Meta-analysisindependent cross-checkfull text checked★ key study I-FAT-03 in-depth review
Butter at usual amounts: little link to heart disease
Is Butter Back? A Systematic Review and Meta-Analysis of Butter Consumption and Risk of Cardiovascular Disease, Diabetes, and Total Mortality.
Overview
Pimpin L et al.. PLoS One 2016 · PMID 27355649 · DOI 10.1371/journal.pone.0158118
Per tablespoon/day: CVD RR 1.00; mortality RR 1.01.
Does the study support the specific claim? Partly · Methodological quality: moderate
Plain-language summary
Study explained simply
- What did the researchers want to find out?
Butter and health.
- How was the study done?
Meta-analysis.
- How many people, animals or samples were studied?
636,151 people.
- What was actually found?
Small or neutral associations.
- How large was the observed effect?
Mortality RR 1.01 per 14 g/d.
- What do the results mean in everyday life?
Small amounts of butter are not a major risk.
- What are the study's limitations?
Observational; only 4 cohorts for CVD, some adjusted for blood cholesterol (which can hide an effect via LDL); mostly modest intakes; ghee not studied.
- What does it say about Elawady's specific claim?
Partly supports the claim.
- Do the results apply to Germany and Europe?
Moderate.
- What can explicitly not be concluded from this study?
That ghee or high butter intakes are safe for the arteries; authors rate unsaturated vegetable oils as healthier than butter.
Scientific in-depth review
1. Full reference
- Authors
- Pimpin L et al.
- Journal
- PLoS One
- Year
- 2016
- Identifier
- PMID 27355649 DOI 10.1371/journal.pone.0158118
- Verification
- verified – PubMed efetch 2026-10-25
- Retracted
- no
- Corrections
- none
- Extent checked
- full text
- Conflicts of interest and funding
Funding obtained by D. Mozaffarian (funding source not named in the available full text; PubMed record lists NHLBI grants). D. Mozaffarian reports honoraria/consulting (Boston Heart Diagnostics, Haas Avocado Board, AstraZeneca, GOED, Life Sciences Research Organization), UpToDate royalties, Elysium Health advisory board, and is co-inventor on a Harvard patent on trans-palmitoleic acid (a dairy-fat biomarker) for metabolic disease.
2. Research question
Butter consumption and mortality, CVD and diabetes.
3. Study design and population
- Study design
- Meta-analysis – Systematic review and dose-response meta-analysis of prospective cohort studies (no RCTs found); MOOSE/PRISMA reporting.
- Population
Adults (mean age 44-71 years) in general-population cohorts in the USA, Netherlands, Sweden, Finland (incl. one cohort of male smokers, ATBC) and 8 European EPIC countries.
- Country
Multinational
- Period
Searches from database inception through May 2015; included studies published 2005-2015; follow-up up to 23 years.
- Sample size
5,770 abstracts screened; 9 publications, 15 country-specific cohorts; 636,151 participants; 6,539,822 person-years; 28,271 deaths, 9,783 incident CVD, 23,954 incident diabetes cases. Mortality: 2 studies (9 cohorts), 379,763 participants; CVD: 4 cohorts, 175,612 participants; stroke: 3 cohorts (5,299 cases); CHD: 3 studies (4,484 cases); diabetes: 4 studies (11 cohorts), 201,628 participants.
- Inclusion and exclusion criteria
Prospective cohorts or RCTs in adults with multivariable-adjusted estimate of total or added butter vs all-cause mortality, CVD, CHD, stroke or type 2 diabetes. Excluded: children, populations with end-stage disease, follow-up <3 months, butter not separable from other fats/dairy, soft endpoints only, crude estimates only. Searched 9 databases incl. grey literature without language restriction.
4. Methods and endpoints
- Exposure / intervention
Butter intake (FFQ/food records/diet history), standardised per 14 g/day (1 tablespoon); median intakes ranged from 4.5 g/day (EPIC) to 46 g/day (Finland). One EPIC estimate combined butter and margarine (margarine reported low). No study of ghee was found.
- Comparator
Per dose
- Endpoints
Mortality, CVD, CHD, stroke, diabetes.
- Statistical methods
Greenland-Longnecker two-stage dose-response; DerSimonian-Laird random effects with fixed-effect comparison; I2; restricted cubic splines for non-linearity; Egger's test, funnel plots, trim-and-fill. Quality via modified Newcastle-Ottawa (0-5). Subgroup analyses planned but not performed (too few studies).
5. Actual results
- Results
Per 14 g/day butter: all-cause mortality RR 1.01 (95% CI 1.00-1.03; P=0.045; I2=0%); any CVD RR 1.00 (0.98-1.02; P=0.704; I2=0%); CHD RR 0.99 (0.96-1.03); stroke RR 1.01 (0.98-1.03); total CVD (2 cohorts) RR 0.99 (0.96-1.02); type 2 diabetes RR 0.96 (0.93-0.99; P=0.021; I2=42.1%). No evidence of non-linearity for CVD (P=0.364) or diabetes (P=0.160). Trim-and-fill for diabetes: RR 0.95 (0.93-0.98).
- Absolute risks
not reported in full text (only event counts per study in Table 1).
- Authors' conclusion
Relatively small or neutral overall associations of butter with mortality, CVD and diabetes; no need for major emphasis in dietary guidelines on increasing or decreasing butter; butter may be healthier than refined grains/potato it is spread on, but oils rich in unsaturated fat (soybean, canola, flaxseed, extra-virgin olive oil) appear healthier than butter.
