Is everything all right if nothing seems wrong? A simple method of assessing the diagnostic value of endoscopic procedures when a gold standard is absent

Abstract

Purpose: We demonstrate a simple yet comprehensive method to evaluate the sensitivity of endoscopic procedures when no gold standard is available.

Materials and methods: In 208 consecutive patients with superficial bladder cancer 328 endoscopies were performed to compare the sensitivity of white light and 5-aminolevulinic acid induced fluorescence endoscopy. Both procedures were performed during the same session.

Results: The maximum interval of observable sensitivity for 5-aminolevulinic acid induced fluorescence endoscopy ranged between 78 and 97.5%, and the best estimate for sensitivity based on realistic assumptions was 93.4% (95% confidence intervals 90 to 97.3). The best sensitivity estimate for white light endoscopy was 46.7% (95% confidence intervals 39.4 to 54.3, maximum range 47.2 to 53%).

Conclusions: This method to determine the maximum possible range of sensitivity estimates for endoscopic procedures without a gold standard is easily applied. Depending on the assumptions a range of reasonable scenarios can be constructed and the corresponding sensitivities can be reported. This approach gives fast and valid results, and could further indicate the diagnostic superiority of 5-aminolevulinic acid induced fluorescence compared to white light endoscopy.