Estimation of sensitivity and specificity of site-specific diagnostic tests.

Clinical trials designed to estimate the sensitivity and specificity of periodontal diagnostic tests often use multiple sites per patient as experimental units of analyses. Since site-specific test results within a patient are dependent observations, a correlated binomial model should be employed to estimate the sensitivity and specificity of these diagnostic tests. Ignoring the within-patient correlation can result in an over- or underestimation of the true standard errors.