Basic Statistics

COMMON TYPES OF BIOMEDICAL STUDIES^7

obtain reliable and unbiased information on events that happened long ago both in
terms of the event and the timing. For example, referring back to Figure 1.1, it may
be difficult to determine that a causal factor occurred during the relevant time period,
particularly if the causal factor occurred in a short time period.
Getting the general population to respond to phone surveys is getting to be an
increasing problem, and the switch to cell phones is making it worse. Even in mail
surveys to patients who have received medical treatment, response rates of about 50%
are common. Obtaining a good response rate takes careful planning (see Groves et
al. [2004]).
However, public health surveys are used by states and counties as a means of
assessing a wide range of information on the health status of the general public. For
example, a survey was done in Washington State on topics such as access to health
care and characteristics of the uninsured, use of health screening procedures such as
mammograms and PSA tests for prostate cancer, and the use of routine dental care.
Such surveys can provide information on where public health workers need to direct
their efforts.

1.3.2 Experiments

Experiments are usually simpler to interpret than surveys. Typically, a single sample
is obtained and then the sample is split into different groups, often called treatment
and control groups. Treatment denotes the new treatment being tested and control
is either no treatment, a placebo treatment, or a standard treatment that is useful for
comparison purposes. The experimental subject or unit could be an animal, person,
tissue culture, and so on. Random assignment of the subject to treatment or control
is the recommended practice. How to perform a random assignment is discussed
in Section 2.4.3. The aim is to have the only difference between the treatment and
control groups be the treatment itself, not what or who is being treated.
In an experiment, the causal factor is given first and there is no question as to
what the factor is since the experimenter makes that decision. The measurement
of the outcome follows the treatment in time so causality can be established if the
experiment is carefully designed. This is illustrated in Figure 1.3 by the absence of a
question mark before cause.

1.3.3 Clinical Trials

In clinical trials using patients, the major objectives are to test efJicacy (whether the

treatment is better than the control in treating the disease) and toxicity (whether the

treatment has fewer side effects or unwanted effects than the control). Decisions

need to be made on who is eligible for the trial, what treatment will be used, and

how the outcomes of the treatments will be assessed. These are complex studies to

perform which involve obtaining human subject consent, decisions on entry criteria

of the patients, under what circumstances the treatment could be modified, or if the

clinical trial should be terminated early. Also, because the patients normally come

for treatment singly over time, it may take several years to accumulate a sufficient