International
conference on
saturated fat

Professor Peter Elwood DSc MD FRCP
with Professor Ian Givens, Dr John Gallacher and Janet Pickering
Department of Primary Care and Public Health, Cardiff University, Cardiff CF14 4XN UK

Perceptions of the risks and benefits of milk and other dairy foods have been hopelessly biased by short term studies of the relationships between consumption and the levels of plasma cholesterol, lipid fractions and other vascular risk factors. Evidence from such studies is unbalanced, and possibly misleading, as it ignores the effects of dairy food consumption on other biological mechanisms relevant to vascular and other diseases. The only appropriate basis for food policy and advice is evidence in which the consumption of a food item is related directly to deaths and to incident disease events.

The best evidence on the benefits or harm of a food item would come from randomised controlled studies with death or disease incidence as the outcome. There have been no such studies of dairy food items and there probably never will be any. The best available evidence comes therefore from large prospective studies in which the long-term consumption of a dairy food item(s) is related to mortality and to incident disease.

We have reported a search of the literature followed by a meta-analysis of all cohort studies in which dairy food items have been evaluated against clinical outcomes (J Amer Coll Nutr. 2008;27:723S-734S). We are at present conducting a further, even more extensive search of the literature and this will enable us to examine relationships of disease with all the dairy food items – as far as the published literature allows.

The most important conclusion reached so far is that apart from milk, the evidence on dairy foods is sparse and, for several dairy items inconsistencies (heterogeneity) in the results of the studies prevents the drawing of conclusions in terms of disease risk with any reasonable degree of confidence.

An exception to this last is milk, for which there is homogeneity between the studies and for which meta-analyses suggest a lower incidence of vascular disease, and of incident diabetes in the subjects with the highest milk consumption - each by about 10-15%..

There are limitations in cohort research studies. In particular, relationships of interest can be confounded by factors other than the food item of interest, such as life-style, smoking, social class and possibly other unknown factors. Serious confounding seems unlikely in the case of most dairy items, but is highly likely in relation to types of milk (whole or fat-reduced). The selection of the type of milk by people is likely to be so highly confounded with other health related behaviours, that conclusions as to whether fat-reduced dairy products are better or worse than the whole item, simply cannot be drawn.

A call should therefore be issued for further evidence from cohort studies before any new advice is given on dairy foods, and before policy relating to dairy foods is changed.