lecture 6: study design i: clinical trials reading gordis - chapters 6, 7 lilienfeld and stolley -...
TRANSCRIPT
Lecture 6:Study Design I: Clinical Trials
Reading Gordis - Chapters 6, 7 Lilienfeld and Stolley - Chapters 8, 9
Types of studies
• Experimental: randomized clinical trials• Non-experimental (observational):
– Prospective (cohort) studies– Retrospective (case-control) studies
Objectives of randomized trials
• Evaluate therapeutic and preventive aspects of medical practice
• Evaluate new approaches to health care delivery
• Evaluate impact of health education on health behavior
Design of a randomized trial
Study Population
RANDOMIZATION
New Treatment Current Treatment
Not Improved
ImprovedImproved Not Improved
Reference population
Randomized control trials (RCT)
• RCT are generally regarded as the most scientifically rigorous method of hypothesis testing available in epidemiology.
• RCT have greatest control over the research setting– manipulation of study factor – random assignment of exposure
Study Population
RANDOMIZATION
New Treatment Current Treatment
Not Improved
ImprovedImproved Not Improved
Reference population
Considerations in selecting a study population
• Source (reference) population: the population for whom the results of the intervention are thought to be applicable.
• Target (study) population: the group in which the intervention is conducted
• Our objective is to generalize the results beyond the study population
Considerations in selecting a study population
• Internal validity is the ability to reach the correct conclusion in those actually studied– Absence of methodologic problems in the
study• External validity is the generalizability of the
study to the reference population– Do studies of physicians REALLY
represent the general population?
New Treatment Current Treatment
Random Allocation
Source population
Eligible
Consent toparticipate
Ineligible
Declineparticipation
Studypopulation
Subject selection: factors to consider in setting eligibility criteria
1. Potential benefits and risks of intervention(s)
a. Who would be expected to reap benefits from the new intervention?
b. For whom would the new intervention pose unacceptable risks? (Example: known allergy).
2. Need to think proactively about internal validity of trial
a. Ability of subjects to provide valid and reliable data: e.g., intact mental status, language fluency
b. Expected compliance with a regimen.
c. Low probability of dropping out: e.g., expectation of continued residence in community, absence of serious comorbidity.
3. Enhancement of the power of the trial
a. May restrict participation to a group with known high incidence of relevant outcome(s).
b. May restrict to a relatively homogeneous study population, in order to reduce variability in outcome.
c. May use matching to reduce variability in outcomes from sources other than exposure.
“Controlled trial of sotalol for one year after myocardial infarction”
Julian et al.(1982) Lancet 1: 1142-47
• A multicenter (25) double-blind randomized study.
• Treatment: 320 mg sotalol once daily compared to placebo in patients surviving acute myocardial infarction (MI)
• Patients between the ages of 30 and 69 years who had survived for five days after MI were evaluated for inclusion
• Treatment started 5-14 days after infarction
• Subjects were excluded (55% of those eligible for the study) for the following reasons:– heart block of >1st degree– heart rate < 54 per minute– women of child bearing years– history of asthma or obstructive airway
disease– IDDM– clinical evidence of infarction at 12th post-
infarction day
Julian et al.(1982) Lancet 1: 1142-47
• MORE exclusions!– Systolic BP persistently < 100 mm Hg– other cardiac or non-cardiac conditions
thought to be serious enough to worsen short term diagnosis
– lack of cooperation by patient or inability to follow up the patient for psychological or geographical reasons (!)
Julian et al.(1982) Lancet 1: 1142-47
New Treatment Current Treatment
Random Allocation
Source population
Eligible
Consent toparticipate
Ineligible
Declineparticipation
Studypopulation
Informed Consent
By federal regulations this means subjects must be told:
1. That they would be participating in research
2. The research procedures involved.
3. The potential risks and discomfort
4. The potential benefits (including those to society)
5. Alternative procedures of possible advantage to subjects
Informed Consent
And that’s not all….
