Ultra-Processed Foods and Incident Cardiovascular Disease in the Framingham Offspring Study
Original Investigation
Central Illustration
Abstract
Background
Ultra-processed foods provide 58% of total energy in the U.S. diet, yet their association with cardiovascular disease (CVD) remains understudied.
Objectives
The authors investigated the associations between ultra-processed foods and CVD incidence and mortality in the prospective Framingham Offspring Cohort.
Methods
The analytical sample included 3,003 adults free from CVD with valid dietary data at baseline. Data on diet, measured by food frequency questionnaire, anthropometric measures, and sociodemographic and lifestyle factors were collected quadrennially from 1991 to 2008. Data regarding CVD incidence and mortality were available until 2014 and 2017, respectively. Ultra-processed foods were defined according to the NOVA framework. The authors used Cox proportional hazards models to determine the multivariable association between ultra-processed food intake (energy-adjusted servings per day) and incident hard CVD, hard coronary heart disease (CHD), overall CVD, and CVD mortality. Multivariable models were adjusted for age, sex, education, alcohol consumption, smoking, and physical activity.
Results
During follow-up (1991 to 2014/2017), the authors identified 251, 163, and 648 cases of incident hard CVD, hard CHD, and overall CVD, respectively. On average, participants consumed 7.5 servings per day of ultra-processed foods at baseline. Each additional daily serving of ultra-processed foods was associated with a 7% (95% confidence interval [CI]: 1.03 to 1.12), 9% (95% CI: 1.04 to 1.15), 5% (95% CI: 1.02 to 1.08), and 9% (95% CI: 1.02 to 1.16) increase in the risk of hard CVD, hard CHD, overall CVD, and CVD mortality, respectively.
Conclusions
The current findings support that higher consumption of ultra-processed foods is associated with increased risk of CVD incidence and mortality. Although additional research in ethnically diverse populations is warranted, these findings suggest cardiovascular benefits of limiting ultra-processed foods.
Introduction
Cardiovascular diseases (CVDs) remain a leading cause of chronic disability and death worldwide (1). Poor diet is a major modifiable CVD risk factor and represents a critical target of cardiovascular prevention efforts (2). Ultra-processed foods (i.e., highly processed industrial formulations made with little or no whole foods) provide 58% of daily calories in the average U.S. diet and are increasingly consumed worldwide (3,4). The production of ultra-processed foods involves a number of novel processing techniques (e.g., extrusion), ingredients (e.g., modified starches, protein isolates), and additives (e.g., emulsifiers, artificial flavors) of exclusive industrial use (5). Processing can alter a food’s health potential by removing beneficial nutrients and naturally occurring bioactive components, introducing nonbeneficial nutrients and food additives, and modifying the physical structure (6,7). Consumption of ultra-processed foods has been linked to overweight/obesity, hypertension, metabolic syndrome, and type 2 diabetes in observational studies (8).
Few studies have investigated the role of ultra-processed foods in relation to CVD risk (9,10). In a large cohort of French adults, higher ultra-processed food consumption was associated with increased risk of CVD, coronary heart disease (CHD), and cerebrovascular events over an average follow-up period of 5.2 years (10). Conversely, ultra-processed food intake was associated with higher risk of all-cause mortality, but not CVD mortality, in a prospective analysis within the Third National Health and Nutrition Examination Study (NHANES III) (9). Given the high burden of CVD and the shift toward ultra-processed foods in the global food supply, it is vital to further elucidate the link between processing level and CVD risk. The current study examines the association between ultra-processed foods and CVD incidence and mortality in the prospective Framingham Offspring Cohort (FOS).
Methods
Study population
Study population
The Framingham Heart Study is an ongoing prospective cohort study in Framingham, Massachusetts (11). The present study used data from the second generation, FOS, which was enrolled in 1971 to 1975 (12). Participants underwent a clinical examination every 4 years. Follow-up data were available until examination 9 (2011–2014) and mortality data until 2017. Details regarding FOS have been published previously (13). Written informed consent was obtained from all participants. The current study was approved by the New York University Institutional Review Board.
Analytical dataset
Examination 5 (1991–1995) was considered baseline for the present analyses (n = 3,712). We excluded participants with preexisting CVD (n = 363) and those missing CVD follow-up data (n = 17). In accordance with Framingham Heart Study criteria (14), participants with invalid dietary data or implausible energy intakes also were excluded (n = 329). The final analytical dataset included 3,003 adults (Figure 1).
Creation of the Final Analytical Dataset From the Framingham Offspring Study
Examination 5 (1991 to 1995) was used as baseline in the present analyses. FFQ = food frequency questionnaire; FOS = Framingham Offspring Cohort.
Assessment of dietary intake
Diet at baseline was assessed by mail, using the validated 131-item Harvard semiquantitative food frequency questionnaire (FFQ) (15). Participants reported the frequency of consumption of specific foods in the previous year, with options ranging from <1 serving per month to ≥6 servings per day. During the in-person examination, trained personnel reviewed the FFQs. The U.S. Department of Agriculture nutrient database was used to compute nutrient intakes from reported dietary intakes (15).
