Screening for tuberculosis infection among newly arrived asylum seekers: Comparison of QuantiFERON^®TB Gold with tuberculin skin test

Enrolment and characteristics of the study group

TST and QFT results

A total of 460 (50%) participants tested positive with TST (≥ 6 mm) while 264 (29%) tested positive with QFT. Among the TST positive participants 50% were QFT negative, whereas 7% of the TST negative participants (< 6 mm) tested positive with QFT. Positive TST and QFT tests were most common among African participants (63% and 43%). European participants more often tested positive on TST (52%) than Asian participants (39%), while QFT-results were similar among European and Asian participants, 19% vs 20% respectively. Positive QFT-results were seen with all sizes of TST indurations, but the proportion of positive QFT tests increased with increasing induration size (Figure 1a and 1b). Among the 1000 asylum seekers initially enrolled in the study, five were diagnosed with culture-confirmed tuberculosis. Out of these, two were among participants later excluded from the study; one (TST 15 mm) did not have a QFT-result and one (TST 17 mm) only had a single positive QFT, without a confirmatory test done. Among the remaining three, one patient (TST 9 mm) had a negative QFT result and two were QFT positive (10 and 18 mm TST). All five reported exposure to and symptoms of tuberculosis and had chest x-rays indicative of tuberculosis.

Agreement between QFT and TST

Predictors of positive tests

Implications for screening

Agreement between QFT and TST

Consistent with other studies, we found a moderate agreement between the results of QFT and TST, although differences in design and populations make comparison between such studies difficult [7, 9, 11, 18, 19]. Importantly, we observed that the proportion of individuals with positive QFT increased with increasing TST induration, just like the expected proportion of tuberculin reactions actually caused by M. tuberculosis would do [20]. A TST reaction of ≥15 mm has been interpreted as high likelihood of true tuberculosis infection [21, 22]. Discordant results in our study were found in the range of 21 – 29% depending on the TST cut-off value, but mainly as positive TST and negative QFT in the range of 6–9 mm (72%) and 10–14 mm (49%). The poor specificity of TST in these intervals is a major challenge as most positive TST results (69%) occurred within this range as well as 48% of the positive QFT-results. False positive TST indurations caused by BCG or non-tuberculosis mycobacteria are expected to be moderate and rarely exceeding 15 mm [10, 21–23]. Boosted TST reactions or differences in the nature of the immune response measured by the two tests might also explain discordant results [8, 24]. In this study the blood sample was drawn at the time when the TST was read. A potential influence of PPD exposure on a QFT test has been investigated, and having a QFT test after TST is advised to be a reliable approach [25]. This is also in agreement with the European Consensus Report, where it is advised that IGRA should be implemented as a supplement following a positive TST [26].

Possible explanations for discordant results

Previous BCG vaccination; a meta analysis reported increased likelihood of a positive TST in BCG-vaccinated persons within the first 15 years after vaccination [27]. We found no significant effect of previous BCG-vaccination on QFT nor on TST-results in multivariate regression analyses, regardless of TST cut-off 15, 10 or 6 mm. The limited effect of BCG observed in our study might be explained by the fact that vaccination is most often offered to newborns and all our participants were older than 18 years. In addition, scars might be underreported either by being missed or being misinterpreted by the observer, or vaccinated persons might not have developed a scar. Non-vaccinated persons in this study were to a large extent from Somalia, a population where the prevalence of latent infection is high. This might explain the better agreement between the tests for non-vaccinated compared to vaccinated participants. As revaccination policies have been common in some European countries (including Russia) the stronger effect of BCG vaccination might explain the lower agreement found among European participants.

Difference in immunological responses measured by the two tests; tuberculin induces a delayed-type hypersensitivity reaction reflecting a memory response, while the blood tests detect cellular INF-γ release reflecting an effector response to an ongoing infection. Due to these differences in immunological responses it is suggested that IFN-γ release assays detect more than TST those with recent and persistent infection [8, 18]. Differences in immunological responses can partly explain discordance and might have important clinical and programmatic implications, as recent infection is more likely to progress to active disease and preventive treatment therefore indicated. A recent publication indicates at least equivalent sensitivity of QFT compared to TST for predicting progression to tuberculosis [28]. The lack of an age associated increase in discrepant results between QFT and TST (cut-off 15 mm) in Table 3 could indicate that many of the participants have recent infection. There is also some ongoing research to determine whether the level of IFN-γ response to the specific M. tuberculosis antigens may be meaningful for predicting the outcome of latent infection [29].

