Section 1
Preview this deck
What are the disadvantages of not blinding?
Front
| 5 | 0 | ||
| 4 | 0 | ||
| 3 | 0 | ||
| 2 | 0 | ||
| 1 | 0 |
Active users
0
All-time users
0
Favorites
0
Last updated
6 years ago
Date created
Mar 1, 2020
Cards (161)
Section 1
(50 cards)
What are the disadvantages of not blinding?
Participant bias. Drop outs if people are not happy with their assignment. Dropouts will decrease denominator (C and D if placebo) (A and B if tx)
Is ABABABABAB... assignment random?
No, because the allocation of the next participant is predictable
What are adaptive procedures?
Changing allocation probability as a study progresses
When can accidental bias arise in RCT?
If random procedure does not achieve balance in risk factors or prognostic covariates
What are disadvantages of blocked randomization?
Analysis is more complicated. Investigator may know assignment. Standard statistical analysis assumes simple randomization.
What is the aim of RCTs that use double blinding?
To try to establish efficacy of a treatment
Fixed allocation randomization
Assigns participant with pre-specified probability (usually equal). Probability is not altered throughout the study.
When does stratified randomization work best?
In small studies. With minimal strata.
Allocation concealment
Not disclosing to participant or others the allocation sequence. Always feasible.
When is it okay not to have equal allocation to groups?
Ex. 2 to 1 to gain patient information. This does reduce sensitivity and may indicate that one treatment is better.
What is an option when it is difficult to blind an investigator?
Sham intervention. May be expensive.
How can simple randomization be accomplished?
Through random digit tables, for example. Algorithms are used for larger studies.
Triple Blinding
participants, data collectors, and data analysts are blinded
Single Blinding
Either the patient or investigator is unaware of the treatment they are receiving. Only one.
What are problems associated with sham interventions.
Sometimes not ethical. Can be costly. Can be time consuming.
Sham intervention
Ex. CPAP, sham surgery, sham machines
What study design is the gold standard?
Randomized Control Trial
Fundamental point of chapter 7
A clinical trial should ideally have a double blind design in order to limit potential problems of bias during data collection and assignment. In studies where such a design is impossible, other measures to reduce potential bias are advocated
When might no blinding be used?
During trials that address surgery, lifestyle changes, learning techniques
Is non blinding protective against concomitant treatments?
Nope
What are double blind studies usually restricted to?
Drug and biologic trials
What are the two types of adaptive procedures?
1. Adjusting according to imbalance in numbers of participants or participant characteristics. 2. Adjusting based on responses of participants to previous allocation.
When can selection bias arise in RCT?
If allocation process is predictable
What are the 3 types of fixed allocation randomization?
1. Simple 2. Blocked 3. Stratified
How does blocked randomization work?
For each block of an even number, half are assigned to 1 treatment and the other assigned to the other. This can be based on permutations of blocks or assigned numbers order.
Double Blinding
Investigator/ those collecting or assessing data and patient do not know of treatment assignment.
What are advantages of stratified randomization?
Makes populations comparable. Increases power. Reduces variability.
What is the appeal of non blinded studies?
Investigators are more comfortable in knowing who is assigned what. Simpler to execute. Less expensive. More accurately reflects clinical practice.
Ethical issues with debriefing of placebo
1. Patient benefiting from placebo may relapse. 2. Patient may have distrust for MD. 3. Patient may have other negative consequences.
What are the advantages to single blinding?
It is simple, and it may allow for better judgement of outcomes
Double Dummy
The use of matching placebo to ensure blinding, when all patients receive multiple formulations (drug and placebo); i.e., active tablet plus placebo liquid compared to placebo tablet and active liquid.
What are the disadvantages of single blinding?
Investigators may affect the outcome of the study
Unblinded studies
Both patient and investigator know to what a patient is assigned
Inadvertent unblinding
One knows what treatment is based on differences in meds, side effects, etc.
Systematic error
Difference between true value and what was found due to causes other than sample variability
What are two common methods for baseline adaptive randomization?
Coin randomization and Urn randomization
When is it hard to blind an investigator?
When implementation of intervention makes it obvious to which arm a participant is randomized (ex. CT v Xray).
What are disadvantages to simple randomization?
At any point there can be imbalance between groups. Reduces ability to find differences in groups. May lead to loss in credibility.
What is another name for simple randomization?
