Clinical trials for the treatment of Heart disease Flashcards
What are the key design features of clinical studies?
Study designs:
- Observational studies - Observing, not interfering with exposure
- Cross-sectional
- Case control
- Cohort (longitudinal)
- Experimental - Do interfere with exposure
- Clinical trial/interventional study
- Randomised controlled trial
What are the strengths of randomised controlled trials?
- If well conducted, difference in outcome b/w intervention and control groups will reflect either:
- Casual relationship
- Chance/random error
What interventions can be tested in experimental studies?
- Medical treatment
- Surgical treatment
- Public health interventions (e.g. healthy lifestyle interventions)
More difficult interventions:
- Long-term interventions (loss to follow-up, expensive, participant burden)
- Interventions to change ‘lifestyle behaviours’
Why are control groups necessary?
- Because some patients will get better by themselves; needed for comparison
- In practice control groups:
- Sometimes receive no active treatment (placebo only)
- Sometimes receive usual care (i.e. existing pattern of treatment, but w/o new intervention)
Why are placebo control groups necessary?
- To take account of ‘placebo effect’ the benefit obtained from receiving apparently helpful treatment (even if treatment has neutral effect)
- Placebo needs to be appropriate (e.g. medical or surgical like the treatment)
How are participants allocated to groups?
Allocation to intervention and control groups needs to be random i.e. every participant has an equal chance of being in each group
Randomisation can be done by:
- Toss of (balanced) coin
- Random number tables
- Computer generated codes
- External randomisation service (CTU)
NOT allocating systematically or by day of week admitted etc.
Why is random allocation important?
- Best way of ensuring characteristics of patients in intervention and control groups are similar - aiming for balanced baseline differences (such as age, sex, disease, severity, disease duration)
- Avoids Bias - specifically allocation bias
- Any differences in health outcomes at end of trial are due to intervention and not differences in characteristics in intervention and control groups
What is blinding?
Keeping participants unaware of which treatment they have been assigned to
This can apply to:
- The patient
- The outcome assessor
- The statistician/analyst
Patient OR outcome assessor blind - Single blind
Patient AND outcome assessor blind - Double blind
Why is blinding important?
Blinding of observers/assessors avoids:
- Selection bias (who to invite/recruit to trial)
- Assessor bias (how outcome is recorded)
- Keeps assessors objective (decisions to withdraw patients/ amount of encouragement)
Blinding of participants avoids:
- Selection bias (whether to participate in trial)
- Response bias (systematic differences in how outcomes reported by participants)
- Can affect compliance and completion of participants
Describe cross-over design trials
Why are parallel and crossover group designs different?
Parallel group design = independent groups → Interested in between group differences
Crossover design = same individuals: Paired data → Interested in within-person changes
What are the advantages and disadvantages of the cross over trial design?
Advantage:
- Need considerably fewer participants (as they acts as their own controls and within-individual comparisons are made)
Disadvantages:
- Cannot be used to assess treatments with prolonged effects (which may carry over into the second period)
- Disease may not remain stable over trial period
- More burden on participants, longer trial period, could result in more drop-outs
Where may parallel and crossover designs be used?
- Usually applied to trials involving individual patients
- Could also use them to randomise groups containing many individuals (cluster randomised trials)
- General practices
- Primary care trust
- Primary schools
- To examine impact of community intervention
Why is an intention-to-treat analysis used?
- Not all patients will do exactly what they’re supposed to
- Some control patients will up being on treatment (often most severely ill)
- Some intervention patients end up coming off active treatment (often most severely ill)
How does an intention-to-treat analysis differ from a per-protocol analysis?
Per protocol analysis could provide biased estimate of treatment effect
Describe intention-to-treat analysis
- Analysis on basis of original randomisation (ignores drop outs/crossovers), so preserves randomisation at baseline (avoiding bias)
- Likely to underestimate effect of intervention (depending on how much lack of adherence there was)
- Provided unbiased estimate of intervention effect - gives an indication of treatment policy rather than specific effect (more relevant for prevention)
Describe per protocol analysis
- Carries out analysis on basis of treatment actually taken
- Less likely to underestimate effect of intervention
- May be biased (depends on number and characteristics of drop-outs)
How is relative risk calculated?
Dividing risk of event in low dose by risk of event in high dose
How is absolute risk difference calculated?
Minus risk of event in low dose by risk of event in high dose
How do you estimate the number needed to treat?
NNT = 1/ Absolute risk difference
NNT - inverse of absolute risk difference, i.e. if 9 given low dose over high, there would be fewer than 1 haemorrhagic event
