Randomised control trials Flashcards
What are RCTs?
RCT are randomised (people are randomly allocated to groups), controlled (there is a control group) and trialed (tested effect)
In general a well conducted RCT will provide the highest level of evidence. Depends on the quality of the study
RCT randomly assign exposure, then follow up to find out outcomes, intervention (experimental) and control groups are randomly allocated
What is randomisation?
Randomisation of individuals in study groups is the only way to ensure that all of the groups are as similar as possible at the start of the study
We want the intervention and control groups to be as similar as possible - exchangeability - so we can just look at the effect of the intervention without any other aspects that may effect a person
What is confounding?
Confounding factor - something that ‘muddles’ up the effect you’re seeing between the exposure and outcome. Key threats to the validity of the study
Why should we use randomisation?
If enough people randomly allocated, should have the same proportion of the confounder (e.g. age) in each group. This applies to known and unknown confounders. Successful randomisation means confounding is an unlikely reason for differences in outcomes between the groups as there is a equal chance for each participant to be in either group
How do we know if randomisation is successful?
If the randomisation is successful then the characteristics of the group will be about the same for the treatment and control group, the differences we see will be due to chance
Randomisation likely to be more successful with a larger sample population
What is random selection?
Random selection is where people are randomly selected from the source population to be in the investigation
What is cluster randomisation?
Cluster randomisation is where groups/clusters are randomised, useful if you have an intervention best done at a group level (schools, households, GP practises)
What is intention-to-treat analysis?
Intention to treat analysis preserves the benefits of randomisation. Means the participants are analysed into the groups they’re randomised into regardless or not if they stuck to that group. Analysed as randomised - groups kept as they were intended, can more accurately reflect real-world effect of intervention
What is per-protocol analysis?
Per-protocol analysis is where people are analysed according to what they actually did e.g. changing groups, benefits of randomisation as lost as people aren’t analysed in the groups they were introduced to. Analyse as treated, tends to be more appropriate for efficacy trials
What is concealment of allocation ?
Important that allocation sequence is concealed and unpredictable otherwise could introduce bias
What is blinding?
Important way to protect against bias, questions how might knowing which group of the study a participant was in affects the researcher or participant. Can be challenging as it can sometimes be obvious which group a participant is in
What is loss of follow up and non-adherence?
Loss of follow up - similar problem as in cohort studies
Non-adherence - participants don’t always do what you ask them to do
Strengths of RCTs
Randomisation maximises the likelihood that potential confounders are evenly distributed between groups
Best way to test an intervention
Can calculate incidence
Temporal sequence is known
Strongest design for determining causation
Limitations of RCTs
Not all exposures are suitable to be randomly allocated (some cannot be randomly allocated, some are very difficult, some should not be random)
Need to have clinical equipoise (genuine uncertainty about benefit or harm of intervention)
Can be difficult to achieve blinding
Potential for loss of follow up and non-adherence
Resource intensive
Highly selective (often highly selected group of people involved in an RCT, this can affect generalisability)
