In the performance of analytical studies epidemiologists move from the demanding chores of collecting accurate information into the realms of designing studies that seek to answer important individual questions about the causes of cancer. In this area they will usually have an idea to test - a hypothesis about some possible causative factor. The focus shifts from whole nations or whole regions to a much more closely defined group of individuals. By collecting a great deal more information about a rather smaller number of people (but not so small that our conclusions might be based on pure chance), it is possible not only to demonstrate links between particular factors and particular cancers but also to look carefully to see if there are any possible alternative links which have to be considered or excluded by careful work. A number of methods of performing analytical epidemiology are recognized and are worth mentioning to give the general flavour of this sort of work: cohort studies, case-control studies and intervention or experimental studies.
Case-control Studies. In this method the group of people about whom information is collected are those who are already suffering from the particular cancer. They are then matched to another group of people who do not have the cancer but who are similar in other aspects such as age, sex and often social group. The group of patients with the cancer (the cases) are then compared to the group who do not have the cancer (the controls) in terms of their previous exposures to all sorts of factors. If the cases have had more exposure to a particular factor than the controls, it suggests that that particular factor is linked to the cancer. Again, it sounds easy bur collecting the information is a laborious task and choosing controls is full of pitfalls. If the groups are not properly matched then misleading links can be suggested. A particular pitfall is choosing groups of people when they come to hospital. Hospital-based control groups may be very unrepresentative of the general population.