Systematic reviewcited by the Tayebat archivefull text checked★ key study A-UPF-13 in-depth review

Umbrella review: ultra-processed food and health

Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses.

Overview

Lane MM et al.. BMJ 2024 · PMID 38418082 · DOI 10.1136/bmj-2023-077310

UPF was linked to 32 of 45 health outcomes; certainty was mostly low.

Does the study support the specific claim? Partly · Methodological quality: high

Original publication

Plain-language summary

Study explained simply

What did the researchers want to find out?: What is the credibility of associations between UPF and health outcomes?
How was the study done?: Umbrella review of 45 pooled analyses with credibility classes and GRADE.
How many people, animals or samples were studied?: 45 pooled analyses; n=9,888,373; Pooled populations
What was actually found?: Class I: CVD mortality RR 1.50 (1.37-1.63; GRADE very low); T2D dose-response RR 1.12 (1.11-1.13; moderate); prevalent anxiety OR 1.48 (low); common mental disorders OR 1.53 (low). Class II: all-cause mortality RR 1.21 (low), heart-disease mortality HR 1.66 (low), T2D OR 1.40 (very low), depression HR 1.22 (low), sleep OR 1.41, wheezing RR 1.40, obesity OR 1.55 (low). GRADE overall: 22 low, 19 very low, 4 moderate.
How large was the observed effect?: CVD mortality RR 1.50; T2D 1.12 per dose
What do the results mean in everyday life?: Consistent links, low certainty.
What are the study's limitations?: All the studies behind it are observational; many results rest on only 2-4 studies; for most outcomes the certainty is low or very low; the mental-health results come mainly from one-time surveys that cannot show what came first.
What does it say about Elawady's specific claim?: Core summary for C-UPF-03 and C-UPF-04.
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: Lane MM, Gamage E, Du S, Ashtree DN, McGuinness AJ, Gauci S, Baker P, Lawrence M, Rebholz CM, Srour B, Touvier M, Jacka FN, O'Neil A, Segasby T, Marx W
Journal: BMJ
Year: 2024
Identifier: PMID 38418082 DOI 10.1136/bmj-2023-077310 · PMC10899807
Verification: verified – PubMed E-utilities efetch, 2026-10-25
Retracted: no
Corrections: none
Extent checked: full text
Conflicts of interest and funding: No funding for this work. Several authors affiliated with the Food & Mood Centre (Deakin University), which received research funding from Be Fit Food, Bega Dairy and Drinks and the a2 Milk Company and philanthropic foundations; other individual disclosures (WHO, FSANZ, Bloomberg Philanthropies). None of the included meta-analyses was funded by UPF-producing companies.

2. Research question

What is the credibility of associations between UPF and health outcomes?

3. Study design and population

Study design: Systematic review – Umbrella review (PROSPERO CRD42023412732, PRISMA) of 14 meta-analyses providing 45 unique pooled analyses of observational studies; re-meta-analysis with random effects, prediction intervals, excess-significance and small-study tests; credibility classes I-V; GRADE; AMSTAR 2 for included meta-analyses.
Population: Pooled populations across the life course (mostly adults; 5 pooled analyses in children/adolescents)
Country: Multinational
Period: 2009-June 2023
Sample size: 14 meta-analyses; 45 pooled analyses (13 dose-response, 32 non-dose-response); 2-9 primary studies per pooled analysis (mean 4); total 9,888,373 participants (overlapping)
Inclusion and exclusion criteria: Systematic reviews/meta-analyses of cohort, case-control and/or cross-sectional studies on NOVA-defined UPF and any adverse health outcome; no pooled RCTs found; when several meta-analyses covered the same outcome, the most recent and/or largest was retained; no language limits.

4. Methods and endpoints

Exposure / intervention: Greater UPF exposure
Comparator: Lower exposure
Endpoints: 32 outcomes incl. mortality, cancer, mental, cardiometabolic
Statistical methods: Random-effects re-analysis of primary estimates; I2; 95% prediction intervals; Egger test; excess significance; largest-study significance; credibility classes I-V; GRADE (observational evidence starts at low); AMSTAR 2 (qualitative).

5. Actual results

Results: 71% (32/45) of pooled analyses showed direct associations. Class I: CVD mortality RR 1.50 (1.37-1.63; GRADE very low); type 2 diabetes dose-response RR 1.12 (1.11-1.13; moderate); prevalent anxiety OR 1.48 (1.37-1.59; low); prevalent common mental disorders OR 1.53 (1.43-1.63; low). Class II: all-cause mortality RR 1.21 (1.15-1.27; low); heart-disease mortality HR 1.66 (1.51-1.84; low); T2D OR 1.40 (1.23-1.59; very low); incident depressive outcomes 1.22 (1.16-1.28; low; 2 cohorts); sleep outcomes OR 1.41; wheezing RR 1.40; obesity OR 1.55 (1.36-1.77; low; cross-sectional). Null or weak: cancer mortality HR 1.00 (0.81-1.24; class V); breast cancer OR 1.15 (0.99-1.34; class V); prostate cancer; ulcerative colitis HR 1.17 (0.86-1.61; class V); Crohn's disease HR 1.71 (1.37-2.14; class IV); overall cancer HR 1.12 (1.06-1.19; class III; very low). Dose-response all-cause mortality RR 1.02 (1.01-1.03) per serving. GRADE overall: 4 moderate, 22 low, 19 very low. Moderate-to-high heterogeneity in 47% of analyses; excess significance in 9/28; small-study effects in 5/28.
Absolute risks: not reported in full text
Authors' conclusion: Greater UPF exposure was associated with higher risk of adverse health outcomes, especially cardiometabolic, common mental disorder and mortality outcomes; evidence for asthma, gastrointestinal health, some cancers and intermediate risk factors remains limited; supports mechanistic research and public-health measures.

