Healthy Eating for the Planet, Cardiovascular Health, and Longevity: An Observational Study

Abstract

Background

There is limited evidence on the impact of the environmentally sustainable diet proposed by the EAT‐Lancet Commission on cardiovascular disease (CVD) and death. We aimed to investigate the association between the EAT‐Lancet diet and cardiovascular morbidity and death in the United States.

Methods

We included 13 444 US middle‐aged adults from the ARIC (Atherosclerosis Risk in Communities) study who were free of CVD at baseline, with dietary data collected at visit 1 (1987–1989) and visit 3 (1993–1995). We assessed adherence to the EAT‐Lancet reference diet using the Planetary Health Diet Index (PHDI), which ranged from 0 to 135, with higher scores indicating greater adherence. Associations between PHDI and risk of CVD (a composite outcome of coronary heart disease, stroke, and heart failure), CVD deaths, and total deaths were evaluated using Cox proportional hazards regression.

Results

After a median follow‐up of 29 years, we documented 5074 total CVD events, 2512 deaths caused by CVD, and 8436 total deaths. The mean PHDI score was 76 (range, 30–113). Participants in the highest versus lowest quintile of PHDI had a 13% lower risk of total CVD (P‐trend <0.001). A 20‐point higher PHDI was associated with 13%, 16%, and 9% lower risk of coronary heart disease, stroke, and heart failure, respectively (all P≤0.02). It was also associated with 13% and 10% lower risk of CVD death and all‐cause death, respectively (both P≤0.003).

Conclusions

A dietary pattern that promotes planetary health was associated with a lower risk of CVD morbidity and death in a general population.

Nonstandard Abbreviations and Acronyms

ARIC

Atherosclerosis Risk in Communities

PHDI

Planetary Health Diet Index

Research Perspective.

What Is New?

• Among middle‐aged, predominantly Black and White women and men in the United States, greater adherence to the EAT‐Lancet diet, measured by the Planetary Health Diet Index, was associated with a lower risk of total cardiovascular disease, coronary heart disease, stroke, heart failure, cardiovascular death, and all‐cause death over nearly 3 decades of follow‐up.

• Our findings suggest that the EAT‐Lancet diet, a dietary pattern designed to promote both human and planetary health, improves cardiovascular health and longevity.

What Question Should Be Addressed Next?

• Future research is needed to better clarify the biological pathways linking the EAT‐Lancet diet to lower cardiovascular disease risk and death.

Diet, planet, and human health are deeply interconnected. ¹ Driven by rising income and urbanization, imbalanced diets high in refined sugars, saturated fats, and red and processed meats, and low in whole grains, fruits, and vegetables, are increasingly replacing traditional diets. ¹ This global nutrition transition leads to excessive energy intake and contributes to rising obesity and noncommunicable disease rates among both children and adults, reducing global life expectancies. ² , ³ , ⁴ In addition, this transition in the food system accounts for a third of anthropogenic greenhouse gas emissions, as well as global land degradation and water pollution from intensive food production. ⁵ , ⁶ With the global population expected to reach nearly 10 billion by 2050, having sustainable access to a healthy diet is a major challenge. ⁷ However, most national dietary guidelines remain incompatible with global health and sustainability targets on food‐related planetary boundaries. ⁸

In 2019, the EAT‐Lancet Commission proposed an environmentally sustainable diet based on existing evidence of the role of diet in both human and planetary health. ⁷ In addition to minimizing food waste and enhancing agricultural practices, adherence to this diet could sustainably feed the growing population, reduce greenhouse gas emissions by 50% by 2050, and prevent 11 million premature deaths each year. ⁷ , ⁹ In response to the EAT‐Lancet Commission's recommendations, several dietary measures have been developed to assess adherence and evaluate health impacts in European, South American, and Asian populations. ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ However, many of these indices are limited by simplistic scoring that reduces sensitivity and power to detect associations, while some fail to prioritize key plant‐based foods like whole grains and legumes. ¹⁰ , ¹² , ¹⁴ , ¹⁵ , ¹⁶ Recently, a novel dietary measure, the Planetary Health Diet Index (PHDI), was developed within 3 large prospective US cohorts to comprehensively quantify adherence to the EAT‐Lancet reference diet. ⁷ , ⁹ The PHDI included 15 food groups, with a scoring range from 0 to 140, and has been inversely associated with the risk of incident cardiovascular disease (CVD), total deaths, and cause‐specific deaths. ⁹ , ¹⁷ However, studies using the PHDI to robustly assess the degree of adherence to the EAT‐Lancet diet within the US population have been limited in number and have primarily focused on highly educated White health professionals, raising concerns of the generalizability of these findings to the broader US population.

The EAT‐Lancet diet differs in several ways from the Dietary Guidelines for Americans, 2020–2025. Due to its emphasis on environmental sustainability, the EAT‐Lancet diet favors high‐quality plant‐based foods, limits the intake of animal‐sourced products such as meat and dairy to low or moderate levels, and restricts the consumption of added sugar and saturated fat. ⁷ In contrast, while the Dietary Guidelines for Americans recommends similar food groups, it has fewer restrictions on animal‐sourced and refined products. ¹⁸ Given the high demand for meat in the typical American diet, applying the PHDI to the broader US population is necessary to assess the potential health impact of the EAT‐Lancet diet in the American context.

