Tuberculin skin test: estimation of positive and negative predictive values from routine data.

OBJECTIVE: To estimate, using routinely available data, the predictive values of tuberculin testing with 2TU RT23 for detection of latent tuberculosis infection (LTBI) in the Dutch population as a basis for recommendations on cut-off values at various levels of infection prevalence. DESIGN: Smoothed distributions of TST reactions among 312 tuberculosis patients and 2848 healthy non-BCG-vaccinated persons were used to estimate the sensitivity and specificity at various cut-off values. RESULTS: Sensitivity was 98.9% at 5 mm, 95.4% at 10 mm and 79.8% at 15 mm cut-off. Specificity with the corresponding cut-offs was 95.3%, 96.3% and 97.1% before and 98.0%, 98.8% and 99.6% after adjustment for presumed LTBI. At 10 mm, the positive predictive value (PPV) was > 75% if the infection prevalence in the tested population was at least 10%, but strongly declined with lower prevalences. For lower prevalences a cut-off of 15 mm was proposed, as this results in a higher PPV without greatly affecting the negative predictive value. CONCLUSION: Estimation of the predictive values of the TST from routine data can be useful for establishing cut-off values for detection of LTBI in different populations with different tuberculin preparations.