12 Experimental study designs Flashcards
Why do we randomize?
Balances known and unknown pronostic (confounding) factors
Eliminate bias
5 sources of bias in RCTs
Inadequate random sequence generation
Inadequate concealment of random allocation
Inadequate blinding
Incomplete data for participants’ outcomes
Violation of the intention-to-treat principle
What is adequate random sequence generation?
Can be a table of random numbers, a computer generated list of random numbers, a roll of a dice or flip of coin
What is adequate concealment of random allocation?
Can use a central randomization office, pre numbered or coded containers, or sequentially numbered, sealed, opaque envelopes
Crossover design
- Assumptions
- Advantages
- Disadvantages
- Chronic or recurrent condition, no carry over effect
- Patients act as own control (less inter patient variation), can have a lower sample size
- Cannot have carry over effect, need chronic disease state, cannot assess death and other major event outcomes
Factorial desgin
Evaluate 2 or more interventions compared with control in a single experiment
2 or more research questions/hypotheses
Randomized cluster/community design
Unit of randomization is a cluster
To avoid contamination
No need for consent in many areas
Issues with cluster design
Intra-class correlation coefficient p Need larger sample size
Intention to treat analysis
Analyze patients in the group to which they were randomized (irrespective of adherence)
Per protocol analysis
Analyze only those who were adherent
As treated analysis
Analyze all those randomized according to what they actually received
Type 1 error
When you conclude something works when it does not
Type 2 error
When you conclude there is no difference between two treatments when there actually is