Sensitivity related issues; among participants with a strongly positive TST (≥ 15 mm) we found that 26% (37 out of 141) had a negative QFT test. High proportions of discrepant results in this group has also been observed in other studies [17, 30, 31]. Among immunocompetent persons the sensitivity of tuberculin is regarded as very high, while sensitivity of QFT for detection of latent infection has been variable (75 – 97%) depending on the study population and design [18, 32, 33]. Among patients with tuberculous disease confirmed by culture, sensitivity measurements for QFT varies (70–90%) [7, 10, 34, 35]. A recent study reported 13% indeterminate and 20% negative QFT results among patients with culture confirmed tuberculosis [34]. The fact that we found one QFT negative individual among four participants with confirmed active disease demonstrates the need for careful evaluation of test results. The main purpose of TST and QFT used in screening is to identify persons with latent infection eligible for preventive treatment. Screening of symptoms and chest x-ray would have identified all five asylum seekers with active tuberculosis in our study. Among participants with negative TST, we found that 7% (32 out of 452) had a positive QFT. This might reflect false negative TST results as the IFN-γ release assays have shown to have higher sensitivity than TST among immunocompromised individuals [34, 36]. The risk of future progression to active disease in TST negative and QFT positive adults should be studied in large, prospective studies. Due to the lack of a positive control for a functional immune response (mitogen stimulation) in the QFT assay available when the study was conducted, the general immune status of the participants were not known. This is a limitation of the study and may in cases of immuno-suppression have lead to false negative results with respect to both QFT and TST. However, recent information on the performance of the QFT test in immuno-compromised individuals demonstrates that only severe immune-suppression with low CD4 counts may lead to indeterminate results and potential false negative results in case of no positive mitogen control [37–40]. Based on surveillance of hiv-infection in Norway we have no reasons to believe that severe immuno-suppression was affecting more than very few cases, if any at all, among our participants. On this basis, we do not think that the lack of a positive mitogen control in this study represents any major problem for interpretation of the results. Exposure to M. bovis might result in a QFT test that is false positive with respect to M. tuberculosis. As clinical management of cases with tuberculous disease is the same for all species included in the M. tuberculosis-complex, this does not affect the screening approach.

Predictors of positive tests

The multivariate analysis showed similar significant risk factors for positive QFT and TST >15 mm. Country of origin and increase in age were both significantly associated with both positive QFT and TST ≥15 mm, reflecting accumulated exposure time and background prevalence in country of origin. Exposure to an infectious case of tuberculosis was significantly related to QFT, but not to TST, as found in other studies [7–9, 41–44]. Previous history of tuberculosis was not a significant predictor for either test as it correlated with exposure.

Implications for screening in Norway

There is no gold standard test for latent infection. Both positive QFT results and TST>15 mm (independent of each other) have been suggested to accurately identify latent infection and also predict progression to active disease [21, 23, 28] We have, for the purpose of assessing the use of QFT in screening, defined latent tuberculosis infection as having a positive QFT test regardless of TST induration or a TST ≥ 15 mm and a negative QFT. In this study 301 participants are defined as having latent tuberculosis infection; the 264 QFT positive participants and the 37 with a TST ≥ 15 mm and a negative QFT. Based on this definition, a more specific test than TST for diagnosis of latent infection is clearly needed as TST with cut-off at 6 mm or 10 mm results in referral of 42% or 27% false positive cases, respectively, while a cut-off of 15 mm will detect less than half of the actual cases. As it is neither feasible nor cost effective to screen all asylum seekers with both tests, the options are to replace TST with QFT (one step approach) or to use QFT to confirm a positive TST (two step approach). A one step approach with only QFT has logistic advantages and referral based on a false positive TST is avoided. The strength of this approach would be the inclusion of the TST negative among the QFT positive participants (7%), a presumable high-risk group for active disease and cases lost in a two step approach. The two step approach allows for referral procedures based on both tests and it is recommended in a European consensus report from 2006 [26]. This strategy follows normal procedures for screening with a first test with high sensitivity followed by a second test with high specificity. The two-step approach identifies and refers all TST positive cases with a positive QFT-result and also the 12% proportion with TST ≥ 15 mm and a negative QFT, which would be lost in a one step approach (Table 4). The approach of using TST ≥ 6 mm as cut-off value, would identify a similar proportion (88–89%) of persons defined as having latent infection even though it would not be the same persons referred. By implementing QFT the number referred for preventive treatment would be reduced by 43%, independent of whether the one- or two step approach is chosen. Adjusting TST cut-off value to 10 or 15 mm would reduce the number referred to 25% and 15%, respectively, but lost cases with latent infection would consequently increase to 25% and 53%. Whether a one step approach with only QFT is feasible depends also on laboratory capacity and the cost implications of the different screening approaches need further study. Based on the present study the Norwegian Institute of Public Health and its Tuberculosis Advisory Committee have recommended to the Ministry of Health the use of QFT in a two step approach for screening of immigrants.