Sensitivity of endoscopic screening for gastric cancer by the incidence method

Abstract

Although radiographic screening for gastric cancer has been conducted in Japan, it is anticipated that endoscopy will become a new screening method because of its high detection rate. The sensitivities of endoscopic and radiographic screening were calculated by the detection method and the incidence method based on the results of community-based screening in Japan. There were 56,676 screenings for gastric cancer using endoscopy and radiography from April 2002 to March 2007 in Yonago, Japan. The target age group was from 40 to 79 years. Screen-detected and interval cancers were investigated based on a screening database linked to the Tottori Cancer Registry. All gastric cancers diagnosed within 1 year after a negative screen were considered interval cancers. Based on the screening history, these were divided into prevalence screening and incidence screening. Prevalence screenings included 7,388 for endoscopic screening and 5,410 for radiographic screening, whereas incidence screenings included 18,021 for endoscopic screening and 11,417 for radiographic screening. The sensitivity of prevalence screening calculated by the incidence method was 0.886 (95% confidence interval [CI] = 0.698-0.976) for endoscopic screening and 0.831 (95% CI = 0.586-0.964) for radiographic screening; however, the difference was not significant (p = 0.626). The sensitivity of incidence screening calculated by the incidence method was 0.954 (95% CI = 0.842-0.994) for endoscopic screening and 0.855 (95% CI = 0.637-0.970) for radiographic screening (p = 0.177). Endoscopic screening for gastric cancer had a higher sensitivity than radiographic screening by the incidence method in both screening rounds. However, further study is needed to evaluate mortality reduction and to estimate overdiagnosis with endoscopic screening for gastric cancer.