Biomedical Engineering Reference
Table 9.10: Study Control Group Types (MHRA, 2011)
Control Group Type
Both groups, and all within, are
treated by the same person.
Gives the study good control but causes an
issue of transferability: Are the results only
applicable to the one clinician? Get over this
by using multicenter studies.
The control group does not receive
treatment from the same person as
the active group.
Less controlled but any differences may be
due to the different clinician.
The subject acts as the control
and crosses from control to active
group under their own volition but
following a transition protocol.
Only useful for devices that treat long-term
or chronic conditions. Subjects cannot
suddenly perform an operative procedure on
A retrospective study against a
group separated by time.
As stated previously, retrospective studies are
This is not a route to be advised; even though all statistical guides say they can
be used, most clinicians do not accept the “do nothing” approach. However, no
individual subject knows which group they are in, so you have to do your best to
make sure that they cannot identify which group they are in! In this way the subject
is effectively “blind” and has no influence on the outcome. Randomization is an
important aspect of a blind trial; subjects are put into groups at random and they are
not selected .
- Double Blind: A double blind study removes the influence of the investigator as
well. In this study nobody knows who is in which group. This is usually achieved by
having an external statistician who controls the subjects' ID and their grouping but
does not share this information. Only when they start to examine the data does the
grouping become apparent. Once again random selection is important. A double blind
trial is the gold standard in clinical investigations.
All of the categories above can be controlled or uncontrolled . Having a control group does
not control the study; it is only control by name, not control by design.
- Uncontrolled : The investigator lets everything happen by chance with no
consideration of the effects of any variables.
- Controlled : One or more variables 4 are controlled. This is normally controlled using
the exclusion/inclusion criteria. However you may wish to control factors that are not
directly subject related but are directly related to the operation of your device and its