Study Designs Flashcards
Clinical Trial/Randomised Controlled Trial
> often used to compare new treatment to control.
gold standard for assessing effectiveness and safety of treatments
What are Complex Intervention trials
> similar to clinical
instead assesses non-drug treatments
psychological or educational etc.
Long-Term Follow-up
> Usually follows a clinical trial
less controlled than actual trial
long-term outcomes
Record Linkage Study
> uses routinely collected data to compare treatments and outcomes
Usually based on routine care
not as strict as clinical trial
Cost-effective
can be lots of variability/missing data
Observational Study
> collecting data on patients throughout their care, without any intervention.
Real-world data
cheaper
less controlled
Cohort Study
> recruit a cohort of a similar demographic anf follow over time
if for an extended period, can collect invaluable data
Case-Control Study
> recruit people with a disease and match to similar people without the disease
Ask them about previous exposures in their lives, related to the disease
retrospective; unreliable
Epidemiological Study
> large population-level studies of disease
assessing transmission and risk factors
Survey/Questionnaire
> sample of people recruited to report their views on a particular subject.
Clinical trial phase 1
> for cancer studies, terminally ill patients
otherwise, healthy volunteers
safety and dose-finding
very small
Phase 2 clinical trial
> safety and dose-finding
pilot/feasibility
relatively small
to assess whether larger trial can be done:
1.recruitment
2. study procedures
3. follow-up
Phase 3 clinical trial
> efficacy and safety
definitive
large
strictly monitored
randomised, controlled, blinded
Clinical trial designs
- superiority
- eqiuvalence/non-inferiority
3.parallel arm
4.crossover
5.cluster
6.stepped-wedge
7.adaptive designs
superiority
> is intervention better than control?
control can be placebo or standard care
H0 = no difference
Equivalence
> intervention effectiveness must be within a certain range of the control
two-sided confidence interval
non-inferiority
> intervention must not be significantly worse than control
more common
one-sided CI
used if new intervention is easier to use or cheaper.
parallel arm
> two+ treatment groups
randomised to experimental or control group
usually blinded
repeated measures
between-group analysis
Crossover clinical design
> treatments are not expected to cure
patients receive 2 treatments, one after the other.
washout period = carryover effects
patients followed up
within patient analysis = less variability
Cluster clinical design
> usually parallel arm
clusters, rather than individuals, are randomised.
between group analysis, adjusted for cluster.
avoids contamination between groups.
stepped-wedge clinical design
> all patients start on control
patients randomised to which wave they start the treatment in.
usually done for logistical reasons
individually or cluster randomised
time may be confounding factor
adjust for cluster and wave in analysis
clinical adaptive designs
> usually parallel arm
dose finding; eliminate less effective does
safety: high toxicity = reduce does or end study.
complex and not often used
why do we use controls?
> avoid placebo effects
adjust for change over time; changes not due to treatment will be present in both groups
why do we randomise patients to treatments?
> ensure balance
avoid placebo effects (equally experienced by both groups) and bias
enable attribution of treatment effect
methods of randomisation
> computer program
blocking
stratification/minimisation
phone/web-based randomisation system
