 Opportunities
are available for you to pursue a thesis project in either basic or clinical
research. Clinical research is less completely "controllable,"
and is therefore more subject to potential confounders and sources of bias.
However, clinical research offers the advantage of more direct clinical
relevance. Whatever topic you choose, you can likely find a qualified advisor
in the medical school. But an advisor knowledgeable in your area of interest
may or may not have a strong background in research methods. On the chance
that they do not, you should be prepared to follow a systematic process
in the development of your project to be sure the results are what you intend.
Be advised
that data do not make a thesis. Without an excellent hypothesis, defined
outcome measure(s), and good methodology, collected data cannot lead to
meaningful insights. Above all, a good thesis depends on an excellent hypothesis.
It is more than coincidence that a "thesis" is derived from the
underlying "hypo-thesis."
The following
sequence is recommended for the development of a clinical research or epidemiology
thesis project. Don't wait for your advisor to bring up each component;
they may never do so. You should be assertive that each item below be addressed
before you begin the actual research. If you and your advisor are uncertain
about or want help with any of the steps below, there are experts in methodology
available for consultation (See section on Secondary Thesis Advisors). Be
sure to obtain such a consult early. There is little that methods can offer
once the data have been collected.
- Start with a good question. In general, "good" or
"excellent" is best defined as a question for which the answer matters either to
other researchers in the field, practicing clinicians, or patients. Remember, you will
likely do just as much work to answer a question which has "below average"
interests to others as you will in answering a question which others will call important.
- Convert the question to a hypothesis by asserting a position.
This will lead directly to a consideration of measures, both of exposure and outcome.
- Generate measures of exposure and outcome. This step is
facilitated by a review of the pertinent literature. How have other researchers defined
/measured the exposure and/or outcome? The effort to generate meaningful measures will
generally require a return to the hypothesis for refinement, and narrowing (i.e., express
the hypothesis in terms of the specific exposure of interest, and the specific outcome
anticipated).
- Once a reasonable hypothesis developed, a protocol should be
constructed. How can the hypothesis be tested? The first requirement is that a comparison
be made. Here, too, a search of the literature for methods is apt to be helpful. Choosing
the right control group is challenging, and subtle. Once the comparison group is chosen,
the magnitude of expected difference should be estimated, as a basis for determining
sample size. In clinical research, standard methods include cohort studies (prospective,
or retrospective; the randomized controlled clinical trial is a sub-category of the
prospective cohort study) that assemble groups on the basis of exposure/intervention, and
follow for outcome; and case-control studies that assemble groups on the basis of the
outcome and assess for previous exposures. We urge you to use one of the three.
- In light of the design you deem most appropriate, revise your
measures of exposure and outcome as required. For example, in a prospective study, you can
choose how to measure factors of interest, whereas in a retrospective study you will need
to rely on measures obtained in the past, or the subjects' recall.
- Once your measures are established, determine the appropriate
sample size and the methods of analysis. A plan for data collection and management should
also be developed. Consultation with a statistician may be helpful at this stage.
- The research should only begin after steps 1-6 are dealt with
successfully. Again, there are no methods that can transform a vague question and data
into a methodologically rigorous study after the fact. Good methods must come first.
|