Sir

The science of dieting, as Declan Butler observes in your News Feature “Slim pickings” (Nature 428, 252–254; 2004), is “painfully thin”. Yes, but he is too kind. The problem is not merely a shortage of large, long-term, well-controlled trials. More fundamental is our inability to measure what people actually eat.

Most trials of free-living populations involve assigning subjects to alternative diets. Their food intake is measured, if at all, by one of the conventional instruments — food questionnaires, diaries, duplicate portions. All these methods share the same weakness: they depend on people honestly telling researchers what they consumed.

In most trials, there is no independent measure of food intake. ‘Independent’, in this context, means a biochemical or physiological indicator of the nutrients and other food components consumed. Yet it is universally recognized that most people underreport how much they eat. And overweight people underreport most of all, sometimes by more than 50%. So we do not know to what degree, if at all, subjects were actually following the diets to which they were assigned.

Experiments in sealed metabolic wards have total control over food portions. But these are small, short-term studies in artificial conditions. Getting free-living volunteers to drink double-labelled water — containing markers which allow their urine to be measured for energy expenditure — is an acceptable surrogate measure, but too expensive to be used in the large, long-term trials that are needed.

In contrast, we have precise measures for the dependent variables, such as weight. But these days, serious trials also attempt to measure the effects of alternative diets on other important biochemical indicators, such as serum cholesterol or triglycerides.

With these measures, researchers sometimes draw conclusions that one diet is better for, say, the heart than the other. Yet they do not know if their subjects were actually eating different diets. They are trying to correlate two variables, without having adequately measured one of them.

Only when we can measure accurately both total energy intake and its component parts will we be able to determine which diets yield sustainable weight loss. These measures must be practical for use in trials on large numbers of free-living subjects over long periods. So they must be cheap, rapid, non-intrusive, painless, self-administrable, and capable of direct data transmission. This will require developments in technology as well as in basic science.

In sum, accurate measurement of food intake is the foundation stone for a true science of dieting.