I hope my previous post on Descriptive Studies was useful.
Doll R, Hill AB. (1954). "The mortality of doctors in relation to their smoking habits". BMJ 328 (7455): 1529. doi:10.1136/bmj.328.7455.1529. PMC 437141.PMID 15217868.
I'm going to have a go at describing two types of analytic studies today - Case Control Studies and Prospective Cohort Studies. When surfing the web for literature, it's useful to keep the various advantages and disadvantages of each study type in mind so that you can properly evaluate what you are reading.
Case Control Studies
These studies work by identifying people with a disease or condition and looking back in time to identify factors that may have caused that disease or condition. The next step is to see how common these risk factors are in the group with the condition and in a control group to see if there is a real difference between them.
A good example of a case-control study was that performed by Herbst and colleagues in 1974. Herbst looked at a cohort of young women with vaginal adenocarcinoma which was unusual in this age group and also very rare. He found that mothers of these young women had been given Diethylstilbestrol during pregnancy - a medication used to prevent miscarriages and premature labour and this was identified as the risk factor responsible for the condition.
Case Control Studies are useful for studying the causes of rare diseases, they are quick to perform and need small sample sizes.
The problem is they rely on a patient's memory and on hospital records which can sometimes be unreliable. They can also be affected by different types of bias such as recall bias - where those with the condition are more likely to think harder about their exposure to certain risk factors than the control group.
Prospective Cohort Studies
This type of study aims to make observations about the association between an exposure or risk factor and the subsequent development of the disease. In other words, someone may have an inkling that a certain risk factor may be causing a disease and so will monitor people who are exposed to that risk factor and see whether they get the disease.
The best example of such a study was the 'British Doctors Study' which began in 1951 and ended in 2001. Here, a group of doctors were followed over a period of time and the first convincing association was made between smoking and the development of lung cancer (yes - it's true - some doctors do smoke!).
This type of study can be either concurrent - i.e you monitor over time or 'non-concurrent' - i.e. the information regarding the risk factor has been taken sometime in the past and you look at it retrospectively.
These studies tend to be more representative than case-control studies because you are following people up in real time rather than getting information at one point in time. Another advantage is that you can look for many different disease outcomes - in the British Doctors Study - they also showed an association between smoking and heart disease.
These studies also have the share of drawbacks - they are expensive, take a long time, require large sample sizes and you run the risk of losing participants to follow up.
I've referenced the two studies below and will post soon about clinical trials and meta-analyses.
Herbst AL, Skully RE. Adenocarcinoma of the vagina: a report of 7 cases including 6 clear
cell carcinomas (so called mesonephromas). Cancer 1970;25:745–57.