Ultra-Processed Food and Heart Risk: Waist Size Explains Half

A prospective cohort analysis from the Multi-Ethnic Study of Atherosclerosis (MESA), published in JACC: Advances, reports that waist size (central adiposity) and inflammation statistically accounted for about 47% of the link between ultra-processed food and heart disease over a median 16.7 years. In 6,531 US adults, higher ultra-processed food intake went with a higher waist-to-height ratio and higher interleukin-6 (IL-6), an inflammation marker, while LDL cholesterol and blood pressure were not associated with ultra-processed food intake.

Key takeaways:

• In a cohort of 6,531 multi-ethnic US adults followed for a median of 16.7 years, waist-to-height ratio and IL-6 inflammation statistically accounted for roughly 47% of the relationship between ultra-processed food consumption and hard cardiovascular events.
• Waist-to-height ratio was the larger statistical mediator: 44% of the association in its own model and 38% when modeled together with IL-6. IL-6 accounted for 13% on its own and 9% in the joint model.
• Ultra-processed food intake was not significantly associated with blood pressure, LDL cholesterol, oxidized LDL, triglycerides, total cholesterol, fasting glucose, or serum insulin.
• After accounting for waist-to-height ratio and IL-6, the direct effect of ultra-processed food intake on cardiovascular disease was no longer statistically significant.

What happened

PubMed lists the study, from the Multi-Ethnic Study of Atherosclerosis (MESA), as published online ahead of print on September 18, 2026; the journal's own page gives an online publication date of September 30, 2026. The full text is open access in JACC: Advances. The paper, titled "Central Adiposity and Inflammation Mediate the Association Between Ultraprocessed Food Intake and Cardiovascular Disease Risk," was written by Haidar and colleagues at UTHealth Houston, UCLA, Harvard Medical School, the USDA/ARS Children's Nutrition Research Center (Baylor) and Wake Forest.

The authors analyzed long-term observational data from 6,531 participants aged 45 to 84 years across six United States communities: Baltimore, Chicago, Forsyth County in North Carolina, Los Angeles County, Northern Manhattan and the Bronx in New York, and St. Paul, Minnesota. At baseline enrollment, which occurred between July 2000 and August 2002, all participants were free of clinically apparent cardiovascular disease.

Over a median follow-up period of 16.7 years (with an interquartile range of 11.2 to 17.5 years), 809 physician-adjudicated incident hard cardiovascular disease events took place. These events included nonfatal myocardial infarction, resuscitated cardiac arrest, coronary heart disease death, nonfatal stroke, and stroke death.

What changed

The study used mediation analysis, a statistical method that estimates how much of a link between an exposure and an outcome runs through other measured factors, here waist size, blood markers and blood pressure.

The researchers found that each additional daily serving of ultra-processed food was associated with a higher waist-to-height ratio (model coefficient beta: 0.006; 95% confidence interval: 0.005 to 0.007), higher levels of the inflammatory cytokine interleukin-6 (beta: 0.038; 95% CI: 0.023 to 0.054), and lower high-density lipoprotein (HDL) cholesterol (beta: -0.256; 95% CI: -0.432 to -0.080). Conversely, blood pressure, LDL cholesterol, triglycerides, total cholesterol, oxidized LDL, fasting glucose, and serum insulin levels were not associated with ultra-processed food consumption in this cohort.

For cardiovascular disease, waist-to-height ratio and IL-6 were the two significant statistical mediators. In individual models, waist-to-height ratio mediated 44% of the total association (beta: 0.012; 95% CI: 0.007 to 0.017), while IL-6 mediated 13% (beta: 0.003; 95% CI: 0.001 to 0.006). In parallel mediation models that evaluated both markers simultaneously, both retained independent indirect effects: adiposity accounted for 38% and IL-6 accounted for 9%, yielding a total combined mediation of roughly 47% (Table 3 combined proportion: 0.47).

After accounting for these two factors, the direct effect of ultra-processed food intake on cardiovascular disease was no longer statistically significant. The overall association was itself small: a total effect per daily serving of 0.027 on the logit scale (95% CI: 0.001 to 0.053), only just statistically significant.

