Randomised Control Trials Flashcards
What are three sources of information for determining efficacy?
Experience of authority figures
Underlying theory explaining why something should be effective
Intervention studies
What are the three types of intervention studies?
Uncontrolled before and after
Controlled before and after
Randomised control trial
In an uncontrolled before and after trial, what factors other than the intervention can change the outcome?
Regression to norm
Seasonal factors
Policy, practice, management changes
What effect would regression to norm have on the results of the intervention?
If the intervention is implemented in response to a chance peak in rate of disease, and the rate than returns to normal, the intervention may look more successful than it is
What effect can seasonal effects have on the results of the intervention?
If this intervention is introduced in winter, when disease rates are higher, and the study concludes in summer, when disease rates are lower, the intervention may seem more effective than it is
What effect can policy/practice/management changes have on the results of the intervention?
Coincidental changes can result in a difference in data measurement between the ‘before’ period and the ‘after’ period
Give some examples of factors which could affect the outcome in a controlled before and after trial (other than the intervention)
Age
Sex
Comorbidity
Health/disease state
What is the problem with the grouping system in a controlled before and after trial?
The grouping is not randomised, so patients or doctors or researchers might be responsible for assigning participants to groups
What is the issue with patients choosing which group they go into in a controlled before and after trial?
Patients who select active treatment typically differ in health behaviours to those who select a control/placebo
What is the issue with doctors choosing which group participants go into in a controlled before and after trial?
Doctors are likely to put frailer patients into the control arm because it is known to be safe
Doctors are likely to put younger/healthier patients into the new treatment arm because they are more likely to tolerate it
What is the issue with researchers choosing which group participants go into in a controlled before and after trial?
Researchers are likely to put the healthiest patients/those expected to do best in the trial arm, so that the research results appear better
TWhat are four key features of RCTs?
Randomisation
Allocation concealment
Blinding
Intention to treat analysis
What is the aim of randomisation?
Balance groups to ensure equivalence of unknown and know prognostic risk factors
Which type of bias does randomisation aim to reduce?
Selection bias
What question does section A of the CASP tool ask?
Are the results valid?
What question does section B of the CASP tool ask?
What are the results?
What question does section C of the CASP tool ask?
Will the results help locally?
What three features must the outcome measure be?
Reliable
Valid
Responsive
What is meant by a reliable outcome measure?
One that can produce consistent, reproducible estimates of the true effect
What is meant by a valid outcome measure?
One that measures what it claims to
What is meant by a responsive outcome measure?
One that can measure changes in the construct to be measured over time
What ethical considerations must be made when randomising?
An individual should not be disadvantaged by being randomised to either arm of the trial
What are some examples of the choice of control in an RCT?
Usual care
No treatment, if that is the usual care
Placebo
Sham treatment
What is the definition of randomisation?
Allocation by chance of individuals or groups
Why is randomisation done?
To reduce risk of selection bias and confounding
What would ideal randomisation look like in terms of confounder distribution?
Equal distribution of known and unknown confounders between the two arms of the trial
What distribution pattern of confounder would cause confounding to take place?
Uneven distribution of known and unknown confounders between the groups
What is allocation concealment?
Ensuring that the recruiter does not known which group patients are assigned to
Why is allocation concealment needed?
Doctors’ beliefs about the suitability of an intervention for patients would influence their allocation
Patient preferences would influence their allocation
What would happen without allocation concealment?
There would be an imbalance in characteristics of patients between the study arms, causing confounding
Selection bias would not be removed
What is the purpose of the Intention To Treat policy?
To preserve randomisation and prevent allocation bias
Why is the Intention To Treat policy needed?
To deal with contamination/crossover of participants from one group to the other
What is the principle of the Intention To Treat policy?
Results are analysed according to the original group participants were assigned to, regardless of whetehr the intervention was received or not
What effect would the Intention To Treat policy have on the estimate of the intervention?
Intention To Treat may underestimate the effect of the intervention, but preserves randomisation
What secondary analysis can be performed if there is contamination/crossover of patients, in order to understand why this occurred?
Per Protocol analysis
What effect does a per protocol analysis aim to estimate, and which RCTs might it be used in?
The effect of adherence in RCTs with low adherence
Why might there be non-adherence to interventions?
Patients don’t like the intervention
Side effects
Patients unable to commit to trial
How does performance bias arise in the control group?
Patients receive the intervention anyway (contamination)
Patients seek complimentary therapies
HCPs give extra care to patients
How does performance bias arise in the intervention group?
The intervention group might receive more frequent exams and tests, because the intervention is new, resulting in falsely improved outcomes
How is performance bias reduced?
Use of blinding
Which types of bias can blinding reduce
Performance bias
Detection bias
How does detection bias differ from performance bias?
Detection bias is systematic difference in the way outcomes are reported, due to knowledge of allocation, whereas performance bias is differences in provision or adherence of the intervention itself due to knowledge of allocation
What is single blinding, and what effect does it have?
Blinds participants to what their treatment is, to avoid disappointment from their usual care, so behaviour does not change
What is double blinding, and what effect does it have?
Blinds participants and HCPS, so that compensatory treatments cannot be given to the control, or extra attention cannot be given to the intervention group
What is triple blinding, and what effect does it have?
Blinds participants, HCPs and statisticians, so that there is no bias in measuring/interpreting the outcomes.
What would introduce detection bias?
If the assessor of the results/statistician is not blinded, therefore they may exaggerate odds ratios, or may use different methods to collect or verify outcome information
When assessing outcomes, potential for bias is influenced by what three things?
Patients’ expectations
Researchers’ expectations
Statisticians’ expectations
What is attrition bias?
Dropout of participants who differ systematically from the remaining participants
Give three factors which contribute to attrition bias
Lost participants belonging to certain demographic groups
Lost participants due to outcome of sickness or death
Lost participants due to contamination of allocation, especially if participants are unblinded
How is precision assessed?
Based on the 95% confidence intervals
How would precision be increased?
Using a larger sample size
What are the components of a sample size calculation?
Expected effect
Variation in measured outcome
Significance level
Power
What does power refer to in an RCT?
The probability that the test correctly rejects the null hypothesis
Which two aspects of the intervention does a power calculation critique?
Whether the reported outcome from the intervention is better than the control or not
Whether the reported outcome is statistically significant or not
What questions need to be asked when applying the results locally?
How does the trial population compare with the real population?
What is usual care like?
What effect does a new intervention have on a population’s health behaviours and attitude?
When might a cluster RCT be used?
For RCTs with an increased risk of contamination
When might risk of contamination be higher in an RCT?
An situation when it is not possible to give intervention to 1 individual at a time