In our previous post discussing methods of recording food intake for nutrition research we discovered several ways that researchers can find out what people are eating. From this they can further calculate a persons consumption of macro nutrients, micro nutrients, calories and much more.
This data by itself, whilst interesting, is not very useful. Only when data is collected and analysed as part of a study can we really begin to make discoveries and reach solid, evidence based conclusions.
There are many methods available when conducting research. These generally fall into two categories, observation and intervention. Let’s start with the observational methods.
Ecological studies are focused not on individuals, but groups or national populations. For example, consumption of olive oil per capita vs heart related illness in that country.
However the data can be misleading. For example, countries that consume the most alcohol per capita have lower rates of traffic related deaths. Clearly it would be wrong to assume that alcohol consumption is somehow a preventative factor for traffic deaths.
Cross sectional study
In a cross sectional study researchers take a group of people and look at the exposure and outcome at the same time. For example, the exposure might be the persons body weight, and the outcome their blood pressure. Both sets of data can be collected at the same time and examined. In this case, people with higher body weight tend to have higher blood pressure.
As with ecological studies, cross sectional studies can be misinterpreted. For example, people who drink large amounts of diet sodas rather than the non-diet versions tend to have higher BMI and body fat levels.
It would be incorrect to assume that the diet soda intake is causing the higher BMI. In this case, participants are drinking the diet soda rather than the non-diet version in an attempt to help them lose weight. The high BMI and body fat levels came first, followed by their choice to drink diet sodas.
With cross sectional studies it is difficult, if not impossible to know for sure which factors are the causes and which are the results. Cross sectional studies can show us interesting correlations, but cannot provide proof of what factors cause what results.
Case control study
Say we are interested in the potential causes of a certain type of disease. In a case control study, we would take a group of people who have the disease, the “cases”. We would ask about their dietary history, for example frequency and quantity of meat consumption.
We would then do the same for a “control” group; people who do not have the disease. By comparing the results we may be able to see a correlation between the amount of meat eaten, and the chance of developing the disease.
Selection bias can play a factor in the quality of case control studies. Both the cases and control groups should be from the same geographical area, be of similar economic status and so on. It is important to normalise as many variables as possible in order to have a fair comparison between the groups. Otherwise external factors such as quality of healthcare and housing, environmental pollution etc can skew the results.
Confounding is another possible limitation to case control studies. We know that people who eat large amounts of meat tend to be heavier. We also know that being heavier can be a risk factor for the disease we are researching in this study. So perhaps any correlation between meat consumption and occurrences of the disease may be due to the cases group being heavier, and not the amount of meat they are eating.
We can attempt to minimise the amount of confounding in a study. In the example above, we may add the BMI of the participants in the data as a confounder.
This is the most powerful observational study. Participants are assessed and continually studied over long periods of time.
Cohort studies can be extremely large, such as the Nurses Health studies in the US and the EPIC study in Europe involving over 100,000 people. They provide info on their dietary intake and food frequency every couple of years.
This data is analysed and linked to the rate that the participants develop diseases or other illnesses.
The main limitation in these studies are the food frequency questionnaires which are used to gather the data. See the previous post for more about the issues with FFQs.
Commonly known and followed nutrition guidelines such as to eat more fruit and vegetables, or to eat less saturated fat all come from cohort studies.
In an intervention study we no longer just watch, as with the observational studies above, but actively intervene in the lifestyle of our subjects. A group of subjects will be randomised and split into two or more groups. One will get the intervention, which for this example is a fish oil pill. The other group will get a pill that they think is fish oil but is actually inert and will not contain any active ingredients. However they will not be told this.
After a period of time the subjects will be reexamined. In this example study, the subjects may have their cholesterol levels taken before and after the fish oil pill intervention. We can compare the results between the groups that did and did not have fish oil pill.
Ideally these studies would run for many years and include many people, however this is very expensive and difficult to do. However even short term intervention studies can yield valuable results.
The advantage of experimental, or intervention studies is that it is possible to establish causality.
In other words, does doing/eating/having X cause Y (weight gain, an illness etc). This is not possible with observational studies, which can only establish if there is a correlation or association between the factors being studied. Observation studies can suggest, whereas intervention studies can prove.
It is very important to find out and analyse what study design was used in any research that we look at. We cannot rely on other people such as press officers, authors or journalists to do this for us. Often press coverage is given to poorly designed or significantly flawed studies simply because the conclusion can provide a catchy headline. “Eating chocolate causes you to lose weight!“ for example.