22 POPULATIONS AND SAMPLES
of as replicates. If animals are used, it is recommended that they be randomly assigned
to the treatment group, and sometimes they are randomly assigned to cage location
if there is thought to be any possibility that this location could affect the outcome.
Random assignment can be accomplished easily using a random number table. Table
A.l was developed by the RAND Corporation and has been tested extensively for
randomness, It is also possible to obtain random numbers from statistical programs.
For example, random numbers can be obtained from Minitab, SAS, SPSS, and Stata.
Suppose that there are two treatments, A and B. Then the researcher can look in a
random number table such as Table A. 1 and assign an animal to A if the number is
an even number and to B if it is an odd number. When dealing with three groups, if
the random number chosen is 1, 2, or 3, the animal can be assigned to A; a random
number of 4,5, or 6 results in the animal being assigned to B; and a random number
of 7, 8, or 9 results in the animal being assigned to C. Zeros would be ignored.
Randomized clinical trials are experiments where the observational unit is the
patient. The purpose is to test the efficacy (does it help) and the toxicity (unwanted
effects) of a new treatment versus a placebo or a standard treatment used currently.
Patients who are used must meet a variety of strictly drawn criteria and sign consent
forms. Patients tend to come in gradually over time and so are not available in a group
as are test animals. They are a sample taken at given medical institutions within a
given time period who meet the criteria for entry and who agree to participate. They
are often chunk samples.
In general, these clinical trials can be separated into two broad groups. The first
group can be called independent trials done by individual or small groups of physi-
cians. For example, a surgeon may decide to compare two methods of reconstructive
breast surgery for women who have had a mastectomy. Patients who meet the criteria
and agree to accept either operation will be randomly assigned to one of two surgical
methods. Here, one recommended procedure is to list A and B in a random order
where A designates one treatment and B the other. Again using oddeven numbers
from the random number table to obtain a random order of the A and B treatments
is advised. The A or B can be put into consecutively numbered envelopes that are
not opened until the admission process is completed. Thus neither the patient, recep-
tionist, nor surgeon can control which treatment the patient gets. Alternatively, the
admission process can be done at a different location and the results communicated
by phone or e-mail to the person delivering the treatment. If the treatments are the
dispensing of one or two drugs, this can be done by a pharmacist. This is called
double-blind since neither the doctor nor the patient knows what treatment is given.
Assigning A and B to every other patient is not recommended. The personnel entering
or evaluating them can determine the assignment procedure if they even know what
one patient received.
Larger more formal studies are performed before drugs are approved for use. These
clinical trials are divided into phases. Phase I trials, which are conducted on healthy
volunteers, are conducted first to assess safety at various doses. If a drug is deemed
safe, a phase I1 trial is conducted to assess the optimum dose and efficacy in patients
with the disease to be treated. If the drug passes this test it goes on to phase 111, where
an experiment is done with patients randomly assigned to the new drug or the current