Development, validation, and evaluation of a risk assessment tool for personalized screening of gastric cancer in Chinese populations

Abstract Background Effective risk prediction models are lacking for personalized endoscopic screening of gastric cancer (GC). We aimed to develop, validate, and evaluate a questionnaire-based GC risk assessment tool for risk prediction and stratification in the Chinese population. Methods In this three-stage multicenter study, we first selected eligible variables by Cox regression models and constructed a GC risk score (GCRS) based on regression coefficients in 416,343 subjects (aged 40–75 years) from the China Kadoorie Biobank (CKB, development cohort). In the same age range, we validated the GCRS effectiveness in 13,982 subjects from another independent Changzhou cohort (validation cohort) as well as in 5348 subjects from an endoscopy screening program in Yangzhou. Finally, we categorized participants into low (bottom 20%), intermediate (20–80%), and high risk (top 20%) groups by the GCRS distribution in the development cohort. Results The GCRS using 11 questionnaire-based variables demonstrated a Harrell’s C-index of 0.754 (95% CI, 0.745–0.762) and 0.736 (95% CI, 0.710–0.761) in the two cohorts, respectively. In the validation cohort, the 10-year risk was 0.34%, 1.05%, and 4.32% for individuals with a low (≤ 13.6), intermediate (13.7~30.6), and high (≥ 30.7) GCRS, respectively. In the endoscopic screening program, the detection rate of GC varied from 0.00% in low-GCRS individuals, 0.27% with intermediate GCRS, to 2.59% with high GCRS. A proportion of 81.6% of all GC cases was identified from the high-GCRS group, which represented 28.9% of all the screened participants. Conclusions The GCRS can be an effective risk assessment tool for tailored endoscopic screening of GC in China. Risk Evaluation for Stomach Cancer by Yourself (RESCUE), an online tool was developed to aid the use of GCRS.