Trends in dietary patterns over the last decade and their association with long-term mortality in general US populations with undiagnosed and diagnosed diabetes

Diabetes mellitus is a leading cause of disability and economic burden worldwide, with an estimated 536 million adults affected in 2021 and projections reaching 783.7 million by 2045. A substantial proportion of diabetes is undiagnosed (about 44.7%), delaying timely care and increasing complication risks. Diet and overall dietary patterns are closely linked to diabetes risk and are key components of lifestyle management. The Healthy Eating Index (HEI) reflects adherence to the Dietary Guidelines for Americans and has been associated with lower cardiovascular disease and diabetes risk, while the Dietary Inflammatory Index (DII) quantifies dietary inflammatory potential and higher values have been linked to increased diabetes risk. However, evidence on how dietary patterns differ between undiagnosed and diagnosed diabetes and how these patterns relate to long-term prognosis is limited. This study aimed to quantify trends in HEI and DII in US adults over the last decade by diabetes status (no diabetes, undiagnosed, diagnosed) and to evaluate their associations with long-term all-cause and cause-specific mortality, informing lifestyle interventions, particularly for undiagnosed individuals.

Prior research indicates that certain dietary patterns and foods influence diabetes risk: Mediterranean diet adherence is associated with a lower risk of diabetes, while processed red meat, sugar-sweetened beverages, and higher dietary inflammatory potential are associated with increased risk. HEI has been linked to lower cardiovascular and diabetes risk, though its relation to diabetes prognosis required further investigation. DII, developed to quantify inflammatory potential of diet, has been associated with obesity, diabetes, and cardiovascular disease, and higher DII predicts greater diabetes risk. Despite established diet–diabetes incidence links, data on dietary patterns in undiagnosed diabetes and on dietary patterns’ association with long-term prognosis in diabetes populations have been sparse.

Design and data source: Retrospective cohort analysis using NHANES 2007–2018, a nationally representative survey with complex multistage probability sampling. Inclusion: adults ≥20 years, non-pregnant, with at least one valid 24-hour dietary recall. For survival analyses, NHANES 2007–2014 participants with mortality follow-up were included. Exposure assessment: Dietary data from validated 24-hour recalls were used. HEI-2015 (13 components: total vegetables, greens & beans, total fruits, whole fruits, whole grains, dairy, total protein foods, seafood & plant proteins, fatty acids, sodium, refined grains, saturated fats, added sugars) was computed using standard HEI scoring algorithms. DII was calculated per established methodology using available nutrients and energy (including B vitamins, vitamin C/E, minerals such as Mg/Fe/Zn/Selenium, carotenoids, omega-3/omega-6 fatty acids, etc.); higher DII indicates more pro-inflammatory diets. Outcomes: Primary endpoint was all-cause mortality; secondary endpoints were deaths due to malignant neoplasms (ICD-10: C00–C97), heart disease (I00–I09, I11, I20–I51), chronic lower respiratory diseases (J40–J47), and cerebrovascular diseases (I60–I69). Mortality was ascertained via National Death Index linkage through December 31, 2019. Diabetes classification: Diagnosed diabetes was defined by self-reported physician diagnosis or diabetes medication/insulin use; undiagnosed diabetes by meeting glycemic thresholds without self-reported diagnosis (fasting glucose ≥7.0 mmol/L, 2-hr OGTT glucose ≥11.1 mmol/L, or HbA1c ≥6.5%). Covariates: age, sex, education, BMI, smoking, hypertension and hyperlipidemia (using standard clinical/lipid thresholds or medication use), and alcohol use (definitions for moderate/heavy and binge drinking specified). Statistical analysis: NHANES sampling weights, clustering, and stratification were applied. Baseline characteristics were summarized as weighted means with 95% CIs or weighted proportions; group comparisons used weighted generalized linear models, chi-square, or logistic regression as appropriate. HEI and DII were analyzed as continuous variables (including per SD increase) and categorized into tertile-like groups: HEI (low ≤44.4; medium 44.4–56.8; high >56.8) and DII (low ≤0.81; medium 0.81–2.65; high >2.65). Weighted Cox proportional hazards models estimated associations with mortality, with age-adjusted and multivariable-adjusted models (adjusted for age, sex, education, BMI, smoking, hypertension, hyperlipidemia, alcohol consumption, and diabetes status). Two-sided p<0.05 denoted significance. Analyses used R 4.0.3.

