Driving Clinical and Translational Research Using Biomedical Informatics - Translational Informatics: Realizing the Promise of Knowledge-Driven Healthcare - page 98

that do not involve direct and study-specifi c interventions (such as observational or
comparative effectiveness studies), and instead rely on data generated during the
course of standard-of-care activities. Therefore, for the purposes of this topic, we
will defi ne a clinical study (which can included clinical trials as well as non-
interventional investigations) as follows:

A scientifi c study in which investigators evaluate one or more aspects of the
effi cacy , safety , cost , and / or performance of a diagnostic or therapeutic
approach to a given disease state , or observe a disease state in order to bet-
ter understand its biological , clinical , and population level implications .

As implied previously, clinical studies are part of a broader research paradigm,
often referred to as translational research [ 7 ]. A common metaphor for this process
is that of bringing a novel therapeutic approach from “bench to bedside”.
Given their role as the “gold standard” for clinical studies, we now focus on the
design and conduct of clinical trials. Such trials can be divided into two primary
types: (1) observational studies and (2) randomized controlled clinical trials.
Observational studies are generally the result of reoccurring clinical phenomena
observed by investigators during the delivery of conventional care. These observa-
tions are then evaluated, generalized and reported upon [ 8 ]. In contrast, randomized
controlled trials are specifi cally designed studies that aim to answer pre-defi ned
scientifi c questions and minimize bias in study results [ 9 ]. The overall design of
randomized controlled trials can be further divided into three major classes [ 8 ]:

￿ Open Trials : Study investigators and participants know the type of treatment

being provided.

￿ Single - blind : Study investigators know the type of treatment being provided, but

participants do not.

￿ Double - blind : Neither study investigators nor participants know what type of

treatment is being provided. A set of records to allow for un-blinding during
subsequent data analysis is maintained by a trusted third party.

A key component of the randomized controlled trial model is the concept of
control , which refers to the comparison of the performance of an intervention in a
test group to the performance of a control group that does not receive the aforemen-
tioned intervention. These test and control groups can be further sub-divided based
on variations in the aims of the study. The process of assigning trial participants to
test or control groups is usually done in a random manner during a process known
as randomization. This random assignment is performed in order to reduce potential
biases in the study outcomes. The groups to which participants are assigned are
often called treatment or study arms.
The conduct of clinical trials can be further defi ned by study phases, where each
phase corresponds to a discrete scientifi c aim of the trial. A complete clinical trial
ideally consists of four phases [ 8 ]:

￿ Phase I : Investigators evaluate the intervention in a small group of participants

in order to assess overall safety. This safety assessment includes dosing levels in