Randomised trials and other evidence Flashcards
To distinguish intention-to-treat and on-treatment analyses.
ITT = ANALYSED BASED ON THE GROUP ALLOCATED, NOT ON WHO ACTUALLY TOOK THAT TREATMENT OT = WHO ACTUALLY TOOK THE TREATMENT
Why are new treatments generally tested in randomised controlled trials?
To determine whether they are better than the existing treatments. To show small effects, studies must avoid bias.
What are single and double blinding?
Single blind – the participant does not know if they are in the treated or control group (a placebo is given)
Double blind – neither the participant nor the doctor
knows
What bias does blinding prevent?
Observer bias - avoids knowledge of treatment influencing:
- patient’s assessment of response
- doctor’s assessment of response
- patient’s decision to withdraw from trial
- doctor’s decision to withdraw patient from trial
What is a cross-over RCT?
Allocate half to group A (treatment group) and the other half to group B (control group). Group A has a treatment period, then control period and group B has a control period, then the treatment period.
When are parallel RCTs done?
When effect of treatment is curative e.g. – antibiotics for infection, chemotherapy for cancer
When are cross-over RCTs done?
When treatment is reversible – for example, analgesics for chronic pain (e.g. osteoarthritis), inhaler for asthma, drug to lower blood pressure
Advantages of cross-over trials over parallel group trials? (4)
Fewer participants
Avoids “scores” for quantifying subjective symptoms (e.g. pain)
Compares 2 drugs in the same individual (not one drug in person A and one in person B)
Can omit non-compliers in the analysis (because each participant is his/her own control)
Name two interventional trial study designs.
Parallel group randomised trial
Cross-over randomised trial
Name two observational trial study designs.
Cohort (or prospective) study Case control (or retrospective) study
Describe a cohort study.
Take a group of people
Categorise each according to the exposure (e.g. smoker or non-smoker)
Follow them up
Identify those who develop the disease
Benefits of cohort studies. (2)
Can generate useful summaries
Rigorous
Describe a case control study.
Take samples of people who have the disease (cases) and who do not (controls)
Determine whether the cases are more likely to have been exposed
Disadvantages of cohort studies. (4)
Long duration (many years)
Large size (thousands of individuals)
Need large numbers (tens of thousands)
Susceptible to confounding
Advantages of case control studies. (2)
Quick
Need small numbers
Disadvantages of case control studies. (2)
Susceptible to bias, especially recall bias
Susceptible to confounding
Advantages of RCTs. (1)
They eliminate bias
Disadvantages of RCTs. (4)
- Expensive
- Maintaining high compliance difficult
- They seldom exceed five years in duration (not well suited to long term effects)
- Unsuited to behavioural exposures (smokers v non smokers, take v not take the contraceptive pill)
What is the most appropriate study design to answer the question?
- Does vitamin D reduce the risk of hip fracture in elderly people?
Parallel group trial
What is the most appropriate study design to answer the question?
- Does the measles/mumps/rubella vaccine cause autism?
Case-control study
What is the most appropriate study design to answer the question?
- Does the contraceptive pill cause breast cancer?
Cohort study
What is the most appropriate study design to answer the question?
- Do margarines with added sterols or stanols lower serum cholesterol?
Cross-over trial
What is the most appropriate study design to answer the question?
- Does mental stress increase a person’s risk of myocardial infarction?
Parallel group - counselling v no counselling after MI
Cohort study - level of mental stress measured at outset
What is the most appropriate study design to answer the question?
- Does driving my car after drinking a pint of beer
increase my chance of an accident?
Case control study
Cross-over trial (with surrogate endpoint)
What is the meaning of P=0.05?
There is a 5% probability that this result (or one more extreme) could have arisen by chance if there was no genuine association at all.
What does P<5% or <0.05 mean?
STATISTICALLY SIGNIFICANT
It does not mean the association is 95% likely to be causal. It does not exclude bias. It only applies to a study testing a prior hypothesis.
What is a 95% confidence interval?
We are 95% certain that the “true” answer lies within the specified limits
The larger the number treated, the _______ the confidence interval.
Narrower
Do different causes of a disease usually interact additively or multiplicatively.?
Multiplicatively
