Bias in the evaluation of DNA-amplification tests for detecting Chlamydia trachomatis.

The purpose of this paper is to show that the sensitivity and specificity estimates obtained by 'discrepant analysis' are biased. Discrepant analysis is a widely used technique that attempts to provide estimates of sensitivity and specificity in the presence of an imperfect gold standard. Many researchers have applied this technique to estimate the sensitivity and specificity of DNA-amplification tests for Chlamydia trachomatis such as the plasmid based ligase chain reaction (LCR) and polymerase chain reaction (PCR) tests. Moreover, the June 1993 package insert of the PCR AMPLICOR Chlamydia trachomatis test contains estimates of sensitivity and specificity based on 'discrepant analysis'. Even if one employs a perfect test to resolve the discrepant results, discrepant analysis estimates of test sensitivity and specificity remain biased. Thus, one should not adopt this technique to evaluate the performance of a diagnostic test.