Higher habitual intakes of flavonoids and flavonoid-rich foods are associated with a lower incidence of type 2 diabetes in the UK Biobank cohort

Type 2 diabetes (T2D) is a major global non-communicable disease, contributing to over 4 million deaths annually, with approximately 90% of diabetes cases being T2D. Diet is crucial for primary prevention given the strong role of modifiable risk factors such as overweight and obesity. Prior research indicates that healthy plant-based dietary patterns may reduce T2D risk beyond effects on energy balance. Flavonoids—polyphenolic compounds abundant in fruits, vegetables, tea, cocoa/dark chocolate, and red wine—comprise several subclasses (flavanones, flavones, flavan-3-ols including proanthocyanidins, flavonols, anthocyanins, isoflavones) with distinct bioavailability and bioactivity. Observational studies have reported inverse associations between flavonoid subclasses and incident T2D, and short-term RCTs suggest that anthocyanins and flavan-3-ols improve insulin sensitivity and lipid profiles. This study investigated whether a flavonoid-rich diet, operationalized using a novel Flavodiet Score (FDS), is associated with incident T2D in the UK Biobank, and explored potential biological mediators. It also examined associations of major flavonoid-rich food contributors and flavonoid subclasses with T2D risk.

Previous prospective studies and meta-analyses have reported inverse associations between several flavonoid subclasses (e.g., flavanols, flavonols, flavan-3-ols, isoflavones) and T2D risk, with risk reductions up to around 14%. Evidence from RCTs indicates that consumption of flavonoid-rich foods such as cocoa, tea, and blueberries (high in anthocyanins and flavan-3-ols) can improve insulin sensitivity, insulin resistance, and lipid profiles. Findings on flavanones are mixed, potentially influenced by fruit juice intake which contributes substantially to flavanone intake and has been associated with higher T2D risk due to glycaemic load. Meta-analyses also suggest inverse associations for tea, berries, apples, grapes, and grapefruit with T2D risk, and mixed or null associations for oranges and onions. Mechanistic literature supports roles for flavonoids in modulating insulin signalling, inflammation, oxidative stress, and potentially liver and kidney function, which are relevant to T2D pathogenesis.

Design and population: UK Biobank prospective cohort of over 500,000 adults aged 40–69 years recruited 2006–2010. For this analysis, 113,097 participants with at least two 24-h dietary assessments (Oxford WebQ) were included. Exclusions: fewer than two dietary assessments, implausible energy intakes, withdrawal of consent, baseline diabetes, CVD, cancer, T2D diagnosed before last dietary assessment, and pregnancy at baseline. Exposure assessment: Flavonoid intakes (total and subclasses) were estimated using USDA flavonoid databases. A Flavodiet Score (FDS) was created from mean servings/day across ten flavonoid-rich food items identified as top contributors (tea [black/green; capped at 4 servings/day], red wine, apples, berries, grapes, oranges, grapefruit, sweet peppers, onions, dark chocolate). The FDS summed servings across items and was categorized into sex-specific quartiles; analyses were also run excluding red wine. Reproducibility and reliability: Intraclass correlation coefficients assessed FDS reproducibility over time; Spearman correlations assessed flavonoid intake reliability over time and agreement between USDA and Phenol-Explorer estimates. Covariates: Adjusted for demographic, lifestyle, clinical, dietary, and genetic factors including sex, age, region, education, BMI, waist circumference, ethnicity, physical activity, smoking, alcohol, energy intake, polypharmacy, multimorbidity, Townsend deprivation index, family history of diabetes, hypercholesterolemia, hypertension, menopausal status, T2D polygenic risk score (PRS), number of dietary assessments, wholegrain, sugar-sweetened beverages, red/processed meat, and coffee; alternatively, a healthful plant-based diet index in sensitivity analyses. Outcome ascertainment: Incident T2D identified via hospital inpatient records using ICD-10 code E11, with follow-up censored at hospitalisation, death, or end of follow-up (to 2021, varying by UK nation). Statistical analysis: Cox proportional hazards models estimated hazard ratios (HRs) and 95% CIs across sex-specific quartiles and per 1-point increment of FDS, flavonoid subclasses, and individual flavonoid-rich foods; linear trends tested. Model 1 adjusted for sex and education; stratified by age (5-year categories) and region. Model 2 further adjusted for the full set of covariates above. Multiple testing for food and subclass analyses controlled by Bonferroni correction. Subgroup analyses evaluated effect modification by smoking, sex, BMI, education, ethnicity, alcohol intake, and genetic risk (PRS tertiles), with likelihood ratio tests for interactions. Sensitivity analyses excluded participants with ≤2 years of follow-up and used the plant-based diet index. Mediation analyses assessed potential mediators (BMI, IGF-1, C-reactive protein, cystatin C, urate, GGT, ALT, and others) on the FDS–T2D pathway; cumulative mediation proportion estimated.

