Table 3

Tests with best-estimated positive predictive values of at least 90% for latent tuberculosis infection among adults over the probable range of prevalence

Prevalence (%)Pre-test oddsCritical PLR*Bivariate random effects modelUnivariate random effects model
Test, PLR (95% CI)Test, PLR (95% CI)
10–201/9–1/481.0–36.0QFT-G/QFT-G-IT (Japan) 48.1 (19.7 to 117.6)QFT-G/QFT-G-IT (Japan) 43.4 (24.8 to 75.9)
40–502/3–1/113.5–9.0QFT-G/QFT-G-IT (non-Japan) 10.8 (5.3 to 21.8)QFT-G/QFT-G-IT (non-Japan) 11.6 (5.3 to 25.6)
50–551.0–11/99.0–7.4T-SPOT.TB 7.9 (3.6 to 17.3)
  • Probable pre-test odds range from 1/9 to 11/9 for latent tuberculosis infection during tuberculosis contact investigation. Japan and non-Japan in brackets denote results based on Japanese and non-Japanese studies, respectively. Tests that are suitable in settings with lower pre-test odds also work in settings with higher pre-test odds for better positive predictive values. TST with cut-off at 5 mm and 10 mm or 15 mm among predominantly BCG-vaccinated subjects are not included under tests for LTBI as their best-estimated PLR (95% CI) by bivariate meta-analysis are all below critical PLR given probable pre-test odds: cut-off at 5 mm, 1.2 (0.6 to 2.4); cut-off at 10 or 15 mm, 2.2 (1.1 to 4.1).

  • * Corresponding to positive predictive values of at least 90% (see mathematical equations in the Methods section).

  • LTBI, latent tuberculosis infection; PLR, positive likelihood ratio; NLR, negative likelihood ratio; QFT-G, QuantiFERON-TB Gold; QFT-G-IT, QuantiFERON-TB Gold in-Tube; TB, tuberculosis.