Cohort studycited by the Tayebat archivefull text checked★ key study A-DRY-01 in-depth review
Swedish study: very high milk intake linked to earlier death in women
Milk intake and risk of mortality and fractures in women and men: cohort studies.
Overview
Michaëlsson K et al.. BMJ 2014 · PMID 25352269 · DOI 10.1136/bmj.g6015
Women drinking 3 or more glasses a day had a 1.93-fold mortality hazard compared with less than 1 glass (17.7 vs 10.7 deaths per 1,000 person-years); no fewer hip fractures. Yogurt/fermented milk and cheese showed the opposite: lower mortality.
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?
Check whether high milk intake is linked with death and fractures.
- How was the study done?
Two large cohorts followed for 11-20 years using diet questionnaires.
- How many people, animals or samples were studied?
61,433 women and 45,339 men in Sweden.
- What was actually found?
More milk was linked with higher mortality, especially in women, and not with fewer fractures; yogurt/soured milk and cheese were linked with lower mortality and fewer hip fractures in women.
- How large was the observed effect?
Women HR 1.93 (1.80-2.06) for >=3 vs <1 glass/day (absolute 17.7 vs 10.7 per 1,000 person-years); per glass HR 1.15 women, 1.03 men.
- What do the results mean in everyday life?
A warning signal for very high milk intake, not proof of harm.
- What are the study's limitations?
Observational; confounding and reverse causation possible; female estimates stronger partly due to repeated measurement; biomarker data cross-sectional; authors ask for caution.
- What does it say about Elawady's specific claim?
Supports the archive figures but not a causal claim.
- Do the results apply to Germany and Europe?
European data; intakes higher than average.
- What can explicitly not be concluded from this study?
That milk causes death or fractures; that yogurt or other fermented dairy is harmful (the study suggests the opposite).
Scientific in-depth review
1. Full reference
- Authors
- Michaëlsson K, Wolk A, Langenskiöld S, Basu S, Warensjö Lemming E, Melhus H, Byberg L
- Journal
- BMJ
- Year
- 2014
- Identifier
- PMID 25352269 DOI 10.1136/bmj.g6015 · PMC4212225
- Verification
- verified – PubMed E-utilities efetch, 2026-10-25
- Retracted
- no
- Corrections
- none
- Extent checked
- full text
- Conflicts of interest and funding
Swedish Research Council grants; funders had no role. One author (EWL) employed by the Swedish National Food Agency. ICMJE disclosure: no support from any organisation for the submitted work, no relevant financial relationships.
2. Research question
Is high milk consumption associated with mortality and fractures?
3. Study design and population
- Study design
- Cohort study – Two prospective population-based cohorts (Swedish Mammography Cohort, Cohort of Swedish Men) with food-frequency questionnaires; biomarker subsamples from two further cohorts.
- Population
61,433 women aged 39-74 (baseline 1987-90) and 45,339 men aged 45-79 (baseline 1997) in three central Swedish counties.
- Country
Sweden
- Period
Women: baseline 1987-90 (second FFQ 1997), follow-up to 31 Dec 2010 (mean 20.1 y, median 22 y, 1,231,818 person-years). Men: baseline 1997, follow-up 1 Jan 1998 to 31 Dec 2010 (mean 11.2 y, median 13 y, 534,094 person-years).
- Sample size
106,772 (61,433 women, of whom 38,984 with 1997 update; 45,339 men); biomarker subsamples: urine 8-iso-PGF2alpha in 892 women (SMC Clinical) and men of ULSAM, serum IL-6 in men of ULSAM (text gives 633 and 700, figure legend reverses them)
- Inclusion and exclusion criteria
SMC: all 90,303 women aged 39-74 invited to mammography screening in Uppsala and Vastmanland counties (74% response; 70% to 1997 questionnaire). COSM: all 100,303 men aged 45-79 in Orebro and Vastmanland invited (49% response; 48,850 responders, 45,339 after exclusions shown in flow chart). Pathological (tumour) fractures excluded; high-impact fractures retained.
4. Methods and endpoints
- Exposure / intervention
Total milk intake (all fat grades; 1 glass = 200 g) from FFQ (96 items; validity r ~0.7 vs food records); categories <200, 200-399, 400-599, >=600 g/d and per 200 g/d; time-updated cumulative averages in women, single assessment in men. Fermented milk/yogurt and cheese analysed as secondary exposures.
