Clinical Trials in Respiratory Disease Flashcards
In epidemiology, how is a given clinical question typically framed?
P - population I - intervention C - comparator / Control O - Outcome T - Timing
- These are the parameters that we are both looking for and by which that validity is measured.
How is evidence found for clinical trials?
Finding the Evidence:
- Literature Search
- Papers are ranked in terms of their reliability and ranking in the scientific community.
- In order to determine if that paper is appropriate to the situation at hand (65yo man), we look at the following elements.
Explain the concept of internal validity?
Internal Validity:
- extent to which the results of a study are valid (accurate, robust etc.) for the sample of patients being studied.
- How well did the study answer the question it set out to answer?
- How well was bias and confounding dealt with during the study.
- Dependent on appropriate study design, data collection and data analysis.
What are the two main mechanisms for dealing with bias and confounding?
- Randomisation
2. Blinding
Give some details regarding the method of randomisation.
Randomisation:
- random allocation of subjects into each arm
- objective: treatment groups identical in all aspects other than the intervention
- even distribution of potential confounders (even those that are unknown)
- primary rationale: reduce confounding
- also reduces selection bias
- for example: tendency of investigators to assign subjects a particular intervention based on bias
- Stratified Randomisation: randomisation stratified by levels of key confounders.
- for example: subjects randomised within stratum to which they belong (eg, country and smoking status)
- purpose: make composition of groups even more similar with respect to key confounders, and hence further reduce potential for confounding
- Judge whether the randomisation was successful by analysing the table provided to give indication of the general features of those participating (are they evenly matched in terms of confounders?)
Explain features of blinding.
Blinding/Masking:
- non-awareness of intervention allocation
- single-blind: subjects unaware
- double-blind: + investigators
- ‘triple-blind’: + outcome assessors
- rationale: reduce information bias
- prejudice about the intervention can influence the outcome or its ascertainment
What are the features surrounding Objective Outcome Ascertainment?
Objective Outcome Ascertainment:
- outcomes determined according to strict, standardised, objective criteria
- multi-centre studies: centralised process
- rationale: reduce information bias
What are the reasons for completing an Intention-To-Treat Analysis?
Intention-To-Treat Analysis:
- assume subjects remained in group to which they were randomised, regardless of actual treatment received, drop-out, loss to follow-up or cross-over.
- rationale: reduce selection bias
- subjects who drop-out, cross-over etc are almost always systematically different from those who don’t
Example: Selection Bias During Follow-up in a RCT
RCT of drug versus placebo:
- sick subjects cease new drug due to side effects
- selects for healthier group in drug group, which experiences less outcomes
- misperception: new drug better than placebo
- always under-estimates the treatment effect (provides conservative estimate)
- reasons:
– less treatment in intervention group than assumed
– more treatment in control group than assumed
What is involved in Interpretation of Statistical Analyses?
Interpretation of Statistical Analyses:
- Statistical significance - p value, confidence interval
- Precision - confidence interval
- Clinical significance
What is the significance of the p value in statistical analysis?
P value
- derived from statistical analyses
- probability that the observed result arose from chance
- that is, there is truly no difference between the groups being compared, and the observed difference was simply a chance finding
- conventional cut-off = 0.05
p
What is a confidence interval in statistical analysis?
95% Confidence Interval
- interval within which there is 95% confidence that the ‘true’ value lies
- if the null value is excluded, result is stat significant
- null value: value if there was no difference between the groups being compared
- null value: 1.0 for ratios (eg HR, RR, OR) and 0 for differences (eg absolute risk differences)
How is precision measured statistically?
Precision
Confidence Interval
- CI also provides indication of the precision of result:
- narrower CI - more precise result
- wider CI - less precise result
- width of CI dependent on sample size of study
- bigger sample size - narrower CI
- smaller sample size - wider CI
- The more power you have, the less likely you are to get a type II error. (Increase sample size)
What is meant by non-statistically significant results?
Non-Statistically Significant Results:
- possible explanation: lack of power
- studies are designed so that there is sufficient power (usually 80% or 90%) to be able to detect a specified difference between the groups in terms of the primary outcome.
- sample size determines power
What is meant by NNT (number needed to treat)?
Number Needed to Treat:
- number of people needed to undergo intervention in order to prevent the outcome in one person.
- marker of the efficiency of the intervention
NNT = 1 / absolute risk or rate reduction
- Affected by:
– relative effect (often constant)
– underlying likelihood of outcome
- Example from TORCH:
– risk(placebo) = 15.2%; risk(S+F)= 12.6%; HR 0.83
– absolute reduction = 2.6%; NNT = 38.5
- hypothetical exmaple:
– risk(placebo) = 1.52%; risk(S+F)= 1.26%; HR 0.83
– absolute reduction = 0.26%; NNT = 385
What is meant by the concept external validity?
External Validity (Generalisability): - Seek concordance between the RCT and the clinical setting in terms of: P - population I - intervention C - comparator / Control O - Outcome T - Timing - There is no substitution for bedside care
