Cohort studycited by the Tayebat archivefull text checked★ key study A-UPF-06 in-depth review
Ultra-processed food and cancer in France
Consumption of ultra-processed foods and cancer risk: results from NutriNet-Santé prospective cohort.
Overview
Fiolet T et al.. BMJ 2018 · PMID 29444771 · DOI 10.1136/bmj.k322
Each 10% more ultra-processed food was linked with 12% more cancer.
Does the study support the specific claim? Yes · Methodological quality: moderate
Plain-language summary
Study explained simply
- What did the researchers want to find out?
Is UPF intake associated with cancer risk?
- How was the study done?
Population-based prospective cohort with repeated 24-h records.
- How many people, animals or samples were studied?
104,980; 2,228 cancer cases; Adults >=18 (median age 42.8)
- What was actually found?
Overall cancer HR 1.12 (1.06-1.18); breast cancer 1.11 (1.02-1.22); robust to nutritional-quality adjustment.
- How large was the observed effect?
HR 1.12 (1.06-1.18)
- What do the results mean in everyday life?
Association only.
- What are the study's limitations?
Observational; only about 5 years of follow-up, which is short for cancer; volunteers, mostly women; no link found for prostate or bowel cancer; may reflect other lifestyle differences.
- What does it say about Elawady's specific claim?
Supports C-UPF-03.
- Do the results apply to Germany and Europe?
High.
- What can explicitly not be concluded from this study?
Causation.
Scientific in-depth review
1. Full reference
- Authors
- Fiolet T, Srour B, Sellem L, Kesse-Guyot E, Allès B, Méjean C, Deschasaux M, Fassier P, Latino-Martel P, Beslay M, Hercberg S, Lavalette C, Monteiro CA, Julia C, Touvier M
- Journal
- BMJ
- Year
- 2018
- Identifier
- PMID 29444771 DOI 10.1136/bmj.k322 · PMC5811844
- Verification
- verified – PubMed E-utilities efetch, 2026-10-25
- Retracted
- no
- Corrections
PubMed lists: ErratumIn (BMJ. 2026 Jul 16;394:e100298. doi: 10.1136/bmj-2026-100298.)
- Extent checked
- full text
- Conflicts of interest and funding
Public funding (French Ministry of Health, InVS, INPES, Région Ile-de-France, INSERM, INRA, CNAM, Université Paris 13, INCa); funders had no role. ICMJE: no competing interests.
2. Research question
Is UPF intake associated with cancer risk?
3. Study design and population
- Study design
- Cohort study – Population-based prospective cohort with repeated 24-h records.
- Population
French adult volunteers >=18 years without cancer at baseline; 78.3% women; mean age 42.8 (SD 14.8)
- Country
France
- Period
Recruitment from May 2009; follow-up to 1 January 2017; median 5 years
- Sample size
104,980; 2,228 first incident cancers (739 breast, 281 prostate, 153 colorectal); 426,362 person-years
- Inclusion and exclusion criteria
Participants without cancer at baseline with >=2 valid 24-h records in the first 2 years; energy under-reporters excluded; basal cell carcinoma not counted.
4. Methods and endpoints
- Exposure / intervention
Proportion (% by weight) of ultra-processed foods in the diet from repeated 24-h records (mean 5.4 per participant), per 10-point increment and sex-specific quarters
- Comparator
Lower UPF
- Endpoints
Overall, breast, prostate, colorectal cancer
- Statistical methods
Cox models (age timescale) adjusted for sex, BMI, height, activity, smoking, number of records, alcohol, energy, family history of cancer, education (+ reproductive factors for breast cancer); further adjustment for lipids/sodium/carbohydrates and Western pattern; mediation analysis; 2-year lag; multiple imputation; stratified analyses.
5. Actual results
- Results
Per 10-point UPF increment: overall cancer HR 1.12 (1.06-1.18); breast cancer 1.11 (1.02-1.22) (postmenopausal P=0.04; premenopausal P=0.2, not significant); prostate cancer 0.98 (0.83-1.16) and colorectal cancer not significantly associated. Highest vs lowest quarter, overall cancer: HR 1.21 (1.06-1.38). Excluding cancers in the first 2 years: overall 1.10 (1.03-1.17), breast 1.15 (1.03-1.29). No mediation by nutritional factors (0-2%). Minimally processed foods: overall cancer HR 0.91 (0.87-0.95) per 10 points. UPF fats and sauces, sugary products and drinks associated with overall cancer.
- Absolute risks
Absolute incidence by exposure not reported in full text; cases by quarter 712, 607, 541 and 368 (lowest to highest; crude counts, high-UPF participants younger). Cohort age- and sex-standardised cancer incidence 786 per 100,000 per year (France 972).
