|Home | About | Journals | Submit | Contact Us | Français|
In late 2006, our hospital implemented use of the QuantiFERON-TB Gold (QFT-G) assay, a whole-blood interferon-γ release assay, for detection of tuberculosis infection. All newly hired healthcare workers (HCWs) with positive Mantoux tuberculin skin test (TST) results were routinely tested with the QFT-G assay, to take advantage of its higher specificity. We then undertook a quality assurance review to evaluate the QFT-G test results in HCWs with multiple risk factors for latent tuberculosis infection (LTBI).
The clinical records for TST-positive HCWs tested with the QFT-G assay were reviewed. HCWs with 2 or more risk factors commonly associated with LTBI were classified as “increased risk” (IR). IR HCWs who had negative QFT-G test results underwent repeat QFT-G testing and were offered testing with a different interferon-γ release assay (T-SPOT.TB) and with extended T cell stimulation assays.
Of 143 TST-positive HCWs tested with the QFT-G assay, 26 (18%) had positive results, 115 (81%) had negative results, and 2 (1%) had indeterminate results. Of 82 IR HCWs, 23 (28%) had positive QFT-G test results, and 57 (70%) had negative results. Of the 57 IR HCWs with negative results, 43 underwent repeat QFT-G testing: 41 had negative results again, and 2 had positive results. These 43 HCWs were also offered additional testing with the T-SPOT.TB diagnostic, and 36 consented: 31/36 tested negative, and 5/36 tested positive. Extended assays using the antigens ESAT-6 and CFP-10 confirmed the positive results detected by the overnight assays and yielded positive results for an additional 7/36 (19%) of individuals; strikingly, all 36 HCWs had strongly positive test results with assays using purified protein derivative.
The extreme discordance between the results of our clinical diagnostic algorithm and the results of QFT-G testing raises concern about the sensitivity of the QFT-G assay for detection of LTBI in our HCWs. Results of extended stimulation assays suggest that many of our IR HCWs have indeed been sensitized to Mycobacterium tuberculosis. It is possible that the QFT-G assay identifies those at higher reactivation risk rather than all previously infected, but, in the absence of long-term follow-up data, we should interpret negative QFT-G results with some caution.
In the United States and other countries with a low prevalence of tuberculosis, healthcare workers (HCWs) constitute one of the largest populations of individuals for whom routine screening for latent tuberculosis infection (LTBI) is generally recommended. The Centers for Disease Control and Prevention (CDC) has stated that QuantiFERON-TB Gold (QFT-G; Cellestis), one of the newly developed interferon (IFN)– γ release assays (IGRAs) for diagnosis of tuberculosis infection, “may be used in all circumstances in which the TST [tuberculin skin test] is currently used.”1(p49) This recommendation for use of the QFT-G assay was based in part on its high specificity and an expectation of similar sensitivity to the TST in healthy individuals. However, the authors of these guidelines left open the possibility that “QFT-G sensitivity for LTBI might be less than that of the TST,”1(p52) while acknowledging that the lack of a confirmatory test would make this difficult to assess. Accordingly, the CDC recommended that “each QFT-G result and its interpretation should be considered in conjunction with other epidemiologic, historic, physical, and diagnostic findings.”1(p52)
In late 2006, our hospital began routinely testing all newly hired TST-positive HCWs (ie, those with an induration of 10 mm or more in size) with the QFT-G assay, to aid in the diagnosis of LTBI. This hospital testing algorithm was developed to take advantage of the increased specificity of the QFT-G assay versus TST, particularly in HCWs with a history of bacille Calmette-Guérin (BCG) vaccination,2 and mirrors the most recent UK recommendations.3 However, a preliminary review of initial testing results suggested major discordance between traditional risk factors for tuberculosis (including large TST reactions) and QFT-G test results. Therefore, a detailed quality assurance review was performed, which included repeat QFT-G testing, and parallel research testing with a different IGRA (T-SPOT.TB; Oxford Immunotec) was performed. Additional in vitro extended proliferation and IFN-γ assays using both purified protein derivative (PPD) and the antigens used in the IGRAs (CFP-10 and ESAT-6) were performed in parallel to further investigate the cause of the observed discordance.
