a person wearing blue jeans and a black shirt

Do Ultra-Processed Foods Cause Colorectal Cancer?

A 19-year study of 461,682 older US adults found no overall link between ultra-processed food and colorectal cancer — but risk split by food type.

A 19-year study of nearly 462,000 adults found no overall link between ultra-processed food and colorectal cancer — but the risk split sharply by food type, and the null is narrower than the headline suggests.

The wire headline was flat: ultra-processed food has no link to colorectal cancer. What the study actually shows is narrower and more interesting — a food category too mixed to yield one number, with its processed-meat content still tracking alongside higher risk.

What this means

Among 461,682 US adults aged 50 to 71 followed a median 18.8 years, those who ate the most ultra-processed food were no more likely to develop colorectal cancer than those who ate the least (highest vs. lowest fifth: HR 0.97, 95% CI 0.91–1.03). But the category is a mixed bag: ultra-processed meat products came with higher risk, while the whole grains, fiber and dairy in ultra-processed foods came with lower risk. This is an association from one diet questionnaire answered in 1995-96, not cause and effect, and the interval still permits a small effect either way.

The headline null, and the split underneath

A National Cancer Institute team analyzed the NIH-AARP Diet and Health Study: 461,682 cancer-free adults aged 50 to 71 at enrollment in 1995-96, followed a median 18.8 years, with 10,075 colorectal cancer cases. Diets came from a single 124-item food questionnaire at the start. Median ultra-processed intake was 293 g per 1,000 kcal/day — about 43% of daily calories.

Comparing the highest fifth of intake with the lowest, the hazard ratio for colorectal cancer was 0.97 (95% CI 0.91–1.03), with a flat trend across fifths (p-trend 0.55). That rules out anything beyond roughly 3% higher risk — but not a small effect in either direction.

The average conceals opposing subgroups. Ultra-processed reconstituted meat and fish products came with higher risk (HR 1.10, 95% CI 1.04–1.18). Going the other way were ultra-processed bread (HR 0.91, 0.86–0.97), breakfast cereals (HR 0.84, 0.79–0.90) and frozen desserts and milk drinks (HR 0.82, 0.77–0.88) — the category’s whole-grain, fiber and dairy carriers. These subgroup estimates are exploratory — the paper states no correction for testing many comparisons — so their sizes are tentative.

Association, not cause

The design supports association only: an observational cohort with one baseline diet questionnaire in 1995-96, tracking outcomes to 2018. Participants were mostly non-Hispanic White older adults recruited through AARP membership, so the findings describe that population and dietary era — not children, younger adults or current food supplies. Questionnaire misclassification was moderate, and unmeasured diet and lifestyle differences could still confound these estimates.

Where this fits

This is the largest single-cohort null so far — not evidence of no effect. Pooled analyses lean mildly positive: three US cohorts found higher risk in men at the highest intake (HR 1.29, 95% CI 1.08–1.53; no association in women), colorectal cancer precursors were more common at high intake (odds ratios 1.18 and 1.20), and two meta-analyses put risk roughly 11% to 13% higher at the top of the range. We have written before about ultra-processed-food-and-heart-risk-what-51-studies-show-and-where-the-certainty-ends and about whether does-cutting-processed-foods-improve-health. What this paper adds is why a single ultra-processed-food number can read as zero: the category averages processed meat against whole grains, fiber and milk.

Terms explained

  • Ultra-processed food (Nova group 4): industrial formulations built mostly from extracted or modified ingredients plus additives — soft drinks, packaged snacks, reconstituted meats, many mass-produced breads and cereals.
  • Hazard ratio: in a group followed over time, the rate of an outcome in one group divided by the rate in another. 1.00 means no difference.
  • 95% confidence interval (CI): the range of values statistically compatible with the data; an interval spanning 1.00 is consistent with no association.
  • Quintile: one of five equal groups ranked by intake. The highest fifth is Q5, the lowest Q1.
  • Food frequency questionnaire (FFQ): a survey asking how often a person ate each of a list of foods, used to estimate usual intake. It measures diet with error.

Sources

  1. Primary source: Abar L, O’Connell CP, Hong HG, et al. “Associations of food processing and dietary risk factors with colorectal cancer incidence in the NIH-AARP Diet and Health Study: A cohort study.” PLOS Medicine 2026;23(10):e1005251. DOI: 10.1371/journal.pmed.1005251 · PMID: 42879326 · PMCID: PMC13671506. Peer-reviewed prospective cohort analysis; full text read (CC0).
  2. Landscape — three US cohorts: Wang L, et al. “Association of ultra-processed food consumption with colorectal cancer risk among men and women: results from three prospective US cohort studies.” BMJ 2022;378:e068921. DOI: 10.1136/bmj-2021-068921. Abstract-level access.
  3. Landscape — precursor lesions: Hang D, et al. “Ultra-processed food consumption and risk of colorectal cancer precursors: results from 3 prospective cohorts.” J Natl Cancer Inst 2023;115(2):155-164. DOI: 10.1093/jnci/djac221. Abstract-level access.
  4. Landscape — meta-analysis: “Ultra-processed food consumption and gastrointestinal cancer risk: systematic review and meta-analysis.” Am J Gastroenterol 2024;119(6):1056-1065. DOI: 10.14309/ajg.0000000000002826. Abstract-level access.
  5. Landscape — latest synthesis: Akkapelli SS, et al. “Ultra-Processed Food Consumption and Colorectal Cancer Risk: a Systematic Review and 2-Stage Mediation Meta-Analysis.” Nutr Rev 2026. DOI: 10.1093/nutrit/nuaf283. Abstract-level access.

Related Articles