Performance of gastric cancer screening by endoscopy testing through the National Cancer Screening Program of Korea.

Recent reports have proposed endoscopy as an alternative strategy to radiography for gastric cancer (GC) screening. The current study presents the first reported population-based data from a large GC screening program that provided endoscopic examinations. A retrospective population-based study was conducted using the National Cancer Screening Program (NCSP) database. We evaluated GC detection rates, sensitivity, specificity, and the positive predictive value of an endoscopic screening program for the average-risk Korean population, aged 40 years and older, who underwent the NCSP from 2002 to 2005. The detection rates of GC by endoscopy in the first and subsequent rounds were 2.71 and 2.14 per 1000 examinations, respectively. Localized cancer accounted for 45.7% of screen-detected GC cases. The sensitivity of endoscopy was 69% (95% confidence interval [CI]: 66.3-71.8). The endoscopic screening was less sensitive for the detection of localized GC (65.7%, 95% CI = 61.8-69.5) than for regional or distant GC (73.6%, 95% CI = 67.4-79.8). In the multiple logistic models for localized GC and all combined GC, the odds ratio (OR) of sensitivity for the undifferentiated type was statistically significantly higher than that for the differentiated type, whereas the OR of sensitivity for the mixed type was lower than that for the differentiated type. The sensitivity of the endoscopic test in a population-based screening was slightly higher for the detection of regional or distant GC than for localized GC. Further evaluation of the impact of endoscopic screening should take into account the balance of cost and mortality reduction.