In this article, we present what is, to our knowledge, the first detailed comparison of QFT-G test results with composite clinical assessment for LTBI (based on standard screening and treatment guidelines4,5) in US HCWs, specifically those identified as being TST positive during routine preemployment screening for tuberculosis.
The Employee and Occupational Health Services (EOHS) of Beth Israel Deaconess Medical Center (BIDMC) screens approximately 3,000 newly hired employees each year with the TST, approximately 15% of whom are found to be TST positive. In October 2006, the QFT-G assay was introduced for routine testing of new employees. The following testing algorithm was approved for routine, nonresearch use by the EOHS.
A baseline TST was done for all newly hired employees according to BIDMC/CDC guidelines.5 A TST (5 tuberculin units) was placed intradermally, and the result read 48 –72 hours later by experienced EOHS nurses. Two-step testing was performed when indicated.5 All newly hired employees who were TST-positive (ie, induration 10 mm or more in size) were tested with the QFT-G assay, typically on the day the TST result was read. New employees with a documented history of a positive TST result were tested with the QFT-G assay without undergoing TST; this group comprised approximately 5% of the TST-positive employees. TST-negative employees and employees who had ever received therapy for 1 month or longer for active tuberculosis or LTBI were excluded from QFT-G testing. All newly hired TST-positive employees underwent chest X-ray, which was interpreted routinely by BIDMC radiology staff. The EOHS required that each employee also complete a basic tuberculosis screening questionnaire, which included information about the employee’s prior TST results, country of origin, immigration, BCG vaccination history, tuberculosis contacts, prior treatment, and chronic medical conditions. Employees who had a positive QFT-G test result were referred to the Infectious Diseases Clinic of the hospital.
Prompted by the preliminary observation of discordance between QFT-G results and standard EOHS LTBI risk assessment, we undertook a quality assurance review. A retrospective review of all clinical records for employees tested with the QFT-G assay, for quality assurance purposes and for identification of subjects for further research testing, was approved by the BIDMC institutional review board. Written informed consent was obtained for testing with the T-SPOT.TB and extended in vitro assays. Available clinical records for all 143 HCWs tested with the QFT-G assay during the period from October to December 2006 were reviewed. We based our assessment of LTBI risk on standard clinical guidelines,4,5 such that we expect that any HCW classified as “increased risk” (IR) by our criteria would unequivocally be offered LTBI therapy at any EOHS setting in the United States (barring any contraindication).
Thus, HCWs who had at least 2 of the following 6 risk factors were considered to be in the IR group: (1) TST induration 15 mm or more in size; (2) abnormal chest X-ray findings consistent with prior tuberculosis; (3) direct daily patient care for at least 1 year, either past or current; (4) long-term residence (5 years or more) in an area where tuberculosis is highly endemic (incidence rate, 40 cases or more per 100,000 persons)6; (5) close contact at any time in the past, in a social or professional context, with a person with known active tuberculosis, without use of airborne precautions; or (6) documented PPD conversion (ie, an increase in induration of 10 mm or more between the first and the second TST, as performed as part of a contact investigation, an annual screening test, or another screening testing, but not as part of a “2-step” procedure). If there was any uncertainty in the PPD conversion data, the employee needed to have 2 additional risk factors to be classified as IR. To avoid misclassification, HCWs who interacted with known tuberculosis patients but wore masks, or who interacted with patients on a regular basis (some of whom could have had undiagnosed tuberculosis), were considered to have had “casual,” not “close,” contact.
Of the 143 HCWs tested with the QFT-G assay, 28 (20%) had clinical records that were insufficient for risk classification, and they were not reachable for clarification. Therefore, they were excluded from further assessment (Figure 1). Because information on the chronic medical conditions of HCWs was often incomplete on the EOHS questionnaire, it was not included in our review.
IR, QFT-G negative employees (57 of the 82 IR HCWs) were asked to present during the period from February to April 2007 for repeat QFT-G testing. Of these 57 HCWs, 45 (79%) responded (Figure 1); 2 were excluded from further testing, because one newly reported a history of LTBI therapy in childhood and the other had not had a close tuberculosis contact as previously reported (and was reclassified as being at baseline risk). Of the remaining 43 HCWs who underwent repeat QFT-G testing, 36 (84%) consented to concurrent T-SPOT.TB and extended in vitro assays (Figure 1). Only 1 HCW reported any potential tuberculosis contact since the initial QFT-G testing.