This study aimed to examine the associations between the PHDI and the risk of total CVD, coronary heart disease (CHD), stroke, heart failure, CVD death, and all‐cause death in the well‐characterized ARIC (Atherosclerosis Risk in Communities) study.

Methods

Data and code will be made available upon request and approval through the National Heart, Lung, and Blood Institute Biologic Specimen and Data Repository Information Coordinating Center or the ARIC Coordinating Center.

Study Design and Population

The ARIC study is an ongoing, community‐based prospective cohort that enrolled 15 792 men and women aged 45 to 64 years, including predominantly Black and White individuals with varied education and income levels. ¹⁹ Participants were recruited from 4 US study sites (Washington County, Maryland; Forsyth County, North Carolina; suburbs of Minneapolis, Minnesota; and Jackson, Mississippi). Visit 1 occurred between 1987 and 1989, subsequently followed by visits 2 (1990–1992), 3 (1993–1995), 4 (1996–1998), 5 (2011–2013), 6 (2016–2017), 7 (2018–2019), 8 (2020), 9 (2021–2022), 10 (2023), and 11 (2024‐2025). At visit 1, participants completed questionnaires regarding demographic, socioeconomic, lifestyle, and clinical information. Dietary intake was assessed at both visit 1 and visit 3. The institutional review board at each participating site approved the study protocol, and all participants provided written informed consent.

At visit 1, we excluded participants with more than 10 missing items on the food frequency questionnaire and reported unrealistic energy intake (men, <600 or >4200 kcal/d; women, <500 or >3600 kcal/d). Then, we excluded participants with missing covariates, missing food components of the PHDI, a self‐reported history of CHD, stroke, or heart failure, and those without outcome information on incident CVD, CVD death, and all‐cause death at follow‐up visits. The final analytic sample consisted of 13 444 participants (Figure S1).

Dietary Assessment

Participants' usual dietary intake was assessed by trained interviewers at visits 1 and 3 using a modified 66‐item semiquantitative Willett questionnaire. ²⁰ Participants reported the frequency of consuming prespecified portions of each food item over the past year, with frequency categories ranging from “almost never” to “>6 per day.” Visual cues, such as measuring cups and glasses in various sizes, were provided to help visualize portion sizes. Nutrient intake was calculated by multiplying the frequency of reported intake by the nutrient content of each food item, based on US Department of Agriculture data.

Planetary Health Diet Index

We assessed adherence to the EAT‐Lancet reference diet by applying the previously developed PHDI using ARIC study dietary data. ⁷ , ⁹ The PHDI reflects the environmentally sustainable diet proposed by the EAT‐Lancet Commission, which emphasizes high‐quality plant‐based foods, limits animal‐sourced products like meat and dairy, while restricting added sugar and saturated fat. ⁹ When adapting the PHDI to the ARIC population, we included all food components except soy foods, which were not captured by the food frequency questionnaire.

Briefly, we classified food and beverages into 14 components: (1) whole grains, (2) vegetables, (3) whole fruit, (4) fish and shellfish, (5) nuts, (6) nonsoy legumes, (7) unsaturated added fat, (8) dairy food, (9) red and processed meat, (10) tubers, (11) chicken and poultry, (12) eggs, (13) saturated fat, and (14) added sugar and sugar from fruit juice (Table S1). Each food component was scored from 0 to 10, except for the nonsoy legumes component, which was assigned a weight of 5. Higher scores represented greater planetary health benefit. Scoring thresholds were proposed in the EAT‐Lancet report, previously modified for the PHDI, and based on the dose–response association between food components and disease risk. ⁷ , ⁹ Theoretical scores for total PHDI in the ARIC study ranged from 0 to 135, with higher scores indicating greater adherence to the PHDI.

We calculated PHDI scores from food frequency questionnaire responses at visit 1 and visit 3 separately. From visit 1 to visit 3, the risk of each outcome was calculated in relation to visit 1 dietary intake. From visit 3 onward, we averaged the PHDI scores at visits 1 and 3, to improve precision and better reflect long‐term intake, for participants who had complete food component data at visit 3. Otherwise, dietary scores calculated from visit 1 were used. ²¹ In a secondary analysis, we treated PHDI as a time‐varying exposure by updating dietary intake at visit 3 without averaging the score across visits 1 and 3.

Outcome Assessments

The primary outcomes of interest were incident total CVD, CHD, stroke, heart failure, CVD death, and all‐cause death. Incident total CVD was defined as a composite outcome of CHD, stroke, and heart failure from baseline to December 31, 2022. CVD events were ascertained through annual telephone calls with participants or their proxies, active surveillance of hospital discharge records and state death registries, and linkage to the National Death Index. An expert panel adjudicated all CHD and stroke cases. Incident CHD was defined as a definite or probable myocardial infarction, which required evidence of heart muscle damage based on clinical symptoms, elevated cardiac enzymes, and abnormal ECG readings, or fatal CHD. ²² Incident stroke was defined as a definite or probable stroke, which required evidence of the sudden onset of neurological symptoms lasting >24 hours or resulting in death, without an alternative cause. ²³ Incident heart failure was defined as the first hospitalization or death related to heart failure, with International Classification of Diseases, Ninth Revision (ICD‐9) code 428 or Tenth Revision (ICD‐10) code I50. All‐cause death was defined as deaths attributable to any cause and was ascertained through annual follow‐up telephone calls (semiannual since 2012) with participants or their proxies, and by linkage to local death records and National Death Index up to December 31, 2022. ²⁴ CVD death was defined as deaths with ICD‐9 codes 390 to 459 or ICD‐10 codes I00 to I99. ²⁵