Exposure ascertainment
The NOVA framework classifies foods according to the extent and purpose of the industrial processing they undergo (5). Based on NOVA, FFQ food items were classified into 5 mutually exclusive categories: 1) “unprocessed or minimally processed foods,” including fresh, dry, or frozen plant and animal foods; 2) “processed culinary ingredients,” including table sugar, oils, fats, salt, and other constituents used in kitchens to make culinary preparations; 3) “processed foods,” including foods such as canned fish and vegetables, simple breads, and artisanal cheeses; and 4) “ultra-processed foods”: industrial formulations made with no or minimal whole foods and produced with additives such as flavorings and preservatives. An additional category, 5) “culinary preparations,” was created as an extension of the NOVA classification to encompass mixed dishes that were indicated to be homemade or assumed to be homemade due to lack of detailed information. For foods of potentially different processing levels, assumptions were made according to the most frequently consumed processing level in NHANES 2001 to 2002 and the literature (4,16,17). For each participant, we calculated energy-adjusted daily intakes (servings per day) of foods within each NOVA processing level, as well as energy-adjusted weekly intakes (servings per week) of specific ultra-processed food groups (e.g., sugar-sweetened beverages), using the residual method (18). One serving corresponds to for example, 1 can of soda, 1 cup breakfast cereals, or 1 oz potato chips. Energy-adjusted ultra-processed food intakes were categorized into quintiles.
CVD case ascertainment
Outcome data were obtained from FOS survival and sequence of events files, which provide information on confirmed incident CVD events, CVD-related deaths, and total mortality. Information about CVD events at follow-up was derived from medical histories, physical examinations at the study clinic, hospitalization records, and communication with participants’ physicians (19). The primary outcomes were incident hard CVD (sudden and nonsudden coronary death, myocardial infarction, and fatal/nonfatal stroke) and hard CHD (sudden/nonsudden coronary death and myocardial infarction). Overall CVD (CHD, fatal, ischemic and hemorrhagic stroke, transient ischemic accident, cerebral embolism, other cerebral cardiovascular disease, peripheral artery disease [defined as intermittent claudication] and congestive heart failure [hospitalized or nonhospitalized, diagnosed on basis of examination or physician notes]), CVD mortality, and total mortality were secondary outcomes. A total of 648 cases of incident CVD were identified, including 251 cases of hard CVD and 163 cases of hard CHD. There were 713 deaths during the follow-up period, including 108 CVD deaths.
Evaluation of covariates
Demographic and lifestyle characteristics
Demographic and lifestyle characteristics
Age, physical activity, smoking, and alcohol intake were self-reported at each examination, and education at examination 2. A physical activity index was calculated by multiplying the average daily number of hours in light, moderate, or vigorous activity with the activity-specific metabolic costs, and then summing the weighted hours (20). Based on this index, physical activity level was categorized as low (<30), moderate (30 to 33), and high (>33) (21). Alcohol intake (g/d) was computed from self-reported consumption-frequency of a standard serving of beer, wine, and cocktail. Participants were classified as current, former (≥1 year abstinence), or nonsmokers.
Anthropometric and clinical measurements
Trained personnel measured weight, height, and waist circumference at each examination, using standardized methods (22). Body mass index (BMI) was calculated from measured height and weight. BMI values of 25 to 29.9 kg/m2 and ≥30 kg/m2 were classified as overweight and obese, respectively (23). Diabetes was defined as fasting glucose ≥126 mg/dl or use of insulin or oral hypoglycemic medications. Blood pressure readings were obtained by 2 physicians using a sphygmomanometer, and the average value was computed (24).
Diet quality
Diet quality was assessed using the Dietary Guidelines for Americans Adherence Index (DGAI) 2010 (25). The DGAI-2010 evaluates intakes of 14 food groups (fruit; dark green vegetables, orange and red vegetables; starchy vegetables; other vegetables; grains; milk; meat, protein, and eggs; seafood; nuts; legumes; sugar; variety in protein choices; and variety of fruits and vegetables) and 11 healthy choice or nutrient intake recommendations (amounts of total fat, saturated fat, trans-fat, cholesterol, sodium, fiber, alcohol; and percentage of lean protein; low-fat milk; whole grains; and fruits consumed as whole fruits) (26). The maximum score is 100 and higher scores indicate greater adherence to the dietary guidelines and higher diet quality.
Statistical analyses
Statistical analyses were performed using Stata/SE 15.1 (StataCorp, College Station, Texas). We assessed demographic, dietary, and clinical characteristics at baseline for the entire sample, and across quintiles of energy-adjusted ultra-processed food consumption (servings per day). Test of differences across consumption quintiles was performed by unadjusted linear regression for continuous variables and by Cochran-Mantel-Haenszel statistic for categorical variables.
We used Cox proportional hazards models with time-varying covariates to determine the association between baseline ultra-processed food consumption (energy-adjusted servings per day; continuous) and risk of incident CVD and/or CHD and CVD mortality during follow-up. Age-adjusted and multivariable-adjusted hazard ratios (HRs) (95% confidence intervals [CIs]) were computed separately for hard CVD events, hard CHD events, overall CVD events, and CVD mortality. We defined the follow-up period as the interval between the date of baseline and the date of the first known CVD event for cases, and between baseline and date of death or last contact for noncases. Participants were considered censored if they died of non-CVD causes, were lost to follow-up, or if the event had not yet occurred at the last examination in which they participated.
For all multivariable analyses, we included potential confounders separately in age-adjusted models. Variables were retained in the final model if risk estimates changed by >10% or if theoretically important based on the literature. Multivariable models were adjusted for age (continuous), sex, education (≤12 years; 13 to 15 years; ≥16 years), smoking status, alcohol intake (g/d) and physical activity level (continuous). In cases of missing covariate data, the last observation was carried forward. We tested potential multiplicative interactions between ultra-processed food intake and sex, age, smoking status, and BMI, respectively, at the 0.10 alpha level.
