OUP user menu

QuantiFERON-TB GOLD ELISA assay for the detection of Mycobacterium tuberculosis-specific antigens in blood specimens of HIV-positive patients in a high-burden country

Chrisna Veldsman, Marleen M. Kock, Theresa Rossouw, Martin Nieuwoudt, Mark Maeurer, Anwar A. Hoosen, Marthie M. Ehlers
DOI: http://dx.doi.org/10.1111/j.1574-695X.2009.00610.x 269-273 First published online: 1 December 2009


Tuberculosis is a life-threatening infection worldwide. Despite improvements in therapy, it results in 2 million deaths and 9 million new cases annually. This study evaluated the use of the QuantiFERON-TB GOLD enzyme-linked immunosorbent assay in a high HIV/TB burden setting in an ARV clinic at the Tshwane District Hospital, South Africa. The sensitivity and specificity of the QF assay in the clinic were 30% (9/30) and 63% (19/30), respectively, when compared with the gold standard culture results. Analysis also suggested that the sensitivity of the QuantiFERON assay is determined by a limiting patient CD4 value between 150 and 200.

  • tuberculosis
  • HIV-positive patients


Until recently, the tuberculin skin test (TST) was the only available screening method for tuberculosis (TB) and is still predominantly used for the diagnosis of latent tuberculosis infection (Bakir et al., 2009). However, the status of a ‘gold standard’ may be challenged due to variability in interpretation and problems with false-positive and -negative results (Kariminia et al., 2009). The TST is a measure of a delayed-type hypersensitivity response to a purified protein derivative (PPD) (Eum et al., 2008). A PPD is a mixture of mycobacterial antigens, some of which are shared between nontuberculous mycobacteria (NTM), Mycobacterium bovis Bacille Calmette–Guerin (BCG) vaccine strains and Mycobacterium tuberculosis (Eum et al., 2008). The BCG vaccination of children at birth responds immunologically to PPD prepared from M. tuberculosis (Kariminia et al., 2009). The PPD has been used to support the diagnosis of tuberculosis in the clinic as well as for screening in national programmes and epidemiological studies (Andersen et al., 2000).

The greatest disadvantage of PPD is that most protein components in this substance are shared between mycobacterial species or with unrelated species of bacteria, which decreases the specificity of the TST significantly (Kariminia et al., 2009). Another reason for the decrease in specificity is due to prior exposure to NTM (Kariminia et al., 2009). The sensitivity of the test may be low in individuals with a compromised immune function (Eum et al., 2008). Sputum microscopy for acid-fast bacilli is often negative in an HIV-positive population with end-stage disease and the ‘gold standard’, sputum culture, is time-consuming and can lead to a delay in tuberculosis diagnosis (Andersen et al., 2000). Limitations such as these emphasize the need for antigen assays specific to M. tuberculosis for the diagnosis of tuberculosis, and such alternatives have been recognized and are currently being investigated (Andersen et al., 2000; Mackensen et al., 2008).

One such method is the QuantiFERON-TB GOLD In-Tube enzyme-linked immunosorbent assay (ELISA) (QFT) (Cellestis, UK), an assay that measures the release of interferon-γ (IFN-γ) in whole blood after stimulation by specific tuberculosis antigens, for example early-secreted antigenic target-6 (ESAT-6) and culture filtrate protein-10 (CFP-10) (Eum et al., 2008; Katiyar et al., 2008). IFN-γ is released from sensitized lymphocytes into whole blood and incubated overnight with a PPD from M. tuberculosis and control antigens (Kariminia et al., 2009). The test is based on DNA strip technology and was approved by the Food and Drug Administration in 2001 as an aid for detecting the M. tuberculosis complex (Pai et al., 2009). This test is attractive because the ESAT-6 and CFP-10 are absent from all BCG vaccine strains and from commonly encountered NTM, except Mycobacterium kansasii, Mycobacterium szulgai and Mycobacterium marinum (Kariminia et al., 2009). Further advantages of the assay are that the test is simple to perform, the results can be obtained within 24 h, it takes only one patient visit and it has no potential for boosting with repeated tests (Miranda et al., 2009). A study by Cho (2007) indicated a higher sensitivity (76% vs. 71%) and specificity (97% vs. 66%) when compared with the TST assay (Cho, 2007).