Covariate Assessment

At visit 1, participants self‐reported sociodemographic information (date of birth, sex, race, education level, health insurance status) and health behaviors (smoking status, alcohol consumption, and physical activity). Leisure‐time physical activity was assessed by a modified Baecke questionnaire, with scores ranging from 1 (least active) to 5 (most active). ²⁶ , ²⁷ Cholesterol‐lowering medication use and health conditions (hypertension status, diabetes status, and kidney function) were also collected. Hypertension was defined as a sitting diastolic blood pressure ≥90 mm Hg, a systolic blood pressure ≥140 mm Hg, or self‐reported use of antihypertensive medication in the past 2 weeks. Diabetes was defined as a fasting blood glucose level ≥126 mg/dL, nonfasting blood glucose concentration ≥200 mg/dL, a self‐reported physician diagnosis, or use of diabetes medication in the past 2 weeks. Current use of cholesterol‐lowering medication was based on self‐report of use in the 2 weeks before the visit. Kidney function was assessed using the estimated glomerular filtration rate, calculated with the 2021 Chronic Kidney Disease Epidemiology Collaboration creatinine‐based equation, excluding race. ²⁸ Trained staff measured participants' weight and height to calculate body mass index.

Statistical Analysis

We examined the means±SDs and proportions for participant characteristics and dietary intake of PHDI food components across quintiles of PHDI. Cox proportional hazards regression models were used to calculate hazard ratios (HRs) and 95% CIs for the association between PHDI and the risk of incident total CVD, CHD, stroke, heart failure, CVD death, and all‐cause death, using quintile 1 as the reference group. Time on study was used as the time metric, with follow‐up time calculated from visit 1 until the occurrence of events or death for each individual outcome or administrative censoring on December 31, 2022. We used Kaplan–Meier survival curves to compare survival across PHDI quintiles for each outcome and assessed differences using the log‐rank test.

We incrementally adjusted the multivariable regression models and used a directed acyclic graph to visualize relationships between variables (Figure S2). Model 1, the minimally adjusted model, included age (continuous); sex (female, male); race‐center, a variable representing race and study center; and total energy intake (continuous). Model 2 additionally adjusted for education level (less than high school, high school graduate, some college or more), health insurance status (yes, no), physical activity (continuous), smoking status (current, former, never), and alcohol consumption per week (continuous). Model 3 further adjusted for clinical factors, including cholesterol‐lowering medication use, hypertension status (yes, no), and kidney function (modeled using 2 linear spline terms with 1 knot at 90 mL/min per 1.73 m²). Model 4, the primary model, also included diabetes status (yes, no) and body mass index (continuous). We tested for trends across PHDI quintiles using the quintile median value and examined continuous associations for each primary outcome per 20‐unit higher in PHDI. The proportional hazards assumption was evaluated using smoothed Schoenfeld residual plots for all covariates. The curves were approximately flat over time, indicating no meaningful violations of the assumption. E value was used to assess the potential impact of unmeasured confounding. We also explored the shape of the associations using restricted cubic splines, with 4 knots at the 5th, 35th, 65th, and 95th percentiles of PHDI to capture potential nonlinearity while avoiding overfitting. The reference was set at the 10th percentile, the median of the first quintile of PHDI. To assess linearity, we conducted likelihood‐ratio tests comparing a linear model to a restricted cubic spline model for each health outcome.

To evaluate the robustness of associations across subgroups, we conducted stratified analyses by sex (women versus men) and race (White versus non‐White). We used likelihood ratio tests to compare model 4 results with and without the inclusion of an interaction term. We also examined the association between individual food components of the PHDI and total CVD, CVD death, and all‐cause death, by comparing the HRs for quintile 5 versus quintile 1 of each specific food component score.

As a sensitivity analysis, we explored the association between PHDI and primary outcomes after excluding events that occurred within the first 5 years of follow‐up, to reduce the potential impact of reverse causation. Additionally, we assessed associations after excluding participants who developed hypertension or diabetes between visit 1 and visit 3, to examine the potential influence of dietary changes following disease diagnosis. To account for the competing risk of death from non‐CVD causes, we applied the Fine–Gray subdistribution hazard model to evaluate the association between PHDI and CVD death. Finally, we treated PHDI as a time‐varying exposure by using only the visit 3 dietary data to estimate the disease risk from visit 3 onward.

All analyses were performed using R Studio version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria) and Stata version 16.0 (StataCorp, College Station, TX).

Results

Participant Characteristics and Dietary Intake

Among the 13 444 participants, PHDI scores ranged from 30 to 113, with a mean score of 76 at baseline. The mean age was 54±6 years, and the mean body mass index was 27±6 kg/m². Compared with participants in the lowest quintile of PHDI, those in the highest quintile were more likely to be older, women, non‐White individuals, former or never smokers, and physically active and have higher education, health insurance, and lower total energy intake (Table 1). They were also more likely to use cholesterol‐lowering medications, have hypertension and diabetes, and consume less alcohol. In terms of usual dietary intake, individuals in the highest quintile of PHDI consumed more whole grains, vegetables, whole fruits, fish and shellfish, nuts, dairy food, and chicken and other poultry. They consumed less red and processed meat, tubers, unsaturated and saturated fat, and added sugar and sugar from fruit juice.

