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.

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.