Latent class modeling to compare testing platforms for detection of antibodies against the Chlamydia trachomatis antigen Pgp3

Abstract

Latent class modeling can be used to combine the results of multiple tests to compare the sensitivity and specificity of those tests in the absence of a gold standard. Seroepidemiology for chlamydia infection may be useful for determining the cumulative risk of infection within a population. Initial studies using the Chlamydia trachomatis immunodominant antigen Pgp3 have shown utility for seroepidemiology of sexually transmitted chlamydia and the eye disease trachoma. We present our latent class modeling results for comparison of antibody data obtained from three different Pgp3-based platforms - multiplex bead array, ELISA, and lateral flow assay. Sensitivity and specificity estimates from the best fitting latent class models were similar to estimates derived from those previously obtained using a nucleic acid amplification test as a gold standard for sensitivity and non-endemic pediatric specimens for specificity, although the estimates from latent class models had wider confidence intervals. The modeling process and evaluation highlighted the importance of including as many antibody tests as possible when fitting a latent class model to ensure that as many patterns as possible are available for evaluation. Future studies designed to evaluate antibody test performance in the absence of a gold standard should utilize as many tests as possible.