- Cohort: 113,097 participants; median follow-up ~12 years (1,358,384 person-years); 2,628 incident T2D cases. - FDS and T2D: Q4 (≈6 servings/day) vs Q1 (≈1 serving/day) associated with 28% lower T2D risk (Model 2 HR 0.72; 95% CI 0.64–0.81; Ptrend <0.001). Per 1-point FDS increment HR 0.94 (95% CI 0.92–0.97). Excluding red wine: Q4 vs Q1 HR 0.74 (95% CI 0.66–0.84; Ptrend <0.001); per 1-point increment HR 0.95 (95% CI 0.93–0.97). - Food-based associations (Q4 vs Q1, Model 2): tea (4 servings/day) HR 0.79 (95% CI 0.70–0.90; Ptrend <0.001); berries (1 serving/day) HR 0.85 (95% CI 0.74–0.98; Ptrend = 0.01); apples (1 serving/day) HR 0.88 (95% CI 0.79–0.98; Ptrend = 0.03). After Bonferroni correction, only tea remained statistically significant. - Flavonoid subclasses (Q4 vs Q1, Model 2): anthocyanins (~19% lower risk), flavan-3-ols (~26%), flavonols (~28%), flavones (~19%), polymers (~26%), and proanthocyanidins (~27%)—all significant after Bonferroni correction. Using Phenol-Explorer data, results were consistent except flavones lost significance after multivariable adjustment. - Mediation: Individual mediators explained small proportions—BMI 5%, IGF-1 2%, C-reactive protein 2%, cystatin C 4%, urate 2%, GGT 2%, ALT 4%; cumulative mediation ~28% of the FDS–T2D association. - Subgroups and sensitivity: No heterogeneity by smoking, sex, BMI, education, ethnicity, alcohol; inverse associations significant within intermediate and high genetic risk strata (PRS) without statistical heterogeneity. Results robust to excluding early cases and adjusting for overall plant-based diet quality.

The study demonstrates that a higher intake of flavonoid-rich foods, operationalized via the Flavodiet Score, is associated with substantially lower risk of incident T2D in a large, well-characterized cohort, independent of numerous sociodemographic, lifestyle, clinical, dietary, and genetic factors. The findings extend prior evidence on individual flavonoid subclasses by showing a diet-level measure (FDS) relates to lower T2D risk and that specific foods—particularly tea, with supportive evidence after multiple-testing correction—contribute to this association. Mediation analyses suggest that benefits may operate partly through lower adiposity and improved metabolic, inflammatory, kidney, and liver function, consistent with mechanistic literature on flavonoids’ effects on insulin signalling, oxidative stress, and organ function. The lack of effect modification by genetic risk underscores the potential for dietary flavonoid intake to reduce T2D risk across genetic backgrounds.

In this large prospective cohort, higher habitual intakes of flavonoid-rich foods, reflected by a higher Flavodiet Score, were associated with a substantially lower risk of incident type 2 diabetes, independent of established risk factors and genetic predisposition. The inverse association may be partly mediated by beneficial effects on obesity/sugar metabolism, inflammation, and kidney and liver function. The results support dietary guidance to increase fruit consumption and highlight a specific role for berries and apples, with tea as a key beverage contributor. Encouraging achievable increases in tea, berry, and apple intake may lower T2D risk.

- Generalisability: UK Biobank participants are predominantly White-European, middle-aged adults, limiting applicability to more diverse populations. - Dietary assessment: Potential recall bias, over-/under-reporting with 24-h recalls; limited food item granularity (e.g., berry types grouped), and possible misclassification when mapping foods to USDA flavonoid codes. - Measurement error: Single biomarker measurements may cause regression dilution; modest mediation proportion (~28%) suggests additional unmeasured mediators. - Residual confounding: Observational design cannot eliminate residual confounding. - Food intake distributions: High proportion of zero intake for dark chocolate may reduce power to detect associations. Sensitivity analyses excluding early cases suggested results were not due to reverse causality.