The overall use of the QFT is rapidly expanding in a low-burden setting and there is some doubt as to whether it can distinguish between latent and active tuberculosis (Pai & O'Brien, 2008). Furthermore, data on the diagnostic ability of the QFT are lacking in high tuberculosis-burden settings (such as South Africa), HIV-infected and immunocompromised populations (Pai & O'Brien, 2008). In India (a high-burden and resource-limited setting), a 15-year follow-up study of 280 000 subjects suggested that the TST remains a useful test and that it continues to serve a useful purpose (Pai, 2005).

QuantiFERON-TB testing may be limited by an elevated rate of indeterminate results in immunocompromised hosts, such as patients with HIV, malignancy or renal dysfunction (Leutkemeyer et al., 2007; Miranda et al., 2009). Indeterminate results do not provide useful information with regard to the likelihood of tuberculosis disease (Miranda et al., 2009). However, these results reflect either impairment of the immune system and/or technical errors during the testing process (Miranda et al., 2009). A study by Miranda (2009) showed a decrease in the rate of indeterminate test results by adding another tuberculosis antigen (TB 7.7) and by vortexing the tubes for 10 s. Researches conducted by Pai (2007) have shown that IFN-γ responses can vary over time within the same individual and that conversions and reversions can occur, making diagnosis even more complicated (Pai et al., 2007).

Thus, more research on the IFN-γ release assay is necessary before these assays can be implemented in diagnostic laboratories (Cho, 2007). In this study, the QFT-TB GOLD ELISA assay was used to detect M. tuberculosis-specific antigens in blood specimens of HIV-positive patients in a high-burden setting due to the lack of information from a low-income, high-burden country.

Materials and methods

Study setting

The study was conducted at an ARV clinic at Tshwane District Hospital in Gauteng (South Africa). This study was approved by the Faculty of Health Sciences Research Ethics Committee, University of Pretoria (45/2008). HIV-positive patients qualifying for ARV treatment according to the South African National HIV Treatment Guidelines (CD4 counts <200 or WHO Stage 4 disease), older than 18 years of age, attending the ARV clinic, who had given informed consent and clinically presenting with tuberculosis symptoms were included in this study.

Clinical assessment

A symptom-screening questionnaire was completed, followed by a physical examination for active tuberculosis. The presence of any one of the following: coughing more than 2 weeks, chest pain, recent weight loss, night sweats, fever, swelling of the lymph nodes, generalized tiredness and not currently on antituberculosis drugs, were the inclusion criteria. Phlebotomists at the ARV clinic (Tshwane District Hospital) drew blood (total volume of 3 × 1 mL each) from voluntary patients according to the manufacturer's protocol and the kits were immediately transported to the research division at the Department of Medical Microbiology, University of Pretoria.

Laboratory assays

Blood samples were processed within 4 h of collection. All patients were allocated a unique number and no patient identities were revealed. Blood specimens (total volume of 3 × 1 mL each) were mixed with antigens and controls for the identification of M. tuberculosis and all three tubes were incubated (Shaking incubator, Labcon) for 16–24 h at 37 °C. The samples were tested for quantitative IFN-γ (IU mL−1) by ELISA (Bio Tek, ELx800, A.D.P., South Africa). The ELISA was performed according to the manufacturer's instructions using standard kits (Cellestis). In addition to the QFT assay, samples were sent to the National Health Laboratory Services (NHLS), University of Pretoria, where routine diagnostics, namely sputum microscopy for acid-fast bacilli and tuberculosis culture, were performed.