Both QFT-G and T-SPOT.TB assays were performed according to the manufacturers’ recommendations.7–9 For QFT-G, an IFN-γ concentration of 0.35 IU/mL in response to either ESAT-6 or CFP-10, relative to the negative control value, was considered positive. For T-SPOT.TB, a sample was considered positive if either the ESAT-6 or CFP-10 spot count, minus the nil control spot count, was 6 or more. For the QFT-G assay, any indeterminate result prompted new sample submission; there were no indeterminate T-SPOT.TB results. The initial and repeat QFT-G tests were done by different technicians. A 4-tube standard curve was used (with 0, 0.25, 1.0, and 4.0 IU/mL of IFN-γ, as recommended in the “QFT-G In-Tube” package insert10) rather than an 8-tube standard curve, as specifically directed by the manufacturer’s technical support team that supervised assay implementation.
Assays were performed essentially as described elsewhere.11 Peripheral blood mononuclear cells were incubated (2 × 105 cells per well) in medium only or in medium containing either PPD (2 μg/mL), ESAT-6 (2 μg/mL), or CFP-10 (5 μg/mL). Plates were cultured for 5 days, pulsed with [3H] thymidine for an additional 18 hours, harvested, and counted. The proliferation index was defined as the ratio of counts per minute in the test well compared with that in the negative control well (medium). The levels of supernatant IFN-γ (obtained 72 hours after culture initiation) were analyzed using a sandwich enzyme-linked immunosorbent assay. A positive result was defined as an IFN-γ concentration of more than 50 pg/mL (all negative control wells had concentrations of essentially zero). PPD for tissue culture assays was prepared as described elsewhere.12 Recombinant full-length ESAT-6 and CFP-10 (supplied by Dr. John Beslile and Dr. Karen Dobos, Colorado State University [Tuberculosis Research Materials contract 1-A125174 from National Institute of Allergy and Infectious Diseases, National Institutes of Health]) were assayed for lipopolysaccharide contamination using a limulus amoebocyte lysate assay; lipopolysaccharide was present at concentrations of 0.08 ng per 1 mg protein and 4.5 ng per 1 mg protein, respectively.
We compared IR, QFT-G negative HCWs with IR, QFT-G positive HCWs. Means for TST size and age were compared by the Student t test, and distributions of dichotomous risk factors were compared by the χ2 test, using Stata/SE, version 7.0 (Stata).
During the period from October to December 2006, 143 newly hired TST-positive HCWs with no history of tuberculosis treatment had been routinely tested with the QFT-G assay, and all had undergone chest X-ray. Approximately 95% were foreign born, and 93% reported a history of BCG vaccination, although few knew at what age they had received it. The 143 HCWs had varying degrees of patient contact, ranging from daily contact (eg, doctors, nurses, physical therapists, nurses’ aides, interpreters, patient-transport personnel) to less contact (eg, food service and custodial personnel) to essentially no contact (researchers). Of the 143 TST-positive HCWs who underwent QFT-G testing, 26 (18%) tested positive (Figure 1).
Of the 143 HCWs who underwent QFT-G testing, 115 (80%) had sufficient risk factor data available to allow LTBI risk classification. Of these 115 HCWs, 82 (71%) were classified as IR on the basis of our criteria. Of these 82 IR HCWs, 57 (70%) tested negative with the QFT-G assay (Figure 1). The distribution of clinical risk factors used for classifying these 57 as IR is shown in Table 1. Of note, ESAT-6 and CFP-10 responses in the QFT-G assays for these 57 HCWs were negligible overall, with only 4 (7%) of the 57 HCWs having responses (ESAT-6-nil or CFP-10-nil) of more than 0.20 IU/mL (0.20, 0.21, 0.27, and 0.34 IU/mL, respectively); the remaining 53 HCWs’ responses were much lower or zero. The mean mitogen-nil response (± standard deviation [SD]) was 25.75 ± 22.29 IU/mL, and the background (nil) response was uniformly low. Of the 115 employees for whom complete clinical data were available, 24 (21%) tested positive with the QFT-G assay, and 23 (96%) of the 24 were considered IR by our criteria (the 1 remaining HCW had a 24-mm TST but was from Iran, and thus did not meet our criteria for IR).
