Randomization and blinding Flashcards
objectives of research design
-goal is to maximize internal validity
-“A well-designed study does not ensure good results,
a poorly designed study guarantees poor results”
Why use a control arm?
-establishes internal validity
-determines effect due to variables other than intervention
-detects variability in envirment
selection of control
-active control should be selected if possible
-ensure compatibility
- Active control often not selected due to expectation of
minimal difference
When is it ethical to randomize?
-clinical equipoise in practice
-randomization after consent
-minimizing of risk to participant
Stratification
-used to create comparable groups based on an important characteristic
-more important in smaller studies
Blocking
-helps ensure equal number of participants in each group
-investigators pick block size and sequence
Allocation concealment
-process of masking randomization from investigators
-used to prevent selection bias and inflated efficacy
Best method for randomization
-call center
-computer generated and centralized
Threats to randomization
-loss of allocation concealment
-use of non random techniques
-error in medication administration
Crossover design
-participants serve as their own control
-can only use for chronic conditions
-need adequate washout period
Factorial design
-can examine 2 research questions at a time
-subjects go through 2 randomization sequences
Masking
Method to prevent detection of group assignment
Why is masking important?
critical when outcomes are subjective and open to
interpretation by investigator or participants
single blind
either investigator or subject unaware of treatment
assignment (not both)
double blinds
both investigator and subject unaware of treatment assignment
triple blind
safety committee and statistician also blinded
* Often may not be blinded for safety or efficacy reasons
double masked
When different devices are used, or it is difficult to make
treatment arms identical, may need to use double-dummy
risks of compromised masking
-bias and flawed results
-attrition(drop outs)