We repeated multivariable Cox proportional hazards models with additional adjustment for: 1) total energy intake; 2) diet quality defined by the DGAI-2010; 3) waist circumference (time-varying); 4) BMI (time-varying); 5) mean systolic blood pressure (time-varying); 6) hypertension treatment (time-varying); 7) lipid-lowering medication (time-varying); and 8) baseline intake of the remaining NOVA processing levels (servings per day). We also evaluated the multivariable association between baseline intake (servings per week) of specific ultra-processed foods (bread; sweets and desserts; ultra-processed meats; salty snack foods; sugar-sweetened beverages; low-calorie soft drinks; fast foods; breakfast cereals; yoghurt and other ultra-processed foods [nondairy coffee whitener, margarine, liquor, and chili sauce]) and the risk of CVD and CHD.
In secondary analyses, we explored the association between ultra-processed food intake and total mortality, as well as the association between the baseline intake of minimally processed foods, processed culinary ingredients, processed foods, and culinary preparations, respectively, and each CVD outcome. The proportional hazards assumption was tested by: 1) interacting time with covariates; and 2) a Grambsch-Therneau test of the scaled Schoenfeld residuals.
Results
Characteristics of the study population
Characteristics of the study population
At baseline, participants were middle-aged (mean age: 53.5 years) and overweight (mean BMI: 27.3 kg/m2); 55.1% were women and most reported a high level of physical activity (52.9%) (Table 1). One-third had undergone 16 years or more of education (33.1%) and two-thirds were either former (47.3%) or current (19.0%) smokers. Mean age, BMI, and waist circumference increased across quintiles of ultra-processed food consumption (p trend <0.001). Conversely, higher consumption of ultra-processed foods was inversely associated with physical activity (p trend <0.001) and education level (p = 0.028). Overall, 5.8% and 19.0% of participants had diabetes and hypertension, respectively, and prevalence was higher among high consumers of ultra-processed foods compared with low consumers (quintile 5 vs. 1; p < 0.001).
| Full Sample (n = 3,003) | Q1 (n = 601) <5.3 s/d | Q2 (n = 601) 5.3–6.4 s/d | Q3 (n = 600) 6.5–7.6 s/d | Q4 (n = 601) 7.6–9.5 s/d | Q5 (n = 600) >9.5 s/d | p Value∗ | |
|---|---|---|---|---|---|---|---|
| Age, yrs | 53.9 ± 9.6 | 52.6 ± 9.6 | 53.7 ± 9.7 | 53.8 ± 9.6 | 54.5 ± 9.9 | 55.1 ± 9.3 | <0.001 |
| Women | 55.1 | 51.9 | 58.2 | 56.5 | 55.6 | 53.2 | 0.178 |
| Mean education, yrs | 0.028 | ||||||
| ≤12 | 39.0 | 36.3 | 38.0 | 37.8 | 39.5 | 43.3 | |
| 13–15 | 27.9 | 29.1 | 25.9 | 30.2 | 29.9 | 24.6 | |
| ≥16 | 33.1 | 34.5 | 36.1 | 32.0 | 30.6 | 32.1 | |
| Body mass index, kg/m2 | 27.3 ± 5.0 | 26.6 ± 4.5 | 27.0 ± 4.8 | 27.1 ± 5.1 | 27.7 ± 5.3 | 28.1 ± 5.0 | <0.001 |
| Waist circumference, inches | 36.3 ± 5.6 | 35.7 ± 5.2 | 35.8 ± 5.5 | 36.1 ± 5.9 | 36.7 ± 5.9 | 37.1 ± 5.5 | <0.001 |
| Smoking status | 0.262 | ||||||
| Never smoked | 33.7 | 34.9 | 35.4 | 35.8 | 33.3 | 29.2 | |
| Current smoker | 19.0 | 17.8 | 18.5 | 19.5 | 17.8 | 21.3 | |
| Former smoker | 47.3 | 47.3 | 46.1 | 44.7 | 48.8 | 49.5 | |
| Physical activity index | 34.9 ± 6.2 | 35.8 ± 6.9 | 35.2 ± 6.6 | 34.3 ± 5.5 | 34.8 ± 6.1 | 34.2 ± 5.8 | <0.001 |
| Physical activity level | <0.001 | ||||||
| Low | 16.3 | 14.8 | 15.2 | 17.9 | 16.2 | 17.6 | |
| Medium | 30.8 | 26.8 | 28.9 | 31.7 | 30.6 | 36.0 | |
| High | 52.9 | 58.5 | 55.9 | 50.3 | 53.1 | 46.4 | |
| Diabetes | 5.8 | 4.0 | 4.0 | 5.2 | 9.1 | 6.7 | <0.001 |
| Hypertension | 19.0 | 17.1 | 14.7 | 17.9 | 21.8 | 23.3 | <0.001 |
Dietary characteristics of the analytical sample at baseline are presented in Table 2. Mean intake of ultra-processed foods ranged from 4.0 servings per day in the bottom quintile to 11.9 servings per day in the top quintile. Greater ultra-processed food consumption was associated with higher fat intake (total, monounsaturated, polyunsaturated, and trans-fat; p trend <0.001), higher sodium intake (p trend <0.001), lower intakes of protein, fiber, and alcohol (p trend <0.001) and lower diet quality as defined by the DGAI-2010 score (p trend <0.001). Carbohydrate intake was similar across quintiles (p trend = 0.861); however, sucrose intake increased with greater ultra-processed food consumption (p trend <0.001).