Complete Randomization
How can an investigator bias a study?
Could want good results. Could want to give concomitant treatment. May not believe that there was equipoise between the trial arms.
How does stratified randomization work?
Prognostic factors are measured. Strata are created. Randomization is performed within strata- either simple or blocked.
What is the benefit of triple blinding?
Results can be evaluated more objectively
Randomization
The process by which each participant has an equal chance of assignment to treatment or control. Used to create two or more relatively comparable study groups.
What is concomitant treatment?
Any non study treatment administered to participants during a trial. It may induce bias if applied unequally among groups.
What are advantages of simple randomization?
It is easy to implement
Fundamental Point of Chapter 6
Randomization leads to produce study groups with respect to known and unknown risk factors, removes investigator bias in allocation of participants, guarantees statistical tests will have valid false positive error rates
Allocation bias
Systematic difference between participants and how they are allocated to treatment
What study design offers the best method for comparability and statistical inference
RCT
What are the advantages of blocked randomization?
Avoids imbalance in groups. Increase in power. More comparable groups are produced. Study will be okay even if recruitment is terminated early.
Biased coin procedure
Try to balance participants in each group based on previous assignment, without accounting for participant response
Section 2
(50 cards)
What are early phase studies?
Phase 1 and Phase 2 clinical trials
Phase 1 Toxicity
Test tolerance to new agent. Assess what a toxic dose is. Estimate MTD.
Effects of gaming the system
Can seem as beneficial intervention even in absence of actual benefit if occurring in a trial with negative outcomes. Can fail to detect benefit of a truly beneficial intervention.
Selection bias
distortions that result from procedures used to select subjects and from factors that influence participation in the study
Per consort statement
A technique used to prevent selection bias by concealing allocation sequences from those assigning participants to intervention groups until assignment
How may investigators anticipate random assignment?
1. If investigator knows baseline status of participants. 2. Investigator can figure out next allocation arm because a block is too small or doesnt alternate.
Central Randomization
A randomization method used in multicenter trials to obtain the desired overall distribution across treatment arms. An outside 3rd party provides allocation.
Phase 4 studies
Broad population use and uncommon adverse events. Effect on whole population. Effectiveness and value in practice. Studies done after intervention has been marketed. Designed to monitor effectiveness of the approved intervention in the general population and any adverse events with widespread use. Can be observational or RCT
What happens if those without outcome are moved form intervention to control?
B to D in 2 by 2 table. Will make denominator smaller, biasing toward the null.
Phase 2 studies
Doses and biomarkers, efficacy of treatment. To evaluate if there is any biological activity or effect. Does drug work at MTD? Compares current control group, historical controls, or pre treated patients. May use different doses or arms. Have more exclusion criteria than P3 trials. Used to decide further development of treatment.
How to avoid gaming the system?
Do not use sealed envelope. Be wary of run in periods that allow participants to identify medications. De post facto, check balance in known potential confounders between arms.
Efficacy and Explanatory Trials
What an intervention accomplishes in an ideal setting.
Random selection
Used to draw a representative sample from a target population.
How can patient expectancies influence outcomes?
1. When blinding fails, we do not know if the outcome is due to the treatment or due to patient expectancy. 2. Patient can make placebo have an effect, which makes treatment look ineffective. 3. Lead to different treatment responses.
How can we avoid differential loss to follow up?
Focus on retention and adherence
Optimal randomization methods
Central computer system. Central call in system. Back up in case systems go down.
Effectiveness and pragmatic trials
What intervention accomplishes in actual practice. Accounts for participants who are not fully compliant.
Why do we use allocation concealment?
To prevent researchers from influencing which participants are assigned to a given intervention group.
Local randomization
A schedule is sent to each site, the individual sites randomize themselves
Are early phase studies a type of clinical trial?
Yes
What is a case control study?
A type of retrospective observational study in which participants are selected on the basis of presence or absence of disease/event/condition of interest. Not truly prospective.
R1 calculation
a/a+b
Crossover trial
a modification of parallel design that uses each participant at least twice
Limitation of case control studies
Treatment is not random, so associations are always influenced by confounders
What are necessary components of clinical trials?
Must employ 1 or more intervention techniques. Should be applied to change an outcome. Must have a control group that must be similar in relevant aspects to the treatment group at baseline so that outcomes can be attributed to treatment.
