Lecture 21 - randomised controlled trials Flashcards
Randomised controlled trials
Analytic study
Intervention studies
Experiment, do something observe the effect
Could the parents’ decision to vaccinate their children have influenced the findings of the study?
Yes
How do we know if randomisation worked?
- Percentages similar
- Age in both groups same
- Gender in both groups same
- Number of medications a day is same
- Balanced similar
Random selection
Randomly select people from source population to become a sample
Randomisation
Already have sample
Randomly assign treatment or control
cross over studies
Each person gets both treatments and control
confounding is effectively eliminated
only be done for long-term conditions and treatments that are not curative (the treatment effect needs to wear-off during the washout)
“per protocol” analysis: Confounding could occur
Lost benefit of randomisation
If participants (or researchers) know which treatment they are on, they may act differently
May affect outcome
Blinding an important way to avoid bias
Making treatment unknown to people
Many exposures should not be randomised:
• Known harmful toxins or procedures
Unethical to do
Baby cot death
No equipoise
Unethical’
how might this introduce bias?
In terms of study design
Find in effects vs treatments
Pharmaceuticals can make their drugs seem more appealing
But disadvantages
Time and money
It can be difficult to achieve blinding
Equipoise (enough uncertainty)
Generalisability? – does it reflect the real-world?
Hierarchy of evidence
whats at top?
RCT
intervention study
RCT
non RCT
analytic study
Essential elements of an RCT
Participants randomly allocated to groups
There is a comparison (control) group
Testing effect of treatments/interventions
What if they didn’t randomise?
Things that may have
influenced parents
cost
community outbreak
polio in family member
What if they didn’t randomise?
Things that may have altered the risk of polio
Polio was more common in wealthy
A community outbreak ? Outbreaks often missed areas
Likely to have been exposed already
in RCT What determines exposure?
confounding factors
factors that determine the exposure may also affect the outcome
In RCT Why do we randomise?
People who decide to take a treatment are often different to those who don’t: confounding
• Age and sex
• Health risks
• Views of the health professionals treating them
• Health beliefs and habits
How do RCTs avoid confounding?
Participants are randomly assigned to intervention or control groups
Randomisation will not affect the outcome
If enough people are randomised,
should there be the same proportion of confounders in each group?
yes
Randomisation / Random Allocation
Both known and unknown
confounders should be balanced
Equal chance for each participant to be in either group (intervention / control)
Randomisation means
confounding is an unlikely reason for differences in outcomes between groups
Randomisation is not
Random Selection
Variants of randomisation
Cluster
Stratified or Block randomisation
may be difficult to randomise individuals so what do you do instead
randomise groups (clusters) of participants
Examples: GP practices, hospital wards, schools
If you want to be certain that important confounders are eliminated. what do you do instead?
Stratified or Block randomisation
Examples: randomise individuals within each age group, sex, or hospital
what is Cluster Randomisation?
Entire practices are randomised to treatment or control
All participants in each practice get the same intervention
GPs don’t have to do different things for different patients.
what is Stratified or Block Randomisation?
Participants are randomised to treatment or placebo in blocks (or strata) at each hospital.
Differences between hospitals will be balanced between treatment and control groups
what are Cross-over studies?
Each person gets both treatments - confounding is effectively eliminated
Can only be done for long-term conditions and treatments that are not curative (the treatment effect needs to wear-off during the washout)
how do you preserve the benefits of Randomisation?
Concealment of allocation:
Intention-to-treat analysis:
in RCT
why do you do Concealment of allocation?
Make sure that people can’t cheat and pick the treatment that they prefer
in RCT
why do you do Intention-to-treat analysis?
Once participants have been randomised, you don’t change the groups
why is it Important that allocation sequence is concealed and unpredictable of participants in RCT?
people could cheat the system and introduce bias
What happens if people withdraw from RCT?
Groups no longer similar
“per protocol” analysis: Confounding could occur
Protect the benefits of randomisation by
analysing people (treated and placebo) as they were randomised – as we Intended To Treat them – regardless of whether they followed the protocol
whats Intention-to-treat analysis?
Analyse participants as randomised
Reflects the ‘real-world’: people often don’t take treatments
Difficult if data are missing – you can’t analyse data you don’t have
whats Per-protocol analysis?
