RCT Flashcards
What are RCT used to evaluate?
Treatment effectiveness
They inform guidelines
Define the components of internal validity of a study
A causal relationship between the two variables must be properly demonstrated.
- Observation
- Explanation of the observation
Eliminate:
1) Bias - the way you assess the outcome could influence the outcome
2) Confounding - another factor in the trial affects it
3) Chance - could it happen by chance/fluke
If not: hypothesis is correct
What is external validity?
Is it applicable to my patient population? Generalisability
Should you treat? Four considerations
- Benefit worth harm?
- Resource costs - ?better use of resources
- Patient preferences
- Ethics
Name three types of intervention studies
1) Before and after intervention - assess the group before treatment and after
2) Intervention groups compared to control - assess before and after but compare to non-intervention group
3) RCT - groups allocated at random and compare to control
What is regression to the mean?
When you observe an extreme value this is partially due to chance. The next observation is likely to be closer to the mean -> can give the appearance of improvement.
Get a control group
What issues do ‘before and after studies’ have?
Assess a group of pts before and after - bus
bias - expectations things would be better
confounding - disease may get better on its own anyway
chance
What issues to intervention and control studies have?
They compare effectiveness with the control group which is good! but…
bias - pt or researcher beliefs can affect outcome measure
confounding - differences between the start groups?, were they treated differently?
chance
Define double-blinded
Participants and assessors do not know what group they’re in
What is randomisation?
Use chance allocation - any differences are allocated by chance.
Confounders are distributed by chance (both known and unknown)
What outcome measures are used in RCTs?
Clinical effectiveness - cure/recovery
Patient experience - QoL, pain score
What are the ethical considerations for RCTs?
Are pts being disadvantaged by being in the control or the intervention group?
What is clinical equipoise?
If you think one treatment is better (placebo or intervention) you cannot ethically randomise.
What are the choices for control?
Usual care
Placebo
Nothing
What is contamination in an RCT (crossover)?
Pts randomised to the control may accidentally receive the intervention or visa versa
How can contamination be managed in an RCT?
1) Analyse according to treatment received (but this causes allocation bias
2) Analyse according to intention to treat - preferred but may underestimate effect
Difficult in information studies
Can reduce by cluster randomisation
What is relative risk, how do you calculate it?
probability of the event in the treatment arm/probability of event in the control arm
What does a relative risk of 0.88 mean?
The risk is less in the treatment arm. The intervention reduces the risk of the measured outcome by 12%
How do you calculate the probability of an event in the treatment arm?
events/total number of people in treatment arm
How do you calculate absolute risk (risk difference)?
probability of an event in the treatment arm - probability of event in the control group
What does an absolute risk (difference) of -0.017 mean?
You are 1.7% less likely to die (experience outcome) on the intervention
What does a number needed to treat of 41 mean?
Treating 41 people would save 1 person from outcome
How do you calculate NNT?
1/absolute risk (ignore the negative sign)
What causes bias in assessing outcome?
pt, researcher and analysts expectations - ideally blind them
Can conceal allocation
