(Photo by Emmy Smith from Unsplash)
In A Nutshell
- Heavy ultra-processed food eaters showed no higher colorectal cancer rate among 461,682 older Americans.
- Packaged bread and cereal were tied to lower risk, reconstituted meat and fish to higher.
- Poor diet quality came with higher risk in some intake groups.
- Authors say findings support more whole grains, fiber, and dairy and less processed meat.
People who ate the most ultra-processed food in a study of 461,682 older Americans developed colorectal cancer (cancer of the colon or rectum) at about the same rate as people who ate the least, even after roughly two decades of follow-up.
That ran against the expectations of researchers with the National Cancer Institute and partner institutions, who had predicted heavy eaters of these foods would fare worse. Instead, the label turned out to hide a split. Packaged bread, breakfast cereal, and frozen desserts and milk drinks were tied to lower cancer rates, while reconstituted meat and fish products, such as sausages and fish sticks, were tied to higher ones. Those opposite patterns may help explain why the overall number came out flat.
Ultra-processed foods supply more than half the calories US adults eat, and some earlier studies have linked them to higher colorectal cancer risk, so the question touches nearly every grocery cart.
No Clear Ultra-Processed Food Link Appeared Across 10,075 Colorectal Cancer Cases
Published in PLOS Medicine, the study used the NIH-AARP Diet and Health Study, which signed up adults ages 50 to 71 in 1995 and 1996. Participants came from AARP’s membership in six states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) and two metro areas (Atlanta and Detroit). Everyone filled out a questionnaire about 124 foods and drinks once, at enrollment, so the results reflect mid-1990s eating habits. After researchers removed people who already had cancer, kidney failure, or poor self-rated health, among other reasons, 461,682 remained: 274,269 men and 187,413 women, roughly 91% of them non-Hispanic white.
To judge how processed each food was, researchers used a ranking system called Nova, which sorts foods by how much industrial processing they go through, from fresh produce at one end to factory-made snacks and packaged meals at the other. A typical participant got about 43% of daily calories from ultra-processed foods, a share the authors say likely understates today’s. Participants were then split into five equal groups by how much ultra-processed food they ate.
Participants were followed through 2018, or until a diagnosis, death, or move out of the study area. State cancer registries tracked new colorectal cancer cases. Half the group was tracked for more than 18.8 years, and 10,075 people were diagnosed. The comparison accounted for age, sex, race, education, smoking, alcohol, exercise, family history of cancer, and self-rated health.
Packaged Bread and Cereal Tied to Lower Colorectal Cancer Risk, Reconstituted Meat to Higher
People in the highest fifth of ultra-processed intake had 3% lower risk than people in the lowest fifth, a gap small enough to be chance. Looking at the upper colon, lower colon, and rectum separately also showed no clear link.
Food by food, the split was clear. Comparing the highest and lowest fifths of intake, risk was 16% lower for breakfast cereal, 18% lower for frozen desserts and milk drinks, and 9% lower for bread, but 10% higher for reconstituted meat and fish.
Nutrients may explain part of that. Ultra-processed foods supplied 82.8% of participants’ whole grains, but only about a third of their fiber and calcium and under a quarter of their dairy. They also supplied 74.3% of the processed meat. Whole grains and fiber from ultra-processed sources were tied to lower risk, dairy and calcium were linked to lower risk from any source, and meat to higher risk from any source.
In a separate model, swapping 10% of the weight of food eaten from ultra-processed sources for minimally processed food, such as fresh produce, was associated with a 2% lower colorectal cancer rate. That estimate comes from comparing what people reported eating, not from a trial in which anyone changed their diet.
Poor Diet Quality Came With Higher Colorectal Cancer Risk at Some Intake Levels
Researchers also scored overall diets on the Healthy Eating Index-2020, a 100-point rating of how closely eating matches US dietary guidelines. People with poor scores (below 51) in two of the five ultra-processed groups, the second-lowest and the highest, had 28% and 22% higher risk than people who ate little ultra-processed food and had good scores. The other groups showed no consistent pattern, and statistical tests found no sign that diet quality and ultra-processed intake worked together.
Results were similar for men and women and across weight categories, and adjusting for diet quality, nutrients, or body weight left the overall result largely unchanged.
Study Supports Eating More Whole Grains and Fewer Processed Meats
Existing dietary advice held up: more whole grains, fiber, dairy, and calcium, and less processed meat. The study cannot prove cause and effect, and it does not show that nutrients from ultra-processed and less processed foods act identically in the body. Participants were older and mostly white, so the findings may not apply to younger or more diverse groups.
Grocery shoppers hunting for a single bad label will not find one here. What is in the package, and what else is on the plate, may carry more weight than the term ultra-processed.
Disclaimer: This article reports on a published research study and is for informational purposes only. It is not medical advice. Readers should consult a qualified health professional before changing their diet or health care.
Paper Notes
Limitations
According to the authors, the questionnaire and food databases were built to estimate nutrient intake, not how processed a food is. Someone who reported eating yogurt, for example, might have eaten a plain or flavored version, which fall into different processing groups, and the questionnaire could not tell them apart. Diet was self-reported and measured only once, at the start of the study in the mid-1990s, so it does not capture changes in eating habits over time or in the food supply. The authors say their estimates likely understate current ultra-processed food intake, and that limited variation across the five intake groups may also have contributed to the null result. Participants were older, mostly white adults, so findings may not apply to younger or more diverse groups. Food additives could not be studied because ingredient-level data were unavailable. Because the design is observational, unmeasured factors such as use of pain relievers like ibuprofen or poorly measured ones such as socioeconomic status could play a role. The authors estimate that an unmeasured factor would need to be linked to both ultra-processed food intake and cancer risk with more than a 2.1-fold higher relative rate to fully explain away their strongest finding, which involved fiber. Analyses of individual foods and nutrients were described as exploratory, and statistically significant results from them should be interpreted with caution because of multiple testing.
Funding and Disclosures
Funding came from the Intramural Research Program of the National Cancer Institute, part of the National Institutes of Health. Seven of the authors are or were employees of the funder. According to the paper, the funder had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript. The authors declared no competing interests.
Publication Details
Title: “Associations of food processing and dietary risk factors with colorectal cancer incidence in the NIH-AARP Diet and Health Study: A cohort study” Authors: Leila Abar, Caitlin P. O’Connell, Hyokyoung G. Hong, Kirsten A. Herrick, Lisa Kahle, Jennifer L. Lerman, Linda M. Liao, Xuehong Zhang, Xinyuan Zhang, Longgang Zhao, Sémi Zouiouich, Rashmi Sinha, Neha Khandpur, Eurídice Martínez Steele, and Erikka Loftfield (corresponding author) Journal: PLOS Medicine, 23(10): e1005251 Published: October 15, 2026 DOI: https://doi.org/10.1371/journal.pmed.1005251