| Full Sample (n = 3,003) | Q1 (n = 601) <5.3 s/d | Q2 (n = 601) 5.3–6.4 s/d | Q3 (n = 600) 6.5–7.6 s/d | Q4 (n = 601) 7.6–9.5 s/d | Q5 (n = 600) >9.5 s/d | p-Trend∗ | |
|---|---|---|---|---|---|---|---|
| Total energy, kcal | 1,875.9 ± 625.0 | 2,116.8 ± 585.0 | 1,766.1 ± 583.6 | 1,714.3 ± 608.1 | 1,776.5 ± 610.9 | 2,005.7 ± 638.1 | 0.008 |
| Intake/processing level, s/d† | |||||||
| Minimally processed foods ‡ | 11.3 ± 3.6 | 12.9 ± 4.3 | 11.3 ± 3.2 | 11.1 ± 3.2 | 10.7 ± 3.3 | 10.3 ± 3.5 | <0.001 |
| Processed culinary ingredients§ | 1.6 ± 1.4 | 1.9 ± 1.7 | 1.7 ± 1.5 | 1.6 ± 1.3 | 1.4 ± 1.2 | 1.3 ± 1.3 | <0.001 |
| Culinary preparations¶ | 0.6 ± 0.5 | 0.6 ± 0.6 | 0.6 ± 0.5 | 0.6 ± 0.5 | 0.6 ± 0.5 | 0.5 ± 0.5 | <0.001 |
| Processed foods|| | 2.0 ± 1.3 | 2.3 ± 1.7 | 2.0 ± 1.2 | 1.9 ± 1.2 | 1.9 ± 1.0 | 1.7 ± 1.1 | <0.001 |
| Ultra-processed foods# | 7.5 ± 2.9 | 4.0 ± 1.2 | 5.9 ± 0.3 | 7.0 ± 0.3 | 8.5 ± 0.5 | 11.9 ± 2.3 | <0.001 |
| % Energy from | |||||||
| Carbohydrates | 50.9 ± 8.4 | 50.6 ± 8.8 | 51.1 ± 7.8 | 51.3 ± 7.8 | 51.2 ± 8.8 | 50.4 ± 8.6 | 0.861 |
| Sucrose | 9.8 ± 4.1 | 9.1 ± 3.3 | 9.8 ± 3.5 | 10.0 ± 4.0 | 10.1 ± 4.8 | 10.0 ± 4.4 | <0.001 |
| Fructose | 5.4 ± 2.8 | 5.6 ± 2.7 | 5.6 ± 2.7 | 5.6 ± 2.6 | 5.4 ± 2.9 | 4.9 ± 2.9 | <0.001 |
| Protein | 16.8 ± 3.3 | 17.5 ± 3.6 | 17.3 ± 3.3 | 16.9 ± 3.2 | 16.5 ± 3.3 | 15.8 ± 3.1 | <0.001 |
| Total fat | 30.1 ± 6.3 | 28.7 ± 6.6 | 29.5 ± 5.9 | 29.9 ± 5.8 | 30.2 ± 6.4 | 32.0 ± 6.5 | <0.001 |
| Saturated fat | 10.5 ± 2.9 | 10.2 ± 3.1 | 10.4 ± 2.9 | 10.4 ± 2.6 | 10.4 ± 2.8 | 11.0 ± 2.9 | <0.001 |
| Monounsaturated fat | 11.1 ± 2.6 | 10.5 ± 2.7 | 10.8 ± 2.4 | 11.0 ± 2.4 | 11.2 ± 2.6 | 12.0 ± 2.7 | <0.001 |
| Polyunsaturated fat | 5.8 ± 1.6 | 5.4 ± 1.6 | 5.6 ± 1.4 | 5.8 ± 1.6 | 6.0 ± 1.8 | 6.3 ± 1.7 | <0.001 |
| Trans-fat | 1.5 ± 0.7 | 1.2 ± 0.5 | 1.4 ± 0.6 | 1.5 ± 0.7 | 1.6 ± 0.7 | 1.9 ± 0.9 | <0.001 |
| Fiber, g/1,000 kcal | 10.3 ± 3.3 | 10.7 ± 4.0 | 10.5 ± 3.4 | 10.3 ± 3.0 | 10.0 ± 3.1 | 9.9 ± 3.0 | <0.001 |
| Sodium, min/1,000 kcal | 1153.2 ± 232.4 | 1095.9 ± 231.1 | 1124.4 ± 220.8 | 1143.5 ± 212.2 | 1182.0 ± 236.3 | 1220.2 ± 240.5 | <0.001 |
| Alcohol, g/d | 10.2 ± 15.2 | 12.8 ± 17.0 | 10.1 ± 14.3 | 9.4 ± 13.8 | 9.9 ± 15.3 | 8.9 ± 15.0 | <0.001 |
| DGAI-2010∗∗ | 59.8 ± 11.5 | 63.2 ± 11.8 | 61.3 ± 11.1 | 59.5 ± 10.9 | 58.4 ± 11.0 | 56.8 ± 11.4 | <0.001 |
Associations between ultra-processed food intake and cardiovascular disease
During a mean follow-up of 18.0 years (53,933.3 person-years), a total of 648 incident CVD events occurred, including 251 cases of hard CVD and 163 cases of hard CHD. Compared with participants consuming the least ultra-processed foods, those with the highest intakes had higher incidence rates per 1000 person-years of hard CVD (3.36 vs. 6.64) and hard CHD (2.00 vs. 4.36) (Figure 2).
