Ultra-processed food and heart disease in France

Overview

Srour B et al.. BMJ 2019 · PMID 31142457 · DOI 10.1136/bmj.l1451

Each 10% more ultra-processed food was linked with 12% more cardiovascular disease.

Does the study support the specific claim? Yes · Methodological quality: moderate

Original publication

Plain-language summary

Scientific in-depth review

1. Full reference

Authors

Srour B, Fezeu LK, Kesse-Guyot E, Allès B, Méjean C, Andrianasolo RM, Chazelas E, Deschasaux M, Hercberg S, Galan P, Monteiro CA, Julia C, Touvier M

Journal

BMJ

Year

2019

Identifier

PMID 31142457 DOI 10.1136/bmj.l1451 · PMC6538975

Verification

verified – PubMed E-utilities efetch, 2026-10-25

Retracted

no

Corrections

none

Extent checked

full text

Conflicts of interest and funding

Public funding (French Ministry of Health, Santé Publique France, INSERM, INRA, CNAM, Université Paris 13; first author funded by INCa and Fondation de France); funders had no role. ICMJE: no competing interests.

2. Research question

Is UPF intake associated with CVD?

3. Study design and population

Study design

Cohort study – Prospective web-based volunteer cohort (NutriNet-Santé; NCT03335644) with repeated 24-h dietary records (mean 5.7 per participant in the first 2 years) classified by NOVA at item level (3,300 items).

Population

French adults >=18 years without CVD at baseline; 79.2% women; mean age 42.7 (SD 14.5); healthier, more educated, less often overweight and diabetic than the French population

Country

France

Period

Recruitment from May 2009; follow-up to 11 January 2018; median 5.2 years (IQR 2.6-7.3)

Sample size

105,159; 1,409 first CVD events (106 myocardial infarctions, 485 angioplasties, 74 acute coronary syndromes, 155 strokes, 674 transient ischaemic attacks); 518,208 person-years

Inclusion and exclusion criteria

Adults with >=2 valid 24-h records in the first 2 years and no CVD at baseline; energy under-reporters (20.0% of the cohort) excluded.

4. Methods and endpoints

Exposure / intervention

Proportion (% by weight) of ultra-processed foods and drinks in the diet, per absolute 10-point increment; also quarters with sex-specific cut-offs

Comparator

Lower proportion of UPF (per 10-point increment; lowest quarter for rates)

Endpoints

CVD, CHD, cerebrovascular disease

Statistical methods

Cox models (age timescale) adjusted for sex, energy, number of records, smoking, education, physical activity, BMI, alcohol, family history of CVD (model 1); further models adding saturated fat/sodium/sugar, healthy pattern, food groups, metabolic diseases and treatments; spline linearity; stratified analyses; lag exclusions (2-5 years); multiple imputation; energy-weighted exposure; secondary analysis of minimally processed foods.

5. Actual results

Results

Per 10-point higher UPF share: overall CVD HR 1.12 (1.05-1.20); coronary heart disease 1.13 (1.02-1.24); cerebrovascular disease 1.11 (1.01-1.21). Angioplasty 1.16 (1.03-1.30); TIA 1.13 (1.03-1.24); CVD excluding TIA 1.12 (1.02-1.23). Linear association. Robust to adjustment for nutritional quality and metabolic disease; excluding first 2 years 1.14 (1.05-1.23); first 5 years 1.13 (1.00-1.28). Energy-weighted UPF share: HR 1.06 (1.01-1.12). Unprocessed/minimally processed foods: HR 0.91 (0.86-0.97) per 10 points. UPF beverages per 100 g/day CVD HR 1.06 (1.02-1.10).

Absolute risks

Age- and sex-corrected CVD incidence 242, 254, 252 and 277 per 100,000 person-years across quarters 1-4 (CHD 109 vs 124; cerebrovascular 144 vs 163 per 100,000 person-years, Q1 vs Q4); i.e. ~35 additional CVD events per 100,000 person-years in the highest vs lowest quarter.

Authors' conclusion

Higher UPF consumption was associated with higher risks of cardiovascular, coronary heart and cerebrovascular diseases; results need confirmation in other populations and causality remains to be established; nutritional composition, additives, contact materials and neoformed contaminants might play a role.

6. Strengths

• Large cohort with repeated, detailed 24-h records and item-level NOVA classification
• Events validated by medical records plus linkage to national health insurance and mortality registry
• Extensive sensitivity analyses (lag exclusions, multiple imputation, energy-weighted exposure)
• Public funding; no COI

7. Weaknesses

• Short median follow-up (5.2 years)
• Almost half of events are transient ischaemic attacks (674/1,409) and many angioplasties; few myocardial infarctions (106) and strokes (155)
• Volunteer cohort (79% women, health-conscious); limited generalisability (authors)
• High UPF consumers were ~11 years younger, smoked more and were less active: residual confounding (authors)
• NOVA misclassification possible (authors)
• Weight-based exposure metric; energy-based HR smaller (1.06)
• Energy under-reporters (20%) excluded, possible selection effects
• Self-reported diet

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.

Full-text review (25 October 2026): Numbers confirmed. Added event composition (674 TIA, 485 angioplasty, 106 MI, 155 stroke), registration, 20% under-reporters excluded, weight-based exposure and energy-weighted HR 1.06, lag analyses, protective association of minimally processed foods (HR 0.91), age/sex-corrected rates by quarter, funding. Domain-level ROBINS-E (overall some_concerns, unchanged).

9. Assessment of the statistical interpretation

Modest relative association (HR 1.12 per 10-point increment) with an absolute difference of about 35 CVD events per 100,000 person-years between extreme quarters. Results are dominated by softer endpoints (TIA, angioplasty). The energy-weighted estimate is smaller (1.06).

10. Relevance to the specific claim

Supports C-UPF-03.

11. Independent contrary findings

None.

12. Transferability to Germany and Europe

High.

13. Overall methodological judgement

Good-quality cohort showing a modest association of UPF share with incident CVD over 5 years; association, not proof of causation; outcome mix and volunteer population limit inference about hard cardiovascular events.

14. Open questions

Causality.

15. Linked fact checks

• Ultra-processed foods raise the risk of cardiovascular disease, type 2 diabetes, cancer and all-cause mortality; an umbrella review found 'convincing' evidence for cardiovascular mortality and type 2 diabetes. – Studies supporting the claim

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 105,159 — each 10% rise in UPF share raised CVD risk by 12%.

A. What was actually studied?

UPF and CVD.

B. What did the authors find?

HR 1.12.

C. What conclusions do the authors themselves draw?

Causality to be established.

D. What does the archive derive from it?

'Raised' CVD risk.

E. Is this derivation scientifically justified?

Number correct; the authors' explicit caveat that causality remains to be established is omitted, and the archive's causal 'raised' overstates the evidence. Nearly half of the events were transient ischaemic attacks.

Adequacy of the interpretation (by the archive)

essentially correct but incomplete