Dairy, the heart and the whole-fat question: what the evidence shows

For heart disease and early death, dairy comes out roughly neutral to modestly favourable in the largest cohorts, and the decades-old advice to avoid whole-fat dairy specifically is not well supported by the outcome data [s1][s2]. That is a narrower claim than either side of the dairy argument usually makes, and it is the one the evidence actually supports. This piece concerns cardiovascular and metabolic outcomes; dairy's better-established role in bone and fracture risk is covered in dairy, calcium and bones.

What the biggest cohort found

The Prospective Urban Rural Epidemiology (PURE) study recorded diet for 136,384 people aged 35 to 70 across 21 countries on five continents, and followed them a median 9.1 years, during which 10,567 died or had a major cardiovascular event [s1]. People eating more than two servings of dairy a day, compared with none, had a hazard ratio of 0.84 for that composite outcome, 0.83 for total mortality, 0.78 for major cardiovascular disease and 0.66 for stroke [s1]. There was no significant association with heart attack, at 0.89 [s1].

The design matters because PURE deliberately set out to test dairy in low- and middle-income countries, where most prior data did not reach, and because it separated the products [s1]. Higher milk intake carried a hazard ratio of 0.90 and yoghurt 0.86 for the composite outcome, while cheese (0.88) and butter (1.09) showed no significant association [s1]. Notably, whole-fat and low-fat dairy were examined separately, and the results did not single out whole-fat dairy as harmful — the finding that most directly challenges long-standing guidance [s1].

Breaking the deaths down, the higher-dairy group had a hazard ratio of 0.77 for cardiovascular mortality and 0.86 for non-cardiovascular mortality, though the confidence interval for cardiovascular mortality (0.58 to 1.01) crossed the no-effect line [s1]. That pattern — a benefit that is consistent in direction but not always statistically firm on the narrower outcomes — is worth noting rather than rounding up, because it is exactly where a single cohort can mislead. Butter, the most saturated-fat-dense product, was eaten in small amounts across the cohort, which limits what the trial can say about it specifically [s1].

The more sober meta-analysis

PURE is one cohort, and its authors' framing drew criticism for leaning toward a favourable reading. A larger pooled analysis lands more cautiously. A 2017 dose-response meta-analysis of 29 cohort studies — 938,465 participants, with 93,158 deaths, 28,419 coronary heart disease cases and 25,416 cardiovascular disease cases — found no association between total dairy or milk and mortality, coronary heart disease or cardiovascular disease [s2]. For fermented dairy it reported a very small inverse association, a relative risk of 0.98 per 20 grams a day for both mortality and cardiovascular disease, and 0.98 per 10 grams a day of cheese for cardiovascular disease [s2]. Those marginal signals weakened further once a single large Swedish study was removed, and the authors' overall conclusion was that dairy's associations with cardiovascular and all-cause mortality are essentially neutral [s2].

Reconciling them

The two results are less contradictory than they look. PURE found a modest benefit; the meta-analysis found broad neutrality with a faint edge for fermented products. Neither found the harm that older saturated-fat logic predicted, and that is the through-line [s1][s2]. The mechanistic assumption — that because whole-fat dairy contains saturated fat it must raise cardiovascular risk — is not borne out when researchers measure disease and death rather than blood lipids. This overlaps with the broader re-examination of saturated fat and of specific foods rather than nutrients, a theme in our coverage of red and processed meat and plant-based diets.

Where the two studies part company is instructive in itself. PURE weighed total dairy at high intakes against none at all, in populations that often ate little dairy to begin with; the pooled analysis stacked dose-response curves from 29 cohorts, most in wealthier countries, and leaned harder on removing outliers [s1][s2]. The faint fermented-dairy signal that survived in the meta-analysis — cheese and yoghurt rather than milk or butter — hints that any real effect may track the product rather than the fat content, but the authors were explicit that even that edge nearly vanished when one Swedish cohort was dropped [s2]. Two large studies both failing to find harm is a firmer conclusion than either finding a benefit.

Two limits keep this honest. Both sources are observational, so residual confounding cannot be excluded, and dairy's effect depends heavily on what it replaces in a diet — swapping cheese for vegetables is not the same trade as swapping it for processed meat. What the evidence does support is modest: dairy is not the cardiovascular villain older guidance implied, and the whole-fat-versus-low-fat distinction carries far less weight for heart outcomes than it was once given. This is informational, not dietary advice.

Sources

• Association of dairy intake with cardiovascular disease and mortality in 21 countries (PURE) — The Lancet, 2018
• Milk and dairy consumption and risk of cardiovascular diseases and all-cause mortality: dose-response meta-analysis — European Journal of Epidemiology, 2017