Incidence Rates of Hard CVD, Hard CHD, and Overall CVD According to Quintile of Ultra-Processed Food Intake in the FOS Cohort
Incidence rates of hard cardiovascular disease (CVD, n = 251), hard coronary heart disease (CHD, n = 163), and overall CVD (n = 648) according to quintile of ultra-processed food intake in the Framingham Offspring Study (FOS) Cohort (N = 3,003). Graphs display rates per 1,000 person-years and 95% confidence levels. Mean intake of ultra-processed foods: quintile 1 = 4.0 servings per day, quintile 2 = 5.9 servings per day, quintile 3 = 7.0 servings per day, quintile 4 = 8.5 servings per day, quintile 5 = 11.9 servings per day.
Controlling for age, sex, education level, smoking status, alcohol intake, and physical activity, a 1 standard deviation increase in ultra-processed food intake (2.9 servings) was associated with 22% and 30% increased risk of hard CVD and hard CHD, respectively (Central Illustration). Each additional serving of ultra-processed food was associated with a 7% increase in the risk of hard CVD (HR: 1.07; 95% CI: 1.03 to 1.12) and a 9% increase in the risk of hard CHD (HR: 1.09; 95% CI: 1.10 to 3.28) in multivariable-adjusted models (Table 3). Additional adjustment for total energy intake, diet quality, waist circumference, BMI, systolic blood pressure, and current hypertension treatment or lipid-lowering medication did not meaningfully change the associations between ultra-processed food consumption and CVD outcomes. Likewise, additional adjustment for intake of all NOVA processing levels did not alter the current findings. No statistically significant interactions were identified.
Ultra-Processed Food Intake and Cardiovascular Disease Incidence and Mortality in the Framingham Offspring Study Cohort
Higher intakes of ultra-processed foods were significantly associated with a higher CVD incidence and mortality in the Framingham Offspring Study (N = 3,003; mean follow-up: 18.0/20.2 years). The diagram shows the increase in CVD incidence and mortality associated with a 1-SD increase in ultra-processed food intake (2.9 servings). CHD = coronary heart disease; CI = confidence level; CVD = cardiovascular disease; HR = hazard ratio.
| Ultra-Processed Foods (s/d)∗ | |||||
|---|---|---|---|---|---|
| Age Adjusted | Multivariable Adjusted† | Multivariable + Total E‡ | Multivariable + DGAI-2010§ | Multivariable + WC¶ | |
| Hard CVD | 1.08 (1.04–1.12) | 1.07 (1.03–1.12) | 1.07 (1.03–1.12) | 1.06 (1.02–1.11) | 1.06 (1.02–1.11) |
| Hard CHD | 1.09 (1.04–1.14) | 1.09 (1.04–1.15) | 1.10 (1.04–1.15) | 1.09 (1.03–1.15) | 1.09 (1.03–1.14) |
| Overall CVD | 1.04 (1.02–1.07) | 1.05 (1.02–1.08) | 1.05 (1.02–1.08) | 1.04 (1.01–1.07) | 1.04 (1.01–1.07) |
| CVD mortality | 1.07 (1.01–1.14) | 1.09 (1.02–1.16) | 1.09 (1.02–1.16) | 1.09 (1.02–1.16) | 1.09 (1.02–1.16) |
Secondary analyses
Higher ultra-processed food intake was associated with increased risk of overall CVD (multivariable-adjusted HR: 1.05; 95% CI: 1.02 to 1.08) and CVD mortality (multivariable-adjusted HR: 1.09; 95% CI: 1.02 to 1.16), but not total mortality (multivariable-adjusted HR: 1.01; 95% CI: 0.99 to 1.04). Mean follow-up for mortality analyses was 20.2 years (60,598.7 person-years).
Intake of bread was associated with increased risk of incident hard CVD, hard CHD, and overall CVD, whereas ultra-processed meat intake was associated with increased risk of hard and overall CVD, but not hard CHD (Table 4). Salty snack foods were associated with increased risk of incident hard CVD and CHD, but not with overall CVD, whereas intake of low-calorie soft drinks was associated increased risk of overall CVD. Conversely, intake of breakfast cereals was associated with decreased risk of overall CVD. No statistically significant associations were observed for sugar-sweetened beverages, yoghurt, sweets and desserts, fast foods and other ultra-processed foods in multivariable-adjusted models.
| Hard CVD (n = 251) | Hard CHD (n = 163) | Overall CVD (n = 648) | |
|---|---|---|---|
| Bread∗ | 1.02 (1.01–1.04) | 1.02 (1.01–1.04) | 1.01 (1.01–1.02) |
| Sweets and desserts† | 1.00 (0.98–1.01) | 0.99 (0.97–1.01) | 1.00 (0.99–1.01) |
| Ultra-processed meats‡ | 1.05 (1.01–1.09) | 1.04 (0.99–1.10) | 1.05 (1.02–1.08) |
| Salty snack foods§ | 1.02 (1.00–1.03) | 1.02 (1.01–1.04) | 1.01 (1.00–1.02) |
| Sugar-sweetened beverages¶ | 1.01 (0.98–1.03) | 1.01 (0.99–1.04) | 1.00 (0.98–1.02) |
| Low-calorie soft drinks|| | 1.01 (0.99–1.03) | 1.01 (0.99–1.03) | 1.01 (1.00–1.02) |
| Fast foods# | 0.98 (0.92–1.05) | 1.00 (0.93–1.08) | 1.03 (0.99–1.07) |
| Breakfast cereals∗∗ | 0.95 (0.90–1.01) | 0.98 (0.92–1.05) | 0.96 (0.93–1.00) |
| Yoghurt | 1.02 (0.95–1.09) | 1.06 (0.98–1.14) | 0.98 (0.93–1.03) |
| Other ultra-processed foods†† | 1.00 (0.99–1.02) | 1.01 (0.99–1.02) | 1.00 (0.99–1.01) |
Each additional daily serving of minimally processed foods was associated with a 3% lower risk of incident overall CVD in age-adjusted models (HR: 0.97; 95% CI: 0.95 to 0.99) but the association was not statistically significant in multivariable-adjusted models (HR: 0.98; 95% CI: 0.96 to 1.01) (data not shown). The remaining processing levels were not associated with CVD outcomes in age- or multivariable-adjusted models.
Discussion
In this large prospective cohort of Caucasian adults, each additional daily serving of ultra-processed foods was associated with a 7%, 9%, and 5% increase in the risk of incident hard CVD, hard CHD, and overall CVD, respectively, and a 9% increase in CVD mortality.
Our findings are consistent with those from the French NutriNet-Santé cohort (N = 105,159), which observed that each 10% increment in weight of dietary ultra-processed foods was associated with a 12%, 13%, and 11% increase in the rates of overall CVD, CHD, and cerebrovascular disease, respectively (10). Our findings are also in line with previous studies noting that healthy dietary patterns with fewer ultra-processed foods are associated with lower cardiovascular risk (27,28), and with epidemiologic evidence linking ultra-processed foods to obesity, hypertension, metabolic syndrome and type 2 diabetes (8). In contrast to previous studies (9,29–31), we did not observe an association between ultra-processed food consumption and total mortality. Ultra-processed food consumption was associated with higher total mortality, but not CVD mortality in NHANES III (9) and with higher total mortality in 3 European cohorts (29–31). Discrepant results between studies may be due to varying methodological factors, including accuracy of cause-specific mortality ascertainment (32); assessment of ultra-processed food consumption, as well as the amounts and types of ultra-processed foods consumed within study populations; sociodemographic, cultural, and lifestyle characteristics of study populations; and duration of study follow-up.
Several biological mechanisms may explain the observed associations between ultra-processed foods and CVD (Figure 3). Experimental evidence suggests that ultra-processed foods may contribute to higher energy intakes and weight gain (33), potentially due to their high energy-density and low satiating potential (34). However, the associations between ultra-processed food intake and CVD risk remained robust when we controlled for total energy intake, waist circumference, and BMI, suggesting the existence of additional biological pathways. Ultra-processed foods are generally high in trans-fats and sodium, and low in potassium and dietary fiber (35), characteristics linked to CVD and CHD (36). Excessive intakes of sugar from ultra-processed foods, particularly from sugar-sweetened beverages, are associated with CVD risk factors including obesity, hypertension, and type 2 diabetes (37). CVD risk associated with ultra-processed foods may partly be attributed to lower consumption of cardioprotective minimally processed foods, such as fruit, vegetables, nuts, fish, whole grains, and legumes (36). Nevertheless, adjustment for diet quality did not alter the observed associations between ultra-processed foods and CVD outcomes in the present study.
Potential Biological Mechanisms Underlying the Association Between Ultra-Processed Foods and CVD
Key mechanisms include altered serum lipid concentrations and glycemic response, insulin resistance, modified gut microbiota and host-microbiota interactions, obesity, inflammation, oxidative stress, and hypertension. Arrows indicate stimulation of a pathway. CVD = cardiovascular disease.
Beyond nutrient composition, processing modifies the physical structure of the food matrix which may alter nutrient bioaccessibility, absorption kinetics and the gut microbiota profile (7,38,39). The large share of acellular nutrients in ultra-processed foods and consequent high nutrient availability in the small intestine may promote an inflammatory gut microbiota associated with cardiometabolic conditions (38,39). Additives in ultra-processed foods, including artificial sweeteners (40) and emulsifiers (41), may disrupt gut microbiota integrity, promoting a proinflammatory status and metabolic dysregulation. The low-calorie sweetener sucralose may also reduce insulin sensitivity by impairing gut-brain regulatory pathways (42). Other additives in ultra-processed foods, such as inorganic phosphate salts, may promote arterial calcification, oxidative stress, and endothelial dysfunction (43).
Study Limitations
Some limitations of the present study should be noted. Although the Harvard FFQ is widely used and has been validated in U.S. populations (15), measurement error is inherent in dietary assessment (44). Dietary measurement error typically attenuates true associations (44,45). Nevertheless, energy-adjusted food intakes, as used in the current analysis, are less affected by measurement error due to cancellation of correlated errors in reporting of energy and food intakes (45). Furthermore, our analytic strategy capitalizes on the ability of FFQs to rank individuals according to relative food intake (45).
Under- and overestimation of ultra-processed food intake due to misclassification of FFQ items cannot be excluded, which if present may have attenuated or strengthened the observed associations. However, misclassification error would likely be random, which would bias the associations toward the null. To minimize misclassification, we based assumptions regarding processing level of FFQ items on the current literature and on data regarding actual consumption among adults in NHANES 2001–2002 with demographics similar to the FOS cohort. Due to the observational design, causality cannot be determined and uncontrolled confounding due to unmeasured lifestyle factors cannot be excluded. Nonetheless, the ample data regarding lifestyle, clinical, and demographic variables collected within FOS permitted the adjustment for multiple potential confounders. Due to the limited number of CHD events and consequent lower analytical power in these analyses, CHD and CVD mortality risk estimates had wide confidence intervals and should be interpreted with caution. As we lack data regarding cause-specific mortality beyond CVD and cancer, we were unable to further examine the lack of association between ultra-processed food and total mortality in FOS. Last, participants in FOS are primarily Caucasian and have higher levels of education and income than the general U.S. population, which limits the generalizability of our findings.
