The Chronic Disease of Needing a Gold Standard for Diagnostic Accuracy: Is There a Simple Cure?

摘要

The belief that a perfect reference test (often termed a gold standard) is required for evaluation of the accuracy of diagnostic tests (measures such as sensitivity, specificity, likelihood ratio of a positive test and area under the receiver-operating characteristic curve) for a designated purpose has been entrenched in the scientific literature for years. Historically, tests such as virus, bacteria, and parasite isolation were considered gold standards when tests such as real-time PCR were validated. When results of a side-by-side comparison in a test validation study showed discordant results (positive-negative or negative-positive), the dilemma became how to interpret PCR-positive results when the isolation results were negative. In a traditional statistical analysis (2x2 table with isolation as the reference test), PCR might be deemed to be less specific and less sensitive than isolation for testing clinical samples even though laboratory experiments may have shown PCR to have a lower limit of detection (analytical sensitivity) than isolation.