Cohort studyindependent cross-checkfull text checked★ key study I-DRY-01 in-depth review
US cohorts: dairy neither lowers nor clearly raises mortality
Associations of dairy intake with risk of mortality in women and men: three prospective cohort studies.
Overview
Ding M et al.. BMJ 2019 · PMID 31776125 · DOI 10.1136/bmj.l6204
Only the highest intake (about 4.2 servings/day) showed 7% higher mortality.
Does the study support the specific claim? Partly · Methodological quality: high
Plain-language summary
Study explained simply
- What did the researchers want to find out?
Dairy and mortality.
- How was the study done?
Three cohorts, up to 32 years.
- How many people, animals or samples were studied?
217,755 adults.
- What was actually found?
Mostly neutral; slight excess mortality at very high intake (about 4 servings/day) and with whole milk; cheese and yogurt not associated.
- How large was the observed effect?
HR 1.07 (1.04-1.10) highest vs lowest.
- What do the results mean in everyday life?
Moderate dairy intake appears neutral.
- What are the study's limitations?
Observational; US health professionals; association disappeared when using most recent intake.
- What does it say about Elawady's specific claim?
Does not support strong harm claims.
- Do the results apply to Germany and Europe?
Moderate.
- What can explicitly not be concluded from this study?
Causal effects.
Scientific in-depth review
1. Full reference
- Authors
- Ding M et al.
- Journal
- BMJ
- Year
- 2019
- Identifier
- PMID 31776125 DOI 10.1136/bmj.l6204
- Verification
- verified – PubMed efetch 2026-10-25
- Retracted
- no
- Corrections
- none
- Extent checked
- full text
- Conflicts of interest and funding
Funded by NIH grants (UM1 CA186107, P01 CA87969, R01 CA49449, R01 HL034594, R01 HL088521, UM1 CA176726, R01 CA67262, UM1 CA167552, R01 HL35464, R01 HL60712); funders had no role. FBH: research support from California Walnut Commission and honoraria from Metagenics, Standard Process and Diet Quality Photo Navigation, outside the submitted work; others none.
2. Research question
Dairy intake and total and cause-specific mortality.
3. Study design and population
- Study design
- Cohort study – Three prospective cohorts with repeated diet measures (NHS, NHS II, HPFS).
- Population
168,153 women (NHS 74,805; NHS II 93,348) and 49,602 men (HPFS), US health professionals (~95% European ancestry), free of CVD and cancer at baseline (1984 NHS, 1989 NHS II, 1986 HPFS)
- Country
USA
- Period
Baseline 1984-1989; follow-up to 31 December 2016 (29-32 years)
- Sample size
217,755 participants; 51,438 deaths (12,143 cardiovascular, 15,120 cancer)
- Inclusion and exclusion criteria
Excluded: CVD or cancer at baseline; implausible energy intake (<500 or >3,500 kcal women; <800 or >4,200 kcal men).
4. Methods and endpoints
- Exposure / intervention
Cumulative-average total dairy and subtypes (skim/low-fat milk, whole milk, cheese, yogurt, other) from validated FFQs every 4 years; butter not included; diet updating stopped at diagnosis of T2D, CVD or cancer.
- Comparator
Lowest fifth of total dairy (mean 0.8 servings/day); lowest serving category for subtypes; substitution with other foods.
- Endpoints
Total, cardiovascular and cancer mortality (death certificates/medical records reviewed by blinded physicians; >98% ascertainment).
- Statistical methods
Age-stratified Cox models with multivariable adjustment (family history, BMI, hypertension, hypercholesterolaemia, physical activity, AHEI-2010, energy, smoking, alcohol, menopausal/hormone status); fixed-effect pooling across cohorts; restricted cubic splines; substitution analyses; Bonferroni correction (alpha 0.001 for 60 tests); sensitivity analyses (time-varying intake, excluding first 4 years).
5. Actual results
- Results
Total mortality, highest (4.2 servings/day) vs lowest fifth: HR 1.07 (1.04-1.10), per serving/day 1.03 (1.02-1.04), P trend <0.001 (significant after Bonferroni). CVD mortality highest vs lowest 1.02 (0.95-1.08); cancer mortality 1.05 (0.99-1.11), but per serving/day 1.03 (1.01-1.05). Nonlinear curve: about 2 servings/day associated with lowest total and CVD mortality. Whole milk per 0.5 serving/day: total mortality 1.11 (1.09-1.14), CVD 1.09 (1.03-1.15), cancer 1.11 (1.06-1.17). Skim/low-fat milk per 0.5 serving/day: total mortality 1.01 (1.00-1.02). Cheese per 0.5 serving/day: total 1.00 (0.99-1.02), CVD 0.99 (0.96-1.03), cancer 1.02 (0.99-1.05); yogurt reported as not associated (estimates not shown). Substituting 1 serving/day dairy with nuts/legumes HR 0.86 (0.78-0.95), whole grains 0.89 (0.84-0.93), red/processed meat 1.05 (1.01-1.09). Time-varying (most recent) intake: no significant associations; excluding first 4 years: unchanged. Association present in non-obese, absent in obese (P interaction 0.03).
