When Gulliver's travels took him from Lilliput he encountered the big enders and the little enders, who argued furiously, to the point of war, over the best end of the egg for extracting the contents. This comes to mind when one contemplates arguments over cholesterol, not because eggs contain it but because in much of the debate facts seem to be used as missiles to defend entrenched positions rather than elements in a solution to a scientific problem of profound importance clinically and for public health.1

Not that there is any shortage of facts. We now have several overviews of the data on plasma cholesterol concentrations and mortality, of which the latest are published in this issue of the BMJ (pp 363-79).*RF 2-4* While some people have suggested that trial data support the findings of epidemiological studies,5 others have suggested that low cholesterol may be associated with an increased risk of non-cardiovascular disease6 and that lowering cholesterol concentrations may not be effective7 and may even be harmful8 in anyone other than men at high risk.9 The big and little enders fought their battles with conventional instruments of war. We fight ours with meta-analysis, regression dilution bias, surrogate dilution bias, risk stratification, confounding, weighting by size of effect, partition of variance, and log-linear models, in addition to the usual multivariate attacks.

The case for the relation between plasma cholesterol concentration and coronary heart disease is well summarised in the three papers by Law and colleagues.*RF 2-4* They put together data from 10 large cohort …