What are problems that arise when implementing randomization?
1. Bad method choices. 2. Design or programming error in implementation. 3. Human error in conduct.
RR calculation
R1/R0
What is a control in a clinical trial?
Best current therapies available. No treatment. Placebo. Standard of care.
Anecdotal case reports
the weakest form of study
Compensory Rx
If investigator wants to give those in control arm an intervention. i.e. prescribing another treatment This will bias the trial by making the denominator smaller, bias toward the null. Move from C to D.
Fixed randomization
Uses predefined schedules to assign treatment
What is population in phase 1 studies?
Generally health volunteers or people who have tried other therapies.
Fundamental point of chapter 1
A properly planned and executed clinical trial is the best experimental technique for assessing effectiveness of interventions. Contributes to identification of possible harms.
What are late phase studies?
Phase 3 and Phase 4 clinical trials
When are phase 1,2,3,4 studies used?
Usually for pharmaceutical studies. Some studies may blend among phases.
Allocation sequence
The order in which participants are allocated to treatment
Phase 1 studies
Usually to assess drug tolerance, interaction, metabolism, toxicity. Focus on drug activity and distribution. Find maximum tolerated dose (MTD) by dose escalation. Test a small amount of people (20-80ish) to determine efficacy and evaluate safety.
Phase 3 studies
Usefulness and safety. Usually longer follow up (unless for chronic conditions). Determine efficacy in large group (hundreds to thousands) by comparing intervention to other standard or experimental treatment. Monitor adverse effects.
Dynamic randomization
Use baseline information on participants, require knowledge of past assignments to make new assignments
Fundamental point of chapter 5
Sound scientific clinical investigation almost always demands that a control group be used against which the new intervention can be compared. Randomization in the preferred way of assigning participants to control and intervention groups.
Differential loss to follow up
Bias that occurs when you lose follow up long term.
How do blinding and allocation concealment relate?
Allocation concealment involves not disclosing to participants the allocation sequence before it occurs. Blinding involves not disclosing to participants and outcome assessors the treatment allocations after it occurs.
R0 calculation
c/c+d
What is a clinical trial?
A prospective study comparing the effects and value of interventions against a control in human beings. Start at time 0.
Can early phase studies be controlled and uncontrolled?
Yes
Why do we use clinical trials?
It is the best method to determine effectiveness and safety. Determine adverse events
Accidental Bias
When randomization fails to balance groups. Usually an issue with small sample sizes. Not usually a problem with large sample sizes.
What happens if those without outcome are moved from control to intervention?
D to B in 2 by 2 table. Will make numerator larger, bias away from the null.
How many people are usually in a phase 2 study
A few hundred. Less than in phase 3
How can participants game the system?
If two sites are closely situated. Participants who arent happy with randomization assignment may travel to other site and reenroll.
Section 3
(50 cards)
Specifying your research question
1.Define question in advance 2. Must be capable of being seen in all participants 3. Unless combined response, participation ends when primary response occurs 4. Response variable should be capable of unbiased assessment 5. Variable should be ascertained as completely as possible
Key assumptions in equivalency/ noninferiority study
1. Proper control arm must be standard treatment, not inferior. 2. Constancy in time and among participants. 3. Availability of data for prior control studies. 4. Assay sensitivity to see true difference, response must be sensitive to effects of control and intervention
Crossover v Case Crossover
Crossover is a clinical trial. Case crossover is an epi study in which there is no intervention- observe participant behavior before illness and use it as a control.
Strengths of historical controls
All participants can receive intervention. Participant may be more willing to enroll with treatment. Recruitment time halved.
Equivalency/ Noninferiority trials
Demonstrate new intervention is not worse than a standard by a predefined margin
Historical controls
New intervention used in participants, results are compared to outcome in previous series of comparable participants. Non randomized and non concurrent. Obtained from medical charts and literature.
Large Simple/ Pragmatic Clinical Trials
For common conditions, can uncover benefits and risk in short term intervention in large population
Large simple trial
Provide clinically relevant data because done in a clinical setting. Common conditions, important outcomes, easily implemented in large.
Advantage of group allocation design
Helpful if concerned about contamination
Assumptions of crossover designs
Must assume effects of first intervention do not carry to the second period
Chapter 3 fundamental point
Each clinical trial must have a primary question. Questions must be carefully selected, clearly defined, and stated in advance.
