3. Overview
An experimental study assigns participants to
intervention and control groups in order to examine
whether an intervention causes an intended outcome.
Because the researcher assigns participants to receive a
particular exposure, the exact timing, dose, duration,
and frequency of the exposure are known.
5. RCTs
In a randomized controlled trial (RCT):
•Some participants are randomly assigned to an active
intervention group
•The remaining participants are assigned to a control
group
•All participants from both groups are followed forward
in time to see who has a favorable outcome and who
does not
6. Describing the Intervention
The research plan should carefully define:
•What the intervention will be
•Where and how participants will receive the
intervention
•When, how often, and for what duration participants
will receive the intervention
•Eligibility criteria for participants
7. Defining Outcomes
Most experimental studies are superiority trials that aim
to demonstrate that a new intervention is “better” than
some type of control.
Because the term “better” can be defined in so many
ways, the researcher must carefully define what
constitutes a favorable outcome for the experiment.
10. Selecting Controls
Experimental studies usually assign some participants to
the active intervention and the remainder to a control
group
•The most typical control is a placebo, an inactive
comparison that is similar to the therapy being tested
•Sometimes the new therapy is compared to some
existing “standard of care” or other therapy
•Sometimes varying doses and durations of a therapy
may be compared to one another
13. Hawthorne bias
Hawthorne effect: participants in a study may change
their behavior for the better simply because they know
they are being observed
This may interfere with the accurate measurement of the
impact of the new intervention.
14. Blinding
Blinding = masking = participants in an experimental
study do not know whether they are in the active
intervention group or the control group
•In a single-blind study, participants are unaware of
their exposure status
•In a double-blind study, neither the participants nor the
persons assessing the participants’ health status know
which participants are in the active and control groups.
Blinding minimizes information bias
15. Randomization
A variety of approaches can be used to randomly
allocate participants to an active intervention group or a
control group, such as:
•Simple randomization
•Block randomization
•Stratified randomization
17. Ethical Considerations
Experimental studies involve a particularly high level of
ethical risk because the researcher assigns participants
to exposures that the participants do not choose and may
have been unlikely to encounter in normal life had they
not volunteered to participate in a research project.
18. Ethical Principles
• Equipoise: experimental research should be
conducted only when there is genuine uncertainty
about which treatment will work better
• Distributive justice: infers that the source population
must be an appropriate and non-exploitative one
• Beneficence (do good) and nonmaleficence (do not
harm): researchers must balance the likely benefits
and risks of the study
19. Ethical Principles
• Respect for persons:
– Participants must volunteer for a study without
being unduly influenced by the prospect of being
compensated for their participation
– Participants must be able to understand what it
means to be a research subject, including the
possibility of being assigned to a control group
instead of the new intervention
20. FIGURE 13- 7 Examples of Ethical
Issues to Consider When Planning an
Experimental Study
21. Analysis
Experimental studies use many of the same measures of
association that cohort studies do:
•Relative rates (RRs)
•Attributable risks (ARs, AR%s)
•Measures of survival
Experimental studies use these measures to examine the
impact of an assigned exposure on the likelihood of
having either a favorable or unfavorable outcome.
22. Analysis: Efficacy
Efficacy: the proportion of individuals in the control
group who experience an unfavorable outcome who
could have been expected to have a favorable outcome
had they been in the active group instead
A high efficacy is an indicator that an intervention is
successful.
23. Analysis: NNT
Number needed to treat (NNT): the expected number of
people who would have to receive a treatment to
prevent an unfavorable outcome in one person (or,
alternately stated, to achieve a favorable outcome in one
person)
A small NNT indicates a more effective intervention.
24. Analysis: NNT
• If a drug is intended to prevent stroke and has an
NNT of 5, then 5 people have to take the drug for one
year (or some other specified time period) to prevent
one of the 5 from having a stroke.
• If a drug has an NNT of 102, it means that 102 people
have to take the drug to prevent one of the 102 from
having a stroke.
25. FIGURE 13- 8 Efficacy and Number
Needed to Treat (NNT)
26. Analytic Frameworks
• Treatment-received approach: limit analysis to the
participants who were fully compliant with their
assigned intervention
• Treatment-assigned approach (intention-to-treat
approach): includes all participants even if they were
not fully compliant with their assigned intervention
27. FIGURE 13- 9 Flow of Participants in
an Experimental Study
28.
29. Screening & Diagnostic Tests
• The goal of some studies is to compare two tests that
are supposed to measure the same thing, such as
comparing the results of an antibody test for cancer to
biopsy results
30. Screening & Diagnostic Tests
• Sensitivity = Of those who have the disease, what %
test positive?
• Specificity = Of those who do not have the disease,
what % test negative?
31. Screening & Diagnostic Tests
• Positive predictive value (PPV) = Of those who test
positive, what % actually have the disease?
• Negative predictive value (NPV) = Of those who test
negative, what % actually do not have the disease?
33. Tests of Agreement
• Tests of inter-observer agreement (also called
concordance) can be used to determine the extent of
agreement between two assessors who are evaluating
the same study participants
• For example, a measurement known as the kappa
statistic can indicate whether two radiologists
examining the same set of X-rays reach the same
conclusion about the presence or absence of a fracture
more or less often than can be expected by chance