Intro to Analytic Studies

Causality and Comparison

In EBM, we want to understand if a particular exposure or treatment cause or cure the disease? In order for this to happen, there has to be some sort of causal relationship. The higher up the hierarchy we move, the stronger the relationship between exposure and disease.

RCT vs Cohort Studies

In a randomized controlled trial, we randomly assign patients to different groups. This will hopefully distribute the confounding variables evenly between both groups. Remember confounding variables are those variables that we may not know exist but also have an impact on the outcome. Hopefully the two groups are exactly the same. Then you control which group gets the treatment and which doesn’t (the only difference between the group).

Now since we are explicitly creating the exposure, we can’t ethically expose some patients to an intervention which we think may be harmful. So RCT’s are often used for treatment studies in which we hypothesize that the treatment will be beneficial.

In a cohort study, the people in the study population already have the exposure. Now we divide the population into those who were already exposed and those who weren’t. There is no random assignment, so there may be confounding variables.

Bradford Hill Criteria for causation

There are nine criteria which provide evidence for causation in a cohort study. We’ll look at the more important of these here.

1. Consistency: a relationship that is observed repeatedly is more likely to be causal. This is why it is important to replicate studies to demonstrate causality.
2. Temporality: causes precede effects. Cross-sectional studies cannot show this because we are looking at only one point in time.
3. Dose-Response Relationship (Biological Gradient): an increasing amount of the exposure increases the risk of the disease.
4. Plausibility: can the relationship be explained by plausible biological events or relationship?
5. Coherence: a causal conclusion should not contradict present biological knowledge. This and the above I consider to be the ones that pass the “common-sense” test. Are the results of this experiment consistent with what we already know?
6. Experimental Evidence: the condition or disease can be altered by an appropriate experimental intervention. This is where an RCT is important, but as we discussed, we can’t always do an RCT.