Disadvantage to non randomized concurrent control studies
Potential non comparability. Bias to allocation.
Equivalency study
Test whether new intervention is equivalent to an established one
Comparative effectiveness research
Compare 1 agent against another. Compare 2 or more interventions commonly used. Commonly non inferiority.
Concerns with factorial design
Possible interaction between intervention and impact on sample size. Added complexity. Impact on recruitment and adherence.
Superiority Trial
Assess whether treatment is different from control (usually treatment proven to be effective)
Pharmacodynamics
what the drug does to the body
Noninferiority trial
Whether an intervention is not worse or as good as an established intervention
Intervention
Manipulation of the subject or subject's environment to modify processes or endpoints
Hybrid designs
Used if a lot of data is historical. Smaller numbers allocated to control and more to the new treatment. Historical data must be fairly recent
Limitation of withdrawal study
A highly selected sample is evaluated. Anyone with adverse events would be taken out of study. Design can overestimate benefit and underestimate toxicity.
Criteria for large simple trial
Unbiased allocation of participant treatment and control. Unbiased assessment of outcomes. Large samples
Biomarkers and Surrogate response variables
It is less expensive/ time consuming to get continuous outcomes rather than events. Acceptable for serious disease , quick dispersion, early phase studies.
Clinical Trials
A research study in which 1 or more human subjects are prospectively assigned to 1 or more interventions (which may include placebo or other control) to evaluate the effects of those intervention on health related biomed and behavioral outcome
Factorial trial design
Employ two or more independent assignments to treatment or control. Possible to have non complete factorial design.
Advantages to non randomized concurrent control studies
do not have researcher pick how patient is treated. Can be easier than convincing patient to randomize
Subgroup analysis
It is likely that the effect of an intervention is the same across subgroups, can have exceptions.
Pragmatic/ Practical Clinical Trials
Conducted in clinical practices rather than in labs. Ask questions relevant to practices. Broad applications. Depend on easy intervention and outcome.
Does stratifying always help?
Nope. can always have unmeasured confounders that we did not adjust for.
Margin of indifference/ non inferiority
Relative risk of new intervention should be close to 1, meaning no difference. 1.2-1.4 is usual. Means 20-40% inferior to standard and still considered equal or noninferior. 1- control < delta
Role of historical controls
Fast Inexpensive Broad scope
Group allocation designs
A group is randomized to intervention or control.
Pharmacokinetics
what the body does to the drug
Assay sensitivity
Ability to show difference if one truly exists
Non randomized concurrent control studies
Controls are participants treated without new intervention at the same time intervention group is treated. Participants allocated to a group, but not randomly
Variations of group allocation designs
Crossover Modified crossover Stepped wedge
Crossover designs
Special type of RCT in which each participant is own control. May have wash out period. Any trend from first period to second can be eliminated in estimate of group difference in response.
Non inferiority/ Equivalence trial
Evaluate if new treatment is no worse than established treatment by margin (delta).
Outcomes
pre specified goal or condition that reflect the effect of 1 or more interventions
Chapter 4 fundamental point
Study population should be derived in advance, stating unambiguous inclusion criteria. Impacts design, generalizability, and participant inclusion.
Withdrawal study
Type of parallel design in which all participants start with treatment, then randomly some continue or stop treatment. Used to evaluate duration of benefit of intervention that is known to be useful
Two main types of exclusion
1. Patient with contraindications to intervention 2. Assign issues impacting adherence
Limitations of historical controls
Vulnerable to bias Shift in population Method of selection may differ Shift in diagnosis criteria Accuracy/ completeness of data issues
Main points of supplemental lecture 2 readings
Not all people know about the study design they are going into, think theyre just getting treated. Need equipoise to run study. No methods to track if people actually understand informed consent. Should make understandable to all.
Equipoise
State of uncertainty on parts of investigator about therapeutic merits of each arm in trial
Composite outcome
Combination response variable with 2 kinds of events. Can be used if 1 event is too infrequent.
Advantages of factorial design
Can be informative and efficient
Superiority trial
Goal to show new intervention is better than placebo or control alone
Study population
Subject of population with condition or characteristic of interest defined by eligibility criteria
Advantages of crossover design
Allows assessment of each participant on all treatments. Variability is reduced because participant is used twice. Smaller sample sizes
Section 4
(11 cards)