- Comparator
<1 glass/day (<200 g/d; mean about 60 g/d women, 50 g/d men)
- Endpoints
All-cause, cardiovascular and cancer mortality (Swedish cause-of-death register); any fracture and hip fracture (national patient register); urine 8-iso-PGF2alpha and serum IL-6 (cross-sectional subsamples).
- Statistical methods
Cox proportional hazards with DAG-selected covariates (age, energy, BMI, height, education, living alone, calcium/vitamin D supplements, cortisone, healthy dietary pattern, physical activity, smoking, Charlson index; women also HRT, nulliparity); restricted cubic splines; MCMC multiple imputation of covariates absent at 1987-90 baseline; sensitivity analyses (baseline-only exposure, mutual adjustment for other dairy, additional nutrients, competing risks, negative-control outcome of vehicle-related non-fracture injury).
5. Actual results
- Results
Women per glass/day: all-cause mortality HR 1.15 (1.13-1.17), CVD mortality 1.15 (1.12-1.19), cancer mortality 1.07 (1.02-1.11), any fracture 1.02 (1.00-1.04), hip fracture 1.09 (1.05-1.13). Women >=3 vs <1 glass/day (mean 680 vs 60 g/d): mortality 1.93 (1.80-2.06), CVD mortality 1.90 (1.69-2.14), cancer mortality 1.44 (1.23-1.69), any fracture 1.16 (1.08-1.25), hip fracture 1.60 (1.39-1.84). Men per glass: mortality 1.03 (1.01-1.04), CVD mortality 1.05 (1.03-1.07), cancer mortality 0.99 (0.97-1.02), any fracture 1.01 (0.99-1.03), hip fracture 1.03 (0.99-1.07); >=3 vs <1 glass (mean 830 vs 50 g/d) mortality 1.10 (1.03-1.17). Single-exposure (non-time-updated) analyses in women gave attenuated estimates similar to men. Fermented milk/yogurt and cheese: in women each serving associated with 10-15% lower mortality and hip-fracture rates (P<0.001); weaker or none in men. Biomarkers: milk positively associated with 8-iso-PGF2alpha in both sexes and with IL-6 in men; fermented milk negatively associated with both markers; cheese not associated. Negative-control outcome null (HR 0.96, 0.90-1.03).
- Absolute risks
Age-standardised rates per 1,000 person-years, women <1 vs >=3 glasses/day: all-cause mortality 10.7 vs 17.7; CVD mortality 3.6 vs 6.2; cancer mortality 2.8 vs 2.8; any fracture 16.0 vs 15.1; hip fracture 3.1 vs 4.2. Men: mortality 18.2 vs 20.7; hip fracture 2.4 vs 2.3.
- Authors' conclusion
Higher milk consumption is not accompanied by lower fracture risk and may be associated with higher mortality; a link with lactose/galactose content is possible but causality needs experimental testing. Results should be interpreted cautiously given the observational design and merit independent replication before use for dietary recommendations.
6. Strengths
- Two large population-based cohorts with complete register-based outcome ascertainment
- Repeated dietary assessment and time-updated exposure in women
- DAG-based covariate selection; multiple sensitivity analyses including a negative-control outcome and checks for reverse causation (comorbidity, prior fracture, family history)
- Separate analysis of milk vs fermented milk/cheese
7. Weaknesses
- Observational; residual confounding acknowledged (socioeconomic nuances; in men high milk intake went with lower education and more smoking)
- Reverse causation not excluded by authors
- Covariates such as smoking and physical activity were not measured at the 1987-90 baseline in women and were imputed from 1997 data
- Low response in men (49%)
- Sex difference largely explained by measurement approach: single-exposure analyses in women attenuated estimates towards those in men
- Biomarker analyses cross-sectional and in small subsamples; IL-6 measured only in men; inconsistent subsample sizes between text and figure
- Galactose mechanism based on animal models; Swedish lactase-persistent population with high intakes
- Minor inconsistencies: men's fractures 5,066 (abstract) vs 5,379 (results); hip fractures 1,166 (text) vs 1,266 (table)
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
Simplified single-reviewer ROBINS-E from the full text; formal assessment needs two independent reviewers and a protocol.
