53 - Clinical Trials in CVD Flashcards
Why are clinical trials considered best to study causation?
Active change of exposure status.
Tightly-controlled study environment
Relative measures of intervention effect
1)
2)
1) Relative risks
2) Hazard ratios
Absolute measures of intervention effect
1)
2)
1) Absolute risk/rate reduction
2) Number needed to treat
Way to deal with confounders in a clinical trial
Randomisation
Way to deal with information bias
Blinding (single- or double-blind).
Reduces overseer bias.
Source of significant selection bias in clinical trials
Cross-over.
EG: sick patients drop out because they aren’t getting batter. Healthy patients stay in, so data appear to suggest greater effect of intervention.
Intention-to-treat analysis
A method of analysis for randomised trials in which all patients randomly assigned to one of the treatments are analysed together, regardless of whether or not they completed or received that treatment.
A way to minimise impact of cross-over
Difference between hazard ratio and relative risk
Hazard ratio applies to entire period of follow-up, is a weighted average of all measurements.
Relative risk is a risk at one point in time.
Survival analysis
1)
2)
3)
1) During follow up, note when occurrence of interest happens. Measure of interest is ‘time to event.’
2) Construct a Kaplan-Meier curve, which plots hazard or survival (1-hazard) vs time.
3) ‘Survival’ is avoidance of event of interest, not death.
Hazard
Continuously-updated ratio of event/no event
Hazard ratio
Hazard intervention:hazard control ratio.
EG: HR=0.5 means that at any point in time in follow up probability of outcome in intervention group is half that of in the control group
Risk/rate reduction
Change of outcome with exposure expressed in relative and absolute terms.
EG: Rate of outcome in control = 10/100
Rate of outcome in intervention = 7/100
Rate ratio = (7/100)/(10/100)=0.7
Absolute reduction = 10/100 - 7/100 = 3/100
Number needed to treat
Number of people that need to undergo treatment to prevent outcome in one.
Calculated by 1/(absolute risk or rate reduction)
EG: if absolute risk reduction is 3/100 per year, then 1/(3/100)=33.3. So 33.3 need to be treater/year to avert one case of outcome.
Big determinant of number needed to treat
How common condition is. If condition is rare, then a lot of people will need to be treated.
Name for NNT when intervention increases risk of developing outcome
Number needed to harm