Lecture 10 (RCTs) Flashcards
RCT (def)
-Subjects randomly allocated to treatment or comparison grp, followed to compare outcomes
Omenn et al: Beta C and Vit A on lung cancer/cardiovasc disease (CARET)
- Conclusion: didn’t reduce risk, may have increased it
- Don’t know which one though
Alternatives to randomization in exp study
- Historical comp grp
- Non-randomized concurrent comp grp ( systematic allocation or matching)
Systematic allocation in exp study
- Allocation according to predetermined rule
- Can –> diffs in confounders even w/o manipulation
Matching in exp study
- Selection such that grps are similar w/respect to potential confounders
- Doesn’t work for unknown confounders (validity of study depends on this)
Randomization
- Allocation by chance (each has fixed prob, can’t be predicted)
- Pros: no selection bias, defends against confounding w/large n
- Can also be stratified randomization
Blinding (purpose)
- Prevents info bias –> diff misclassification
- More subjective the outcome, more important
Benefits of placebo
- Blinds participants and researchers
- Psych benefits
- Helps keep participants assigned to comp grp
2x2 factorial design
- Evaluates 2 diff treatments in single study
- n needed = considerably smaller
Avoiding effect mod in 2x2 factorial design
- Treatments have independent mechanisms of action
- Treatments act on 2 separate outcomes
Albanes et al: a-tocopherol, beta C on lung cancer
- Double blind, placebo
- Baseline: very similar characteristics
- Conclusions: neither reduced risk, Beta C may have ^ it (stopped before end)
Crossover design (RCT)
- Upon completion, switched to other treatment
- Each patient serves as own control
What does non-adherence do in crossover design
-Biases estimate towards Ho (no effect)
Adherence in CARET
-15-20% of certain grps stopped, probably –> underestimation of effect
Run-in period
Prior to randomization, assessing adherence to determine who will be included
-Increases prob that subjects will adhere
