Prediction algorithm for gastric cancer in a general population: A validation study.

Area of Science:

• Oncology
• Epidemiology
• Biostatistics

Background:

• Gastric cancer is a significant global health concern, ranking high in incidence and mortality worldwide.
• Predicting gastric cancer risk is crucial for early detection and intervention strategies.
• A readily available clinical data-based scoring system is needed for risk stratification.

Purpose of the Study:

• To develop and validate a scoring system for predicting gastric cancer risk.
• To utilize easily accessible clinical data for risk assessment.
• To apply the system to a large Chinese population.

Main Methods:

• A retrospective study involving over 6 million subjects who underwent upper digestive endoscopy in Hong Kong (1997-2018).
• Binary logistic regression was used to identify gastric cancer predictors in a derivation cohort.
• Model performance was evaluated using receiver operating characteristic (ROC) curve analysis and area under the curve (AUC) in a validation cohort.

Main Results:

• Key predictors identified include age, male gender, Helicobacter pylori infection history, proton pump inhibitor use, and non-use of aspirin, NSAIDs, and statins.
• A scoring system was established, categorizing individuals into 'average risk' (≤8) and 'high risk' (≥9).
• The high-risk group showed a significantly higher prevalence of gastric cancer (1.81%) compared to the low-risk group (0.096%). The AUC in the validation cohort was 0.834.

Conclusions:

• A validated, simple, and accurate scoring algorithm for gastric cancer prediction has been developed.
• The scoring system demonstrates high discriminatory capability.
• This tool can aid in risk stratification for individuals suspected of gastric cancer, enabling prioritized investigations.