Who is affected

Participants were US adults aged 45 to 84 from six communities, and their diet was recorded in 2000 to 2002, so the results describe that group and era. Intake was measured in servings per day, broken down into quartile means of 1.31, 2.89, 4.65, and 8.67 daily servings.

Rather than assessing specific chemical formulations, the study classified intake using a baseline 120-item food-frequency questionnaire categorized under the Nova system. The specific items classified as ultra-processed in this analysis included:

• Breakfast foods and grains: pancakes, hot cereal, cold cereal, white bread, dark bread, muffins, biscuits, and instant breakfast.
• Snack items and sweets: chips, crackers, white donuts, chocolate donuts, pies, pudding, candy, ice cream, frozen yogurt, and tofu dessert.
• Convenience entrees and sides: French fries, oriental noodles, dumplings, chow mein, refried beans, pizza, hamburger, hamhocks, fried chicken, and fried fish.
• Beverages and spreads: margarine on rolls, sweet milk, soy milk, soda, diet soda, hot cocoa, and liquor.

Because Nova labels whole food categories as ultra-processed, items as different as candy, diet soda and soy milk were counted in the same group. The data reflect collective dietary patterns rather than the risk profile of any single product.

What it means for shoppers

The JACC: Advances paper (PubMed) suggests that, in this group, the link between ultra-processed food and heart disease ran largely through waist size and inflammation rather than LDL cholesterol or blood pressure. The authors stress that mediation models do not prove cause and effect.

Nova is a research classification for broad food types, not a regulatory safety rating and not a verdict on any single ingredient. The MESA list put dark bread and soy milk in the same group as candy, soda and fried chicken. A shopper cannot judge heart risk from processing status alone. For how regulators are trying to define the term, see US to Define 'Ultra-Processed Food': What the Proposal Means, and for the wider evidence, Ultra-Processed Food and 8.8 Million Adults: What the Review Found.

When reviewing packaged goods, practical steps include:

• Checking the Nutrition Facts panel for fiber, added sugars, and saturated fat rather than assuming a product is wholesome based on marketing terms.
• Reading the ingredient deck to see whether whole-food components make up the majority of the formulation.

This is an observational study of overall diet. It cannot say what eating any one packaged food does to a person's heart risk.

The background

MESA is funded through contracts from the National Heart, Lung, and Blood Institute (NHLBI), National Center for Advancing Translational Sciences (NCATS) grants, and US EPA STAR agreements under MESA Air. MESA began enrolling participants in 2000. The authors reported no relationships relevant to the paper.

The analysis asked whether the link between Nova-classified ultra-processed food and heart disease runs through standard risk factors such as blood pressure and LDL cholesterol, or through waist size and inflammation.

The authors noted crucial limitations inherent to the study design. Dietary intake was assessed at a single baseline examination between 2000 and 2002 using a 120-item food-frequency questionnaire, so dietary changes during the median 16.7-year follow-up were not captured. Candidate mediators were likewise measured only at baseline. Furthermore, residual confounding cannot be ruled out, mediation models rely on assumptions of no unmeasured confounding, and statistical indirect effects do not prove biological causality. Additionally, while indirect effects were statistically significant, the per-mediator proportion-mediated 95% confidence intervals in Table 4 crossed zero (waist-to-height ratio: -0.01 to 0.88; IL-6: -0.02 to 0.29).

The role of specific ingredients is unknown. The questionnaire was coarse, so items like "cold cereal" or "diet soda" were captured as single broad variables. As a result, the data cannot establish whether specific sweeteners, preservatives, or processing techniques drive these associations.

What to watch next

In the paper's Perspectives section, the authors call for intervention studies to test whether cutting ultra-processed food reduces central adiposity and inflammation.

They also call for research on other possible pathways, including food additives, gut dysbiosis and oxidative stress. Food additives were not tested in this analysis.

If you want to look past the ultra-processed label to the actual ingredients, Aurascan is a free food-label scanner app in beta (iOS TestFlight, with a waitlist for other platforms). It pulls product data from Open Food Facts, USDA and barcode databases and explains every ingredient one by one with cited evidence, using per-ingredient verdicts (Clean, Mixed or Flagged) instead of a single score. Citations are checked by an automated verifier, and the app has no sponsors. Scan your next label to see what is in it.