Transcript Lectures 14 Experimental Methods

Experimental Studies
Types of Experimental Studies
• Multiple experimental groups
• Blinds
 single, double, triple
Public Health & Clinical
Objectives
• Modify natural history of disease and
express disease prognosis
Prevent or delay death or disability
Improve health of patient or population
• Need to use best preventive or
therapeutic measures
Randomized trials are ideal design to
evaluate effectiveness and side effects
of new forms of intervention
Historical Perspectives
• Sir Francis Galton (1883) - ruminated
over the influence of prayer
• Joyce and Welldon (1965) found no
benefit of prayer
• R. C. Byrd (1988) - suggested
positive benefits
• Washington Post Parade article
(2003) - also suggested positive
benefits
Recent Perspectives
• Effect of:
coffee on CHD
carotene on cancers
hormonal therapy on breast cancer
drug-lowering cholesterol on CHD
Randomized Trials
• Historically, were done accidentally,
in other words, “unplanned trials”
Ambroise Pare (1510 - 1590) discovered
new treatment for war wounds when
original therapy was unavailable
James Lind (1747) studying scurvy
• Subjects assigned to groups using a
non-biased procedure
Design of a
Randomized Clinical Trial
Selection of Subjects
• Well-designed
• Eliminate subjectivity
• Promote reliability
Replicable, as with laboratory
experiments
Accurate
Selection of Subjects:
Studies without Comparison
• Question: If we administer a drug
and the patient improves, can we
attribute the improvement to the
administration of that drug?
• Answer: Results can always be
improved by omitting controls.
- Prof. Hugo Muensch
Harvard University
Selection of Subjects:
Studies with Comparison
• Historical controls (comparison group
from past)
Data must be abstracted from records not
kept for research purposes
Differences may be due to quality of the data
May not be able to substantiate differences
Can be useful for drugs developed against
fatal diseases
Selection of Subjects:
Studies with Comparison (cont.)
• Simultaneous Non-Randomized Controls
 May introduce bias
Example - BCG vaccination study in NYC in
1975
• Investigators introduced selection bias in the
experimental group and controls
• A change in the study design that eliminated
selection bias, although still not randomized,
also eliminated differences observed in final
results
Selection of Subjects (cont.):
Randomization
•
•
•
•
Best approach
Uses tables of random numbers
Must still eliminate physician bias
Can achieve non-predictability
Effect of Comparability
Not
Randomized
Randomized
Selection of Subjects (cont.):
Stratified Randomization
• Useful when concerned that certain
variables may affect the outcome
 For example, when the prognosis may be
much worse for older patients
• Want two treatment groups to be
comparable in terms of the variables of
concern
• Initially stratify (layer) the study
population according to each variable of
concern and then randomize participants
to treatment groups within each stratum
Selection of Subjects (cont.):
Stratified Randomization
Data Collection on Subjects:
Potential Variables
• Treatment:
 that was assigned
 that was received
• Outcome
 Explicit criteria required
 Comparable measurements required
• Prognostic Profile at Entry
 If risk factors for a bad outcome are known,
assure that treatment groups are reasonably
similar for these factors
 Data for prognostic factors obtained upon
enrollment in study
• Masking (Blinding)
Data Collection on Subjects (cont.):
Masking (Blinding)
• Attempt to eliminate biases & preconceptions
• Single-blind
 Subject masking
 Use of placebo
• Double-blind
 Subject masking and researcher masking
• Data collectors and data analysts
• Triple-blind
 Subject masking, researcher masking and study
sponsor masking