Effect of endoscopic screening for non-cardia gastric cancer: a 12-year report of a population-based randomized trial.

Abstract

The benefit of endoscopy for non-cardia gastric cancer (GC) prevention lacks evidence from randomized controlled trials (RCTs). This study aimed to evaluate the effect of endoscopic screening for non-cardia GC through an RCT. We conducted a cluster RCT in rural Hua County, northern China (ClinicalTrials.gov, NCT01688908). A total of 668 villages were randomly selected and assigned to either undergo upper gastrointestinal endoscopic screening (screening group) or no screening (control group). Permanent residents aged 45-69 years were enrolled from January 2012 to September 2016. Non-cardia GC mortality between the two groups was compared in intention-to-treat, per-protocol, and subgroup analyses. Poisson regression fitted with generalized estimating equations was used, adjusting for baseline characteristics and accounting for village clustering. The study enrolled 17,104 participants in the screening group, of whom 15,165 (88.7%) underwent endoscopy, and 16,743 in the control group. After a maximum follow-up of 12 years, non-cardia GC mortality was 13.1 per 100,000 person-years in the screening group and 13.5 per 100,000 person-years in the control group. Intention-to-treat analysis showed a 15% lower non-cardia GC mortality in the screening group compared with the control group (adjusted rate ratio [aRR], 0.85 [95% CI: 0.49-1.50]), although the difference was statistically non-significant. In the per-protocol analysis, the aRR for non-cardia GC mortality was 0.70 (95% CI: 0.38-1.29). A subgroup analysis of the screening group restricted to participants who complied with screening protocol and received timely treatment for detected malignancies yielded similar results, with an aRR of 0.66 (95% CI: 0.35-1.22); neither of these analyses reached statistical significance. In this trial, endoscopic screening was associated with a non-significant reduction in non-cardia GC mortality; larger trials in higher-incidence settings are needed to confirm a benefit.