Table 1.

Baseline (1987–1989) Characteristics and Dietary Intake According to Quintile of PHDI in the ARIC Study^*

	Overall population (N=13 444)	Quintile 1 (N=2689)	Quintile 2 (N=2689)	Quintile 3 (N=2689)	Quintile 4 (N=2689)	Quintile 5 (N=2688)
PHDI, median (range)	76.0 (29.6–113.3)	61.8 (29.6–66.7)	70.2 (66.7–73.2)	76.0 (73.2–78.7)	81.7 (78.8–84.8)	89.3 (84.8–113.3)
Age, y	53.9±5.7	53.1±5.7	53.7±5.6	54.0±5.8	54.3±5.7	54.7±5.8
Female sex	7539 (56.1)	1079 (40.1)	1332 (49.5)	1558 (57.9)	1730 (64.3)	1840 (68.5)
Race
White	10 029 (74.6)	2153 (80.1)	2005 (74.6)	1960 (72.9)	1955 (72.7)	1956 (72.8)
Black	3372 (25.1)	530 (19.7)	675 (25.1)	722 (26.9)	724 (26.9)	721 (26.8)
Other†	43 (0.3)	6 (0.2)	9 (0.3)	7 (0.3)	10 (0.4)	11 (0.4)
Center
Forsyth County, NC	3536 (26.3)	668 (24.8)	630 (23.4)	688 (25.6)	728 (27.1)	822 (30.6)
Jackson, MS	2958 (22.0)	463 (17.2)	594 (22.1)	641 (23.8)	643 (23.9)	617 (23.0)
Minneapolis, MN	3577 (26.6)	721 (26.8)	780 (29.0)	725 (27.0)	691 (25.7)	660 (24.6)
Washington County, MD	3373 (25.1)	837 (31.1)	685 (25.5)	635 (23.6)	627 (23.3)	589 (21.9)
Education level
Less than high school completion	2928 (21.8)	704 (26.2)	620 (23.1)	581 (21.6)	534 (19.9)	489 (18.2)
High school or vocational school	5550 (41.3)	1177 (43.8)	1111 (41.3)	1098 (40.8)	1111 (41.3)	1053 (39.2)
Some college or more	4966 (36.9)	808 (30.0)	958 (35.6)	1010 (37.6)	1044 (38.8)	1146 (42.6)
Health insurance	12 202 (90.8)	2446 (91.0)	2418 (89.9)	2408 (89.6)	2455 (91.3)	2475 (92.1)
Smoking status
Current smoker	3437 (25.6)	929 (34.5)	764 (28.4)	695 (25.8)	571 (21.2)	478 (17.8)
Former smoker	4239 (31.5)	804 (29.9)	847 (31.5)	825 (30.7)	867 (32.2)	896 (33.3)
Never smoker	5768 (42.9)	956 (35.6)	1078 (40.1)	1169 (43.5)	1251 (46.5)	1314 (48.9)
Body mass index, kg/m2	27.5±5.2	27.2±5.0	27.4±5.2	27.7±5.3	27.6±5.2	27.4±5.4
Physical activity index, 1–5	2.4±0.8	2.3±0.8	2.4±0.8	2.4±0.8	2.5±0.8	2.6±0.8
Total energy intake, kcal/day	1626±605	1827±674	1669±622	1606±589	1552±563	1477±507
Cholesterol‐lowering medication use	325 (2.4)	34 (1.3)	43 (1.6)	61 (2.3)	79 (2.9)	108 (4.0)
Alcohol consumption, g/week	43.1±95.3	56.9 (117.0)	48.5±101.7	44.0±92.8	35.8±81.6	30.2±75.4
Hypertension status	4191 (31.2)	721 (26.8)	868 (32.3)	888 (33.0)	866 (32.2)	848 (31.5)
Diabetes status	1357 (10.1)	187 (7.0)	254 (9.4)	277 (10.3)	298 (11.1)	341 (12.7)
eGFR, mL/min per 1.73 m2	102±13	103±13	102±13	102±12	102±13	102±12
Food group intake per day, median (25th–75th percentile)
Whole grains, g	36.4 (15.5–80.9)	15.5 (2.4–36.0)	28.4 (12.7–51.6)	36.4 (16.8–67.1)	51.6 (28.4–92.5)	80.0 (43.9–110.3)
Vegetables, g	105.7 (65.6–163.8)	79.1 (46.2–123.8)	91.1 (58.0–137.5)	104.9 (67.6–162.6)	116.6 (76.7–173.1)	147.1 (95.1–210.0)
Whole fruits, g	166.4 (74.1–283.5)	64.1 (26.6–135.3)	119.8 (54.2–217.8)	167.4 (85.3–271.4)	212.0 (126.4–323.8)	267.5 (190.0–374.8)
Fish and shellfish, g	25.2 (14.9–45.8)	11.9 (5.6–24.1)	20.5 (9.2–34.5)	25.2 (14.9–45.8)	31.5 (20.5–56.2)	40.1 (25.2–65.8)
Nuts, g	3.9 (1.1–10.8)	2.9 (1.1–6.2)	2.9 (1.1–8.7)	3.9 (1.1–8.7)	4.1 (1.8–12.0)	6.9 (2.2–16.6)
Nonsoy legumes, g	2.4 (0–5.2)	2.4 (0–5.2)	2.4 (0–5.2)	2.4 (0–5.2)	2.4 (0–5.2)	2.4 (0–5.2)
Unsaturated fat, % of TEI	17.7 (15.2–20.2)	17.9 (15.4–20.3)	18.2 (15.7–20.6)	17.8 (15.6–20.2)	17.7 (14.9–20.2)	17.0 (14.2–19.6)
Dairy food, g	217.8 (48.7–292.8)	156.9 (32.6–314.0)	201.8 (37.0–295.2)	209.3 (48.8–287.3)	241.6 (66.9–287.2)	244.7 (87.4–287.3)
Red/processed meats, g	78.0 (42.3–122.2)	106.1 (72.1–145.7)	92.8 (53.4–130.8)	82.2 (46.6–123.9)	67.6 (38.7–111.3)	42.6 (24.4–72.8)
Tubers, g	69.0 (35.0–125.2)	118.9 (50.8–168.1)	107.5 (36.8–127.2)	65.1 (35–123.3)	52.7 (35.0–116.8)	44.3 (28.5–107.5)
Chicken and other poultry, g	33.6 (16.8–51.6)	24.7 (16.8–51.6)	33.6 (16.8–51.6)	33.6 (16.8–51.6)	33.6 (16.8–51.6)	51.6 (16.8–51.6)
Eggs, g	7.0 (3.3–21.5)	7.0 (3.3–21.5)	7.0 (3.3–21.5)	7.0 (3.3–21.5)	7.0 (3.3–21.5)	7.0 (0–21.5)
Saturated fat, % of TEI	12.0 (10.1–13.9)	12.8 (10.9–14.8)	12.6 (10.8–14.4)	12.3 (10.6–14.0)	11.8 (10.0–13.6)	10.4 (8.5–12.3)
Added sugar and sugar from fruit juice, % of TEI	11.3 (6.8–17.4)	17.5 (11.0–24.4)	13.5 (8.5–19.4)	11.5 (7.1–16.6)	9.8 (6.2–14.4)	7.5 (4.5–11.2)