6. Extent to which information gathered will remain confidential
7. Medical services available in the event of an injury, if more than minimal risk involved.
8. Whom to contact with questions about research.
9. Voluntary nature of participation and right to withdraw without penalty or loss of benefits
New Treatment Current Treatment
Random Allocation
Source population
Eligible
Consent toparticipate
Ineligible
Declineparticipation
Studypopulation
Randomization
• Randomization:– Equal probability of participants to be
assigned to intervention or control group– This is the basis for statistical inference
• By randomly assigning the treatment - reduce the risk of confounding because:“confounders - known and unknown” are randomly distributed among the treatment groups
• By blinding (masking) assessment of outcome - reduce the risk of information bias
Methods of randomization: simple random allocation
• Subjects assigned on entry to a treatment group based on some pre-determined allocation plan– table of random numbers– flipping a coin
• If the study is small, this method MAY result in differences in group size or composition by chance
Methods of randomization: stratified randomization
• Randomization increases the likelihood of comparability, but does not guarantee it
• Stratify by variables that are known to be important (such as age) and randomize within each strata
• Guarantees groups are comparable with regard to the important variable
Example: stratified randomization
1,000 Patients
600 males
400 females
360 young 240 old 300 young 100 old
180 + 120+ 150 + 50
= 500
180 + 120 + 150 + 50
= 500
New treatment Current treatment
Stratify by sex
Stratify by age
How would an investigator know if the randomization procedure was
successful?
Check baseline characteristics between the two groups. For example,
Sotalol group Placebo group 54% male 53% male95% Caucasian 96% Caucasianmean age = 54.5 mean age = 55.2
Blinding
• Blinding is when the observers and/or subjects are kept ignorant as to the group to which the subjects are assigned
• Single blind = subjects are ignorant• Double blind = observer and subject are
ignorant• Triple blind = observer, subject and analyst
are ignorant
Why use blinding?
• Use blinding to eliminate the effect of knowledge of treatment- this introduces bias– If a patient is aware of assignment,
knowledge of therapy may affect outcome (use a placebo)
– Observer may differentially recognize and record information regarding the patient based on the treatment group of the subject
What are typical comparison or control groups in RCTs?
1. Placebo
a. A biologically inert intervention which is used to elicit any non-specific psychological effects of the test intervention.
b. Designed to resemble the test intervention as closely as possible, except for the presumably active component.
2. Using a specific alternative intervention
a. For example, comparing Vaccine A versus Vaccine B.
b. Sometimes used to test a hypothesis of equal efficacy (or bioequivalence), possibly comparing a new intervention to the best existing alternative.
3. “Usual care” comparison group
a. Used chiefly in trials that are investigating methods of health care delivery or health education.
b. Requires careful attention to what is actually provided to the control group, so that findings are interpretable.
Sample size in clinical trials
• How many people do we have to study to obtain a statistically significant difference between the groups?
• Sample size is calculated before the study starts.
• Calculate the number that guarantees the power (1-) of the study.– See Gordis pp 110-115 for details
Planned crossover
Group 1
Group 1
Group 2
Group 2
Group 2
Group 1
Planned crossover
• Each patient can serve as his/her own control, thus holding constant the variation between individuals.
• Cautions:– Carryover effects: need a washout period
to eliminate the effect from one treatment to another.
– Order of treatments: may because of psychological responses
– Not feasible for some treatments, e.g., surgery
Unplanned crossover
Randomized
Surgicalcare
Medicalcare
Randomized
Refusedsurgery
Requiresurgery
Surgery Nosurgery
Surgicalcare
Medicalcare
Original design Reality
Unplanned crossover
• Analyze the unplanned crossover: intention to treat- according to the original randomized assignment.
Noncompliance (dropout)• Patients may agree to be randomized, but
they may not comply with the assignment treatment.
• The effect of noncompliance will be to reduce observed differences- non-ignorable missingness.
• How to know the non-ignorable missingness?– Record the demographic characteristics of
noncompliance– Compare compliers and noncompliers in
these characteristics
Phases in new drug testing in US• Phase I: Researchers test a new drug or treatment in a
small group of people for the first time to evaluate its safety, determine a safe dosage range, and identify side effects.
• Phase II: The drug or treatment is given to a larger group of people to see if it is effective and to further evaluate its safety.
• Phase III: The drug or treatment is given to large groups of people to confirm its effectiveness, monitor side effects, compare it to commonly used treatments, and collect information that will allow the drug or treatment to be used safely.
• Phase IV: Studies are done after the drug or treatment has been marketed to gather information on the drug's effect in various populations and any side effects associated with long-term use.