Science

A 17,024-man analysis links ultra-processed food to prostate cancer — in data that cannot show which came first

Medically reviewed by Longevity Benchmark Editorial Last reviewed 2026-08-28

An FAU-led analysis of NHANES 2003–2023 reports 24% to 31% higher prostate cancer risk in men eating more ultra-processed food. The design is cross-sectional, and no confidence intervals were published.

August 28, 2026 · Science · Nutrition · Cancer screening

What was reported

A team led by Charles H. Hennekens at the Florida Atlantic University College of Medicine, with co-authors at FAU's Department of Population Health and a doctoral student at the University of Oxford, published an analysis titled "Consumption of Ultra-Processed Foods and Increased Risks of Prostate Cancer in a Random Sample of United States Adults" in The American Journal of Medicine. FAU announced it on August 12, 2026; ScienceDaily and the trade press picked it up in the last week.

The analysis covers 17,024 U.S. men aged 18 and older who took part in the National Health and Nutrition Examination Survey between 2003 and 2023. Each participant's diet came from two 24-hour dietary recalls, scored with the NOVA classification to give the percentage of daily calories from ultra-processed foods, and the men were split into four groups running from under about 20% of calories to more than 44%. Against the lowest group, the three higher groups combined showed a 29% greater rate of prostate cancer and the two highest showed 30%. After adjustment for age, race and ethnicity, smoking and poverty status, the release reports increases ranging from 24% to 31%. The single highest-intake group showed a possible 34% increase that did not reach statistical significance, which the authors attribute to the smaller number of cases in that group.

The short version

The exposure is plausible and the sample is large, but the design does the deciding here. This is cross-sectional survey data: diet and cancer history were captured at the same visit, so the analysis cannot show that the eating preceded the diagnosis. No confidence intervals, p-values or case counts were published in any source we could read, and the paper itself is paywalled. Treat it as a hypothesis generator that the authors themselves say needs prospective studies and trials — not as a finding that changes what you eat, what you test, or when you screen.

What it means for you

The practical risk with a story like this is not that readers will start eating differently. It is that a cancer headline lands on an audience already being marketed multi-cancer blood tests and whole-body scans, and converts diffuse worry into a purchase. That is worth naming plainly, because the study offers no support for it whatsoever. Nothing in an analysis of dietary recalls speaks to whether a screening test will find a cancer earlier, or whether finding it earlier will help you. Those are separate questions with their own evidence, and we cover them on how accurate cancer screening blood tests are and in our Galleri review.

The second thing worth holding onto is the asymmetry in what gets measured. The exposure in this study costs nothing to assess: the share of your calories coming from industrially formulated packaged food is something you can estimate from an honest look at a normal week. None of the platforms we benchmark measure it. They measure blood markers — dozens of them, at real expense — and the panels have grown considerably faster than the evidence that any given marker changes what a person does. A risk signal that is free to observe and absent from every commercial panel is a useful reminder of what those panels are actually optimized for, which is breadth that photographs well in a results dashboard.

Where this does connect to testing is at the level of expectation-setting. Readers frequently arrive at our multi-cancer blood test comparison having read a study like this one and wanting a test that will tell them whether the damage is done. No available test answers that question. Prostate screening decisions in particular turn on age, family history, ancestry, prior PSA values and symptoms — set against the well-documented harms of overdiagnosing cancers that would never have caused trouble — and they are a conversation with a clinician, not an inference from a dietary survey. On the diet side, the evidence base our nutrition hub is built on is much stronger for cardiometabolic outcomes than for cancer incidence, and that is where the measurable, near-term returns from cutting ultra-processed intake actually show up.

What this cannot tell you

This is a cross-sectional analysis of a national survey, as ScienceDaily describes it. NHANES records a participant's dietary recalls and their reported disease history at the same visit, so the data cannot establish that the diet preceded the diagnosis or rule out that a diagnosis changed the diet. Two 24-hour recalls stand in for years of eating. No confidence intervals, p-values or prostate cancer case counts appear in the institutional release or in any coverage we could access, and the journal article is paywalled — so none are reported here. The authors describe an association and explicitly call for randomized trials; an association is not causation. Nothing above indicates whether any screening test or dietary change is appropriate for you, which is a question for a clinician who knows your history.