ARIC indicates Atherosclerosis Risk in Communities; eGFR, estimated glomerular filtration rate; PHDI, Planetary Health Diet Index; and TEI, total energy intake.

^*Values for continuous variables are mean±SD and for categorical variables are n (%) unless otherwise indicated.

^† Other race includes Asian American and American Indian or Alaskan Indian.

In comparison with participants in the lowest PHDI quintile, those in the highest quintile consumed a higher percentage of energy from total and plant protein, carbohydrates, and polyunsaturated fat, as well as more fiber, phosphorus, potassium, magnesium, iron, vitamin A, vitamin C, and folate. They also consumed a lower percentage of energy from total, saturated, and monounsaturated fat, with less intake of cholesterol, sodium, calcium, vitamin B₁₂, and zinc (Table S2).

PHDI, Cardiovascular Disease, and All‐Cause Death

There were 5074 incident CVD events over a median follow‐up of 25 years, 2512 deaths caused by CVD over a median follow‐up of 29 years, and 8436 deaths from any cause over a median follow‐up of 29 years. Unadjusted Kaplan–Meier survival curves showed that individuals with higher PHDI had greater survival for total CVD and all‐cause death compared with those with lower adherence (log‐rank P≤0.003; Figure S3). There was no significant difference in survival rates across PHDI quintiles for CVD mortality (log‐rank P=0.14). Individuals with higher PHDI scores had a lower risk of total CVD, CVD death, and all‐cause death (Figure 1, Table S3). In model 1, after adjusting for age, sex, race‐center, and total energy intake, participants in the highest quintile of PHDI had 22% lower risk of total CVD, 21% lower risk of CVD death, and 20% lower risk of all‐cause death, compared with those in the lowest quintile (all P‐trend<0.001). In model 2, after additional adjustment for education level, health insurance status, physical activity, smoking status, and alcohol consumption, the associations were slightly attenuated for total CVD (HR, 0.94 [95% CI, 0.85–1.03]; P‐trend=0.07), CVD death (HR, 0.99 [95% CI, 0.86–1.13]; P‐trend=0.43), and all‐cause death (HR, 0.97 [95% CI, 0.90–1.05]; P‐trend=0.05), when comparing the highest and lowest quintiles. In model 3, which further adjusted for cholesterol‐lowering medication use, hypertension status, and kidney function, the inverse associations with incident CVD and all‐cause death were robust (both P‐trend≤0.04) but attenuated for CVD death (P‐trend=0.28). In model 4, which further adjusted for diabetes status and body mass index, the inverse associations remained significant and robust across all outcomes (all P‐trend≤0.02). However, for CVD death, the association was no longer statistically significant after accounting for competing risk of non‐CVD causes (P‐trend=0.54; Table S4).

Figure 1. Risk of total CVD, CVD death, and all‐cause death associated with PHDI, analyzed according to quintiles and continuously in the ARIC study.*^,† .

ARIC indicates Atherosclerosis Risk in Communities; CVD, cardiovascular disease; and PHDI, Planetary Health Diet Index. *Model adjusted for age, sex, race‐study center, total energy intake, education level, health insurance status, physical activity, smoking status, alcohol consumption, cholesterol‐lowering medication use, hypertension status, kidney function (2 linear spline terms with 1 knot at 90 mL/min per 1.73 m²), diabetes status, and body mass index. ^† E‐values: 1.39 (total CVD), 1.43 (CVD death), and 1.35 (all‐cause death).

