Healthy dietary patterns and the risk of individual chronic diseases in community-dwelling adults

The global population is aging rapidly, with individuals aged 65 years and older expected to reach 1.5 billion by 2050. Aging is a key risk factor for non-communicable diseases (NCDs) including cardiovascular disease (CVD), diabetes, cancers, and neurodegenerative conditions, contributing substantially to mortality and healthcare burden. Psychiatric/neurological and digestive disorders also impose major morbidity and economic costs. Healthy lifestyle factors, particularly diet, are actionable targets for prevention. Prior studies indicate that adherence to healthy dietary patterns such as the Mediterranean Diet, Healthy Eating Index, and DASH is associated with reduced risks of CVD, diabetes, and some cancers, though evidence for neurodegenerative diseases and many other chronic conditions remains inconsistent or sparse. This study aims to evaluate, in the UK Biobank cohort, the associations of four commonly used healthy dietary scores with the incidence of a wide range of 48 individual chronic diseases, and to assess which dietary score best predicts major chronic disease risk.

Previous research has linked healthier dietary patterns (e.g., Mediterranean diet, Healthy Eating Index, DASH) to lower risks of cardiometabolic diseases, diabetes, and some cancers. Evidence is inconsistent for neurodegenerative outcomes (dementia, Parkinson’s disease). Some studies suggest benefits for bone and musculoskeletal health, but evidence is limited for respiratory, digestive, renal, dermatologic, endocrine, and ophthalmic diseases. Inflammation-related indices (e.g., dietary inflammatory index) have shown mixed associations with certain outcomes like stroke. Overall, comprehensive evaluations across many individual diseases within one cohort have been lacking, motivating the present study.

Design and population: Prospective cohort analysis using the UK Biobank, a population-based cohort of >500,000 participants aged 39–70 years at enrollment (2006–2010). Ethical approvals were obtained and informed consent collected. For this analysis, participants were excluded if they had no diet data (n=295,101), only one dietary assessment (n=83,413), or total energy intake in the highest/lowest percentile (n=2,478). Final sample: 121,513 participants (55.9% female), aged 30–75 years (mean 59.0±7.9). Dietary assessment and exposure scores: Diet was assessed using the validated Oxford WebQ web-based 24-hour recall administered on multiple occasions between April 2009 and June 2012. Participants with ≥2 assessments were included; average intake across assessments was used. Four dietary scores were computed: - Alternate Mediterranean Diet (AMED): 9 components (whole grains, vegetables excluding potatoes, fruits, nuts/seeds, legumes, fish, monounsaturated-to-saturated fat ratio, red/processed meat [reverse], alcohol 5–15 g/day). Each component scored 0/1 (sex-specific medians); total 0–9, higher is healthier. - Empirical Dietary Inflammatory Index (EDII): Based on 18 food groups scored by associations with inflammatory biomarkers (IL-6, CRP, TNF-α receptor 2, adiponectin). To harmonize direction with other scores, the Anti-EDII (AEDII) was used by reversing EDII so higher is more anti-inflammatory. - Alternate Healthy Eating Index-2010 (AHEI-2010): 10 components (e.g., whole grains, vegetables, fruits, nuts/legumes, SSB/fruit juice, red/processed meat, fish/long-chain n-3, PUFA, added salt [sodium], alcohol); each 0–10, total 0–100. Trans fat not available. - Healthful Plant-based Diet Index (HPDI): 17 food groups; plant foods positively scored, animal foods inversely, and less healthful foods inversely; each component scored 1–5; total 17–85, higher is healthier. Vegetable oil not included due to unavailability. Outcomes: Incident cases of 48 chronic diseases across categories: cardiometabolic disorders (CVD and subtypes, hypertension, diabetes), cancers (overall and specific types), psychological/neurological disorders, digestive disorders, respiratory diseases, chronic kidney disease (CKD), dermatologic, endocrine, ophthalmic, musculoskeletal, hematologic, and other conditions. Baseline diseases were identified via self-report and inpatient records; incident events were ascertained through inpatient hospital records (available since 1997) and mortality registries. ICD codes for each outcome are provided in supplementary materials. Follow-up was from dietary baseline to incident diagnosis, death, or end of follow-up (Dec 31, 2020 for England/Wales; Jan 31, 2021 for Scotland). Covariates: Age, sex, ethnicity, education, household income, smoking, alcohol consumption, physical activity (IPAQ short form), sleep duration, BMI, total energy intake, and a genetic risk score (GRS) for longevity (78 SNPs). For lung cancer, additional smoking metrics (pack-years, age stopped smoking, current cigarettes/day) were included. Statistical analysis: Cox proportional hazards models estimated hazard ratios (HRs) and 95% CIs per quintile increment of each dietary score. Model 1 adjusted for age, sex, total energy intake; Model 2 additionally adjusted for ethnicity, education, income, BMI, smoking, sleep, physical activity, and GRS for longevity. Multiple imputation addressed missing covariates (10 imputed datasets). Benjamini–Hochberg false discovery rate (FDR) was controlled at 5% across multiple comparisons. Sensitivity analyses excluded cases occurring within first 4 years and restricted to participants with ≥3 dietary assessments. Moderation analyses assessed effect modification by age and metabolic status for selected associations.

