Associations between food and disease are commonly reported in the nutritional literature. For example, it has been shown that many ingredients from recipes in a cookbook are associated with cancer (here). This explains the high frequency with which lifestyle sections of newspapers and magazines report a particular food being associated with a particular disease. While it is important to investigate associations between food and disease, the research should always be taken in context because cause and effects cannot be ascribed in such research. For example in a recent paper in the Journal of Nutrition, researchers reported on the associations between breakfast consumption and nutrient intakes1. The results of this study showed that those individuals who consumed breakfast had the highest nutrient intakes. From this it might be assumed that breakfast was the cause of the improved nutrient intake and that eating breakfast is beneficial in this regard. However, is this really what the data is showing or is there more to understand?
Remember that an association is a link between two variables (things that change). When one variable goes up, the other variable moves in concert, either up or down. In associations, the variables are locked together so that changing one changes the other. However, just as it would be ridiculous to state that fire engines cause fires, there does exist an association between fire engines and fires. However, the fires and the presence of the fire engines are both caused by a third unseen variable, in this case electrical faults or matches. When we consider our original association between breakfast cereal and improved nutrient intake, we could assume that because vitamin and minerals are added to breakfast cereals, that those consuming them might have better nutrient intakes because of the act of consuming breakfast. But we must look deeper into the data and then interpret that data as relevant information before we can draw any valid conclusions. In fact this paper contained some interesting information that sheds a better light on the headline association.
In particular, the authors reported on the demographics of the subjects in the study. Of the participants, 11 % did not consume any breakfast, 20 % consumed ready to eat breakfast cereal, whereas 69 % consumed another type of breakfast. Of these groups, the first group (no breakfast) was most likely to be young, male and single, the least likely group to take vitamin or mineral supplements and the most likely to smoke. The second group (breakfast cereal consumers) was mostly likely to be the oldest, had the most food security, was the most likely to use vitamin and mineral supplements and least likely to smoke. The third group (other breakfast consumers) were least likely to speak English, were of intermediate age, smoked but also used supplements. This data is important because it shows that the association between eating breakfast and nutrient intake is much more complex than at first reported. From this, it appears that the breakfast is just a marker for other variables that could be the cause of the association.
If we take liberties and make some assumptions from this data, we can hypothesise that the reason that non-breakfast consumers had lower nutrient intakes was nothing to do with the breakfast, but because these individuals had unhealthy lifestyles. The two groups who consumed breakfast may have had higher nutrient intakes because they had healthier lifestyles which included them being more likely to take vitamin and mineral supplements. Was this the real cause of the association between breakfast and nutrient intake? It is hard to say, but it likely played a strong contributory role. Interestingly the other breakfast group were the least likely to speak English, suggesting that they were French Canadians (the study was done in Canada). This explains why they did not eat ready to eat breakfast cereal, but chose other breakfast option. it also may suggest that they were eating more traditional French diets, which tend to be higher in nutrients, fat and cholesterol. And indeed this group had the highest cholesterol intakes of all.
These observations are important because it is paramount in epidemiological research to go beyond the superficial level and view associations at a deeper levels. Taking the association between breakfast cereal and nutrient intake at face value is damaging because it could lead to modulation of behaviours that are counter productive. For example, someone modifying their diet to include ready to eat breakfast cereals in order to increase their nutrient intake would be misguided. In fact, if that breakfast cereal contained sugar, they may detrimentally damage their health in the long-term. This highlights the difference between those that can read and repeat data, and those who can read and understand the data, and then convert it to useful information. The data is important, but without a broader context it can lead to suggestions that fire engines cause fires. Because science editors in mainstream media often do not understand what they are reading, they are forced to repeat the data at the most basic superficial levels, allowing them to be mislead.
RdB
