The perfect randomised controlled dietary prevention trial of coronary heart disease has never been done. The best we can do is to look at all the trials together. Dietary trials should be separated from drug trials because they have different characteristics. Fourteen dietary trials which had disease or death as the end point are collected in this review for a meta-analysis. Three of the trials had two parts (male/female or low fat/increased fish), making a total of 17 trials. All were randomised trials, except the Finnish mental hospital trial which was a 12-year crossover in two hospitals. The trials were primary or secondary, diet only or multifactorial; numbers of subjects range from 52 to 57,460. For total deaths the ratio of intervention/control in all 17 trials is 0.94 (significantly less than 1.00) and for coronary events the pooled odds ratio is 0.87. But in the seven trials with most effective cholesterol lowering the odds ratios are 0.89 for all deaths and 0.70 for coronary events. There is thus no indication of excess all causes mortality in the dietary trials. Four recent secondary prevention trials had angiographic end points. There were a total of 275 subjects; trials were in Holland, USA, Germany and UK. In all trials plasma cholesterol was effectively lowered and coronary narrowing regressed a little, or progressed less in the diet group but significantly compared with controls. These angiographic trials strongly support the results of the major prevention trials. Lastly, a set of ten trials with fish oil after coronary angioplasty are reviewed. In some there appeared to be lower rates of restenosis, but not in all. The mechanism here is different from the major trials with plasma cholesterol-lowering diets for longer periods.