L4 & 5 - Experimental Health Research (RCTs) Flashcards
What is maturation/spontaneous recovery?
People just improve over time with no direct intervention
What is the practice effect? (AKA learning effect)
Improvement on an assessment following exposure to one previously (the same or similar test)
What is the placebo effect?
The placebo effect is when a fake treatment is given, and a bigger improvement is seen in their condition/pain/QoL than no intervention/treatment
What is the dose response effect?
Increased amount of placebo = increased impact
What are key elements of placebo?
- Dose response effect
- Branding on the packaging
- Route of administration - e.g. IV over pill
When is placebo ethical?
- Not if you are withholding effective treatment
- If there is no proven, effective treatment
Why randomise participants to research groups?
- To avoid selection bias
- In large enough groups, randomisation is likely mean confounders are distributed evenly across both groups
What is selection bias?
Bias due to the selection of participants - it is not properly randomised, meaning the sample achieved doesn’t truly represent the populalation it aims to analyse.
How should the comparison variable be chosen?
- Use the proven effective treatment
- If one doesnt exist, use placebo
Why does allocation/assessment need to be blind to participants and researchers?
It is well established within research that assessment is biased if not blinded.
What is allocation concealment?
- Allocating participants to groups thorugh the use of a pre-generated random sequence
- Allocation concealment means researchers should not be able to predict which group a participant will be allocated to
What is the basic principle of randomisation in a trial as carried out by a statistician/CTU?
- The CTU generates a group allocation sequence (A, B, B, B, A, A, B, A, B, A, A etc) in advance, and allocates each participant accordingly to group A or group B as they enter the study
What is fixed randomisation?
Group allocation has a prespecified probability of going into which group (if its a 2 arm trial, probs 50/50)
What are the three types of fixed randomisation?
- Simple randomisation
- (permuted) block randomisation
- Stratified randomisation
What is simple randomisation?
- The computer generates a random sequence for all 30 participants
What are the potential problems with simple randomisation?
- Failure to recruit enough participants may lead to unequal group sizes
- Trials can be stopped early by the data monitoring committee
What is the problem with unequal group sizes?
- Reduces statistical power
Why may trials be stopped early?
- Trial can be stopped early by the data monitoring committee who monitor data as trial is ongoing to look out for serious, unexpected events
- Pre-specified rules as to when trial would be stopped:
- if treatment is so beneficial it’d be unethical to withold it from all patients
- if there is a clear and serious adverse events
- Interim analysis is difficult if/when group sizes are unequal
What is block randomisation?
- Randomises the participants into small blocks to prevent large differences in group sizes at any point
- Blocks should have an even number of participants
- The number of people allocated to each group within a block must be the same - e.g. all blocks have 4 people in them
- in 4-block groups, there would be a max difference of 2 if there was a failure to recruit
When is block randomisation useful?
- When the participants characteristics vary throughout recruitment
- Monitoring if the study needs to be terminated early
- Failing to recruit the target number of participants
What are the problems with block randomisation?
- Sometimes researchers and participants can work out which group they have been allocated to (e.g. if there is an anticipated side effect)
- If the length of the block/previous allocations are known, it could be worked out the group of person 4 in the block. This could affect recruitment
- Power needs to be high - the ability to detect a difference between the groups if one really exists.
What is permuted block randomisation?
- Block size is varied randomly to prevent predictability
- e.g. mix between 4/6
What is stratified randomisation?
- Obtain comparable groups
- larger N (sample size) means this is more likely, but can’t be guaranteed
- Randomly allocates people based on important risk/prognosis factors that may affect the outcome
- e.g:
- gender
- age
- severity of cmmn impairment
- e.g:
- Recruit equally to each sub group
Stratified Randomisation - Worked Example:
What is a clinical endpoint?
- A clinical endpoint is an event or outcome that can be measured objectively to determine if the results of the study are beneficial
What is a surrogate outcome?
- In clinical trials, a surrogate endpoint is a measure of effect of a specific treatment that may correlate with a real clinical endpoint but does not necessarily have a guaranteed relationship.
How to manage adverse events?
- Predict potential adverse events
- Monitor participants for unexpected adverse events - ensure they have an opportunity to feed back about this
Are adverse events possible in SLT RCTs?
- Adverse events are always a possibility
- e.g. PD study - negative emotional impact of therapy was linked to their preconcieved expectations
What is a primary outcome?
- The most important outcome of the study
- The outcome that the study is based around
- There should only be one
- Has a real world meaning to patients
- Should include measures of PRO/QOL (physical and/or social functioning) where possible
How do you calculate the sample size needed?
- Test varies depending on the planned statistical test
- need:
- Alpha (p value) - the probibility the study will find a difference if it doesn’t exist
- Power - the probability the study will find a difference if one exists
- Effect Size -
- Pooled standard deviation - measure of the spread of data
- need:
What is the alpha/p value?
- typically 0.05
- the liklihood we will find a difference if one doesn’t really exist
When do you do a chi-squared test?
- Categorical data
- Looking to prove a correlation
When do you do a t-test?
When the data is continuous, and normally distributed
When do you do a Mann-Whitney U-test?
Continuous data, not normally distributed.
What statistical test do you do for:
Categorical data?
chi-squared test
What statistical test do you do for:
Normally distributed, continuous data?
t-test
What statistical test do you do for:
Not normally distributed, continuous data?
Mann-Whitney U-test
What are quality criteria for reading papers?
- Was the allocation sequence randomly generated?
- Was the allocation sequence adequately concealed?
- Was there a blinding of participants where possible?
- Was there a blinding of outcome assessors?
- Was there a sample size calculation?
