Epidemiology Lecture 1 -- Randomized Control Trials Flashcards
Define RCTs
The gold standard study design from which evidence on efficacy is assessed. As close to a scientific experiment as is possible in study of humans
3 reasons to do an RCT
1) To test an intervention in patients
2) Field or prevention studies in healthy patients (ie. vaccines)
3) Community trials where groups of people are studied
3 disadvantages of RCTs
1) Costly and time consuming
2) Infeasible in many instances (ethics)
3) Many questions of medical interest cannot be addressed (i.e. cannot withhold a therapy that is proven to be active)
How is randomization carried out and why
Use a chance mechanism (computer generated now) so that neither patient nor physician know in advance which therapy will be assigned
2 goals of randomization
Ensures that every person has the same and independent chance as any other person in receiving either intervention
Ensures than the intervention groups are similar on average at baseline
What is randomization highly dependent on?
Size of the study
What is the purpose of stratified randomization?
Ensure equal numbers of two groups when an important known and measurable risk factor is taken into account (i.e. males have more risk for CVD than women, so make sure study isnt skewed towards men by ensuring equal numbers of men and women in both groups)
What limits stratified randomization
Small sample size
Large number of strata
2 advantages of randomization
Eliminate conscious bias (physician and patient self selection)
Balance unknown biases among treatment groups (i.e. non-measurable/non-stratifiable factors or unknown factors affecting outcome)
2 disadvantages of randomization
Patient or physician may not care to participate in experiment involving a chance mechanism to decide treatment
May influence patient-physician relationship (i.e. if pressure to recruit)
Define the “control” aspect of RCTs
Manipulation and measurement of exposure (treatment)
5 examples of control in RCTs
Who gets what
The exact timing of when drug is administered
Control over when follow-up measurements are made
Same data collection from both groups
Same outcome assessment using validated tools
Purpose of control in RCTs
Provides the best means of avoiding bias and confounding, and quantifying the uncertainty in a statistical procedure
Define blinding
Process of hiding treatment allocation to people involved in the study in an effort to avoid bias (= impossible to distinguish between interventions)
4 levels of blinding
Treatment allocation
Patient
Investigator
Measurement of outcome
Levels of blinding most often involved in double blinding
Patient
Investigator
Levels of blinding most often involved in triple blinding
Patient
Investigator
Measurement of outcome
2 ways that blinding can be impractical
Side effects may be too obvious to blind
Increases pill burden
Define open-label
No blinding (i.e. if the drug is clearly unblindable)
Problem with using placebos
Increases pill burden when using active comparator (i.e. not testing how drug will actually be administered = issue for compliance)
Define equipoise
A state of genuine uncertainty about the benefits or harms that may result from each of two or more regimens. A state of equipoise = an indication for an RCT since there are no ethical concerns about one regimen being better for a particular patient
Define the reference population
Everyone who can benefit from the drug
Define the experimental population
Everyone you can access and recruit for the experiment
Define the study population
People who fit the criteria for the study
Define non-participants
People who do not fit the criteria of the study (i.e. have a comorbid disease)
NOTE: must be documented
When must compliance be ensured?
At recruitment stage and during entire study
How can compliance be ensured during the study?
Make it easy for the patient to comply, such as providing transportation, easily accessible prescription, professionalism/courtesy, provision of medical info during course of the trial, etc.
Some examples of how to measure compliance
Pill counts
Biochemical markers
Self-report
Define the Hawthorne effect
Participation in the trial may affect the outcomes since the patient may be so well-cared for that the effect is disproportionately positive
Who conducts interim analyses?
Data and Safety Monitoring Board (DSMB), who are blinded to treatment allocation
What is a criterion for early termination of a study?
A strong statistical difference between groups or no difference at all
4 aspects of a statistical analysis plan
Outline what is the primary endpoint to be studied
What tests will be used to define the endpoint
How missing data/visits will be handled
What secondary analyses will be done
4 statistical issues
1) Sample-size and power
2) Analyses using intention-to-treat (ITT) vs. per protocol (evaluable; PP) populations
3) Interim analyses monitoring (safety and/or efficacy; stopping rules)
4) Losses to follow-up and missing data
Define Type I error (or alpha)
The study results show an association, but there is none in reality
Define Type II error (beta)
Study shows no effect, but there is an association in reality
Define power
When there is an association in reality and the study shows it
Calculate power
1 - beta
Arbitrary % for sufficient power and what it means
80% = we accept a 20% chance of missing a true effect
Relationship between sample size and power
Larger sample size = greater power
Relationship between loss to follow up and power
Loss-to-follow-up must be taken into account in study design. May need to inflate sample size, for example by 2, if you expect a 50% dropout rate
Define efficacy
Measure of how active the drug is when taken according to exact protocol = maximum estimate of a drug’s potential activity
What population is included in measuring efficacy?
Includes only those who have adhered to the protocol (on treatment or pro-protocol analysis)
Define effectiveness
Measure of how the therapy will function in real life
What population is included in measuring effectiveness?
Everyone as they were originally randomized (thus preserves randomization)
2 types of outcomes
1) A proportion with disease or achieving a pre-determined target level of a surrogate marker for disease (i.e. reduced cholesterol level, HIV viral load)
2) Time-to-event or survival (i.e. for screening)
Define number needed to treat (NNT)
The number of patients who need to be treated in order to prevent one additional bad outcome
Calculate NNT
The inverse of the absolute risk reduction (ARR)
NNT = 1/ARR ARR = |CER - EER|
What do CER and EER stand for?
CER = control group event rate EER = experimental group event rate
What is a generally good NNT?
<10
What is a perfect NNT?
1 (everyone benefits from the treatment without any adverse effects; everyone gets better with treatment and no one gets better with control)
What is an example of an event rate?
Mortality rate (an adverse event)
Define phase 1 trials
First trial using human subjects (after in vitro and animal studies)
Study population of phase 1 trials
Generally done in healthy volunteers
Purpose of Phase 1 trials
Determine pharmacokinetics and pharmacological effects (basically to determine is the drug is safe)
Drug used in phase 1 trials
Often use preparation of medication that is not in final marketed form
Population of phase 2 trials
In a limited number of subject having the disease of interest
Purpose of phase 2 trials
To determine drug effectiveness and side effects (look for optimal dose to use in later trials)
Look at short term effects
Population of phase 3 trials
Larger number/spectrum of subjects with the disease of interest; generally ideal (i.e. study population as defined earlier)
When are phase 3 trials carried out?
After phase 2 and 1 have demonstrated effectiveness and short-term safety
Which phase is highly regulated by the FDA and HPB? Why?
Phase 3 since the drug is in marketed form. This type of trial is necessary for a drug or intervention to be licensed
2 ways to carry out a phase 3 trial
Compare new drug to no therapy (placebo)
Compare new drug with existing standard of care
Traditional question of a superiority trial
Is the new treatment better than no (or the old) treatment?
Goal of a non-inferiority trial
Show that new treatment is not worse than the standard treatment (active control) by more than a pre-specified amount (the “non-inferiority” margin), thus establishing an acceptable efficacy profile for the new treatment
Caveat of non-inferiority trials
For validity, non-inferiority threshold must be set BEFORE trial is initiated
When is phase 4 carried out?
Post-marketing
Purpose of phase 4
To obtain additional information than what was studied in phase 3 (i.e. different dosages, durations, diseases and subpopulations)
