The study of the pharmaceutical effect of a drug is always done in reference to a population of prospective patients, e.g. the clinical dose to be recommended for an older patient is often different from that for a younger patient. Thus, the target population for the study must be well defined in advance. Obviously, it is impractical to study the entire patient population of interest. Fortunately, this is also not necessary. Statistical sampling methodology enables us to draw conclusions from a sample to the population from which the sample had been drawn, to any desirable degree of accuracy and confidence. However, there is one important caveat to this ability: the sample must be 'representative' of the population of interest; meaning that it must preserve all the relevant characteristics of the population. That is, it must have the same proportion of females and males, the same racial distribution, the same percentage of hypertensives, and so on. Clearly, the creation of an exact replica of the population on a small scale is an impossible task. However, statistical sampling methods can produce very close to representative samples with very high probability. These are the methods utilized by pollsters to make highly reliable predictions and inferences on the population from relatively small samples.
In clinical research, the random selection of subjects to be included in the trial from the target population is not practical. Subjects are usually selected from the patient pools available to the investigators participating in the trial. This, in and of itself, is problematic. The patient pool available to a particular center usually reflects the population in the geographical area where the center is located, which may not represent the general potential patient population. To complicate things even further, some of the patients available at a given center may not be suitable for enrollment in a trial with an experimental drug. The investigator may wish to exclude certain patients because of certain known or unknown risks. The possible effects of drugs on the unborn fetus are often unknown, and thus pregnant or lactating women are usually excluded. Patients may be excluded if they are taking another medication which can potentially interact with the study drug. Also, some patients may refuse to participate in the trial for one reason or another. Finally, for the purpose of studying the efficacy of a drug, it is desirable to enroll only patients who are most likely to have a measurable response to treatment. Thus, every trial protocol contains a list of inclusion and exclusion criteria defining the subject population to be studied. Obviously, such a population is hardly ever fully representative of the target population. This raises a question regarding the generalizability of the trial's conclusions.
When defining a set of inclusion and exclusion criteria for a trial, the issue of generalizability must be kept in mind. The rule is that the more restrictive the criteria, the less generalizable the results. On the other hand, setting criteria for eligibility to participate in the trial provides the investigator with an important tool for controlling the variability. Thus, the choice of eligibility criteria must guided so as to balance the efficiency of the trial design against the need to ensure that the result are generalizable. Some of the guiding principles for defining subjects' eligibility are listed below.
Was this article helpful?