Interventional Study Design Flashcards
Interventional Retrospective
Post hoc analysis
Interventional Prospective
Controlled trials
Uncontrolled trials
Interventional
Investigator ASSIGNS TREATMENT
Observational
Treatment is predetermined
Ultimate goal of clinical trials
Establish causality
If X then Y
If you give the pt this antibiotic then their infection will resolve
If you do not give the antibiotic then their infection will not resolve
Research Question
FINER
Feasible, interesting, novel, ethical, relevant
PRIOR to start of study
Background and significance
Why is it worthwhile? What is already known? What is unclear? What needs to be answered? Is it beneficial?
Placebo controlled trials indicated when:
No standard therapy available
Pts aware of placebo control and odds of receiving placebo
Placebo may be added to the standard therapy
Effective Controls
Proper matching
Same appearance, smell, taste
Comparable dose, regimen, duration
Unmasking: accidental or deliberate (safety concern)
Placebo effect
Some kind of intervention = real or preceived positive negative effects “nocebo”
Effects are also possible even if they know they are receiving placebo
Linked to EXPECTATIONS
Hawthorne effect (observer bias)
Subjects enrolled in a study change their behavior in response to the study itself (exercise when they normal don’t, etc)
Pygmalion effect (expectancy advantage)
Investigator’s expectations of an advantage in one group vs another affects subject response
Why Randomize
Minimize bias
Comparable groups
Uncertainty of benefits of a therapy
Randomization Methods
Fixed randomization methods
Adaptive randomization methods
Fixed Randomization Methods
Simple, block, stratified
Adaptive randomization methods
Allocation probabilities change as study progress in order to address imbalances that arise
Simple Randomization
Participants randomly and equally assigned
Simple and easy
Potentially imbalanced
Block Randomization
Avoids large imbalances from occurring throughout the study period
Can lead to selection bias
Double bind = no bias
Stratified Randomization
Smaller study
Stratify pts by variables of concern at baseline ensures comparability btwn grps for that factor
A: ensures even distribution, increases power, balancing only 2-3 baseline variables
D: Large numbers
Define Selection Bias
Sample not representative of the population due to pt selection methods
Define Sample Bias
Sample does not reflect the spectrum of characteristics in the target population
Define Informational bias
Differences in the way that information collected in study occurs
Define study design/analysis bias
Unfair comparisons
Fail to take into account confounders
Control bias
Randomization
Blinding
Parallel Group Pro
Well controlled
Well defined population
Bias and error limited
Well known statistical tools
Parallel Group Cons
Expensive
Need a large number of subjects
Volunteer bias
Loss to followup
Crossover Design Pros
Difference based upon within subject comparisons (less variability and fewer subjects)
Easier
Subjects must be on meds all times
Good for short-acting drugs, stable/chronic diseases, bioequivalent studies
Crossover Design Cons
Not useful for acute or permanent endpoints
Carry over effect
Exaggerated efficacy of second treatment