At this stage in our analysis, the high proportion of negative QFT-G test results (57 [70%] of 82 HCWs) in the group with highest clinical pretest probability of LTBI raised concerns about the sensitivity of the test in this population. To eliminate the possibility of laboratory error, the 57 IR, QFT-G negative HCWs were asked to present for repeat QFT-G testing; 43 (75%) of the 57 received repeat testing in the period from February to April 2007 (Figure 1). Only 1 of the 43 HCWs (who was given the identification number R-17) reported any tuberculosis contact after the first QFT-G assay (and followed airborne precautions). Forty-one of 43 HCWs had negative QFT-G test results on repeat testing, and 2 had positive results, one by ESAT-6-nil = 0.38 (R-25; initial value 0.13) and the other by CFP-10-nil = 0.37 (R-26; initial value 0.00) (Table 2).
At the time of QFT-G retesting, the 43 HCWs were offered additional research testing with T-SPOT.TB and extended in vitro assays measuring IFN-γ production (at 3 days) and proliferation (at 5 days) in response to ESAT-6, CFP-10, and PPD. Of these 43 HCWs, 36 (84%) consented to additional testing; 5 (14%) of the 36 tested positive with T-SPOT.TB (Figure 1 and Table 2). Of these 5 HCWs, 4 tested negative on their repeat QFT-G test, and 1 tested positive (the other HCW who tested positive on repeat QFT-G test did not consent to T-SPOT.TB testing). Of the 36 HCWs who consented to additional testing, 12 (33%) had positive results of proliferation assays using ESAT-6 and CFP-10; a positive test result was defined as a proliferation index of 5 or more times the negative control value (Table 2). Strikingly, all 36 HCWs had positive results of assays using PPD as the stimulation antigen (most responses were overwhelmingly high, with a proliferation index of more than 100). All patients who tested positive with either the QFT-G assay (2 HCWs, identified as R-25 and R-26) or T-SPOT.TB (5 HCWs, identified as R-8, R-13, R-25, R-31, and R-33) also tested positive with a proliferation assay using ESAT-6 or CFP-10, if performed. There were 7 HCWs (R-16, R-17, R-30, R-32, R-34, R-36, and R-38) who tested positive with the proliferation assay but not with either the QFT-G assay or T-SPOT.TB. Extended assays measuring IFN-γ response to ESAT-6 and CFP-10 (Table 2) were positive for only 2 (6%) of the 36 HCWs (R-31 and R-32), whereas all 36 HCWs tested positive in response to PPD. There was no correlation between any of the 6 clinical risk factors used for LTBI risk assessment and positive test results for any of the assays used.
Because, on an individual level, there did not appear to be any way to predict which IR HCWs would test negative and which positive with the QFT-G assay, aggregate characteristics of the 2 groups (ie, the 43 IR HCWs who were QFT-G negative and the 23 IR HCWs who were QFT-G positive) were compared (Table 3). Notably, there was no significant difference in the average TST size between the 2 groups (mean ± SD, 16.3 ± 3.2 mm for QFT-G negative vs 17.4 ± 3.9 mm for QFT-G positive; P = .28), and both mean values were greater than 15 mm, which per guidelines has historically been the cutoff for considering a TST result to be a “true positive.”5 There was also no difference in mean age between the 2 groups. All 43 HCWs who were QFT-G negative reported having been vaccinated with the BCG vaccine, and 21 (91%) of the 23 HCWs who were QFT-G positive reported having been vaccinated. HCWs who were QFT-G positive were significantly more likely to have spent 5 years or more in a high-risk area (P = .038) and were also more likely to have recently immigrated (ie, within 5 years) from a high-risk area (P = .069). In contrast, there was no significant difference (P = .105) between the 2 groups in the proportion of HCWs who reported casual or close contact with tuberculosis patients.