| Domain | Judgement | Rationale |
|---|---|---|
| Confounding | some concerns | Broad DAG-based adjustment incl. comorbidity and dietary pattern; residual socioeconomic and lifestyle confounding plausible (authors acknowledge); covariates imputed for the women's baseline. |
| Exposure measurement | some concerns | Validated FFQ (r ~0.7), repeated in women; first women's FFQ used prespecified categories; self-report error. |
| Selection of participants | some concerns | Population-based invitations but response 74% (women) and 49% (men); exclusions only in flow chart. |
| Post-exposure interventions / changes in exposure | some concerns | Time-updating in women may capture diet change; authors checked that comorbidity or fracture did not change milk intake, but reverse causation cannot be ruled out. |
| Missing data | some concerns | Multiple imputation of covariates not assessed at baseline; extent of missing data not reported in main text. |
| Outcome measurement | low | National registers with high accuracy and complete linkage. |
| Selection of reported result | some concerns | Hypothesis-driven and analyses largely reported (many in supplement); fermented-milk/cheese results only in supplementary tables and not in abstract. |
Full-text review (25 October 2026): Added absolute rates (women mortality 10.7 vs 17.7 /1,000 PY; hip fracture 3.1 vs 4.2), cause-specific HRs, hip fracture HR 1.60 for >=3 glasses, response rates, covariate imputation, sensitivity and negative-control analyses, and the opposite (inverse) associations for fermented milk/yogurt and cheese (10-15% lower mortality and hip fracture per serving in women; inverse biomarker association). IL-6 only measured in men. Noted internal count inconsistencies. Bias overall unchanged (some_concerns).
9. Assessment of the statistical interpretation
Large, precise relative hazards in women driven partly by the time-updated exposure; absolute difference in all-cause mortality between extreme categories about 7 per 1,000 person-years (10.7 vs 17.7), hip fracture about 1.1 per 1,000 person-years (3.1 vs 4.2). In men effects are small (HR 1.03 per glass) and fractures null. The opposite associations for fermented milk/yogurt and cheese argue against simple 'dairy is harmful' interpretations; observational associations cannot establish causation.
10. Relevance to the specific claim
Main archive evidence for C-DRY-03 and partly C-DRY-02.
11. Independent contrary findings
I-DRY-02 (neutral), I-DRY-03 (milk not associated with hip fracture), I-DRY-01 (small excess only at highest intake).
12. Transferability to Germany and Europe
Directly European (Sweden) but intake levels high.
13. Overall methodological judgement
Well-conducted large cohort with a strong but isolated signal for high non-fermented milk intake in Swedish women; associations are observational, partly method-dependent, and are reversed for fermented milk/yogurt and cheese. It does not establish that milk causes death or fractures.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | moderate |
| Risk of bias | some concerns |
| Precision of results | high |
| Directness | direct |
| Reproducibility | not replicated |
| Consistency with the overall evidence | inconsistent |
| Regional transferability | moderate |
| Adequacy of the interpretation (by the archive) | partly correct but over-interpreted |
14. Open questions
Why Swedish results differ from pooled analyses; role of galactose in humans; replication with repeated measures in men and in lactase-non-persistent populations.
15. Linked fact checks
- Dairy products harm health and cause inflammation. – Studies with mixed or unclear results
- High milk consumption raises mortality (women drinking 3+ glasses/day had 93% higher mortality) and increases rather than prevents hip fractures. – Studies supporting the claim
- Fresh milk, yogurt (incl. laban rayeb, ayran, labneh) and white/cottage cheese are harmful and must be avoided, whereas aged cheeses (cheddar, mozzarella, romi), processed cheese triangles, cream (qeshta), butter and ghee are allowed. – 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
Women drinking 3+ glasses of milk daily had 93% higher mortality and more hip fractures, with oxidative stress and inflammation markers rising in parallel.
- A. What was actually studied?
Association of self-reported milk intake with mortality and fractures in Swedish adults.
- B. What did the authors find?
HR 1.93 for mortality (women, >=3 vs <1 glass); hip fracture HR 1.09 per glass and 1.60 for >=3 glasses in women; no fracture association in men; positive biomarker associations for milk; but fermented milk/yogurt and cheese associated with 10-15% lower mortality and hip-fracture rates per serving in women and inversely with oxidative-stress/inflammation markers.
- C. What conclusions do the authors themselves draw?
High milk intake associated with higher mortality and fractures; cautious interpretation recommended.
- D. What does the archive derive from it?
Presents milk as causing death and fractures and as part of an inflammatory mechanism ('a fluid loaded with growth and inflammatory signals').
- E. Is this derivation scientifically justified?
Numbers are correct; the causal reading, omission of the authors' explicit caution, of the small absolute differences and of the opposite (favourable) associations for fermented milk/yogurt and cheese make it partly overinterpreted. The study gives no support for banning yogurt or labneh.
- Adequacy of the interpretation (by the archive)
- partly correct but over-interpreted