6. Strengths
- Comprehensive search of 9 databases incl. grey literature without language restriction
- Duplicate full-text review, extraction and quality scoring; author contact for missing data
- Dose-response analysis using all categories
- Low heterogeneity for mortality and CVD
7. Weaknesses
- Observational only; no RCTs; residual confounding acknowledged (butter eaters have worse diets/lifestyles)
- Few independent cohorts per outcome (CVD 4, CHD 3, mortality 2 publications); subgroup analyses not possible; Egger's test underpowered
- Some included CVD estimates were adjusted for blood cholesterol (LDL-C in one CHD study, serum total/HDL cholesterol in the ATBC stroke study), i.e. for a mediator of butter's effect, which can bias towards null
- Self-reported intake, single baseline assessments and long lag to outcomes may attenuate associations
- Mostly low-to-moderate intakes (median 4.5-46 g/day); ghee not studied; comparison food (substitution) not specified in main analysis
- Title/abstract screening by one investigator; no registered protocol reported; funding sources of included studies not reported
8. Risk of bias
- Orienting tool
- AMSTAR 2
- 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
Single-reviewer AMSTAR 2-oriented appraisal from the full text (formal assessment needs two independent reviewers and the protocol). No critical flaw, several critical-item weaknesses (no registered protocol, crude RoB tool, limited publication-bias assessment); confidence in the review itself moderate, while certainty of the underlying observational evidence for CVD is low (few cohorts, mediator adjustment, residual confounding).
| Domain | Judgement | Rationale |
|---|---|---|
| Item 1 PICO components | low | Population, exposure, outcomes and designs clearly defined. |
| Item 2 (critical) a-priori protocol | some concerns | A study protocol is mentioned in acknowledgements but no registration or published protocol; deviations cannot be checked. |
| Item 3 choice of study designs explained | low | Prospective cohorts and RCTs; rationale given. |
| Item 4 (critical) comprehensive search | low | 9 databases incl. grey literature, no language restriction, hand-searching, author contact. |
| Items 5-6 duplicate selection and extraction | some concerns | Title/abstract screening by one investigator; full-text selection and extraction in duplicate. |
| Item 7 (critical) list of excluded studies with reasons | low | Reference list of excluded publications provided (S4 File). |
| Item 8 description of included studies | low | Table 1 details population, exposure, outcome ascertainment, adjustment. |
| Item 9 (critical) risk of bias of included studies | some concerns | Modified 5-item Newcastle-Ottawa score in duplicate; crude scale that does not capture overadjustment for mediators or residual confounding. |
| Item 10 funding of included studies | high | Not reported. |
| Item 11 (critical) appropriate meta-analytic methods | low | Appropriate dose-response pooling with random and fixed effects. |
| Item 12 impact of RoB on results | some concerns | No sensitivity analysis by quality score (all scored 4-5). |
| Item 13 (critical) RoB considered in interpretation | some concerns | Residual confounding and measurement error discussed; mediator adjustment in some CVD estimates not discussed. |
| Item 14 heterogeneity | low | I2 reported and discussed. |
| Item 15 (critical) publication bias | some concerns | Funnel plots, Egger, trim-and-fill done but underpowered (2-4 studies per outcome); authors note unpublished null cohorts likely. |
| Item 16 COI of review authors | low | COI declared. |
Full-text review (25 October 2026): n broken down per outcome; intake range (4.5-46 g/d) and absence of ghee studies added; results with P values, I2, total CVD and trim-and-fill; COI of senior author added (industry honoraria, dairy-fat biomarker patent); new weaknesses (mediator adjustment for cholesterol in some CVD estimates, few cohorts, single-reviewer screening, no registered protocol); full AMSTAR 2 domains; overall bias unchanged (some_concerns); directness direct -> indirect for the ghee claim.
9. Assessment of the statistical interpretation
Precise near-null relative associations per 14 g/day at the low-to-moderate intakes observed; a null association in cohorts partly adjusted for blood cholesterol and without specified substitution cannot exclude harm via LDL-C, nor does it address high intakes or ghee. The diabetes inverse association is modest and may reflect residual confounding.
10. Relevance to the specific claim
Partly supports C-FAT-01 (butter not strongly associated with CVD at usual intakes).
11. Independent contrary findings
12. Transferability to Germany and Europe
Moderate.
13. Overall methodological judgement
Well-conducted review finding near-null associations of butter at usual intakes with CVD in few observational cohorts; some estimates were adjusted for blood cholesterol, intakes were mostly moderate and no ghee studies exist. Not evidence that liberal ghee/butter intake is safe for arteries; the authors themselves rank unsaturated oils as healthier.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | moderate |
| Risk of bias | some concerns |
| Precision of results | high |
| Directness | indirect |
| Reproducibility | unknown |
| Consistency with the overall evidence | consistent |
| Regional transferability | moderate |
| Adequacy of the interpretation (by the archive) | not applicable |
14. Open questions
Effects of high intakes and of ghee; substitution-specific effects (butter vs unsaturated oils vs refined carbohydrates); unpublished cohort data.
15. Linked fact checks
- Natural fats such as baladi ghee and butter are not a cause of clogged arteries (contrary to the previously dominant view); 'good fats do not stick'. – Studies with mixed or unclear results