Quick Answer: Adults with the highest ultra-processed food intake had a 47% higher prevalence of self-reported heart attack or stroke compared to those with the lowest intake, in a cross-sectional analysis of 4,787 U.S. adults aged 18 and older from NHANES 2021-2023 (RR = 1.47; 95% CI: 1.06-2.04; P = 0.02), after adjusting for age, sex, race/ethnicity, income, and smoking.
Last reviewed: 2026-06-08
Ultra-processed foods now account for more than half of daily calories for the average American adult. That share has risen steadily for decades, and so has the research linking it to poor health outcomes.
A 2026 cross-sectional study published in The American Journal of Medicine adds a specific finding to that body of evidence. Among 4,787 U.S. adults, those in the highest quartile of ultra-processed food intake had a 47% higher prevalence of self-reported heart attack or stroke compared to those in the lowest quartile.
That number needs context. Here is what the study found, what it can and cannot tell us, and what it means for how you eat.
What Did the 2026 Study Find?
Researchers Willett Y, Yang C, Dunn J, and colleagues analyzed dietary and health data from 4,787 U.S. adults aged 18 and older who participated in NHANES 2021-2023, one of the most comprehensive ongoing surveys of American health and nutrition.[1]
They used the NOVA food classification system to quantify each participant's ultra-processed food intake, then divided participants into four groups (quartiles) based on that intake level. The primary outcome was self-reported heart attack or stroke: whether a physician had ever told the participant they had experienced a myocardial infarction or stroke.
Key finding: participants in the highest quartile of UPF intake had a 47% higher prevalence of heart attack or stroke compared to participants in the lowest quartile (RR = 1.47; 95% CI: 1.06-2.04; P = 0.02). The analysis controlled for age, sex, race/ethnicity, poverty-income ratio, and smoking status.
Two methodological details matter here.
The outcome was self-reported. Participants answered a questionnaire; no medical records were verified. Self-reported data for major events like heart attack and stroke tend to be reasonably reliable, but they cannot confirm clinical severity, timing, or cause.
The design was cross-sectional. The study took a snapshot in time. It does not follow people forward to see whether UPF intake predicts future events. This means the association could reflect reverse causation (people in poorer health changing their diet) or unmeasured confounders. Cross-sectional studies identify correlations and generate hypotheses; they do not establish that ultra-processed foods caused higher cardiovascular event rates.
What the data does say: at the population level, eating more ultra-processed food is associated with a meaningfully higher prevalence of reported cardiovascular events. The association was statistically significant and held after controlling for several major confounding variables.
Source: Willett Y, Yang C, Dunn J, et al. The American Journal of Medicine. 2026;139(5):664-667. DOI: 10.1016/j.amjmed.2026.01.012. PMID: 41587677.
What Are Ultra-Processed Foods, and How Does NOVA Classify Them?
Ultra-processed foods are not just foods that have been processed. Almost all food is processed in some way. The NOVA system, developed at the University of Sao Paulo and now widely used in public health research, groups foods by the extent and purpose of industrial processing:
- NOVA Group 1 (unprocessed or minimally processed): fresh fruits and vegetables, plain meat, eggs, dried legumes, plain yogurt.
- NOVA Group 2 (processed culinary ingredients): oils, butter, sugar, flour, salt used in cooking.
- NOVA Group 3 (processed foods): canned fish and vegetables, cured meats, simple aged cheeses, salted nuts.
- NOVA Group 4 (ultra-processed foods): soft drinks, packaged chips and crackers, reconstituted meat products, instant noodles, flavoured breakfast cereals, many packaged breads, flavoured dairy drinks.
Group 4 is defined by ingredients that have no home-kitchen equivalent: hydrolyzed proteins, modified starches, hydrogenated oils, artificial flavours and colours, emulsifiers, high-fructose corn syrup. These are products formulated for palatability, convenience, and shelf stability, not for nutritional quality.
In the NHANES data Willett et al. analyzed, participants in the highest UPF quartile were getting a substantial share of their daily calories from these products.
How Does This Study Fit the Broader Evidence?
The 2026 finding is consistent with a pattern across multiple large epidemiological datasets.
A 2019 analysis in the BMJ, drawing on 105,159 French participants in the NutriNet-Sante cohort, found that a 10% increase in UPF intake was associated with a 12% higher risk of cardiovascular events.[2] A 2021 study in the European Heart Journal found higher UPF consumption associated with greater cardiovascular disease risk in a large European cohort.[3]
None of these studies, including the 2026 AJM paper, establish causation. The consistency across different countries, populations, and study designs does strengthen the overall signal. When multiple independent research teams find similar associations using different datasets and methods, the likelihood that the finding reflects a real relationship increases.
Proposed biological mechanisms include chronic inflammation from pro-inflammatory dietary patterns, disruption of gut microbiome diversity, the metabolic effects of emulsifiers and artificial sweeteners, and the cumulative effect of poor nutrient density in high-UPF diets. These are hypotheses under active investigation, not established pathways.