Analyse as treated (not necessarily as randomised)
Lose the benefit of randomisation
Can be appropriate for efficacy trials (does the drug work if you take it?)
Bias occurs when a study is conducted in a way that
leads to systematic errors
Potential sources of Bias
Lack of Blinding
Loss to follow-up
Non-adherence
whats lack of blinding
If participants (or researchers) know which treatment they are on, they may act differently
whats Loss to follow-up?
If people withdraw because of side effects, we may underestimate the harms of treatment
whats Non-adherence?
If people don’t take the treatment, we will not learn about its true benefits and harms
whats Blinding?
not knowing what treatment a participant was taking influence:
avoids bias
A research team member?
The participant?
The participant’s clinician?
if Participant knew (or guessed) that you were taking a placebo
• Would you be more or less likely to report a benefit from treatment?
• How about side effects?
may act differently
may affect outcome
what does “Single blind” mean?
participants
what does “Double-blind” mean?
participants and the researchers
Who is blinded?
- The participant
- The researchers who give the treatment
- The researchers who collect the data
- The data analyst
Blinding Important, but can be challenging to achieve in practice…
You can usually get a matching placebo pill
But what about surgical or physiotherapy treatment?
Safety and ethical concerns
Intention to Treat analyses:
Analyse participants in the groups they were randomised to
Loss to follow-up
You don’t know what happened (did they get better or worse?)
You can’t analyse data that you don’t have
Confounding and Bias may occur
Non-adherence
Participants often don’t do what you ask them to:
Non-adherence
Participants often don’t do what you ask them to:
- Only take some treatment or stop it altogether
- Don’t turn up for appointments
- Take alternative treatments (including the intervention or control)
If there is too much non-adherence,
difficult to interpret the study
Strengths of RCTs
- best study design to test an intervention
- Well conducted studies should eliminate confounding and bias
- can calculate Incidence, RR, and RD
- strongest design for testing cause-and-effect associations
Many exposures can not be randomised: such as
- Low birth weight
* Whether you develop a disease
Many exposures should not be randomised such as
Known harmful toxins or procedures
Many exposures are very difficult to randomise such as
- Long-term behavioural changes
* Common exposures in the community (how do you avoid them?)
what does Need to have clinical equipoise mean?
Genuine uncertainty about benefit or harm of intervention
what is unethical?
- Give known harmful interventions to people
- Give interventions known to be less effective than current treatments
- Waste resources and risk harm if we already know the answer
how can RCTs be very expensive?
- may need large numbers of participants
- ensure complete follow-up
- can take a long time
RCT Often funded by pharmaceutical companies:
- potential for big profits if the drug works
* unlikely to fund studies of cheap treatments with little chance of profit
Participants in RCTs are often not representative:
- They need to meet all the inclusion criteria
- AND be willing to participate
This can affect generalisability:
Like cohort studies, RCTs are not efficient for rare outcomes
Rare adverse drug effects are often not found by RCTs
usually found by post-marketing surveillance
Cohort vs. RCT
Cohort
- Ascertain exposure status, then follow-up to find out outcome(s)
- Observational study
Randomised Controlled Trial
- Randomly assign exposure, then follow-up to find out outcome(s)
- Interventional (experimental) study
RCT strengths
random allocation to exposure
- Low risk of bias and confounding
- Can calculate incidence, relative risks, and risk differences
- Best way to test interventions
Protecting Randomisation
Large numbers Conceal allocation Blinding Complete follow-up Intention-to-Treat
what do large numbers in RCT do?
Better balance of confounding between groups
what do Conceal allocation in RCT do?
Prevents cheating during randomisation process
what do Blinding in RCT do?
Reduces chance of bias during the study
what do Complete follow-up in RCT do?
Preserves randomisation groups
what do Intention-to-Treat in RCT do?
Preserves randomisation groups
hierarchy of evidence
from low chance of bias and confounding to high chance of bias and confounding
RCT cohort case control cross sectional case study ideas, experts, opinions, editorials anecdotal
what are some limitations of RCT?
Time and money
It can be difficult to achieve blinding
Equipoise
Generalisability? – does it reflect the real-world?