We have found extreme discordance between the results of a clinical diagnostic algorithm and the results of QFT-G testing for diagnosis of LTBI in our HCWs. Specifically, of 82 TST-positive HCWs (most of whom had a TST induration size of 15 mm or more) who had increased pretest probability of LTBI according to clinical criteria that were based on standard and widely accepted guidelines4,5—all of whom would accordingly have been offered treatment according to preexisting EOHS protocol—57 (70%) tested negative with the QFT-G assay. A repeat QFT-G test confirmed these negative test results in 41 (95%) of 43 HCWs tested, and a different IGRA, T-SPOT.TB, generated only 5 positive test results among the 36/43 HCWs who received this additional testing. We could reasonably conclude either that our clinical algorithm incorrectly identified these HCWs as being at increased risk of having LTBI when they actually do not have LTBI, or that IGRAs are insensitive for diagnosing LTBI in our population. (Of note, a recent study of the use of the QFT-G assay in US Navy recruits13 also questioned the sensitivity of the QFT-G assay for a subset of recruits who were born in countries with high tuberculosis prevalence and who had TST induration sizes of 15 mm or more.) However, a third, intriguing hypothesis would be that our clinical algorithm identifies all persons with LTBI (both recent and remote), whereas IGRAs detect a specific subset of that group—perhaps those at highest risk of reactivation. The fact that the CDC has stated that the QFT-G assay can be used in place of the TST for screening of HCWs and other populations1 makes it imperative that we consider these very different conclusions.
It is taught that tuberculosis reactivation risk is concentrated in the first 2 years after infection.14–16 Certainly there is also no doubt of the potential for remote infection to reactivate in individuals with immunosuppression or with other well-known medical risk factors.4 In contrast, there is little guidance regarding the risk of reactivation of remote infection in an immunocompetent, otherwise healthy individual. Notably, at least half of the TB cases among foreign-born persons occur more than 5 years after arrival.17
If our clinical criteria did indeed identify a population with a high prevalence of LTBI, one explanation for the discordance between the TST and IGRA results seen in our study could be that the 2 antigens (CFP-10 and ESAT-6) used in the IGRAs are insufficient for comprehensive detection of LTBI, compared with the complex mix of antigens in the PPD. However, in contact investigations, the IGRAs have appeared to detect recent tuberculosis infections with a sensitivity at least equal to that of the TST and to have better correlation with gradient of exposure to Mycobacterium tuberculosis.18–24 Although the latter has been ascribed to higher specificity of the IGRAs, some of the excess positive TST results could also be due to remote tuberculosis infections. Studies of the sensitivity of IGRAs for detection of active tuberculosis likewise suggest sensitivity similar to that of the TST.23 The integration of this data with ours might suggest that the IGRAs are in fact quite sensitive for detection of recently acquired LTBI, but less so for remotely acquired LTBI.
Of course, an alternative explanation for the observed discordance could be that our IR HCWs are TST positive only because of remote BCG vaccination, rather than because of LTBI.2 However, the fact that all of our IR, QFT-G negative HCWs (36/36) reacted very strongly to PPD in extended (3-and 5-day) in vitro T cell stimulation assays, and that 12 (33%) of the 36 reacted strongly to ESAT-6 and CFP-10 in these assays (including 7 HCWs who tested negative with both the QFT-G assay and T-SPOT.TB), could suggest that many of our IR HCWs were indeed infected with tuberculosis at some point in the past. Such a difference in “sensitivity” between overnight and longer-incubation IGRAs in individuals with suspected LTBI has been noted before.25,26 However, extended stimulation with ESAT-6 and CFP-10 only partially corrected the discordance we observed between the results of commercial IGRAs and the TST, and this may have a number of explanations. First, we would not necessarily expect ESAT-6 and CFP-10 alone to provide 100% sensitivity for detection of LTBI even in extended assays, given the limitations of the sensitivity of IGRAs using ESAT-6 and CFP-10 for detection of active tuberculosis. Second, it may be that waning of immunologic memory in the context of remote (or cleared) infection would result in decreased response to ESAT-6 and CFP-10, even in extended assays. Third, it is possible that not all individuals have the immunologic recognition repertoire to respond to ESAT-6 or CFP-10, even if they do have LTBI; this concept is supported by data from mouse models.27,28 Finally, 14 (39%) of the 36 samples were not tested with ESAT-6 in extended assays due to unavailability of the antigen, and some of these samples might have generated additional positive test results. We do not have an explanation for why we had more positive test results with the proliferation assays than with the extended IGRAs when ESAT-6 and CFP-10 were used, in contrast to the strong responses in both assays to PPD. Of note, it is unlikely that response to ESAT-6 and CFP-10 in extended IGRAs was due to nonspecific stimulation by lipopolysaccharide, as very low levels of lipopolysaccharide were present and only 2 of the 36 samples tested positive with the extended IGRAs.