When examined as a continuous variable, each 20‐point higher PHDI was consistently associated with a lower risk of total CVD, CVD death, and all‐cause death across all models (all P≤0.02), except for CVD death in model 2 (P=0.14) and model 3 (P=0.09). A roughly linear association was observed between PHDI and risk of total CVD (Figure 2A), CVD death (Figure 2B), and all‐cause death (Figure 2C).

Figure 2. Association between PHDI and risk of (A) total CVD, (B) CVD death, and (C) all‐cause death represented by restricted cubic splines in the ARIC study.*^,† .

ARIC indicates Atherosclerosis Risk in Communities; CVD, cardiovascular disease; HR, hazard ratio; and PHDI, Planetary Health Diet Index. *Models were adjusted for age, sex, race‐study center, total energy intake, education level, health insurance status, physical activity, smoking status, alcohol consumption, cholesterol‐lowering medication use, hypertension status, diabetes status, body mass index, and kidney function (2 linear spline terms with 1 knot at 90 mL/min per 1.73 m²). ^†The black solid line shows the adjusted HR. The black dashed lines represent the 95% CI. The gray histogram represents the distribution of the planetary health diet index in the study population. The reference level was set at the 10th percentile (score, 62), and 4 knots were set at the 5th, 35th, 65th and 95th percentiles (scores 58, 72, 80 and 93, respectively).

The associations between PHDI and total CVD incidence, CVD death, and all‐cause death did not vary significantly by sex or race (all P for interaction>0.05) (Table S5).

Analyses of Food Component Scores of PHDI

Participants with higher scores for whole grains and fish and shellfish, representing higher intake of these components, had a lower risk of total CVD in model 4 (Figure 3). Those with higher scores for red and processed meat and eggs, representing lower intake, had a lower risk of total CVD.

Figure 3. Risk of total CVD, CVD death, and all‐cause death associated with scores for food components of the PHDI.*^,†,‡,§ .

CVD indicates cardiovascular disease; HR, hazard ratio; and PHDI, Planetary Health Diet Index. *HRs for events were calculated for highest score (quintile 5) versus lowest score (quintile 1) of each of the food components of PHDI. ^†Models adjusted for age, sex, race‐study center, total energy intake, education level, health insurance status, physical activity, smoking status, alcohol consumption, cholesterol‐lowering medication use, hypertension status, kidney function (2 linear spline terms with 1 knot at 90 mL/min per 1.73 m²), diabetes status, and body mass index. ^‡The scores for moderation components were reverse coded; that is, higher scores indicated lower consumption. The scores for adequacy components were positively scored; that is, higher scores indicated higher consumption. ^§Food groups are ordered according to the direction and strength of their association with incident cardiovascular disease, separately within adequacy and moderation components.

For CVD death, those with higher scores for whole grains, nuts, vegetables, and whole fruits, representing higher intake, and those with higher scores for eggs, representing lower intake, had lower risk. Individuals with higher scores for tubers, representing lower intake, had a higher risk of CVD death.

The food components that contributed to the overall inverse association of PHDI with all‐cause death included higher scores for whole grains, fish and shellfish, nuts, and whole fruit, representing higher consumption, and eggs, representing lower consumption. Participants with higher scores for tubers, representing lower intake, had a higher risk of all‐cause death.

PHDI, Coronary Heart Disease, Stroke, and Heart Failure

There were 2181 cases of CHD over a median follow‐up of 27 years, 1390 cases of stroke over a median follow‐up of 28 years, and 3431 cases of heart failure over a median follow‐up of 27 years. Unadjusted Kaplan–Meier survival curves indicated greater CHD survival among individuals with higher PHDI (log‐rank P<0.001), with no significant differences for stroke (log‐rank P=0.09) or heart failure (log‐rank P=0.10) across PHDI quintiles (Figure S3). In model 1, controlling for demographic characteristics and total energy intake, individuals in the highest quintile of PHDI had a 22% lower risk of CHD, a 26% lower risk of stroke, and a 18% lower risk of heart failure (all P‐trend<0.001) compared with those in the lowest quintile of PHDI (Figure 4, Table S6). After additionally adjusting for socioeconomic status and health behaviors in model 2 and clinical factors in model 3, the associations were slightly attenuated for stroke and no longer statistically significant for CHD and heart failure. In model 4, after accounting for diabetes and body mass index, PHDI was significantly and inversely associated with CHD, stroke, and heart failure (all P‐trend≤0.04). Similar results were observed when examining PHDI as a continuous variable per 20‐point increment (Figure 2, Table S6), and the associations were roughly linear (Figure S4).

Figure 4. Risk of incident coronary heart disease, stroke, and heart failure, associated with PHDI, analyzed according to quintiles and continuously in the ARIC study.*^,†,‡ .

ARIC indicates Atherosclerosis Risk in Communities; and PHDI, Planetary Health Diet Index. *Model adjusted for age, sex, race‐study center, total energy intake, education level, health insurance status, physical activity, smoking status, alcohol consumption, cholesterol‐lowering medication use, hypertension status, kidney function (2 linear spline terms with 1 knot at 90 mL/min per 1.73 m²), diabetes status, and body mass index. ^†Heart failure subgroup excluded 182 individuals with missing heart failure status at follow‐up. ^‡E‐values: 1.44 (coronary heart disease), 1.57 (stroke), and 1.24 (heart failure).

