what is an RCT
An experimental study where participants are randomly assigned to one or more intervention
groups or a control group.
* Designed to test whether an intervention causes a change in outcomes
why are RTCs gold standard
Highest internal validity for determining causality.
* Minimizes bias through careful design
types of RCTs by intervention goal
By intervention goal
* Efficacy: Can it work under ideal
conditions?
* Effectiveness: Does it work in real-
world settings?
* Superiority: Test if one intervention
is better than another.
* Equivalence/Non-inferiority: Test if
interventions are similar.
types of RTCs by how the intervention is delivered
Parallel: Groups receive different
interventions simultaneously.
- Crossover: Participants receive
multiple interventions in sequence. - Factorial: Tests multiple
interventions in combination.
Randomized controlled trials
parallel RCTs
Most common (~80% of RCTs).
- Participants randomized to one of two or more arms.
- Groups receive different interventions throughout the study.
- Suitable for long-term or permanent effects.
cross over trial
Participants receive multiple interventions in sequence.
* Includes a washout period to minimize carryover effects.
* Each participant serves as their own control.
* Best for short-term, reversible effect
factorial design
Participants randomized to combinations of two (or more) interventions and controls.
* Creates multiple groups (e.g., A only, B only, A+B, neither).
* Efficient for studying effects and interactions simultaneously.
* Less common design.
core design elements: intervention and control groups
well defined and comparable
core design elements: allocation method
randomization with allocation concealment
core design elements: outcome measurement
valid, reliable, and blinded assessment
core design elements: follow up
minimize loss to follow up, minimizes bias
core design elements: blinding
participants, providers, accessors - minimizes bias to allocation of intervention
core design elements: data analysis
have a pre-specified plan
The effects of an exposure can only be judged in comparison to what would
happen ___
in its absence
randomization
Allocation to treatment groups by chance,
avoiding predictability
* Equal probability for each participant to be
assigned to any group
Purpose of randomization
Prevent allocation bias
Achieve comparability between groups
(though not guaranteed)
Balance potential confounders
Ensure differences at trial completion are due to the intervention or chance
effect of randomization on confounders
Balances both measured and
unmeasured confounders
If randomization is correct, chance is
the intervention is the only explanation for observed
differences
units of randomization
Individuals
Health providers
Communities
two steps of intervention allocation
- generate the allocation sequence -> anyone has the same chance of getting into any one of the groups (randomization)
- conceal the allocation -> shields upcoming assignments from investigators/patients (could bias allocation of following people)
why concealment matters
Prevents manipulation of who gets the next assignment.
Ensures groups remain comparable.
More critical than randomization alone in avoiding bias.
Ways to generate allocation sequence
Simple randomization (flip a coin, table of random numbers, computer generated) -> doesn’t ensure balance in small trials
Restricted (Permuted) Block Randomization
-participants randomized in small blocks by fixing a ratio in each block
ways to conceal allocation
Methods:
* Sequentially numbered opaque sealed envelopes (SNOSE)
* Central randomization service
* Pharmacy-controlled allocation
* Always feasible in RCTs (unlike blinding).
Allocation Concealment vs. Blinding
Allocation Concealment
* Hides upcoming assignments during enrollment
* Prevents selection bias before
participant entry
Blinding
* Masks group assignment after
allocation
* Prevents measurement and
performance bias during trial
Randomized controlled trials
data analysis: Intention to treat (ITT)
Analyze as randomized
participants are included in their originally assigned
groups, regardless of compliance or protocol deviations.
Maintains benefits of randomization
