Data-Driven Discovery of Plant-Based Dietary Subtypes and Their Association With Cancer, Cardiovascular Diseases, and Type 2 Diabetes in Postmenopausal Women: Insights From the Women's Health Initiative.

backgroundPlant-based diets have been associated with lower risks of chronic diseases; however, not all plant-based diets confer equal health benefits. Data-driven methods provide an opportunity to empirically identify distinct plant-based dietary subtypes and examine their associations with health outcomes.

objectiveTo categorize subtypes of plant-based dietary patterns using a data-driven approach and evaluate their associations with cardiovascular disease (CVD), type 2 diabetes (T2D), and cancer incidence among postmenopausal women in the Women's Health Initiative (WHI).

designProspective cohort study using principal component analysis and k-means clustering to empirically derive dietary subtypes. PARTICIPANTS/

settingA total of 823 postmenopausal women from the WHI Observational Study who were healthy at baseline and reported low meat or poultry consumption during screening (1993 to 1998) were included and followed prospectively for incident chronic disease outcomes during WHI follow-up and extension periods (2023). A stratified comparison group of 823 higher meat consumers, also drawn from the WHI Observational Study, was constructed using stratified random sampling on dietary pattern position and key demographic variables.

main outcome measuresIncident cases of CVD, T2D, and any type of cancer (except nonmelanoma skin cancer), verified through adjudicated clinical outcomes. STATISTICAL ANALYSES PERFORMED: Cox proportional hazards models estimated associations with CVD, T2D, and cancer outcomes, adjusting for sociodemographic, lifestyle, and clinical factors.

resultsThree dietary clusters were identified: Lower Meat: Fruit and Vegetable, Lower Meat: Dairy and Solid Fat, and Lower Meat: Fish and Meat (classified as lower meat based on screening but reporting relatively higher fish and moderate meat intake on the food frequency questionnaire). No statistically significant associations were observed between any lower meat dietary pattern and overall cancer incidence. Compared with higher meat consumers, the Fish and Meat cluster was associated with a significant 67% lower risk of T2D (HR 0.33; 95% CI, 0.14 to 0.81; P = .02). The Dairy and Solid Fat and Fruit and Vegetable clusters did not reach statistical significance for T2D risk (HR 0.77; 95% CI, 0.34 to 1.75 and HR 1.10; 95% CI, 0.46 to 2.61, respectively). For CVD, the Dairy and Solid Fat cluster was associated with a statistically significant higher risk (HR 2.63; 95% CI, 1.05 to 6.56; P = .04), whereas results for the Fruit and Vegetable cluster did not reach statistical significance (HR 0.47; 95% CI, 0.13 to 1.73; P = .26).

conclusionsThis combined PCA-clustering approach revealed meaningful heterogeneity within lower-meat dietary patterns, offering a novel framework for characterizing plant-based diets and improving precision in future diet-disease investigations.