Background and study aims: The degree of accuracy of gastroscopy for the detection of gastric cancer is poorly understood. The aim of this retrospective study was to determine the accuracy of gastroscopy by using cancer registry records.
Patients and methods: Gastroscopic examinations (n = 37094) conducted between 1984 and 1989 were studied by linking them with hospital-based and population-based (Fukui Prefecture, Japan) cancer registry records between 1984 and 1992. False-negative gastroscopies that had been carried out within the three years preceding the diagnosis of gastric cancer were identified.
Results: The numbers of true-positive, false-positive, and false-negative examinations carried out were 659, six and 155, respectively. The sensitivity, specificity, positive predictive value, and negative predictive value were 81.0%, 100.0%, 99.1%, and 99.6%, respectively. The overall diagnostic accuracy of gastroscopy was 99.6%. There was little difference in sensitivity results between the patient groups with regard to reason for referral, type of endoscope used, experience of endoscopist, or location of gastric cancer. The percentage of tumours representing early gastric cancer, identified after false-negative gastroscopy, was lower for those situated in the cardia or gastric body than for those in the angular notch or the antrum.
Conclusions: The accuracy of gastroscopy in the detection of gastric cancer is satisfactory, but false-negative results are sometimes obtained. We emphasize the importance of repeated endoscopic examination for the detection of gastric cancer.