10 INITIAL STEPS
The follow-up period should be long enough so that the exposure could possibly
cause the disease. In other words, the length of the induction and latency periods must
be considered. It is sometimes easier to make inferences from prospective studies if
the exposures are severe and short in duration. For example, a high-level exposure to
radiation or toxic waste that took place in a known short interval would be easier to
study than a low-level exposure over a considerable time period. Examples of a short
interval would be the Chernobyl accident in Russia or the 1999 nuclear accident in
Japan. However, if the exposure is chronic, one can investigate cumulative exposure.
For example, exposure to smoking is often measured in pack-years (number of years
smoked times the average number of packs per day).
1.3.6 Case/Control Studies
In casekontrol or caseheferent studies, the investigator begins with cases who already
have the disease diagnosed (outcome) and looks back earlier in time for possible
causes (see Schlesselman [ 19821). Before discussing this type of study, some mention
should be made of single-sample case studies (no controls without the disease are
used). Here the investigator typically searches a medical record system for all the
cases or patients who have a particular disease outcome in a fixed time period, say the
last 2 years. Then, a search is made through the records to see if some prior exposure
occurred, more than would be expected considering the group of patients involved.
One difficulty with this type of study is that it is difficult to evaluate the levels of the
exposure factor and decide what is high or low, since only cases are studied.
If both cases and controls can be studied simultaneously, an opportunity exists to
utilize a very efficient study design. A group of cases is taken and simultaneously a
group of controls should be taken from the population from which the cases devel-
oped (see Section 2.4.5 for sampling considerations). Case/control studies have the
advantage of no risk to the patients, low cost, and feasibility even for rare diseases.
Since most diseases are uncommon in the general population, casekontrol studies
are often the only cost-effective approach. They are not necessarily a good choice if
the prevalence of the exposure factor is very rare or if there is a possibility of major
differences in ascertaining the exposure factor between cases and controls.
For example, in Los Angeles County, the University of Southern California Cancer
Center obtains information from hospital pathologists on essentially all newly diag-
nosed cases of cancer in the county. Caselcontrol studies using these newly diagnosed
cases and neighborhood controls have been successful in extending the knowledge of
risk factors for several relatively rare types of cancer.
1.3.7 Other Types of StudiesOther types of studies are also used in biomedical research, such as the use of historical
controls. For example, in surgical practice, if a surgeon decides that a new surgical
technique should be used, often all subsequent patients are given the new treatment.
After using the new treatment for a year or two, they may decide to compare the cases
treated with the new treatment to controls treated with the previous surgical treatment