Observational studies are those in which, as its name suggests, the researcher merely observes what happens. Well, to observe and analyze, but has no active role on the exposure or intervention under study. Within these observational studies, we all know cohort studies and case-control studies, the most commonly used.
The basic designs
In a cohort study, a group or cohort is subjected to an exposure and followed over time to compare the frequency of occurrence of the effect in comparison with an unexposed cohort, which acts as control. On the other hand, in a case-control study we begin with two population groups, one of which suffers the effect or disease under study and its exposure to a particular factor is compared with that of the group that have not the disease and that acts as control.
The cohort study is the sounder of the two from the methodological point of view. The problem is that it often requires longer periods of follow-up and large cohorts, especially when the frequency of the disease studied is low, leading to the need to manage all covariates of this entire large cohort, which increases the costs of the study.
Well, for those cases in which neither the case-control nor cohort studies fit well to the needs of researchers, epidemiologists have invented a series of designs that are halfway between the two and can mitigate their shortcomings. These hybrid designs are the cohort nested case-control study and the case cohort study.
Let’s start with the nested case and controls. Suppose we have made a study in which we used a cohort with many participants. Well, we can reuse it in a nested case-control. We take the cohort and follow it over time, selecting those subjects who develop the disease and assigning them as controls subjects from the same cohort who have not yet presented the disease (although they may do so later).
Thus, cases and controls come from the same cohort. It is desirable to match those considering variables that are confounders and time-dependent, for example, the years they have been enrolled in the cohort. Thus, the same subject can act as a control on several occasions and end up as another case, which must be taken into account in the statistical analysis of the studies.
As we are detecting how cases arise, we do the sampling based on incidence density, which allow us to estimate relative risks. This is an important difference with conventional case-control studies, which usually provide odds ratios, only comparable to relative risks when the frequency of the effect is very low.
Another difference is that all the information is collected on the cohort at baseline, so there is less risk of occurrence of the information bias characteristics of the classic case-control studies, which are of retrospective nature.
The other type of hybrid observational design that we will deal with is that of the case cohort study. Here, we also start from a large initial cohort, from which we select a more manageable sub-cohort to be used as a comparison group. Then, we follow the sub-cohort to detect over time the subjects that develop the disease in comparison with the sub-cohort (whether or not they belong to the sub-cohort).
As in the previous example, to detect the cases over time allow us to estimate de density of incidence of cases and no-cases, calculating from them the relative risks. As you can imagine, this design is more economical than conventional studies because it greatly reduces the volume of information from healthy subjects to be handled without losing efficiency when studying rare diseases. The problem that arises is that the sub-cohort has an overrepresentation of cases, so the analysis of the results may not be as the traditional cohorts, but has its own methodology rather more complicated.
And here we will leave this topic for today. To sum up a little, shall we say that nested case-control study is more like the classical case-control study, while the case cohort study is more like the conventional cohort study. The fundamental difference between the two is that in the nested study we sample controls by incidence density and matching, so we have to wait to have occurred all the cases to have selected the entire control population. This is not so in the case cohort study, much easier, wherein the reference population is selected at the beginning of the study.
The drawback of these studies, as we have said, is that the analysis is a bit more complicated than the conventional observational studies because it is not enough to do the analysis of the raw data results. Instead, results must be adjusted by the possibility that a participant can act as a case and as a control (in nested studies) and by the overrepresentation of cases in the sub-cohort study (in the case cohort study). But that’s another story…