The 2026 Willett et al. paper draws on a recent, nationally representative U.S. sample (NHANES 2021-2023). Much earlier research on this topic relied on European cohorts or older American data. The NHANES 2021-2023 dataset reflects contemporary eating patterns in the United States, which adds to its relevance.
What Should You Actually Do With This Information?
The 47% figure is striking but it comes with real uncertainty. The confidence interval runs from 1.06 to 2.04. At the lower bound, the true association could be as small as 6% higher prevalence. The study is cross-sectional and relies on self-reported outcomes. These are not reasons to dismiss the finding, but they are reasons not to treat it as a precise causal estimate.
The evidence does not say that every ultra-processed item in your grocery cart is shortening your life by a calculable amount. What it does consistently show is that diets anchored in whole and minimally processed foods are associated with better cardiovascular outcomes across a wide range of studies.
Practically, that means: cook from minimally processed ingredients more often, read ingredient lists, and treat products with long lists of industrial additives as occasional rather than daily items. You do not need to calculate your NOVA quartile. You need a general sense of where your diet sits.
If you want to check the processing level of packaged foods you buy regularly, the NoJunk app evaluates grocery items by their ingredient list, so you can see where a product lands and find alternatives. For a deeper look at what distinguishes ultra-processed from processed, see our explainer on ultra-processed foods.
Frequently Asked Questions
What does "cross-sectional" mean, and why does it matter for interpreting this study?
A cross-sectional study measures both the exposure (diet) and the outcome (heart attack or stroke history) at the same point in time, using a single survey rather than following participants forward. This design is efficient and useful for describing patterns across a population, but it cannot establish that one thing caused another. In the Willett et al. study, it is possible that people who already had cardiovascular disease changed their diets before the survey, or that lower income (associated with both UPF consumption and cardiovascular risk) explains part of the association. The cross-sectional design cannot rule these out. The finding is still meaningful, but "associated with" is the right language, not "causes."
Is a 47% higher prevalence a large effect in nutrition research?
It is meaningful but not enormous. A relative risk of 1.47 means the highest UPF consumers had a 47% higher prevalence of reported cardiovascular events compared to the lowest consumers, not that 47% of high consumers had a heart attack or stroke. The confidence interval (1.06 to 2.04) is fairly wide, reflecting genuine uncertainty in the estimate. At the lower bound, the true effect could be just 6% higher prevalence. For context, smoking roughly doubles cardiovascular mortality risk (RR around 2.0). A 47% difference across dietary quartiles, while significant and consistent with other research, is a weaker signal than major established risk factors like smoking or hypertension.
Should I stop eating all ultra-processed foods?
No study, including this one, supports complete elimination. The comparison in the Willett et al. study is between the highest and lowest quartiles of intake, not between people who eat zero UPFs and everyone else. The practical implication of the evidence is that reducing high-UPF intake, particularly from daily to occasional consumption, is a reasonable goal. Eliminating all packaged food is unrealistic and unnecessary. A diet that mostly consists of whole and minimally processed foods, with room for convenience items, is what the research describes as lower-risk.
What is the NOVA food classification and who developed it?
NOVA was developed by researchers at the University of Sao Paulo's Center for Epidemiological Studies in Health and Nutrition, led by Carlos Monteiro. First published in 2010 and updated since, it classifies foods into four groups based on the extent and purpose of industrial processing rather than on nutrient content alone. NOVA Group 4 (ultra-processed) is defined by the use of industrial ingredients (like hydrolyzed proteins and emulsifiers) that have no home-kitchen equivalent. NOVA is now used in dietary guidelines in several countries and in a substantial body of published nutrition research.
How were heart attacks and strokes measured in the NHANES study?
NHANES uses structured interviews in which trained staff ask participants whether a doctor or health professional has ever told them they had a heart attack (myocardial infarction) or stroke. Participants answer yes or no. This is a self-reported measure: it captures diagnosed events that participants can recall and report, but it does not include medical record verification, clinical severity, or information about when the event occurred relative to dietary patterns. Self-report for major cardiovascular events is generally considered a reasonably reliable method in epidemiological research, though under-reporting and recall errors can occur.
Sources
| # | Source | Date accessed |
|---|---|---|
| 1 | Willett Y, Yang C, Dunn J, et al. The American Journal of Medicine. 2026;139(5):664-667. DOI: 10.1016/j.amjmed.2026.01.012. PMID: 41587677. | 2026-06-08 |
| 2 | Srour B, Fezeu LK, Kesse-Guyot E, et al. Ultra-processed food intake and risk of cardiovascular disease: prospective cohort study. BMJ. 2019;365:l1451. DOI: 10.1136/bmj.l1451. | 2026-06-08 |
| 3 | Bonaccio M, Di Castelnuovo A, Costanzo S, et al. Ultra-processed food consumption and cardiovascular risk. European Heart Journal. 2021. | 2026-06-08 |
Last reviewed: June 2026