Background: Tuberculin skin tests (TSTs) may be negative in the absence of latent tuberculosis infection (LTBI) or in the presence of active tuberculosis (TB). The same phenomenon likely holds for interferon-gamma release assays (IGRAs).
Methods: A mathematical model of LTBI test specificity was developed using elementary probability theory. Implications for IGRA study design, and data collection and analysis are presented.
Results: The model demonstrates that the specificity of any test for LTBI is theoretically affected by its sensitivity to active TB, and by the distribution of TB, LTBI and healthy controls. Moreover, changing cut-off points to increase the sensitivity for active TB will cause a decline in the specificity for LTBI. Published data are used to demonstrate these counterintuitive results.
Conclusions: As a result of the markedly dissimilar health statuses associated with a negative LTBI test, IGRA specificity for LTBI and sensitivity to active TB depend upon TB and LTBI prevalence. A trade-off exists between the specificity of IGRAs to LTBI and sensitivity to active TB. Field studies of IGRAs will require standardized collection of symptoms of active TB to distinguish LTBI suspects from TB suspects. Different cut-off points for IGRA use in TB suspects and LTBI suspects will likely be needed.