Abstract
SETTING: Serial screening for latent tuberculous infection (LTBI) is commonly performed in certain populations, such as health care workers. The high apparent conversion rate in some studies of interferon-gamma release assays is puzzling given the claimed high specificity of these tests. OBJECTIVE: To understand how test-retest variability, specificity, and underlying LTBI prevalence affect observed outcomes of repeated testing for LTBI. DESIGN: Mathematical model assuming constant test sensitivity and specificity over time and no new infections. RESULTS: Test-retest variability had a large effect on the observed proportion of conversions (initial negative test, followed by a positive test) and reversions (initial positive test, followed by a negative test). For example, a test with 70% specificity and 5% test-retest variability would be associated with a conversion rate of 3.7% and a reversion rate of 7.7%, while a test with 95% specificity but 10% test-retest variability would be associated with a conversion rate of 5.5% and a reversion rate of 57%, assuming that both tests are 80% sensitive and underlying LTBI prevalence was 5%. CONCLUSION: Test-retest variability is a key parameter that should be reported for tests used for serial screening for LTBI. Reducing test-retest variability can reduce false-positive and false-negative results.