- Diabetes prevalence among US adults increased over the last decade, reaching 15.55% in 2017–2018; overall weighted diabetes prevalence was 14.47%, including 4.54% undiagnosed. - Individuals with diabetes (both undiagnosed and diagnosed) were older, had higher BMI and waist circumference, worse metabolic profiles (higher SBP, HbA1c, triglycerides; lower HDL-C), and more hypertension/hyperlipidemia than those without diabetes. - HEI scores declined over recent years across all groups. Undiagnosed diabetes had significantly lower HEI than diagnosed diabetes: undiagnosed 50.85 (95% CI 49.79–51.36) vs diagnosed 51.59 (95% CI 50.93–52.25). - Participants with undiagnosed or diagnosed diabetes had higher DII scores than those without diabetes, indicating more pro-inflammatory diets; DII tended to increase after 2010. - Mortality incidence (weighted): all-cause 9.36%; malignant neoplasms 2.31%; heart disease 2.21%; chronic lower respiratory diseases 0.58%; cerebrovascular diseases 0.46%. - Higher DII was associated with increased mortality: per 1 SD increase associated with higher all-cause mortality (adjusted HR 1.055, 95% CI 1.002–1.083) and per 1-unit increase associated with higher heart disease mortality (adjusted HR 1.076, 95% CI 1.021–1.135). Compared with low DII, medium DII had adjusted HR 1.194 (95% CI 1.068–1.336) and high DII HR 1.253 (95% CI 1.115–1.407) for all-cause mortality. - Higher HEI was associated with reduced mortality: per 1 SD increase, all-cause mortality adjusted HR 0.903 (95% CI 0.859–0.950) and heart disease mortality HR 0.891 (95% CI 0.800–0.992). High vs low HEI was associated with lower all-cause mortality (adjusted HR 0.804, 95% CI 0.708–0.913). - Associations of HEI and DII with mortality were consistent across diabetes diagnostic groups (no significant interactions; p for interaction >0.05).

The study addressed two key gaps: characterizing dietary patterns in undiagnosed versus diagnosed diabetes and linking these patterns to long-term mortality in a nationally representative US cohort. The findings show that undiagnosed diabetes is associated with poorer overall diet quality (lower HEI) and more pro-inflammatory diets (higher DII) than diagnosed diabetes, possibly reflecting the impact of clinical counseling after diagnosis. Declining HEI trends across all groups suggest waning adherence to dietary guidance, paralleling reported declines in glycemic and cardiometabolic risk factor control in US adults with diabetes. Importantly, better adherence to dietary guidelines (higher HEI) was independently associated with lower all-cause and heart disease mortality, while higher dietary inflammatory potential (higher DII) was linked to greater mortality risk, regardless of diabetes status and after adjusting for major confounders. These results underscore the clinical significance of dietary quality and inflammatory potential in influencing prognosis and support reinforcing dietary management as part of comprehensive diabetes care. They also highlight the need for earlier detection and targeted lifestyle interventions for undiagnosed individuals to mitigate progression and complications.

This study, using nationally representative NHANES data, is among the first to compare dietary patterns in undiagnosed and diagnosed diabetes and to link these patterns to long-term mortality. Undiagnosed diabetes exhibited poorer diet quality and higher dietary inflammatory potential than diagnosed diabetes. Higher HEI was associated with lower all-cause and heart disease mortality, while higher DII predicted higher mortality, with consistent effects across diabetes groups. Given the rising diabetes prevalence and declining dietary quality, strengthening dietary counseling and interventions is essential. Future research should test DII-guided dietary interventions, incorporate broader dietary pattern assessments (e.g., Mediterranean diet), and include inflammatory biomarkers to elucidate mechanistic pathways.

- Dietary indices (HEI and DII) were derived from single 24-hour dietary recalls, which may not reflect habitual intake and could underestimate true dietary patterns. - Other dietary patterns (e.g., Mediterranean diet) were not analyzed due to statistical limitations. - Inflammatory biomarkers (e.g., IL-6, TNF-α) were unavailable, limiting direct assessment of diet–inflammation pathways. - As an observational retrospective cohort, residual confounding and measurement error may persist despite adjustment and weighting.