Study strengths
Important strengths of the present study include its prospective study design and long follow-up of nearly 2 decades. Study participants were free from CVD at baseline, which decreases the risk of diet modification in response to disease. Data regarding dietary, anthropometric, and other factors were collected by trained personnel using validated questionnaires and protocols. CVD outcomes were ascertained from medical records and pathological reports by an expert panel, which reduces the risk of misclassification. Another key strength of our study is the determination of food-processing level according to standardized and objective criteria. NOVA constitutes a novel indicator of diet quality and healthy dietary patterns. In contrast to traditional nutrient-focused metrics, NOVA considers the overall quality of foods. From a public health policy perspective, NOVA is advantageous, as evidence is derived from studies ascertaining intake of foods rather than nutrients. Therefore, such evidence can be directly translated to food policy and dietary guidance. Recommendations to limit ultra-processed foods are potentially more user-friendly for the public than nutrient-specific recommendations.
Implications
Although the observed associations need to be studied in ethnically diverse populations and other settings, our study has potential implications for cardiovascular prevention. From a public health perspective, our study suggests the need for increased efforts to implement population-wide strategies. These strategies may include fiscal measures, such as taxation of sugar-sweetened beverages and other ultra-processed foods, and recommendations regarding processing level in national dietary guidelines. Another promising strategy is to require front-of-package warning labels on ultra-processed products (46). Importantly, policies should be designed to simultaneously increase the availability, accessibility, and affordability of nutritious minimally processed foods, especially in disadvantaged populations.
From a clinical practice perspective, our findings provide support to current American College of Cardiology/American Heart Association Guideline on the Primary Prevention of Cardiovascular Disease, which recommends minimizing the intake of processed red meats and refined carbohydrates (2). Our study further expands the evidence base, by recognizing the clinical importance of a broader range of ultra-processed foods. Clinicians play a pivotal role in providing evidence-based nutrition guidance to patients (47). Given the convenience, omnipresence, and affordability of ultra-processed foods, careful nutrition counseling is needed to design individualized, patient-centered, heart-healthy diets.
Conclusions
In this large prospective population-based study of men and women, consumption of a wide range of ultra-processed foods was associated with an increased risk of cardiovascular events and mortality, independent of other cardiovascular risk factors. This relation was consistent across sex and age groups. Our study provides support for nutrition counseling and public health promotion of heart-healthy dietary patterns minimizing the intake of ultra-processed foods.
Perspectives
COMPETENCY IN MEDICAL KNOWLEDGE: Greater intake of ultra-processed foods is associated with a higher risk of cardiac events, independent of energy intake and adiposity.
TRANSLATIONAL OUTLOOK: Future studies should evaluate the association between ultra-processed foods and cardiovascular disease in ethnically diverse populations to develop culturally appropriate interventions for prevention of cardiovascular disease.
Funding Support and Author Disclosures
The current analyses were unfunded. The Framingham Heart Study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with Boston University (Contract No. N01-HC-25195). Funding support for the Framingham Food Frequency Questionnaire datasets was provided by ARS Contract #53-3k06-5-10, ARS Agreement #'s 58-1950-9-001, 58-1950-4-401, and 58-1950-7-707. This manuscript was not prepared in collaboration with investigators of the Framingham Heart Study and does not necessarily reflect the opinions or views of the Framingham Heart Study, Boston University, or NHLBI. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Abbreviations and Acronyms
| BMI | body mass index |
| CHD | coronary heart disease |
| CVD | cardiovascular disease |
| DGAI | Dietary Guidelines for Americans Adherence Index |
| FFQ | Food Frequency Questionnaire |
| FOS | Framingham Offspring Study |
| NHANES | National Health and Nutrition Examination Study |
References
1. "Global, regional, and national burden of cardiovascular diseases for 10 causes, 1990 to 2015". J Am Coll Cardiol 2017;70:1-25.
2. "2019 ACC/AHA guideline on the primary prevention of cardiovascular disease: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines". J Am Coll Cardiol 2019;74:e177-e232.
3. "Ultra-processed products are becoming dominant in the global food system". Obes Rev 2013;14:Suppl 2: 21-28.
4. "Ultra-processed foods and added sugars in the US diet: evidence from a nationally representative cross-sectional study". BMJ Open 2016;6:e009892.
5. "Ultra-processed foods: what they are and how to identify them". Public Health Nutr 2019;22:936-941.
6. "Dietary and policy priorities for cardiovascular disease, diabetes, and obesity: a comprehensive review". Circulation 2016;133:187-225.
7. "Current food classifications in epidemiological studies do not enable solid nutritional recommendations for preventing diet-related chronic diseases: the impact of food processing". Adv Nutr 2015;6:629-638.
8. "Consumption of ultra-processed foods and health outcomes: a systematic review of epidemiological studies". Nutr J 2020;19:86.
9. "Ultra-processed food intake and mortality in the USA: results from the Third National Health and Nutrition Examination Survey (NHANES III, 1988–1994)". Public Health Nutr 2019;22:1777-1785.
10. "Ultra-processed food intake and risk of cardiovascular disease: prospective cohort study (NutriNet-Sante)". BMJ 2019;365:l1451.