- Cohort: 121,513 participants; mean follow-up varied by disease (≈8.4–8.6 years). Incident cases ranged from 94 (multiple sclerosis) to 9,815 (dyspepsia). - Overall breadth: AMED inversely associated with 32 of 48 diseases; AHEI-2010 with 29; HPDI with 23; AEDII with 14 inverse associations and 2 positive associations (alcohol use disorder, psychoactive substance abuse). Cardiometabolic disorders (CMDs): - All AMED, AHEI-2010, and HPDI were inversely associated with all individual CMDs after FDR control. AMED yielded the lowest HRs for most CMDs (except diabetes). - Per quintile increment (Model 2): AMED—CVD HR 0.94 (0.93–0.95), hypertension 0.94 (0.92–0.96), diabetes 0.96 (0.93–0.99). AHEI-2010—CVD 0.96 (0.94–0.97), hypertension 0.93 (0.91–0.95), diabetes 0.98 (0.95–1.01, not significant). HPDI significantly associated with several CMDs. AEDII showed inverse associations, notably strongest for diabetes HR 0.89 (0.86–0.92). Cancers: - All cancers per quintile (Model 2): AMED HR 0.93 (0.91–0.95), AHEI-2010 0.95 (0.93–0.97), HPDI 0.95 (0.93–0.97); AEDII 1.00 (0.98–1.02, null). - Specific cancers: AMED—lower risk of lung cancer 0.89 (0.84–0.94), oesophageal cancer 0.87 (0.78–0.96). AHEI-2010—lower risk of non-melanoma skin cancer 0.95 (0.93–0.98), lung cancer 0.92 (0.87–0.97), breast cancer 0.95 (0.92–0.99). HPDI—lower risk of ovarian cancer 0.89 (0.81–0.98) and colon cancer (per narrative), and other cancers. Psychological/neurological disorders: - Dementia: AMED 0.92 (0.87–0.98), AHEI-2010 0.93 (0.88–0.98). Parkinson’s disease: AMED 0.91 (0.85–0.98). Epilepsy: AMED 0.88 (0.82–0.95), AHEI-2010 0.92 (0.86–0.99), HPDI 0.92 (0.85–0.99). - Depression: AMED 0.92 (0.88–0.97), AEDII 0.95 (0.90–0.99), AHEI-2010 0.96 (0.92–0.99), HPDI 0.95 (0.91–0.99). Anxiety: AMED 0.92 (0.89–0.94), AEDII 0.95 (0.92–0.97), HPDI 0.96 (0.94–0.99). - Substance outcomes: Alcohol use disorder—AMED 0.81 (0.77–0.86), AHEI-2010 0.73 (0.69–0.78), HPDI 0.93 (0.88–0.98); AEDII positively associated 1.18 (1.12–1.25). Psychoactive substance abuse—AMED 0.81 (0.72–0.92), AHEI-2010 0.75 (0.66–0.85); AEDII positively associated 1.14 (1.01–1.28). Digestive disorders: - All four scores inversely associated with dyspepsia, treated constipation, diverticular disease, irritable bowel syndrome, and chronic liver disease. - Examples (Model 2, per quintile): Dyspepsia—AMED 0.95 (0.94–0.96), AHEI-2010 0.95 (0.93–0.96), HPDI 0.95 (0.93–0.96), AEDII 0.96 (0.93–0.98). Chronic liver disease—AMED 0.86 (0.78–0.94), AHEI-2010 0.91 (0.83–0.99), HPDI 0.88 (0.80–0.97), AEDII 0.87 (0.79–0.96). Other chronic diseases: - Respiratory: COPD—AMED 0.83 (0.81–0.85), AEDII 0.84 (0.82–0.87), AHEI-2010 0.83 (0.80–0.85), HPDI 0.86 (0.84–0.89). Asthma—AMED 0.95 (0.92–0.98), AHEI-2010 0.94 (0.91–0.97), AEDII 0.94 (0.91–0.97). - Renal: CKD—AMED 0.91 (0.89–0.93), AEDII 0.91 (0.89–0.93), AHEI-2010 0.94 (0.92–0.96), HPDI 0.91 (0.89–0.93). - Endocrine/Dermatologic/Ophthalmic/Hematologic: Thyroid disorders—only AEDII significant 0.93 (0.90–0.96). Psoriasis/eczema—AMED 0.92 (0.88–0.96). Cataract—HPDI 0.96 (0.94–0.99). Pernicious anemia—AMED 0.75 (0.66–0.86), AEDII 0.80 (0.70–0.92), AHEI-2010 0.80 (0.69–0.93), HPDI 0.73 (0.63–0.84). Prostate disorders—AMED 0.97 (0.95–0.99), AEDII 0.95 (0.93–0.97), AHEI-2010 0.97 (0.95–0.99), HPDI 0.96 (0.94–0.99). AMED components analysis: - Of 450 tested associations, 155 significant after FDR; 154 inverse. Recommended intakes of whole grains, vegetables, fruits, nuts, legumes, fish, higher MUFA:SFA ratio, and moderate alcohol associated with lower risks of many diseases; lower red meat intake associated with lower risks of diabetes, CKD, diverticular disease, osteoporosis, but a higher risk of Meniere’s disease. Moderation and sensitivity: - Stronger inverse associations in subgroups: e.g., AMED with IBS, osteoporosis, dyspepsia, and cataract among those with hypertension/dyslipidemia; AEDII more predictive of diabetes/CKD in younger individuals; AHEI-2010 associations with cancer/cataract modified by age/metabolic status; HPDI-hypertension association stronger in younger participants. Sensitivity analyses excluding early events and restricting to ≥3 diet assessments yielded similar results.