Unfortunately, previous studies of the performance of the QFT-G assay in Japanese, Indian, and Russian HCWs29–31 do not inform the interpretation of our results, partly because their HCW populations (unlike ours) were essentially ethnically homogenous and overall rates of QFT-G positivity varied widely with study location. Notably, many of our HCWs have previously worked in healthcare facilities in areas where tuberculosis is endemic.
We recognize that some limitations inherent to the TST may have contributed to the misclassification of some of our HCW as IR. First, for a minority of positive TST results, we had to rely on documentation from outside readers. We also did not know the timing of BCG vaccination (eg, infancy or childhood) for most of our HCWs, so cannot assess its potential contribution to TST size.2
American Thoracic Society guidelines state that “targeted tuberculin testing programs should be designed for one purpose: to identify persons at high risk for TB who would benefit by treatment of LTBI.”4(pS233) Perhaps, then, our focus in screening programs (in nonimmunocompromised populations) should not be on the identification of any tuberculosis infection, but rather of those with higher reactivation risk.32 If the IGRAs are indeed able to distinguish these individuals, then using IGRAs for tuberculosis screening could focus our attention on a group most worthy of treatment and surveillance. This is supported by our finding that QFT-G positive IR HCWs were more likely to have spent 5 years or more in an area where tuberculosis is highly endemic and to have recently immigrated, suggesting that they could have been more recently infected. However, our findings show that if we base our treatment decisions on IGRA results alone, a sizable proportion of individuals with clinical risk factors historically considered suggestive of true LTBI will suddenly be exempt from treatment that we otherwise would have recommended—and some of these risk factors have historically been associated with increased reactivation risk.
It is likely that, by using established US clinical guidelines, we have been historically overidentifying individuals as “increased LTBI risk” and thus overtreating. However, there are many data underlying our assessment of the positive predictive value of the TST for the risk of development of active tuberculosis,33–35 and moreover, the TST has been used to define LTBI in studies that show the benefits of LTBI treatment.15,36 The critical question is whether LTBI, if not detected by an IGRA, could still reactivate, especially in patients with immunosuppression and other medical risk factors. At the current time, studies of the positive predictive value of IGRAs are few (eg, see Richeldi et al.37 and Doherty et al.38), and studies of their negative predictive value are in a relatively early stage (reviewed in Andersen et al.39).
In summary, the marked difference between the performance of the QFT-G assay in our IR HCWs, many of whom may have remote TB infection, and its performance in the contact setting suggests that the test may perform quite differently in these 2 common testing circumstances. This contrast may be telling us that the TST, which appears to be sensitive to both recent and remote infection, may be greatly overestimating the very high risk subpopulation that we need to target for LTBI treatment. When screening HCWs for LTBI, the finding of fewer positive test results with IGRAs may be their strength, rather than their weakness. However, in the absence of the long-term follow-up data necessary to definitively answer these questions, we feel that we cannot at this time replace the TST with the QFT-G assay for screening of our HCWs, for fear of missing IR, QFT-G negative individuals who may be at higher risk of tuberculosis reactivation. In the frustrating absence of a gold standard for the diagnosis of LTBI, our data suggest that, for now, we must continue to interpret negative QFT-G results with some caution.
We thank the BIDMC EOHS team for their efforts in setting up QFT-G testing and for their flexible response to the data obtained in this study. We also thank the members of the Massachusetts Committee for Elimination of Tuberculosis for their review and discussion of the data.
Financial support. QuantiFERON-TB Gold kits were provided by Cellestis; T-SPOT.TB kits were provided by Oxford Immunotec. The project was supported by a grant from the Pittsfield Anti-Tuberculosis Association (to N.P.). N.P. is supported by the Clinical Investigator Training Program, which is a cooperative effort between the BIDMC and the Harvard–Massachusetts Institute of Technology Division of Health Sciences and Technology, in collaboration with Pfizer and Merck & Co.
Potential conflict of interest. E.N. has served as a past consultant to Oxford Immunotec. S.M.B. has received technical support from Oxford Immunotec for performance of T-SPOT.TB assays in his laboratory. All other authors report no conflicts of interest relevant to this article.