In observational studies, there are no intervention – we just observe what already exists. The case-control study is different in that we start with the cases (just as the name starts with “case”). Then we need to find controls of people who do not have the disease (this can be tricky as you have to match the criteria in both groups, ie age, gender, etc). Then we compare exposure in cases and controls. One major limitation is that it doesn’t measure incidence of disease.
These are good for
- investigating outbreaks (like HIV/AIDS from the previous video) and
- studying rare diseases or outcomes.
Data collection needs to be done in the same way for both groups. Recall bias can also affect the results, as people are trying to think back and explain why they have a particular rare disease.
Case-Control Studies & Odds Ratios
Remember that
relative risk = (incidence in exposed) / (incidence in unexposed),
but since we don’t have incidence, we cannot use this in case-control studies. Instead we have the likelihood of exposure in the cases and the likelihood of exposure in the control groups. So instead we use the odds ratio,
odds ratio = (odds of exposure in cases) / (odds of exposure in controls) = (a/c) / (b/d) = ad/bc
Asking how much higher is the odds of exposure in the cases than in the controls. This can be calculated from the 2×2 table.
Disease CASES |
No Disease CONTROLS |
|
---|---|---|
Exposed | a | b |
Unexposed | c | d |
OR = 1 — no effect
OR > 1 — risk factor
OR < 1 — protective factor
Case Control Example
This video goes over an example of a case-control studies of bicycle related head injuries presenting to a Seattle emergency department.
- Case: head-injured bicyclist
- Control: non head-injured bicyclist
- Exposure: were they wearing a bike helmet
Cohort vs Case-Control
More on the difference between cohort studies and case-control. Remember in case-control, you are starting with a group of people who already have the disease (they fit the case-definition) and then need to search out similar people without the case-definition (the controls).
We’ll review cohort studies more in the next video.
Cohort Study Design
In the cohort study, we start with a population and divide them based on whether they have been exposed or not. Then we look to see who will develop the disease. Because we are measuring the occurrence of disease, incidence, we can use relative risk. At the beginning of the study, no one had the disease, we are measuring who gets the disease.
These patients are followed over time, and so tend to be expensive, so multiple outcomes and multiple exposures are often studied.
relative risk = (incidence in exposed) / (incidence in unexposed),
Types of Cohort Study Design
There are differences between case-control and cohort studies.
Case-control | Cohort |
---|---|
start with disease, look for exposure | start with exposure, look for disease |
good for rare diseases | common diseases associated with rare exposures |
susceptible to recall and selection biases | longer follow-up increases risk of drop-out |
less expensive | more expensive |
So when would a cohort study make more sense?
- Common exposures: single geographic populations (Frahmingham, Massachussetts), single professional populations (the Nurses Health Study)
- Rare exposures: create a special population, like those exposed to asbestos, or retired football players
Retrospective Cohort Studies
In the prospective cohort study, the patients have not yet developed the disease at the start of the study. You have two groups, the exposed-to-a-particular-risk-factor cohort and the unexposed-to-a-particular-risk-factor cohort, and then you look which will develop the disease in the future.
In the retrospective cohort study, the patients have already developed the disease at the time of the study. You have two groups, the exposed cohort and the unexposed cohort, and then you see which have the disease and which don’t.
People often get this confused with the case-control study design. With case-control, we start with the disease and look who has been exposed and who hasn’t. With retrospective cohort studies, we start with two groups (cohorts), those who have been exposed and those who haven’t, and see who develops the disease.
Attributable Risk
We can get relative risk, but also attributable risk. That is, how much of the disease that’s occurring in a person is attributable due to that exposure – and by converse, how much of the disease risk can be eliminated by removing the exposure.
This is calculated by subtracting the incidence in the unexposed group from the incidence in the exposed group. It is also often expressed as the attributable risk proportion. Or, stated mathematically:
- AR = Ie – Iu
- AR proportion = (Ie – Iu) / Ie
- AR proportion = (RRe – RRu) / RRe = (RRe – 1.0) / RRe
Odds Ratios Vs Risk Ratios
There are two main points covered in this video that are important to the interpretation of case-control studies. This is because in case-control studies we start with the OUTCOME and then look for EXPOSURE. This means two things: we cannot calculate risk (but we can calculate odds) and we technically calculate the odds of an exposure given an outcome (which is the reverse of what we want to know). Luckily, two things are true:
<
ul>
So though we cannot calculate risk in case-control studies, we can get a pretty good estimate.