To the Editors:
We read with interest the study by van Leeuwen et al. 1 concerning the use of the T-SPOTTM.TB (Oxford Immunotec, Oxford, UK) interferon-γ release assay (IGRA) to rule out the diagnosis of active Mycobacterium tuberculosis infection. We disagree, however, with the use of the IGRA tests for ruling out active M. tuberculosis infection, especially in immunocompromised subjects. Sensitivity of T-SPOTTM.TB in immunocompromised subjects, although most certainly higher than that of tuberculin skin test (TST), is clearly <100%. The best sensitivities reported for HIV-infected subjects with active tuberculosis (TB) are 90% 2.
A Bayesian analysis of the cases presented illustrates the limitations of relying on IGRA tests to rule out TB 3, 4. In case A, a young female refugee from Bosnia develops a lingular infiltrate and has acid-fast bacilli (AFB) on examination of bronchoalveolar lavage (BAL). Incidence of TB in Bosnia is 52×10−5 5, almost eight times that of the Netherlands. Clinical presentation is compatible with rapid progression of TB after a recent infection; HIV status is not specified. Reported sensitivity for the T-SPOTTM.TB ranges 83–100% and specificity is in the 96–100% range 6. We would consider the probability of TB in this case as at least intermediate (0.25–0.75) or high (>0.75). Post-test probability of TB, if T-SPOTTM.TB is negative, would be 5–34% for a sensitivity of 83% and a specificity of 98%, and 2–13% for a sensitivity of 95%; if the pre-test probability is high, post-test probability increases markedly. In both cases, a negative IGRA test definitely cannot be used to rule out active TB.
Case B is that of an immunosuppressed 54-yr-old subject with an atypical radiological presentation for TB but with AFB on BAL smears. In this case, sensitivity of the T-SPOTTM.TB assay is unknown but is, at best, 90% based on available data in HIV-infected subjects 2. The same Bayesian approach, for an intermediate pre-test probability (i.e. 0.25–0.75), yields a post-test probability of disease, with a negative T-SPOTTM.TB, of 3–23%. In an immunosuppressed individual, these values are too high to rule out active TB and therapeutic decisions must rely on the identification of the organism involved by PCR and cultures.
Cases C and D are also clinical presentations with at least an intermediate probability of M. tuberculosis infection. In case C, nonspecified mycobacteria grow on culture media, and, in case D, AFB were found on biological samples; thus the negative T-SPOTTM.TB results in these settings at most suggest the possibility of an alternative diagnosis.
van Leeuwen et al. 1 mention that speed to rule out TB is of great importance in terms of infection-control measures. However, with the exception of immunosuppressed individuals or young children in whom immediate clinical work-up is mandatory, IGRA tests and TST take 6–8 weeks to convert after infection: no large-scale contact-tracing procedure is warranted for the first 2 months following diagnosis and treatment of the index case.
Sensitivity, likelihood ratio in the event of a negative test and pre-test probability are of major importance to rule out active disease. Furthermore, sensitivity varies according to the IGRA test used and according to the population tested. Ferrara et al. 7 studied both IGRA tests in routine clinical practice and reported a 16–25% false-negative rate (a sensitivity of 75–84%) in 24 subjects with active TB, using both IGRA tests available. Sensitivity reported for T-SPOTTM.TB is 83–100% and is 64–97% for QuantiFERON-TB GOLD (Cellestis, Carnegie, Victoria, Australia); this is on average slightly lower than that of T-SPOTTM.TB 8–10.
The Centers for Disease Control and Prevention 2005 guidelines state that, for reasons of suboptimal sensitivity, a negative QuantiFERON-TB GOLD test cannot be used to exclude the diagnosis of active TB 11. Although the sensitivity of the T-SPOTTM.TB is probably better on average than that of the QuantiFERON-TB GOLD, there are too many clinical situations, such as all forms of immunosuppression, severe comorbidity, HIV infection, older subjects, in which its sensitivity is probably decreased and thus Centers for Disease Control and Prevention guidelines appear reasonable for both IGRA tests. T-SPOTTM.TB may at the very best reasonably rule out Mycobacterium tuberculosis infection in immunocompetent individuals, without any risk factor for exposure to TB, and with a low clinical pre-test probability of Mycobacterium tuberculosis infection (<0.25).
- © ERS Journals Ltd