L5: Randomised controlled trials part 2 Flashcards
elements of design?
Use of placebo/control group
Blinding
Randomization
Early stopping
Parallel groups vs. cross-over trials
Testing for superiority/equivalence
use of a control group?
To obtain the information about mechanisms not related to
treatment
Use of non-treated controls may be unethical and
problematic
– Unethical if an effective treatment exists. If treatment available for cancer patient but you want to try a new one, if not effective or if placebo given you’re letting the person die.
– Problematic if the knowledge of treatment can affect
evaluation of treatment effect
use of a placebo?
Placebo effect
Measurable and clinically significant effect of placebo
Recognised in general, though still controversial
blinding?
Concealing the treatment identity to prevent bias in
treatment outcome evaluation.
Open-label trials: both patients and clinicians know the
assigned treatment
Single-blinded trials: the patient does not know the
treatment (but the clinician does)
Double-blinded trials: neither the patient nor the clinician
knows the treatment.
randomisation?
Eliminates all sources of bias accept accidental bias (accidental bias; chance imbalances with respect to confounders may result with randomisation)
Tends to ensure balance among treatment groups with respect to known and unknown confounders
E.g: equal men and women, equal distribution of age, for cancer distribution of cancer grades. Do not want high grade harder to treat cancers in one group and low grade easy to treat cancer in another.
Guarantees the distributional assumptions of the test statistics and estimators
Simple Randomisation vs Stratified
Stratification factor is a variable which divides the patient population according to its levels i.e gender, age levels
early stopping?
The Helsinki Declaration states:
“Physician should cease any investigation if the hazards are found to outweigh the potential benefits”
Trials with serious, irreversible endpoints should be stopped if one treatment is prove to be superior
Possibility for early stopping should be formalised in the design
Implementation of data and safety monitoring board
parallel groups vs cross over?
Parallel group – each patient is assigned to a group and each group receives an intervention
Cross-over – each patient receives an intervention in a random order
Cluster – pre-existing groups are randomly selected to receive an intervention
Factorial – each patient is randomly assigned to a group that receives a particular set of interventions
superiority and equivalence trials?
A significance test aims at rejecting the null hypothesis, not confirming it.
In superiority trials, rejecting the null means higher efficacy of one of the treatments.
In equivalence trials, due to the different formulation of the null, rejecting the null means similar efficacy of the treatments.
Superiority trials try to prove one is better
Equivalence trials try to prove they are about the same
3 main statistical concerns in clinical trials?
3 main statistical concerns in clinical trials
Results unbiased
E.g: ensure stratification etc.
Power sufficient
To answer hypothesis
Results generalisable
conduct of a trial and what data is published?
All trials follow a protocol which serves as a manual on how to conduct the trial
Trials are continuously monitored including:
Accrual
Toxicity
data toxicity
Possibility of early stopping in case of “promising results” needs to be formally planned
Case report forms- data collection points to fill in on patients ?
Only data relevant to the trial question(s) should be collected
Data or collected on Case Report Forms (CRFs) – electronically or on paper
Analyses are pre-planned – appropriate methods reflecting the design
publication>
Description of the analysis should be detailed enough to allow it to be replicated
Both positive and negative results should be submitted for publication
This it to avoid publication bias
Trial protocols are increasingly being published in peer-reviewed journals (eg. Trials)
ITT and AT?
- Intention-to-Treat (ITT):
Participants are analyzed in the group they were originally assigned, even if they didn’t follow the treatment or dropped out.
Why? It reflects the real-world situation, where some patients might not take the treatment properly.
Prevents bias by maintaining the benefits of randomization.
Used in most clinical trials.
- As-Treated (AT):
Participants are analyzed based on the treatment they actually received, regardless of their original group.
Can introduce bias because those who switch or drop out might be different from those who follow through.
Less commonly used, but some trials report both ITT and AT results.
Key Issue: Missing Data & Non-Compliance
If people drop out due to side effects or lack of efficacy, AT analysis could bias the results.
ITT analysis helps preserve the original randomization and provides a more conservative, realistic estimate of treatment effectiveness.