- Absolute risks
Death counts by fifth per cohort are tabulated (e.g. NHS 4,755 deaths in lowest vs 5,035 in highest fifth), but no absolute rates or risk differences are reported.
- Authors' conclusion
Data do not support an inverse association between high total dairy intake and mortality; about 2 servings/day was associated with the lowest cardiovascular mortality, higher intake with slightly higher mortality, especially cancer mortality; health effects depend on the foods used to replace dairy.
6. Strengths
- Very large cohorts, 51,438 deaths, 29-32 years follow-up
- Repeated validated FFQs with cumulative averaging; diet updating stopped at intermediate diagnoses
- Blinded physician cause-of-death classification; >98% death ascertainment
- Bonferroni correction for multiple testing; several sensitivity analyses; substitution analyses
7. Weaknesses
- Observational; residual confounding acknowledged (whole milk may mark unhealthy lifestyle)
- Mostly white US health professionals; limited generalisability
- Positive total-dairy association disappeared with time-varying intake (authors suggest reverse causation)
- Yogurt results not shown; cancer-site-specific results (e.g. prostate) only in supplementary table
- Self-reported FFQ intake; butter excluded
- No absolute risks
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 judgement from full text; a formal ROBINS-E needs two independent reviewers and a protocol.
| Domain | Judgement | Rationale |
|---|---|---|
| Confounding | some concerns | Extensive adjustment including diet quality; higher-dairy groups smoked less and exercised more; residual confounding possible, especially for whole milk. |
| Exposure measurement | some concerns | Repeated validated FFQs; self-report error; result sensitive to exposure definition (cumulative vs time-varying). |
| Selection of participants | low | Baseline CVD/cancer excluded; implausible energy excluded. |
| Post-exposure interventions / changes in exposure | some concerns | Dietary change after diagnoses handled by stopping updates; authors note reverse causation may persist (e.g. skim milk). |
| Missing data | low | >98% of deaths ascertained; high follow-up. |
| Outcome measurement | low | Mortality from vital records/NDI; cause classified by blinded physicians. |
| Selection of reported result | some concerns | Bonferroni applied and many results reported, but yogurt estimates withheld and site-specific cancer results only in supplement. |
Full-text review (25 October 2026): Added per-serving estimates (total mortality 1.03, cancer mortality 1.03 per serving/day, significant despite NS extreme-fifth comparison), subtype results (whole milk 1.11/1.09/1.11 per 0.5 serving; cheese null; yogurt not shown), substitution HRs, Bonferroni correction, loss of association with time-varying intake, BMI interaction, cohort sizes and full NIH grant list. Bias overall unchanged (some_concerns).
9. Assessment of the statistical interpretation
Small relative associations (3% per serving/day for total mortality; 7% at ~4.2 servings/day), precise but not robust to an alternative exposure definition; whole milk shows the most consistent positive association (11% per half serving/day); cheese and yogurt neutral. Absolute effects not reported.
10. Relevance to the specific claim
Mixed for C-DRY-03 (small excess at very high intake); limiting for C-DRY-04 (cancer mortality NS).
11. Independent contrary findings
I-DRY-02 neutral.
12. Transferability to Germany and Europe
Moderate.
13. Overall methodological judgement
Dairy is neither protective nor clearly harmful for mortality at usual intakes; a small excess appears at very high intake and for whole milk, with residual confounding and exposure-definition sensitivity limiting causal interpretation; replacement foods matter.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | high |
| Risk of bias | some concerns |
| Precision of results | high |
| Directness | direct |
| Reproducibility | partly replicated |
| Consistency with the overall evidence | mixed |
| Regional transferability | moderate |
| Adequacy of the interpretation (by the archive) | not applicable |
14. Open questions
Yogurt estimates; prostate-cancer mortality details (supplement); generalisability to non-US populations; causal inference (e.g. Mendelian randomisation).
15. Linked fact checks
- High milk consumption raises mortality (women drinking 3+ glasses/day had 93% higher mortality) and increases rather than prevents hip fractures. – Studies with mixed or unclear results
- Dairy consumption increases prostate cancer risk (7% per 400 g/day; 34% for more than 2.5 servings/day). – Studies with mixed or unclear results