Public evidence · free forever

Follow the evidence.

Study findings, limitations and source records are shown separately from ATLAS policy. The same versioned food facts underlie both views.

Public evidence stays free forever.

Different diets. Different questions.

Three randomized trials, read in context. Liver fat and visceral fat are different outcomes; these studies are not a league table.

Petersen et al. · 2026

Who was studied: Obesity, prediabetes, fatty liver.
Follow-up: Until approximately 10% weight loss.
What differed: Keto / Mediterranean / low-fat; protein 23% / 15% / 15% of energy.
Observed finding: Relative liver-fat change: −67% / −45% / −45%.
Interpretation limit: Small sample; protein unmatched.

Read the paper ↗

Study funding reported by article

National Institutes of Health · Foundation for Barnes-Jewish Hospital

Article disclosure (English):NIH grants P30 DK56341, P30 DK20579, UL1 TR002345 (KL2 TR002346), and S10 OD027006.

Author interests

Article disclosure (English):S.K. reports advisory-board fees from Abbvie, 89Bio, and Boehringer Ingelheim; an investigator-initiated Merck grant; and support for an industry-initiated Viking Therapeutics trial.

Disclosure source: ACKNOWLEDGMENTS / DECLARATION OF INTERESTS ↗

Source record

DOI: 10.1016/j.cmet.2026.07.020

Randomized: 55
Completed follow-up: 42

Sommersten et al. · CARBFUNC · 2022

Who was studied: Obesity / central adiposity.
Follow-up: 6 and 12 months.
What differed: Carbohydrate patterns differed; protein similar (16–17% of energy).
Observed finding: No significant between-group visceral-fat difference.
Interpretation limit: High dropout; not a ketogenic liver-fat replication.

Read the paper ↗

Study funding reported by article

Trond Mohn Foundation · University of Bergen · Kostfonden

Author interests

Authors report no conflicts

Disclosure source: Funding statement / Conflict of Interest ↗

Source record

DOI: 10.1016/j.clnu.2022.08.028

Randomized: 193
Completed follow-up: 78 / 57

Xu et al. · 2020

Who was studied: Severe obesity; before bariatric surgery.
Follow-up: 3 weeks.
What differed: Hypocaloric diets; protein 30% versus 10% of energy.
Observed finding: Higher protein reduced liver fat more.
Interpretation limit: Small, short study; not animal versus plant protein.

Read the paper ↗

Study funding reported by article

European Association for the Study of Diabetes · German Center for Diabetes Research · German Diabetes Association

Article disclosure (English):The article separately reports BMBF grant 01GI0925 as support for J. Machann.

Author interests

Not verified

Disclosure source: Funding information ↗

Source record

DOI: 10.1111/liv.14596

Randomized: 19
Completed follow-up: Not extracted

ATLAS interpretation · separate from findings

Carbohydrate restriction, higher protein and their combination remain competing explanations. A diet-label comparison cannot isolate these mechanisms. No significant difference does not establish equivalence.

Our starting point is animal-first. We make protein and nutrient coverage explicit, consider carbohydrate strategies in context, and revise assumptions when evidence changes.

Food preference: animal-first is an explicit editorial default, not a result established by these trials.
Protein: inspect quantity and amino-acid coverage separately. Composition does not establish digestibility or clinical outcomes.
Carbohydrate strategy: compare reduced-carbohydrate and ketogenic constraints explicitly. A preference does not select a treatment.

Explore policy assumptions →Inspect protein composition →Compare explicit quantities →

Educational evidence remains free. These summaries do not predict your response or change optimizer settings. Long-term cardiovascular outcomes and personal treatment decisions are outside this comparison.

Editorial review: 2026-09-21 · atlas-study-comparison/1; Declarations are not independent audits. 2026-09-24 · atlas-study-disclosures/1

How strong is this comparison?

One liver-fat question, three randomized trials with different interventions. Compare methods and missing data before interpreting direction.

Provisional method screen · not a formal RoB 2 or GRADE assessment

RCT

Petersen et al. · 2026

Comparison: Ketogenic vs Mediterranean vs very-low-fat; weight loss matched.
Design: RCT
Participant blinding: Whole-diet intervention was open-label
Randomized: 55
Liver-fat follow-up: 42
Liver-fat outcome: Secondary / exploratory
Outcome-assessor blinding: Not verified
Reported liver-fat result: Relative change −67% / −45% / −45%; between-arm P<0.05.
Why it is not the same test: Protein differed (23% vs 15% of energy). Only 42 of 55 completed; ketosis cannot be isolated from other diet differences.
Article-reported support: National Institutes of Health · Foundation for Barnes-Jewish Hospital
Author interests: Outside interests reported in article

Original article ↗Trial registry ↗Article-reported support ↗

Include in evidence export

RCT

Haufe et al. · 2011

Comparison: Moderately reduced carbohydrate vs reduced fat; hypocaloric.
Design: RCT
Participant blinding: Not verified
Randomized: 170
Liver-fat follow-up: 102
Liver-fat outcome: Secondary / exploratory
Outcome-assessor blinding: Not verified; Liver fat is absent from the current registered outcome list; historical versions still need review.
Reported liver-fat result: Relative change −47% vs −42%; no significant between-arm difference.
Why it is not the same test: Not ketogenic. Liver fat measured in 102 of 170 randomized participants; complete-case analysis.
Article-reported support: BMBF · European Commission · German Obesity Network; Joint project with metanomics GmbH and Charité; company role not independently verified.
Author interests: Authors declare no pertinent conflicts; not independently audited

Original article ↗Trial registry ↗Article-reported support ↗

Include in evidence export

RCT

Holmer et al. · 2021

Comparison: Low-carb high-fat vs 5:2 fasting vs standard advice.
Design: RCT
Participant blinding: Whole-diet intervention was open-label
Randomized: 74
Liver-fat follow-up: Not verified
Liver-fat outcome: Primary
Outcome-assessor blinding: Outcome-assessor masking is listed in the trial registry; separate confirmation in the article has not been verified.
Reported liver-fat result: LCHF −7.2 vs 5:2 −6.1 percentage points; between-arm difference −1.3 (95% CI −3.6 to 1.1); P=0.41.
Why it is not the same test: Comparator is fasting, not Mediterranean. Ketosis was not verified; no difference is not equivalence.
Article-reported support: Stockholm County Council · Kostfonden · Skandia Research Foundation · Åke Wiberg Foundation
Author interests: Authors declare no pertinent conflicts; not independently audited

Original article ↗Trial registry ↗Article-reported support ↗

Include in evidence export

Professional evidence export · preview

Includes methods, reported results, limits and source links. Provisional editorial screen; no clinical grade or winner.

No winner score. A non-significant difference does not prove equivalence.

Other evidence needs its own question: observational studies need confounding checks; meta-analyses need review and heterogeneity checks; animal studies can test mechanisms but need a bridge to human outcomes.

2026-09-24 · atlas-liver-fat-methods/3

Knowledge-pack provenance

Signed food-data pack: atlas-usda-composition-5e9c608795c66aba

usda-foundation-2026-04-30

SHA-256: 186e988ec542e913f51ef62b86a47758e8cdd0d1dc3889e7b055581f3c09c77a

Original source ↗

usda-sr-legacy-2018-04

SHA-256: b80817294b8850530aaedf2e515c02593b1824f763a0ff356e5c2081643e6fd0

Original source ↗

Back to everyday ATLAS →