Researchers in human (and veterinary) medicine are always on the lookout for new drugs, medical procedures, and life-style changes that will improve their ability to bring better health to patients. For each one, they must establish whether it truly represents an improvement from what was used before.
In a typical retrospective study, the health profiles of the subjects in a particular "case" group (e.g. smokers) are compared with those in a control group that has been selected to be as similar as possible to the "case" group (similar spread of ages, sex, etc.). Retrospective studies are also called "case-control" studies. Retrospective studies run the risk of investigator bias; that is, the investigator picks subjects that are most likely to show the effect that prompted the study in the first place.
A prospective study selects a population in good health and meeting any other desired criteria (e.g., smoking habits) and follows it over a period of years to see what happens to its members. Prospective studies are also known as "cohort studies".
Where it is possible to do them, clinical trials represent the "gold standard" for evaluation. In performing a clinical trial (e.g. of a new, and possibly better, drug), the investigator
- assembles a group of subjects picked to be as similar as possible in their characteristics (to avoid "confounding variables")
- assigns them randomly to the experimental group and a "control" group
- performs the experiment on the first group. It is best to do this in a "double-blind" fashion; that is neither the subject nor the experimenterknows who is getting the treatment and who is not.
- the subjects in the dark avoids the placebo effect; a response (often quite powerful) that is not due to the treatment but to the expectations of the subject;
- the experimenter in the dark avoids bias in the evaluation of the results. So, for example, a microscopist examining a slide of tissue should not know whether it came from an experimental subject or a control.
Analyzing the Data
The data acquired in any type of clinical study must be evaluated to see if any effect seen is significant.