Abstract 1037: Deep learning-derived dietary patterns and gastric cancer risk in women: A large-scale prospective cohort study

Abstract Background & Aims: Machine learning can address limitations of principal components analysis by capturing nonlinear relationships, enhancing dietary pattern identification. This study aimed to identify dietary patterns using deep learning and cluster analysis and to assess their association with gastric cancer incidence in a prospective cohort study. Methods: Data were obtained from participants of the large-scale prospective cohort study, Health Examinees-Gem (HEXA-G) study, with baseline surveys conducted between 2004 and 2013. Cancer diagnosis information was retrieved from the Korea Central Cancer Registry up to December 31, 2018. The final analysis included 84, 725 women. Dietary intake was assessed using a validated semiquantitative food frequency questionnaire at baseline. Dimensionality reduction was performed using an autoencoder, followed by k-means clustering. The optimal cluster number was determined using the silhouette coefficient. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for gastric cancer risk associated with the dietary patterns. Results: Ten dietary patterns were identified: ‘Common’, ‘Low calorie’, ‘Rice’, ‘High calorie’, ‘Traditional’, ‘Dairy’, ‘Meat’, ‘Processed’, 'Noodles' and ‘Salty’ patterns. A total of 558 cases of gastric cancer were identified during median 9.4 years of follow-up. Compared to the ‘High calorie’ pattern (reference), the ‘Noodles’ (HR 1.97, 95% CI 1.05-3.70), and ‘Low calorie’ (HR 1.91, 95% CI 1.23-2.97) were associated with higher incidence of gastric cancer after adjusting for potential confounders. When stratified by age, ‘Salty’ pattern was associated with 4.34 times higher risk of gastric cancer (HR 4.34, 95% CI 1.49-12.60) among participants aged 40-49. Conclusion: Our results suggest that relative to the ‘High calorie’ pattern, the ‘Noodles’, ‘Low calorie’, and ‘Salty’ patterns are associated with increased risk of gastric cancer among Korean women. While Helicobacter pylori infection—a major risk factor for gastric cancer—was not accounted for in this study, our findings highlight the importance of dietary interventions to reduce high-risk patterns, such as salty or noodle-based diets, to lower the risk of gastric cancer. Citation Format: Hyobin Lee, Dongseok Heo, Sukhong Min, Sinyoung Cho, Bongwon Suh, Daehee Kang. Deep learning-derived dietary patterns and gastric cancer risk in women: A large-scale prospective cohort study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1037.