A network approach using Gaussian graphical models to evaluate dietary patterns and colorectal cancer risk in the UK Biobank

Background : Current research on the link between dietary patterns and colorectal cancer (CRC) has relied on methods that summarize intake without accounting for the structural dependencies among food groups. Gaussian graphical models (GGMs) derive dietary patterns by capturing how foods are consumed in relation to one another, offering a network-based view of overall eating behavior. Objectives : We aimed to identify dietary patterns using GGMs and to examine their associations with incident CRC in the UK Biobank. Methods : This prospective cohort study included 105,676 UK Biobank participants with 24-hour recall dietary data. Dietary patterns were identified using GGMs constructed from 43 food groups, and participants were classified into tertiles of each pattern score. Cox proportional hazards models were applied to estimate hazard ratios (HRs) and 95% confidence intervals (CIs), adjusting for demographic, lifestyle, and clinical factors. Results : Over a median follow-up of 10.1 years, 1,026 cases of CRC were newly diagnosed. The network identified four dietary patterns consisting of different food groups. In sex-stratified analyses, the processed food and dessert pattern was associated with higher CRC risk in male, with the highest tertile showing increased risk (HR = 1.23; 95% CI: 1.00-1.50; P -trend = 0.048). In female, the balanced pattern was inversely associated with CRC, with the highest tertile showing lower risk (HR = 0.82; 95% CI: 0.60-0.93; P -trend = 0.041). No significant associations were observed for the other dietary patterns, including the protein-based and beverage patterns, in the total or sex-stratified analyses. Conclusions : Network-based dietary patterns identified through GGMs showed different sex-specific associations with CRC risk. By addressing pairwise correlations between food variables while controlling for indirect effects from other foods, this approach captures dietary structures with greater specificity than traditional pattern methods. These results contribute to a clearer understanding of how sex-specific dietary patterns relate to CRC risk within existing epidemiologic evidence.