AI-Based Risk Prediction Model for Upper Digestive Tract Cancer

National Taiwan University

Status: Not yet enrolling

Conditions: Dysplasia Stomach; Atrophic Gastritis; Gastric Cancer (GC); Gastric Intestinal Metaplasia; Esophageal Cancer (EsC); Premalignant Lesion

Study type: Observational

Funder types: Other

Identifiers

NCT07605312

202105028RINC

Details and patient eligibility

About

Upper digestive tract cancers are often preceded by pre-malignant lesions, but there is limited evidence regarding optimal risk prediction models and screening strategies for disease progression and cancer development. This prospective multicenter cohort study aims to establish a longitudinal database integrating clinical information, endoscopic findings, pathology, genetics, epigenetics, and gastrointestinal microbiota data from subjects undergoing upper digestive tract endoscopy.

The study will develop explainable artificial intelligence (AI)-based risk prediction models to identify factors associated with disease progression, treatment response, and cancer development. Participants will be followed longitudinally to evaluate changes in lesion severity and clinical outcomes.

Full description

Objectives:

There is no solid evidence about the risk prediction model and screening duration for upper digestive tract pre-malignant lesions and its progression. There is also no longitudinal study combining multi-omic approach, endoscopic and pathologic images and the association with disease development. Hence we design a prospective cohort targeting upper digestive tract disease progression and cancer development, with standardized clinical data collection, quality control and explainable AI (artificial intellegence) model for better reliability of risk prediction model.

Aims:

We aim to develop risk prediction model for the progression of upper digestive tract disease and cancer development.

Methods:

The study is disigned as a multi-center prospective cohort, targeting subjects undergoing upper digestive tract endoscopy. The development of AI risk prediction models will combine endoscopic pre-malignant lesion, pathology, genetics, epigenetics, oro-gastro-intestinal microbiota, and follow-up longitudinally with change in lesion severity, medication response, cancer development.

Outcome measurement:

Primary endpoints: upper digestive tract cancer development. Secondary endpoints: progression in pre-malignant lesions, recurrent colon polyps, other cancer developement, metabolic and cardiovascular disease, response to medication in gastro-esophageal reflux and dyspepsia population.

Enrollment: 10,000 estimated patients

Sex: All

Ages: 40+ years old

Volunteers: Accepts Healthy Volunteers

Inclusion criteria

• Patients undergoing upper gastrointestinal endoscopy.

• Patients with at least one of the following conditions or indications:

  • Previous or current Helicobacter pylori infection (confirmed by serology, histopathology, urea breath test, rapid urease test, or stool antigen test);
  • Dyspeptic symptoms;
  • Gastroesophageal reflux disease;
  • History of oral, oropharyngeal, or hypopharyngeal squamous cell carcinoma;
  • Barrett's esophagus;
  • Gastric premalignant lesions (intestinal metaplasia or atrophic gastritis);
  • Gastric subepithelial lesions.

Exclusion criteria

-