The manufacturer's guidelines for the interpretation of results are as follows: the test consists of a negative control (a ‘nil’ well; whole blood without antigens or mitogen), a positive control (a ‘mitogen’ well; whole blood stimulated with the mitogen phytohaemagglutinin) and two sample wells (whole blood stimulated with ESAT-6 or CFP-10). QFT-TB GOLD values are based on the amount of IFN-γ released in response to the antigens. The IFN-γ level of the nil well is considered the background value and is subtracted from the mitogen- and antigen-stimulated well values. The test result is considered positive and tuberculosis infection is suspected if the IFN-γ level in the sample well after stimulation with the specific antigens is >0.35 IU mL−1 and >25% of the nil value. The result is indeterminate if the IFN-γ level in the nil tube is >8.0 IU mL−1 and resulting in a value for the sample well that is ≤0.35 IU mL−1. A negative test result and no tuberculosis infection is suspected when the sample well value is ≤0.35 and <25% of the nil value.

Statistical methods

Microsoft Excel 2007 was used as the spreadsheet, and statistix 8 (Analytical software, Tallahasee, FL) and spss 17 (SPSS inc, Chicago, IL) were used as the data analysis packages for all data. To examine the validity of the manufacturer's assignment of all indeterminate values to negatives, receiver operating curves (ROC) were drawn for all QFT results, i.e. all true and false positives and negatives, relative to the sputum cultures. All data were converted to binary values and totals and percentages per category were calculated. Patient CD4 counts were stratified in two different ways to test the QFT results in terms of sensitivity and accuracy/specificity, respectively, firstly, as values ≤ or >100, the same for 150, 200, 250 and 300 and secondly, as values, ≤100, between 100 and 150, 150–200, 200–250, 250–300 and >300. After cross-tabulation, Fischer's exact test comparisons were performed of the QFT results with the first set of ranges, or, in the second set of ranges, CD4-value stratified Mantel–Haenzel tests comparing the sputum culture with the QFT results. Multivariate and univariate logistical analyses were conducted to examine whether age, gender or prior ARV exposure were covariates determining the outcomes of the QFTs.

Results and discussion

In this study, 60 HIV-positive volunteers were consecutively recruited to evaluate the use of the QFT-TB GOLD ELISA assay at the ARV clinic, Tshwane District Hospital. The number of patients presenting with positive, negative and indeterminate QFT values relative to the sputum culture results are as presented in Table 1. The observed sensitivity and specificity of the QFT-TB GOLD assay in the ARV clinic were 60% (9/24) and 59% (16/27), respectively, when compared with the gold standard culture results as indicated in Table 1. Twenty-five percent (15/60) culture confirmed tuberculosis cases were reported as a QFT-negative test result and 18% (11/60) of QFT-positive results were negative for culture. However, when all indeterminate QFT results were labelled negative, the sensitivity decreased significantly to 30% (9/30) with a slight increase noted in the specificity, 63% (19/30). The rates of indeterminate QFT results were 15% (9/60) and supported the high rate of indeterminate results (11%) revealed by a study by Miranda (2009). ROC curves demonstrated that the manufacturer's assignment of indeterminate values to negatives was justified in each case. Interestingly, a Fischer's exact test comparing sputum culture to the QFT results demonstrated that the relative proportions in the two groups were similar, despite there being 30 positive and negative cultures and 20 positive and 40 negative QFT results. After stratification of all data by CD4 range, it was evident that the majority of patients in this study fell into the low-count categories (Table 2).

View this table:
Table 1

A comparison between the culture and QFT results and an overview of the positive, negative and indeterminate outcomes of the assay

Sputum culture
View this table:
Table 2

Percentage sensitivity, error and specificity/accuracy of the QFT test relative to the sputum culture method

CD4 ranges
NPatients51.7 (31)11.7 (7)8.3 (5)11.7 (7)3.3 (2)13.3 (8)100 (60)
SensitivityAll negatives62.5 (25)12.5 (5)5.0 (2)2.5 (1)5.0 (2)12.5 (5)100 (40)
All positives30 (6)10 (2)15 (3)30 (6)0 (0)15 (3)100 (20)
neg ≤ >CD4: 20080 (32)20 (8)
pos ≤ >CD4: 20055 (11)45 (9)
ErrorFalse negatives52.4 (11)19 (4)9.5 (2)0 (0)9.5 (2)9.5 (2)100 (21)
False positives36.4 (4)18.2 (2)9.1 (1)18.2 (2)0 (0)18.2 (2)100 (11)
neg ≤ >CD4: 20081.0 (17)19.0 (4)
pos ≤ >CD4: 20063.6 (7)36.4 (4)
Total46.9 (15)18.8 (6)9.4 (3)6.3 (2)6.3 (2)12.5 (4)100 (32)
Total % of N48.885.76028.61005053.3
Specificity/accuracyTrue negatives73.7 (14)5.3 (1)0 (0)5.3 (1)0 (0)15.8 (3)100 (19)
True positives22.2 (2)0 (0)22.2 (2)44.4 (4)0 (0)11.1 (1)100 (9)
neg ≤ >CD4: 20078.9 (15)21.1 (4)
pos ≤ >CD4: 20044.4 (4)55.6 (5)
Total57.1 (16)3.6 (1)7.1 (2)17.9 (5)0 (0)17.9 (4)100 (28)
Total % of N51.614.34071.405046.7
  • Values given as % (actual).