How this fits what we already know

Ultra-processed intake keeps turning up on the unfavorable side of associations across categories, which is part of why the finding is easy to believe and hard to bank. Our briefing on the gut microbiome health index noted the same exposure tracking with lower index scores and higher subclinical inflammation in a separate cohort. The consistency is real. What stays missing across all of it is the design that would settle the question: people whose diets were recorded before anyone got sick, followed forward.

That gap is the same one we documented in epigenetic clocks underperforming conventional risk factors — a field can accumulate a great many suggestive associations while the question a buyer is actually asking goes unanswered. Hennekens frames the clinical takeaway around access rather than individual choice, noting that "healthcare providers should recognize that many patients face barriers to accessing healthier food options." That is a more honest reading of a population-level association than any personal instruction drawn from it would be, and it is closer to what this study can support.

Sources

Florida Atlantic University, "Ultra-Processed Foods Linked to Notably Higher Prostate Cancer Risks," FAU Newsdesk, August 12, 2026, fau.edu (accessed August 28, 2026). Institutional release; source for the author list, sample size, NHANES years, NOVA grouping, every risk figure quoted above and the author quotation.
Florida Atlantic University via EurekAlert!, news release 1139706, eurekalert.org (accessed August 28, 2026). Source for the full study title; confirms no confidence intervals, p-values or case counts were released.
"Eating more ultra-processed foods may raise prostate cancer risk by 30%," ScienceDaily, August 26, 2026, sciencedaily.com (accessed August 28, 2026). Source for the cross-sectional description of the design.
Hennekens C.H. et al., "Consumption of Ultra-Processed Foods and Increased Risks of Prostate Cancer in a Random Sample of United States Adults," The American Journal of Medicine, 2026. DOI 10.1016/j.amjmed.2026.07.043 (paywalled; not accessible to us at the time of writing, so no figures are taken from it).

Frequently Asked Questions

What did the study actually find?

In an analysis of 17,024 U.S. men aged 18 and older drawn from NHANES surveys between 2003 and 2023, men in the higher ultra-processed food consumption groups had a greater reported rate of prostate cancer than men in the lowest group. The institutional release puts the figure at 29% for the three higher groups combined and 30% for the two highest, and states that after adjustment for age, race and ethnicity, smoking and poverty status the increase ranged from 24% to 31%. A possible 34% increase in the single highest-intake group did not reach statistical significance.

Does this mean ultra-processed food causes prostate cancer?

No, and the authors do not claim it does. ScienceDaily describes the analysis as cross-sectional, and NHANES collects a participant's dietary recalls and their disease history at the same visit. That design cannot establish that the diet came before the diagnosis — a man diagnosed several years earlier reports the diet he is eating now, which may have changed because of the diagnosis. The authors call for large-scale observational studies and randomized trials, which is the appropriate next step for a hypothesis at this stage rather than a settled finding.

How was ultra-processed food defined?

By the NOVA classification system, applied to two 24-hour dietary recalls per participant, expressed as the percentage of daily calories coming from ultra-processed foods. Participants were split into four groups, described in the release as ranging from under roughly 20% of calories to more than 44%. Exact cut points for each group were not published in the sources available to us. Two days of recall is the standard NHANES instrument, and it is a snapshot rather than a record of long-term eating.

Should this change whether I get screened for prostate cancer?

Nothing here bears on that decision. Prostate screening decisions run on age, family history, ancestry, prior PSA results and symptoms, weighed against the known harms of overdiagnosis, and they belong with a clinician who knows your history. A dietary association measured in survey data does not enter that calculation, and no screening guideline uses diet as a criterion. If you are weighing a test, the criteria your clinician will actually use are the ones worth asking about.

Do any of the testing platforms you review measure this?

None of them measure ultra-processed food intake, and it is worth noticing that the exposure in this study is free to assess and requires no panel at all. The share of your calories coming from packaged, industrially formulated food is something you can estimate from your own week without ordering anything. That asymmetry — a widely covered risk signal that no commercial test captures, against dozens of blood markers that are heavily marketed — is a recurring feature of this category.