): Domínguez J, Ruiz-Manzano J, De Souza-Galvão M, et al. Comparison of two commercially available gamma interferon 964 Special Feature