Abstract
OBJECTIVE: To prospectively examine associations between five low-carbohydrate diets (LCDs), differentiated by macronutrient quality, and type 2 diabetes (T2D) risk. RESEARCH DESIGN AND METHODS: This cohort study included 199,006 U.S. adults from the Nurses' Health Study (NHS) (1984-2018), NHSII (1991-2019), and Health Professionals Follow-up Study (1986-2018); free of T2D, cardiovascular disease, and cancer at baseline; and followed over 30 years. Diet was assessed every 2-4 years using validated food frequency questionnaires since baseline. Five LCD scores were derived based on intakes of protein, fat, and carbohydrates from contrasting food sources. The primary outcome was incident T2D. RESULTS: During 4,987,761 person-years of follow-up, 20,452 T2D cases were documented. After adjustments for baseline BMI and other covariates, higher overall LCD score was associated with higher T2D risk (hazard ratio comparing highest vs. lowest quintile 1.31 [95% CI 1.25-1.37]; P-trend < 0.001). An animal-based LCD emphasizing animal protein and fat and an unhealthy LCD score further deemphasizing whole grains and other high-quality carbohydrates were associated with higher T2D risk (1.39 [1.32-1.45] and 1.44 [1.37-1.51]; both P-trend < 0.001). In contrast, a vegetable-based LCD emphasizing plant protein and fat was associated with a 6% lower T2D risk (0.94 [0.90-0.98]; P-trend = 0.004), and a healthy LCD further deemphasizing refined carbohydrates was associated with a 16% lower T2D risk (0.84 [0.81-0.88]; P-trend < 0.001]). Associations for overall, animal-based, and unhealthy LCDs were stronger among participants with lower baseline BMI and were partially mediated by weight change. CONCLUSIONS: LCDs may not be beneficial for primary prevention of T2D unless they prioritize plant-based protein, healthy fats, and high-quality carbohydrates.