Sensitivity Analyses

After excluding cases that occurred within the first 5 years of follow‐up, in model 4, higher PHDI remained significantly associated with a lower risk of total CVD and all‐cause death in categorical analyses (both P‐trend≤0.01; Table S7). For CVD death, the association was attenuated and no longer statistically significant (P‐trend=0.06). In continuous analyses, the inverse association remained significant across all 3 outcomes (P≤0.01). Inverse associations remained significant for all 3 outcomes in model 4, after excluding hypertension and diabetes cases that developed between visits 1 and 3, except for the attenuated association with CVD death after excluding incident diabetes cases (Table S8). When treating PHDI as a time‐varying exposure by using only visit 3 diet to calculate disease risk at visit 3 onward, the inverse associations with total CVD, CVD death, and all‐cause death remained largely consistent with the primary analysis (Table S9).

Discussion

In a community‐based cohort of middle‐aged US men and women, greater adherence to the EAT‐Lancet diet, as indicated by a higher PHDI score, was associated with a lower risk of total CVD, CVD death, and all‐cause death over nearly 3 decades of follow‐up. The strongest contributors to this inverse association were higher scores for whole grains, nuts, eggs, and red and processed meats. Similarly, a higher PHDI was associated with a lower risk of CHD, stroke, and heart failure.

Our study findings align with previous literature. Evidence from 3 large US prospective cohort studies, the Nurses' Health Study, Nurses' Health Study II, and the Health Professionals Follow‐Up Study, reported that individuals in the highest quintile of PHDI had 17% lower risk of incident CVD, 14% lower risk of CVD death, and 23% lower risk of all‐cause death than those in the lowest quintile. ⁹ , ¹⁷ The study similarly identified whole grains and nuts as key contributors to the inverse association between PHDI and all‐cause death. ⁹ However, since the participants in these cohorts were predominantly educated, non‐Hispanic White health professionals, our study, which consisted of a multicenter, community‐based cohort of predominantly Black and White individuals with diverse geographic and sociodemographic backgrounds, strengthens the external validity of the inverse associations between a planetary health diet, CVD morbidity, and death.

The inverse association has been observed in other populations using different scoring systems to reflect adherence to the EAT‐Lancet reference diet. Several studies developed similar continuous scoring schemes to assess intake from 14 dietary components proposed by EAT‐Lancet, proportionally rewarding higher consumption of planetary healthy foods. In the European Prospective Investigation Into Cancer and Nutrition–Netherlands study, which followed 35 496 adults for 15 years, participants in the highest quartile of the Healthy Reference Diet score (80–117 points) had 14% lower risk of total CVD and 12% lower risk of CHD than those in the lowest quartile (32–66 points), though no association was found for stroke. ²⁹ In a prospective study of 118 469 adults from the UK Biobank followed for 9 years, those with the highest adherence (90–120 points) to the PHDI was associated with 14% lower risk of total CVD, 12% lower risk of myocardial infarction, and 18% lower risk of stroke, compared with the lowest adherence (21–71 points). ³⁰ Another UK Biobank study, which followed 10 071 participants for nearly a decade, reported 21% lower risk of total CVD, 27% lower risk of ischemic heart disease, 10% lower risk of atrial fibrillation, and 12% lower risk of heart failure, when comparing individuals with the highest adherence (67–110 points) to the lowest adherence (10–44 points) for the Planetary Health Diet. ³¹ In 57 078 adults followed for 23 years in the Singapore Chinese Health Study, those with the highest adherence (65–71 points) to the Planetary Health Diet had 15% lower risk of all‐cause death and 21% lower risk of CVD death than the lowest adherence group (40–46 points). ³²

In contrast, several studies categorized adherence to the EAT‐Lancet diet, which yielded mixed results for CVD morbidity and death. A study of Canadian adults found no association between EAT‐Lancet Reference Diet score (0–14 points, 1 point per recommendation met) and risk of CVD over 80 529 person‐years of follow‐up. ³³ However, in the Malmö Diet and Cancer cohort study with 23 877 Swedish adults followed for nearly 25 years, those with the highest adherence (≥23 points) had 20% lower risk of coronary events than the lowest adherence group (≤13 points), using a graded score from 0 to 3 for adherence to each dietary component. ¹⁰ In terms of total deaths, no association was detected per 20% increase in EAT‐Lancet score (0–14 points) among 147 642 participants in 21 countries from the Prospective Urban Rural Epidemiology study over 9 years of follow‐up. ³⁴ The inconsistent associations might be partially explained by the limited discriminatory power of scoring systems that dichotomize the overall adherence score and each dietary component. ⁹ , ¹⁷ In addition, factors such as combining refined and whole grains, using a single dietary measurement, and relatively short follow‐up periods may have reduced the likelihood of detecting associations. ³³ , ³⁴

