The inherent flaws of food intake data

The inherent flaws of food intake data

Measuring our dietary patterns and linking it to patterns of disease is at the core of modern nutritional epidemiology and such data drive national and global food and nutrition policy. There is, however, a serious and inherent flaw in the measurement of food intake which modern nutritional epidemiology tends to forget. That flaw is energy under-reporting. Our energy requirements are composed of several factors, the most important of which is resting metabolism which accounts for about 85% of energy needs in a normal adult following a typical sedentary western lifestyle. These energy needs are to keep our hearts beating, our lungs breathing, our kidneys filtering, our brains remembering and so on. We can directly measure this as a person’s resting metabolic rate (RMR) using a calorimeter and there are also a number of ways of doing so indirectly, some of which are  extremely accurate. We can also calculate our RMR using a number of equations and you’ll find plenty of calculators on the internet. My RMR is 2,030 calories. Because I’m sedentary, except for golf on a Saturday morning, I need to up that figure by about 15% to 2,335 calories to take account of my daily ohysical activity. A very sporty person would have a higher multiplier of RMR.  If I was a volunteer in a dietary survey and I reported an energy intake of 1,900 calories, then ocviously I must be dieting. If I say I’m not dieting and that this is a typical dietary intake, then I’m under-reporting. There never has been and there probably never will be a large survey, large enough to be of value to epidemiology, which does not have some element of under-reporting. And the level of under-reporting is huge - anywhere from 30% to 50%. We know this to be so using both simple equations to measure RMR and also using very sophisticated stable isotopes.

Why do people under-report? We know it is higher among females amd we know it increases with increasing body weight. My explanation, which is not based on any experimental data but on supposition is as follows. Most people with a western sedentary lifestyle, have at some time sought to lose weight. They inevitably start on Monday morning. Come Thursday, something happens, good or bad and the dieting pattern is gone. Its back to normal to start all over again next Monday morning. This cyclical pattern is familiar to many people. So, when asked to take part in a dietary survey and when pressed to be truthful in every way to report their habitual intake, which days do they deem to be “typical?. I’m afraid that 30-50% of people deem the dietary restrictive days of Monday, Tuesday and Wednesday to be normal. Thus they don’t deliberately lie but they do under-report their food intake. In effect, food intake data are flawed and we have to live with that for now until we come up with some smart way of overcoming this problem.

Because under-reporting is higher among the over-weight and obese, many assume that the foods that are under-reported tend to be the so called “guilty” ones: foods high in sugar and fats such as fast food, soft drinks, savoury snacks and so on. This assumption is of course false since obesity is associated with ALL foods (see blog of November 6th: “Taxing the fat and sweet”). Not surprisingly, when we examine food intake data in those with plausible energy intakes against those under-reporting food intake, we find all food categories under-reported.

This issue of energy under-reporting is dismissed by nutritional epidemiology on the grounds that all their propsed statistical associations of diet and disease are adjusted for all of those factors of importance in under-reporting (body weight, energy intake, gender, age etc). However, there is an increasing number of researchers who are showing that this statistical adjustment is flawed when it comes to under-reporting food intake. Basciaclly, an average daily intake of a food is composed of three elements. Firstly, the population average embraces  both consumers and non-consumers of the food in question. Some people who under-report energy intake may simply deny eating one or more foods. That is the first route of under-reporting. The second is that they admit reporting but under-report the frequency of consumption. The third is that they admit eating the food, are truthful about the frequency of intake but are untruthful with the portion size they report. Of course any combination of these is possible. There is simply no way in which statistical jiggery-pokery can unravel this web of deceit. So we have only one option. We create a cut off point (RMR + 15% of RMR) for energy requirement and anyone falling below this is excluded from the analysis. Its painful to lose subjects in this way when statistical power is dependet on adequate numbers.

Without doubt the area of greatest concern over the distorting impact of energy under-reporting is in relation to obesity. Firstly, the scale of under-reporting rises considerably with rising body weight. Secondly, obesity is such a hot topic as regards candidate foods for taxation or labeling. How can we be so confident in shaping public health nutrition policy in obesity when (a) we know that food under-reporting is generally a problem but particularly a problem in obesity and (b) when there is no hope of any statistical trick separating out the three lines of mis-reporting: denying ever eating the food in question, not accurately reporting frequency of the intake of a target food and finally, under-reporting portion size. It bothers me a lot but its a mere nuisance to the high priests of public health nutrition who know both  the problem and the solution.