Bayes' theorem-based assessment of VDRL syphilis screening miss rates

Interlaboratory differences (and conflicting practices) in syphilis screening strategies (serial versus parallel test combinations) prompted us to determine an estimate of the diagnostic miss rate in the detection of (1) infected persons and (2) infected but untreated persons potentially affected by late active syphilis.

We set out to establish the most efficient syphilis screening strategy for two routine tests (VDRL and Treponema pallidum hemagglutination assay [TPHA]) with regard to our tested public health population (average VDRL+ TPHA+ and VDRL- TPHA+ reactor ages being 59.9 years and 50.5 years, respectively.

Retrospective analysis covered the results of a routine parallel VDRL and TPHA testing on 24,863 persons done in four public health laboratories. Nosologic sensitivity of the VDRL test (18.32%) was determined using the TPHA test as reference. The percentage of VDRL nonreactors among infected persons (TPHA reactors) was considered as the VDRL false negative rate (81.68%). We estimated the proportion of persons infected but untreated with potential late active syphilis using our own Bayes theorem-based procedure.

The Bayes theorem-based estimate showed a significantly higher value for persons at risk of active late syphilis than the number of suspected cases obtained using the classical approach (25.1% versus 18.32%, or 83 persons versus 61 of the 330 infected).

In screening an older population, the VDRL test alone (or as the first of a series with TPHA as a confirmation test) may produce a diagnostic miss rate higher than the syphilis detection rate. Another miss in such a population is detected by the Bayes theorem-based method.