Monday, April 28, 2014

Epidemiological Research Designs

Epidemiological research is conducted to study the risk factors leading to disease or prevention (here, I use the word disease generically, as it could also mean a condition, such as low back pain). Epidemiologic designs are used because some questions cannot be answered by conducting clinical trials. For example, the best way to determine if smoking causes lung cancer would be to devise a clinical trial in which we randomize people into 2 groups, one who will smoke 3 packs of cigarettes per day for 5 years and one that does not smoke. At the end of the study, we could then compare cancer rates. But of course this is not ethical in any possible way. Rather, the epidemiological designs observe people as they live their lives, looking to see whether or not a risk factor may lead to a disease (or perhaps, in some case, to prevention. For example, does being active reduce the risk of heart disease).

There are three designs that are used in such studies. In order of increasing complexity, they are cross-sectional studies, case-control studies and cohort studies. They each work differently.
Cross-sectional studies: These look at the association of risk to disease. This kind of study is conducted at one point in time, the present. The most common form of cross-sectional study is a survey, in which we might ask a group of people if, for example, they have low back pain. Some percentage will say yes. We can also ask if they have a short leg. Some will say yes. What we cannot say is that these two are related in any way; this would require a more complex study. But we can demonstrate an association.

Case-control studies: Here, we begin with two groups of people in a population: those who have a condition of interest (i.e. cirrhosis) known as cases, and those who do not, known as controls. We then go “backward in time” reviewing their medical history to see if there is a higher exposure to a risk factor (i.e. heavy drinking) in the cases compared to the controls. In such studies, we can calculate an odds ration; that is, a comparison of the odds of developing a condition in the exposed group to the unexposed group. We cannot calculate risk because risk involves newly diagnosed cases, and we begin here with people already diagnosed.
Cohort studies: In the final and most complex kind of study, a cohort study follows a population of people forward in time to determine whether or not an exposure to a risk factor leads to a higher rate of a condition of interest. An example of a cohort study is the famous Framingham study, in which the people of Framingham, MA have been followed for nearly half a century to study the risk factors for cardiovascular disease. At the outset, no one has the condition; over time, as people live their lives, some will develop it. We then look at exposure to risk and compare the rate of disease in the group that was exposed to the group that was not. Because we are now looking for new diagnoses of disease, we can calculate actual risk rates and thus also a risk ratio: the risk of developing the condition in the exposed group divided by the risk in the unexposed group. This also allows us to determine incidence (number of new cases per population).

All of these designs are elegant in their own way, are less often seen in chiropractic, but are powerful tools for studying the relationship between risk and disease.

No comments: