The Summit emphasized the need for risk-stratification models that combine demographic, clinical, and biomarker data to identify people at higher risk, so that screening resources can be focused more efficiently.
Variables such as age, sex, Helicobacter pylori infection status, family history, lifestyle factors (e.g. smoking, alcohol), body mass index, and serum biomarker panels (e.g. pepsinogen I/II ratio) were proposed as inputs in predictive models.
Some models developed (outside the U.S., e.g. in Korea) achieved moderate to good discriminative ability (Area Under Curve ~0.8), meaning they could meaningfully separate low vs. high gastric neoplasm risk cohorts.
Pilot studies and trial designs highlighted during the Summit also addressed how to validate these tools in real-world settings, including internal validation (bootstrapping) and external/clinical validation in diverse populations.
The tools are seen not just for selecting who should get endoscopy, but also for implementing sequential screening strategies—using less invasive or cheaper prescreen tests followed by more invasive ones only in selected high-risk people.
A particular challenge noted is ensuring that predictive tools are equitable, i.e. calibrated across racial, ethnic, and immigrant populations, so that high-risk minority groups are not under-identified. Also, tools must be feasible in terms of implementation, cost, patient acceptability, and access.