Imperial College London OXFORD, England, United Kingdom
Objectives: Previous research has identified dietary patterns (DPs) and linked them to diabetes and cardiovascular diseases, but their associations with other diseases remain unclear.
Methods: We included UK Biobank participants with two or more 24-h online dietary assessments. DPs were derived using reduced rank regression with 50 food groups as predictors and four nutrients as responses (energy density, free sugar, saturated fat and fiber). Cox models examined the associations between DPs and 623 Phecode-based diseases, covering all disease categories, after excluding the participants with baseline relevant diseases. Incident outcomes were confirmed through death registry, cancer registry and hospitalisation datasets. Models were adjusted for sex, age, socioeconomic status, education, ethnicity, smoking and physical activity. False discovery rate correction addressed multiple comparisons.
Results: In 125095 participants (56.3% females, mean age 59.5 years), we derived three DPs, explaining 75.1% of variance in response nutrient variables (45.3%, 19.5% and 10.3% for DP1, DPD2, and DP3). DP1 featured higher intake of chocolate confectionary, fat spread, desserts and red meat, with lower fruits and vegetables intake. DP2 featured higher fat and red meat, with lower sweetened beverage, fruit juice, free sugar and alcoholic beverage. DP3 featured higher sugar, fat and vegetable, with lower alcoholic beverage, bread and pasta/rice.
During 9.2-year follow-up, DPs were associated with various diseases across categories. Specifically, DP1 was associated with 41 Phecode outcomes (e.g., obesity, circulatory diseases, colorectal cancer, mineral metabolism disorders, anxiety, pneumonia, respiratory insufficiency and failure, non-alcoholic fatty liver disease, cholelithiasis, renal failure). DP2 was associated with 14 Phecode outcomes (e.g., obesity, diabetes, pneumonia, gastritis and duodenitis, spondylosis). DP3 was associated with 29 Phecode outcomes (e.g., diabetes, hypercholesterolemia, gout, anaemia, septicaemia, non-Hodgkins lymphoma, anxiety).
Conclusions: The derived DPs exhibited associations with a broad spectrum of human diseases.