This study systematically evaluated four healthy dietary scores against 48 individual chronic diseases in a large, well-characterized cohort. Higher adherence to AMED, AHEI-2010, and HPDI was broadly protective across cardiometabolic, cancer, psychological/neurological, digestive, respiratory, renal, and several other conditions. AMED most consistently yielded the lowest risks across many outcomes, whereas AEDII provided the strongest inverse association for diabetes but showed positive associations with alcohol use disorder and psychoactive substance abuse, likely reflecting alcohol’s role in the inflammatory index. The components analysis suggests benefits linked to higher intakes of whole grains, vegetables, fruits, nuts, legumes, fish, favorable fat quality (higher MUFA:SFA), and moderate alcohol, alongside lower red/processed meat intake. These findings reinforce dietary guidelines emphasizing plant-forward, high-quality dietary patterns for primary prevention and healthy aging, and extend evidence to under-studied outcomes such as digestive disorders, COPD, CKD, and some ophthalmic and hematologic conditions.

Greater adherence to healthy dietary patterns—particularly the Alternate Mediterranean Diet—was associated with lower risks of numerous chronic diseases, including all cardiometabolic disorders, several cancers, multiple psychological/neurological and digestive disorders, respiratory diseases, CKD, osteoporosis, eczema, prostate disorders, cataract, and pernicious anemia. These results support public health recommendations promoting high-quality, plant-forward diets. Future research should clarify causality (e.g., via randomized trials or Mendelian randomization), refine dietary assessments, evaluate disease subtypes (e.g., aggressive vs non-aggressive prostate cancer), explore mechanisms (microbiome, inflammation, oxidative stress), and assess generalizability across diverse ethnic groups.

- Dietary intake measured via self-reported 24-hour recalls, susceptible to measurement error (likely biasing toward the null). - Observational design precludes causal inference and residual confounding may persist. - Incident diseases ascertained from inpatient and mortality data may underestimate true incidence and introduce detection bias (e.g., cancer screening variability, cataract severity). - Potential reverse causation for psychological outcomes (diet could be influenced by stress/anxiety prior to diagnosis). - Uniform confounder adjustment across outcomes may be suboptimal for some diseases. - Inability to distinguish aggressive from non-aggressive prostate cancers. - Prodromal phases (e.g., dementia) could precede diet assessment. - Broad disease scope limits mechanistic depth for any specific condition. - Predominantly Caucasian cohort limits generalizability to other populations.