We observed that the magnitude of association between the overall PHDI and total CVD, CVD death, and all‐cause death was largely driven by certain food component scores. Scores for whole grains were strongly inversely associated with all 3 outcomes, and scores for nuts were inversely associated with CVD and total deaths. Scores for eggs were inversely associated with all 3 outcomes, and scores for red and processed meat were inversely associated with total CVD. Our results aligned with findings from 3 large US cohorts, which also identified whole grains, nuts, eggs, and red and processed meats as key score components that contributed to the inverse association with all‐cause death. ⁹ The inverse association of whole grain consumption with CVD and death is well established, largely due to the high fiber content, which improves glycemic control. ³⁵ , ³⁶ Accumulating evidence also supports an inverse association of nuts with cardiometabolic conditions and death, likely due to nuts' anti‐inflammatory and antioxidant‐rich properties. ³⁷ , ³⁸ Eggs were reverse coded, indicating that higher scores corresponded to lower intake. Consistent with prior literature, a dose–response association has been observed between egg consumption and higher CVD risk and all‐cause death, supporting dietary recommendations to limit dietary cholesterol. ³⁹ Red and processed meats were also reverse coded. Previous literature has shown that red and processed meat was positively associated with adverse health outcomes, consistent with our findings. ⁴⁰ Interestingly, we noticed positive associations with component scores for tubers, which were reverse coded. Tubers, primarily potatoes, which may vary also in preparation methods and ingredients/seasoning (salt, butter, sour cream), have shown inconsistent associations with CVD events and deaths across various populations in longitudinal studies. ⁴¹ , ⁴²

The inverse associations between the Planetary Health Diet and all outcomes attenuated in model 2, which adjusted for socioeconomic factors and health behaviors in addition to demographic factors and total energy intake. We observed considerable variation in education level, physical activity level, smoking status, and alcohol consumption across participants with differing adherence to the EAT‐Lancet diet, that is, those with greater PHDI adherence were more educated, physically active, and were less likely to smoke or drink compared with those with lower PHDI adherence. The attenuation in model 2 suggests that, beyond the EAT‐Lancet diet, other health behaviors significantly contribute to the onset of CVD events and premature deaths, which aligns with previous literature. ⁴³ The inverse associations remained robust even after controlling for clinical factors and potential mediators, including cholesterol‐lowering medication use, hypertension status, kidney function, diabetes status, and body mass index, highlighting the impact of the EAT‐Lancet diet on cardiovascular risk independent of these factors.

Several biological pathways may explain the inverse association between higher adherence to the EAT‐Lancet diet and lower risk of CVD outcomes and death. The EAT‐Lancet diet has strong anti‐inflammatory potential, due to its restriction of saturated fat and added sugar intake, with greater consumption of antioxidant‐rich foods, such as vegetables and whole fruits. ⁴⁴ A proteomic analysis from the Malmö Diet and Cancer–Cardiovascular Cohort study, which followed 4742 Swedish adults for 25 years, identified 8 plasma proteins inversely associated with the greater adherence to the EAT‐Lancet diet and positively associated with risk of heart failure. ⁴⁴ These proteins, all inflammatory, such as vasodilatory peptide and inflammation‐associated hormones, were implicated in impaired ventricular relaxation and oxidative stress response. ⁴⁴ Lower levels of these heart failure–related inflammatory proteins among those adhering to the EAT‐Lancet diet underscores its anti‐inflammatory properties. Moreover, the impact of the EAT‐Lancet diet on CVD outcomes and mortality prevention might be partially mediated by improvements in cardiometabolic risk markers, including lower body mass index, total cholesterol, low‐density lipoprotein cholesterol, non–high‐density lipoprotein cholesterol, blood pressure, and waist circumference. ¹⁵ , ⁴⁵ , ⁴⁶ , ⁴⁷

Our study is strengthened by its prospective design; an extended follow‐up period of nearly 3 decades; and a well‐characterized, community‐based cohort of predominantly Black and White US adults, with repeated dietary measurements to better capture usual intake. However, several limitations should be considered when interpreting our findings. First, soybean and soy foods, an adequacy food component of the EAT‐Lancet diet, were not assessed in the ARIC study. Since soy products are rich in high‐quality protein, polyunsaturated fats, and phytoestrogens, with various health benefits, the health impact of the EAT‐Lancet diet might be underestimated. ⁴⁸ Second, diet was self‐reported, which can introduce measurement error. Nonetheless, we averaged dietary responses from 2 visits to improve accuracy, and the food frequency questionnaire, administered by trained staff, has high reproducibility. ⁴⁹ Third, as dietary intake was assessed in the late 1980s and mid‐1990s, it might differ from modern consumption patterns for certain food components. Finally, although we adjusted for a wide range of demographic, socioeconomic, health, and clinical factors that correlate with dietary intake and health outcomes, residual and unmeasured confounding cannot be completely ruled out.

In summary, greater adherence to the EAT‐Lancet diet, designed to promote both human and planetary health, was associated with a lower risk of total CVD, CHD, stroke, heart failure, CVD death, and all‐cause death in a predominantly Black and White US adult population over nearly 3 decades of follow‐up.

Sources of Funding

The ARIC study was funded in whole or in part with federal funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, Department of Health and Human Services (75N92022D00001, 75N92022D00002, 75N92022D00003, 75N92022D00004, 75N92022D00005). C.M.R. was supported by a grant from the National, Heart, Lung, and Blood Institute (R01 HL153178) and the National Institute of Diabetes and Digestive and Kidney Diseases (U54 DK137331). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Disclosures

None.

Supporting information

Tables S1–S9

Figures S1–S4