Purpose: To investigate socio-economic disparities in the association between dietary diversity and type 2 diabetes (T2D).

Methods: Prospective secondary analysis study of 10,363 incident cases of T2D over 12 years and a random sub-cohort of 13,937 individuals in the EPIC-InterAct case-cohort study. We used baseline self-reported diet data (1991–1998) to construct scores for diversity between food groups (range: 0–5) and diversity within subtypes of vegetables (0–4), and plant protein sources (0–5). Prentice-weighted Cox regressions estimated multivariable-adjusted hazard ratios (HR) and 95% confidence intervals of incident T2D for each diversity score stratified by education, employment, and marital status.

Results: Overall group differences were evident for education in the association between plant-protein diversity and incident T2D (p for group effect = 0.04). University-educated participants consuming three plant-protein subtypes had 35% lower T2D incidence (HR 0.65 [95% CI: 0.45, 0.95]). Greater vegetable diversity (3–4 subtypes vs 0–1) was associated with 61%, 22% and 18% lower T2D incidence for, respectively, those who were single, had primary education or were employed. Total diet diversity (5 food groups versus ≤ 3) was associated with 43%, 30%, 20% and 15% lower T2D incidence for, respectively, those who were single, employed, had primary education or were married.

Conclusion: Education modified the association between plant-protein diversity and T2D onset in European adults, and certain socioeconomic groups specifically benefited from higher diversity reducing T2D risk. Promoting diet diversity overall may support T2D prevention across populations, while tailored messages of plant-protein diversity may improve relevance and equity in nutrition and diabetes.