The sum of evidence-based nutrition has to be more than a Cochrane-type meta-analysis of randomised controlled trials (RCTs). Most of the evidence base in nutrition is observational, especially cohort studies. RCTs of diet change through to disease outcome are uncommon and the change has usually been addition or removal of only a single food component. Trials with whole diets through to disease outcome are rare and dietary changes made by individuals are unlikely to be an exact copy of the prescription. It is hard to even imagine a trial in which half (randomised) of a large group of middle-aged people agree to avoid vegetables for 5 y and be followed up to see who will develop cancer. Most of the USA's health claims, permitted by the Food and Drug Administration, are not supported by RCTs. But where controlled trials of nutritional change and disease outcome have been achieved, they must be reviewed very carefully. Two Cochrane reviews on diet and cardiovascular disease (CVD), published in the widely read British Medical Journal (BMJ), were criticised after their publication and the conclusions have not been subsequently adopted by expert committees. The first of these reviews was 'Dietary fat and prevention of CVD: a systematic review'. The second was 'Systematic review of long term effects of advice to reduce dietary salt in adults'. A critique of these two Cochrane reviews is presented here as a contribution to our discussion of the potential of Cochrane methodology to the reliability of knowledge about diet and disease.