Role of clinical trials Flashcards
What are the strengths of RCTs?
- can determine causality
- quantification of chance or random error
only if RCT has been well-conducted
BIAS and confounders are generally not seen in RCTs
What kind of interventions can be tested in experimental studies?
[experimental = RCTs]
many:
- medical Rx
- Surgical Rx
- health promotion advice
- dietary change
What are the more difficult interventions to test via RCT?
- long term interventions
- ‘lifestyle’ change (other than advice)
What features would the ideal experimental study have?
- (placebo-) controlled
- randomised
- double blind
- large
- analysed by ‘intention to treat’ method
Why do patients tend to get better (with or w/o Rx)?
- natural tendency to recover
- probability: regression to the mean seen for individuals at the extremes of distribution
- placebo effect
- effect of Rx
Why are controls needed?
- because some patients will get better by themselves, need a comparison group
Why is a placebo control group needed?
- take account of the placebo effect
What are the nature of control groups in practice?
- sometimes receiving no active Rx
- sometimes receiving usual care (without new Rx)
- may be receiving placebo for ‘active’ intervention
How are individuals allocated to each group in RCTs?
Allocation must be RANDOM
[ reduces allocation bias and confounders ]
Why is random allocation in RCTs good?
- avoids allocation bias
- avoids confounding
- simplifies interpretation of differences between groups
- facilitates blinding process
What are the issues with assessing outcomes in RCTs?
- vested interest in trial outcomes
ASSESSMENT BIAS”
subjective outcomes:
- patient-assessed
- observer-assessed
What is the solution to making outcome assessment in RCTs robust?
keep ‘key people’ blinded to randomisation code
- patient
- outcome assessor
- statistician/analyst
What are the different levels of blinding in RCTs?
SINGLE BLIND
=> patient or outcome assessor
DOUBLE BLIND
=> patient AND outcome assessor
TRIPLE BLIND
=> patient AND outcome assessor AND statistician
What trial outcomes may be used in RCTs?
- Sx
- clinical sign
- biochemical sign
- clinical disease onset
- death
- objectively measured ideally
- may have several outcomes
What are the 3 basic design for RCTs?
RANDOMISED
- parallel group design (SINGLE INTERVENTION)
- Crossover trial design
(SINGLE INTERVENTION) - Factorial design
(>1 INTERVENTIONS)
What is the crossover trial design of RCTs?
- v. efficient: every participant acts as both intervention and control
- more precise estimate of effect (than in parallel group)
- only works for RAPIDLY ACHIEVED and REVERSIBLE outcomes
When can crossover RCT design NOT be used?
NOT for irreversible events
e.g. cancer, DM, death
What is the factorial RCT design?
2 interventions
4 equal sized randomised groups:
- None (control)
- A
- B
- A+B (dual intervention)
efficient design
all groups are involved in examining effect of A and B
What are RCTs usually applied to?
individual patients
can also apply to groups using many individuals (= CLUSTER RCTs)
used to examine impact of community intervention
What is the importance of trial size in RCTs?
RCTs are widely used, but sample size is often too small
- limited statistical power
- fail to detect Type 2 stats error (failing to see intervention effect)
- trial effect estimates are imprecise
What are the advantages of using a large sample size in trials?
- higher statistical power
- more precise effect size estimate
- can look at sub-categories for impact of intervention
What does a large trial sample size NOT achieve?
does NOT make trial more representative at large
=> dependent on SOURCE of trial population
What are the presumptions that patients make on entering trials?
- they will not come to any harm
- will not be excluded from the usual effective Rx
- fully informed re: prospects of benefit/risk
What are the implications of trials in practice?
- intervention must benefit rather than cause harm
- control group: should have usual care rather than no care
- informed consent
- monitoring safety, adverse effects
What are the challenges in trials?
- not all patients will do what they’re meant to
- some control patients will end up being on Rx
- if analysis made on those who actually Rx: then imbalance of groups and biased estimate of Rx effect
What is ‘intention to treat analysis’?
- carried out on basis of original randomisation (crossovers are ignored)
- underestimate of intervention effect
- UNBIASED estimate of intervention effect
- this should be the main analysis in most situations
What is ‘per protocol analysis’?
- analysis based on Rx actually taken, takes crossover between groups into account
- less likely to underestimate intervention
- May contain BIAS (dependent on number of drop-outs)
- Generally a subsidiary analysis
What are the main considerations when analysis a trial?
- intervention effect?
- chance impact?
- limitations in trial design?
- consistent results?
- are results generalisable/representative?
- recruitment rate?
What are the main considerations in effective interventions?
- cost-benefit
- cost-effectiveness
should clinical practice be altered?
Which trials looked at whether tight BM control reduced DM complications?
UKPDS
recent T2DM Dx
randomised to Rx with either insulin or sulphonylurea or usual care
aim <6 mmol/L
outcomes: DM-relate, deaths and hypos
Which trial looked at whether cholesterol reduction with statin in DM reduced CVS risk?
randomised 10mg atorvastatin vs placebo
outcome: CVS disease after 4 years
Which trial looked at whether T2DM can be prevented in high risk individuals?
Finnish Diabetes Prevention Study
pre-diabetic middle aged overweight
randomised to I:
- weight reduction
- fat intake reduction- - fibre intake
- increased exercise
outcome: DM onset at annual OGTT