11. "Epidemiological approaches to heart disease: the Framingham Study". Am J Public Health Nations Health 1951;41:279-281.
12. "An investigation of coronary heart disease in families. The Framingham Offspring Study". Am J Epidemiol 1979;110:281-290.
13. "The Framingham Offspring Study. Design and preliminary data". Prev Med 1975;4:518-525.
14. "Whole- and refined-grain intakes are differentially associated with abdominal visceral and subcutaneous adiposity in healthy adults: the Framingham Heart Study". Am J Clin Nutr 2010;92:1165-1171.
15. "Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals". Am J Epidemiol 1992;135:1114-1126.discussion 27–36.
16. "National health and nutrition examination survey: plan of operations, 1999-2010". Vital Health Stat 2013;56:1-37.
17. "Trends in energy intake in U.S. between 1977 and 1996: similar shifts seen across age groups". Obes Res 2002;10:370-378.
18. "Total energy intake: implications for epidemiologic analyses". Am J Epidemiol 1986;124:17-27.
19. "General cardiovascular risk profile for use in primary care: the Framingham Heart Study". Circulation 2008;117:743-753.
20. "Some health benefits of physical activity: the Framingham Study". Arch Intern Med 1979;139:857-861.
21. "Physical activity and life expectancy with and without diabetes: life table analysis of the Framingham Heart Study". Diabetes Care 2006;29:38-43.
22. "Anthropometric Standardization Reference Manual". Champaign, IL: Human Kinetics Press, 1988.
23. WHO. "Obesity and overweight: World Health Organization; 2020". Available at: https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight. Accessed
August 4, 2020 .24. "Does the relation of blood pressure to coronary heart disease risk change with aging? The Framingham Heart Study". Circulation 2001;103:1245-1249.
25. "Diets that follow the 2010 Dietary Guidelines for Americans (DGA) are associated with higher intakes of nutrients of concern". FASEB J 2012;26:S1: 267.1.
26. "Endothelial function, arterial stiffness and adherence to the 2010 Dietary Guidelines for Americans: a cross-sectional analysis". Br J Nutr 2015;113:1773-1781.
27. "Association between healthy eating patterns and risk of cardiovascular disease". JAMA Intern Med 2020;180:1090-1100.
28. "Healthful and unhealthful plant-based diets and the risk of coronary heart disease in U.S. adults". J Am Coll Cardiol 2017;70:411-422.
29. "Association between ultraprocessed food consumption and risk of mortality among middle-aged adults in France". JAMA Intern Med 2019;179:490-498.
30. "Association between consumption of ultra-processed foods and all cause mortality: SUN prospective cohort study". BMJ 2019;365:l1949.
31. "Consumption of ultra-processed foods and mortality: a national prospective cohort in Spain". Mayo Clin Proc 2019;94:2178-2188.
32. "Comparison of expert adjudicated coronary heart disease and cardiovascular disease mortality with the National Death Index: results from the REasons for Geographic And Racial Differences in Stroke (REGARDS) study". J Am Heart Assoc 2017;6:e004966.
33. "Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake". Cell Metab 2019;30:226.
34. "Minimally processed foods are more satiating and less hyperglycemic than ultra-processed foods: a preliminary study with 98 ready-to-eat foods". Food Funct 2016;7:2338-2346.
35. "Ultra-processed food consumption and chronic non-communicable diseases-related dietary nutrient profile in the UK (2008(-)2014)". Nutrients 2018;10:587.
36. "Etiologic effects and optimal intakes of foods and nutrients for risk of cardiovascular diseases and diabetes: systematic reviews and meta-analyses from the Nutrition and Chronic Diseases Expert Group (NutriCoDE)". PLoS One 2017;12:e0175149.
37. "Dietary sugars intake and cardiovascular health: a scientific statement from the American Heart Association". Circulation 2009;120:1011-1020.
38. "Comparison with ancestral diets suggests dense acellular carbohydrates promote an inflammatory microbiota, and may be the primary dietary cause of leptin resistance and obesity". Diabetes Metab Syndr Obes 2012;5:175-189.
39. "The Western diet-microbiome-host interaction and its role in metabolic disease". Nutrients 2018;10:365.
40. "Reshaping the gut microbiota: impact of low calorie sweeteners and the link to insulin resistance?"Physiol Behav 2016;164:Pt B: 488-493.
41. "Dietary emulsifiers directly alter human microbiota composition and gene expression ex vivo potentiating intestinal inflammation". Gut 2017;66:1414-1427.
42. "Short-term consumption of sucralose with, but not without, carbohydrate impairs neural and metabolic sensitivity to sugar in humans". Cell Metab 2020;31:493-502.e7.
43. "Assessing the health impact of phosphorus in the food supply: issues and considerations". Adv Nutr 2014;5:104-113.
44. "A comparison of a food frequency questionnaire with a 24-hour recall for use in an epidemiological cohort study: results from the biomarker-based Observing Protein and Energy Nutrition (OPEN) study". Int J Epidemiol 2003;32:1054-1062.
45. "Addressing current criticism regarding the value of self-report dietary data". J Nutr 2015;145:2639-2645.
46. "Nutrient warnings on unhealthy foods". JAMA 2020Oct1. [E-pub ahead of print].
47. "A clinician's guide for trending cardiovascular nutrition controversies: part II". J Am Coll Cardiol 2018;72:553-568.
Footnotes
Listen to this manuscript's audio summary by Editor-in-Chief Dr. Valentin Fuster on JACC.org.
The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the Author Center.