Regarding the sensitivity of the test, Fischer's exact test comparisons of the QFT results demonstrated significant differences at ≤ or >100 (0.05), ≤ or >150 (0.02), but not at ≤ or >200 (0.11), ≤ or >250 (0.7) or ≤ or >300 (1.0). This suggested that the likelihood of a diagnostic hit (sensitivity) of the QFT assay was determined by a limiting CD4 value lying between 150 and 200. That is, it was more likely that the QFT assay would produce a result of any kind above 200 than below. Unfortunately, a more accurate resolution of this limit was impossible in that only five patients had CD4 counts in this range. Because these diagnostic hits were independent of the results of the sputum culture method, these results indicate nothing of the accuracy/specificity of the QFT assay.

In terms of the accuracy/specificity, examination of the raw data demonstrated that the majority of diagnostic errors by the QFT assay were found in the lower ranges (Table 2 and Fig. 1). Mantel–Haenzel tests were used to compare the QFT results with the culture sputum results after stratifying the data on CD4 ≤100, 100–150, 150–200, 200–250, 250–300 and >300 ranges. No significant differences were found, suggesting that the QFT was inaccurate relative to the culture method independent of the CD4 range. In fact, the overall accuracy of the QFT assay was only 46.7% (Table 2). These results would seem to suggest that the QFT assay indicated the presence of the tuberculosis antibody rather than active disease. Following binary coding, a comparison of the QFT results individually or in combination with the independent variables of gender, age (≤ or >40) and prior exposure to ARV revealed no significant relationships under any circumstance. This was taken as an indication that these variables did not impact the QFT results. Persistent bacteria, even in low numbers, might keep T cells activated and induce detectable IFN-γ responses (Pai et al., 2007). Thus, longer treatment and follow-up may be necessary to completely demonstrate declining T-cell responses (Pai et al., 2007).

Figure 1

Percentage incidence of QFT diagnostic errors relative to the sputum culture method stratified by CD4 value range.

Additional studies are needed to understand the reproducibility and the relative accuracy of the test before its utility in prospective screening programmes can be defined (Perry et al., 2008). This study had two limitations. First, the numbers of events were small. A more robust study needs to be conducted within the setting. No HIV-negative, tuberculosis-positive patients were included and thus the real effect of immune suppression could not be determined. In a resource-poor country, such as South Africa, the cost-effectiveness of the QFT assay is most likely low and until there are more solid data available on the use of QFT in a high-prevalence TB/HIV setting, clinicians should rely on a careful medical history as well as culture results.


Despite the limitations, our data offer some interesting insights for the use of the QuantiFERON-TB GOLD ELISA assay. In the high HIV/TB prevalence setting in this study, the QFT had limited sensitivity and an inability to distinguish between acute and latent infection. This would suggest that in terms of the routine diagnosis of tuberculosis by the clinician in the high HIV prevalence setting, the test should be used with caution. It may be advantageous to detect anti-ESAT-6/CFP10-specific T cells in blood from individuals with tuberculosis infection without the need for IFN production.


This research was financially supported by the Department of Medical Microbiology, University of Pretoria, South Africa. We thank the Diagnostic Division of the National Health Laboratory Services (NHLS), Pretoria, South Africa, as well as Shaheed Vally Omar for his contribution.


  • Editor: Willem van Leeuwen


View Abstract