Abstract
BACKGROUND: With today's rapid advances in technology and understanding of disease, more screening and diagnostic tests have become available in a variety of sociodemographic and clinical settings. This analysis quantifies the impact of varying prevalence rates on test performance for given sensitivity and specificity values. METHODS: Using a worked example of latent tuberculosis infection, we compared true-positive (TP) and false-positive (FP) results when varying prevalence and test sensitivity and specificity. We used estimates from published literature to estimate two tests' sensitivity (81%, QuantiFERON®-TB Gold In-Tube; 88%, T-SPOT®.TB) and specificity (99%; 88%), and we used World Health Organization data to estimate disease prevalence in five countries. RESULTS: Varying sensitivity impacted outcomes most in high-prevalence settings; change in specificity had greater impact in low-prevalence settings. In switching from QuantiFERON-TB to T-SPOT.TB (higher sensitivity, lower specificity), trade-offs between increasing case identification (TPs) and decreasing unnecessary treatments (FPs) varied dramatically with prevalence. Lower-prevalence settings paid a greater "price" of more FPs for each TP gained, with 37.7 FPs per TP in the United States (5% prevalence) versus 2.5 in the Ivory Coast (55% prevalence). CONCLUSIONS: Prevalence affects test performance for given sensitivity and specificity values. To optimize test performance, disease prevalence should be incorporated in testing decisions, and sensitivity and specificity should be set locally, not globally. In lower-prevalence settings, using highly specific assays may optimize outcomes.