If you want to create an animated discussion, start talking about what you should eat (or not) to protect yourself from cancer, cardiovascular disease, diabetes etc… Humans have been wondering which foods to eat since ancient times, but we still have more questions than answers!
Several large studies have shown that we should avoid excess calories, eat plenty of fruit and vegetables, reduce the amount of refined sugars and stay away from processed food, but everything else is still open for discussion. Why is this so?
Nutritional science is based on epidemiological studies, from which researchers try to determine what the effects of different foods or diets are. They can then perform interventional studies, whereby they ask half of the study participants to have more of a specific food or change their diet, and compare the effects with the other half of the participants (= the control group). In the Predimed Study for example, researchers examined the importance of olive oil and nuts in this way. Without good epidemiological studies however, researchers do not know what to look for.
I have just stumbled upon an article by Farin Kamangar and Parisa Karimi explaining how difficult it is to conduct such epidemiological studies, and here comes their list of possible errors and problems.
The first problem Kamangar and Karimi mention is the way they obtain information about what you are eating. Traditionally this is done via questionnaires and interviews assessing your dietary habits over one year. It is clear that it is difficult to recall accurately what you have eaten over a year. Moreover, you might have changed your diet, and what you have eaten last year is therefore not necessarily reflecting your lifelong habits.
Some researchers ask you to write down what you have eaten for the last 24 hours. Even though you will be able to recall this accurately, it is not sure that you are having something similar each day.
Most questionnaires are really long (80 to 200 items), which can discourage even the most motivated participant.
These problems will hopefully be minimised in the future, as researchers are starting to use online self-assessment questionnaires.
Epidemiologists try to find an association between risk factors and diseases. Confounders are factors that independently affect the risk of developing the disease, and therefore falsify the results of the study.
A classical example is the association between coffee drinking and lung cancer. Previous studies suggested this, but as coffee drinkers were more likely to be cigarette smokers, the risk of developing lung cancer had nothing to do with coffee. Another example is the association between lying in bed and dying. As most people die in bed, you would conclude that beds are the most dangerous places on earth.
As confounders are usually unknown, it is easy to see how they can falsify the results. Researchers try to avoid them by comparing populations that are very similar except for the factor they want to study. With this in mind, Beth Taylor and her colleagues compared marathon runners with their non-competitive spouses to examine the effect of strenuous exercise on your arteries.
Variability of food products
Foods are typically grown and/or prepared in different ways in different parts of the world, which might affect their nutritional value. Kamangar and Karimi cite the example of brown rice, which is a healthier choice than white rice since it contains more fibre, but in the United States it has a higher arsenic concentration than white rice. Another example is the difference between grass-fed and grain-fed beef: grass feeding typically results in leaner meat.
The correct control group
As already mentioned above, finding a good control group is a challenge: it has to resemble the study group except for the factor the researchers want to study.
In nutritional science this is very difficult, as the study participants avoiding a specific food usually replace it by something else. For example, imagine what the study results would be if participants abstaining from a specific, probably unhealthy food use to replace it by something containing plenty of simple sugars.
Foods contain chemical substances that influence each other’s absorption and actions, and different foods can contain the same substance in different amounts and availabilities. These interactions are very complex, and researchers need large study groups to try to avoid them.
Epidemiological studies about nutrition are typically large and expensive. Researchers want to use all the data and they therefore test several hypotheses at once. However, there is always a chance to come to a false conclusion by coincidence. Statistically, the probability to obtain such a false positive result is 20%, which means that if you test 20 hypotheses, you will get at least one false positive one.
Nutritional science is fascinating, but difficult. Keep informed, stay critical and be prepared to change your mind.
F Kamangar and P Karimi. The state of nutritional epidemiology: why we are still unsure of what we should eat? Arch Iran Med. 2013; 16(8):483-486.
BA Taylor, AL Zaleski, JA Capizzi et al. Influence of chronic exercise on carotid atherosclerosis in marathon runners. BMJ Open. 2014; 4(2): e004498. doi: 10.1136/bmjopen-2013-004498.
ME Van Elswijck and SH McNeill. Impact of grass/forage feeding versus grain finishing on beef nutrients and sensory quality: the US experience. Meat Sci. 2014; 96(1):535-40.