Screening for latent tuberculosis infection was conducted among 505 health care workers in a high TB incidence country in Sub-Sharan Africa. The study assessed the level of agreement between TST and IGRA, identified factors associated with discordance between test results and compared sensitivity and specificity of the tests using a latent class model.Methods
Three tests for TB infection were employed: Tuberculin skin test, Quantiferon-gold-in-tube and TSPOT.TB. Agreement was measured between test outcomes. Factors associated with discordance were analysed using a multinomial logistic regression model. Latent class analysis using a fixed effects model that allowed for conditional dependence between QFT-GIT and TSPOT.TB was used to fit the data and allow for a direct comparison of test sensitivity and specificity.Result
There was fair agreement between TST and QFT – GIT (κ=0.28) and T-SPOT.TB (κ=0.25), respectively. Marked discordance was noted between the TST and IGRA results in HIV positive individuals OR: 6.35, those who reported previous TB treatment OR: 3.00 or were symptom screen positive for TB, OR: 2.95. TST had the highest sensitivity (93%) and lowest specificity (55%) of the three tests. IGRAs displayed equivalent sensitivity (79%–84%) and higher specificity (94%–97%) with marginal change following the application of a latent class model.Discussion
Marked discordance between TST and TSPOT.TB outcomes in HIV infected individuals reflects potentially greater sensitivity of this assay in immunocompromised persons. In high TB incidence settings IGRA assays do not display significantly greater sensitivity or specificity in diagnosis of LTBI among health care workers following the application of latent class analysis model that allows for conditional dependency between IGRAs.