- Authors' conclusion
A 10% increase in the proportion of UPF in the diet was associated with a >10% increase in risks of overall and breast cancer; results should be confirmed in other populations; further studies are needed on nutritional composition, additives, contact materials and neoformed contaminants.
6. Strengths
- Large cohort with repeated detailed 24-h records and item-level NOVA classification
- >90% of cancer cases with medical records; linkage to national health insurance and mortality registry
- Extensive sensitivity and mediation analyses
- Public funding; no COI
7. Weaknesses
- Short median follow-up (5 years) relative to cancer latency (authors)
- Volunteer, health-conscious, mostly female cohort (authors)
- High-UPF consumers younger, more often smokers and less active: residual confounding (authors)
- NOVA misclassification possible (authors)
- Limited power for site-specific cancers; prostate and colorectal null
- Possible detection/ascertainment differences; exhaustive case detection not guaranteed (authors)
- The text reports a colorectal Q4 HR '1.23 (1.08 to 1.40), P trend 0.07', identical to the all-cancer model 4 estimate - an apparent reporting inconsistency
8. Risk of bias
- Orienting tool
- ROBINS-E
- Type of assessment
- structured assessment based on the full text by a single reviewer (no formal double review)
- Full formal assessment
- no
- Overall risk of bias
- some concerns
- Rationale
Structured single-reviewer judgement from the full text; a complete formal ROBINS-E assessment requires two independent reviewers and a protocol.
| Domain | Judgement | Rationale |
|---|---|---|
| Confounding | some concerns | Broad adjustment, but UPF intake clusters with younger age, smoking and inactivity; residual confounding acknowledged. |
| Exposure measurement | some concerns | Repeated validated 24-h records; NOVA classification by dietitians may misclassify; weight-based metric. |
| Selection of participants | some concerns | Volunteer cohort; exclusion of under-reporters; dropout 6.7%. |
| Post-exposure interventions | low | No individual feedback; short follow-up. |
| Missing data | low | Low covariate missingness; multiple imputation and complete-case analyses similar. |
| Outcome measurement | some concerns | Self-report validated by records (95% confirmed) and linkage; possible differential detection by health-seeking behaviour. |
| Selection of reported result | some concerns | Several cancer sites and subgroups; null sites reported; one internal inconsistency in colorectal estimate. |
Full-text review (25 October 2026): Headline numbers confirmed. Added inclusion criteria, person-years, quarter HR 1.21 (1.06-1.38), 2-year-lag results, null prostate/colorectal and premenopausal breast results, mediation 0-2%, protective minimally processed food association, crude case counts per quarter, cohort incidence, funding/COI, and an internal inconsistency in the reported colorectal estimate. Domain-level ROBINS-E (overall some_concerns, unchanged).
9. Assessment of the statistical interpretation
Modest relative associations (HR 1.12 per 10 points; 1.21 top vs bottom quarter) after only ~5 years, a short period for cancer development, which raises the possibility of residual confounding or detection effects; no association for prostate or colorectal cancer.
10. Relevance to the specific claim
Supports C-UPF-03.
11. Independent contrary findings
Cancer associations in umbrella review mostly weaker grades (A-UPF-13).
12. Transferability to Germany and Europe
High.
13. Overall methodological judgement
Early, supportive signal for overall and postmenopausal breast cancer from one French volunteer cohort; association only, short follow-up, site-specific findings mostly null.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | moderate |
| Risk of bias | some concerns |
| Precision of results | moderate |
| Directness | direct |
| Reproducibility | partly replicated |
| Consistency with the overall evidence | mixed |
| Regional transferability | high |
| Adequacy of the interpretation (by the archive) | essentially correct but incomplete |
14. Open questions
Causality; longer follow-up; which UPF components (additives, contaminants, nutrients) matter; replication in other cohorts.
15. Linked fact checks
Use by the Tayebat archive
Note: according to the archive, the studies were compiled by the archive operators. Whether Dr. Elawady himself cited this study is not established.
- What the archive says about the study
NutriNet-Santé cohort of 104,980 — each 10% rise raised overall cancer risk 12% and breast cancer 11%.
- A. What was actually studied?
UPF and cancer.
- B. What did the authors find?
HR 1.12 and 1.11.
- C. What conclusions do the authors themselves draw?
Further studies needed.
- D. What does the archive derive from it?
Causal 'raised'.
- E. Is this derivation scientifically justified?
Numbers correct; causal wording ('raised') goes beyond the authors, who call for confirmation; prostate and colorectal cancer were not associated.
- Adequacy of the interpretation (by the archive)
- essentially correct but incomplete