The object of most studies is
to prove a casual relationship between two variables; that is, that a
change in one variable "causes" a change in the other. For example, one
might find a causal relationship between amount of exercise and weight
loss. Another type of relationship besides the casual one exits--that
of 'association.' An associative relationship is not necessarily
causal, but can be explained by the presence of other 'un-seen'
variables to which the two variables being studied are themselves
separately linked. For instance, using our previous example, one might
observe that people who eat energy bars tend to lose weight at a greater
rate that those who do not. Although one might suspect a causal
relationship between eating energy bars and weight loss, it could also
be argued that people who eat energy bars also tend to exercise more and
be health-conscience, which is the actual cause of the weight loss. In
this case energy bars and weight loss have an associative relationship,
but not a causal one.In scientific studies, the 'un-seen' variables
are known as confounding variables and their influence in eliminated by
the presents of a control group. For example, in our above study of the
relationship between weight loss and energy bars, exercise and healthy
lifestyle would be the confounding variable. To control for this
confounder, we would try to make sure that our experimental group and
our control group each have the same number (or at least, statistically
insignificant differences in the number) of exercisers/healthy
lifestyles. Thus, any observed difference in weight loss between the
two groups at the end of the trial could be attributed to energy bars
and not exercise/lifestyle. Control is one of the greatest advantages
of the scientific method over other means of investigation that humans
perform.
The difficulty with case-control studies is that they do not usually
provide convincing evidence of causal relationships. The reasons are as
follows: Case-control studies are so-called "retrospective" studies,
which means that they rely on data collected by looking over population
records, usually several years after cases have developed the disease,
in order to elucidate a relationship between two variables . Data, for
instance, is frequently collected by means of a questionnaire, e.g. asking
people to estimate how many herbal treatments they used over a certain time
period, several years ago. The data collection itself is subject to several
biases--e.g. errors in recall, personal prejudice. Secondly,
retrospective studies are not based on a designed study, and as such they
lack a control group (although this difficulty can be overcome by artificially
creating a control group by matching the 'test' subjects with a similar population
on potential confounders like age, sex, and lifestyle). Even this matched group,
however, wouldn't be placebo-controlled, so retrospective studies are not suitable
to study the effects of a drug. Often, case-control studies at best can entice curious researchers to design better
studies to investigate a relationship more vigorously.
Researchers often do attempt to investigate more vigorously, and the
next level of rigor is either the "cohort" or the "experimental" (also
known as the randomized-controlled) study. Both types are
so-called "prospective studies". In a prospective
study, the study is designed before the data is collected, which allows
the researchers several advantages over the retrospective study.
Researchers can design a study with a control group (the principle means
of generating a control group is randomization of people between the
groups...this procedures eliminates many confounding variables),
administer a placebo (making the study so called "placebo controlled"),
design internal devices to eliminate bias (e.g. "double-blind"), and
follow subject throughout the length of the study to observe development
of the outcome in question. The data generated from these studies is
often considered stronger than data from retrospective studies,
largely because of the avoidance of confounding variables. The
down-sides of prospective studies is that they are more expensive and
time-consuming to design and execute than retrospective studies, and are
difficult to use to study rare disorders as the number of subjects is
often too low to form different groups within the study.