6. Strengths

Pre-registered (PROSPERO); comprehensive search without language limits
Duplicate screening and extraction; full re-analysis from primary estimates with open data and code
Two complementary certainty frameworks (credibility classes and GRADE) plus AMSTAR 2
Reports null and weak outcomes transparently

7. Weaknesses

All underlying evidence observational; confounder adjustment and sensitivity analyses of primary studies not assessed (authors)
Many pooled analyses based on 2-4 studies, often cross-sectional (e.g. anxiety, common mental disorders, obesity)
Only one meta-analysis per outcome retained; primary-study overlap across outcomes
Different effect metrics (HR, OR, RR) treated as equivalent in some included meta-analyses
Credibility class and GRADE diverge strongly (CVD mortality class I but GRADE very low)
Seven of 14 included meta-analyses rated low confidence by AMSTAR 2 (unclear protocol)
Excess-significance bias in 9 of 28 testable analyses

8. Risk of bias

Orienting tool: AMSTAR 2 (umbrella review), ROBIS-oriented
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: Changed from low: the review is well conducted, but its conclusions rest on observational primary studies whose risk of bias was not assessed, and the credibility classification overstates certainty relative to GRADE. Structured single-reviewer judgement; a complete formal AMSTAR 2/ROBIS assessment requires two independent reviewers.

8. Risk of bias
Domain	Judgement	Rationale
Item 2 (critical) protocol registered	low	PROSPERO CRD42023412732; scope refined after pilot (no RCT pooled analyses found).
Item 4 (critical) comprehensive search	low	MEDLINE, PsycINFO, Embase, Cochrane, reference lists; 2009-June 2023; no language limits.
Items 5-6 duplicate selection and extraction	low	Duplicate screening (Covidence) and duplicate extraction.
Item 7 (critical) excluded studies justified	some concerns	PRISMA flow given; study-level exclusion list not visible in the main text.
Item 9 (critical) risk of bias of included reviews/primary studies	some concerns	AMSTAR 2 applied to included meta-analyses, but risk of bias/confounder adjustment of primary studies not assessed (authors' limitation).
Item 11 (critical) appropriate synthesis	some concerns	Re-analysis from primary data with random effects; but pooling of heterogeneous metrics and very few studies per outcome.
Item 13 (critical) risk of bias considered in interpretation	some concerns	GRADE reported, but headline emphasises 'convincing' class I even where GRADE is very low.
Item 15 (critical) publication bias	low	Egger, excess significance and largest-study tests applied where >=3 studies.
Item 16 conflicts of interest	low	Detailed disclosures; work unfunded; some centre funding from food companies (not UPF-related claims).

Full-text review (25 October 2026): Headline estimates confirmed. Added: 14 meta-analyses, 2-9 studies per pooled analysis, rule of retaining one meta-analysis per outcome, null/weak outcomes (cancer mortality HR 1.00, breast cancer, ulcerative colitis class V, Crohn's class IV), cross-sectional basis of mental-health class I findings, heterogeneity and excess-significance results, AMSTAR 2 of included reviews, funding/COI details. AMSTAR 2 overall changed from low to some_concerns. Interpretation adequacy set to partly_overinterpreted (consistent with the archive-use assessment).

9. Assessment of the statistical interpretation

Credibility classes reflect statistical robustness (significance, heterogeneity, small-study effects), not freedom from confounding; GRADE (mostly low/very low) is the better guide to certainty. The 'class I' anxiety and common-mental-disorder findings are from cross-sectional studies (prevalent outcomes), so direction of effect cannot be established. Cancer mortality, breast and prostate cancer and ulcerative colitis showed no credible association.

10. Relevance to the specific claim

Core summary for C-UPF-03 and C-UPF-04.

11. Independent contrary findings

Low GRADE certainty.

12. Transferability to Germany and Europe

High.

13. Overall methodological judgement

Best available overview of observational UPF evidence: consistent associations with mortality, cardiometabolic and mental-health outcomes, but certainty mostly low or very low; some outcomes (cancer mortality, several cancers, IBD) show weak or no evidence.

Multidimensional assessment
Multidimensional assessment
Authenticity of the source	verified
Methodological quality	high
Risk of bias	some concerns
Precision of results	high
Directness	direct
Reproducibility	replicated
Consistency with the overall evidence	consistent
Regional transferability	high
Adequacy of the interpretation (by the archive)	partly correct but over-interpreted

14. Open questions

Causality; dose thresholds.

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
Ultra-processed foods increase the risk of depression and common mental disorders. – Studies supporting the claim
Higher ultra-processed food intake raises the risk of inflammatory bowel disease (PURE study: HR 1.82 for >=5 servings/day). – Studies contradicting or limiting 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: Umbrella review of 45 meta-analyses (~9.8 million) — convincing evidence linking UPF to CVD mortality, T2D, mental disorders and obesity.
A. What was actually studied?: Credibility of UPF-outcome associations.
B. What did the authors find?: Class I for CVD mortality (GRADE very low), T2D dose-response (moderate), prevalent anxiety and common mental disorders (low; cross-sectional); obesity class II (low); cancer mortality, breast cancer, ulcerative colitis class V.
C. What conclusions do the authors themselves draw?: As above.
D. What does the archive derive from it?: Lists obesity among 'convincing' outcomes (headline) and omits GRADE.
E. Is this derivation scientifically justified?: Partly overinterpreted: obesity was class II, GRADE ratings (very low for CVD mortality) are omitted, mental-health 'convincing' evidence is cross-sectional, and the null/weak findings for cancer mortality and IBD are not mentioned.
Adequacy of the interpretation (by the archive): partly correct but over-interpreted