Trials Flashcards
Experimental studies are the best methods to
Support cause-effect relationships
Differences with observational studies
Researcher introduces / manipulates the exposure
Much more validity if the assignment of the exposure is random and the sample size sufficiently large
Exposure doesn’t exist before starting the trial
Disadvantages of the experimental studies
Can only answer very limited range of questions (strict rules, no variability) about variety of doses, combinations of doses / exposures, long-term exposures, special effects in subgroups of peculiar patients
Features of experimental studies
Last steps in research
Measure efficacy (ideal circumstances) not effectiveness (real circumstances)
Types of trials
Clinical trials
Field trials
Cluster trials
Clinical trials
Participants randomly allocated are patients with a previous disease
Field trials
Participants are initially free of disease
Cluster trials
The unit of randomization is a group (it should be a large number size)
Phase I trials tests
If a new treatment is safe and look for the best way to give it
Phase I trials focus
Safety and tolerance
Phase I trials participants
20-80 healthy volunteers
Phase I trials features
No control group
Phase II trials test
Response of the disease to a new treatment
Phase II trials provide
Preliminary information on efficacy, dose-response and add information on tolerance
Phase II trials participants
100-200 patients with disease
Phase IIa trials
Pilot study
Small sample
Healthy and diseased
Phase IIb trials
Usually controlled
Test efficacy
Only patients
Phase III trials test
Is a new treatment is better than a standard one
Phase III trials provide
The definitive answer for efficacy and safety
Phase III trials features
Large sample size
Needs sufficient duration
Always controlled
Randomized
In medical journals
Phase IV trials provide
Information about long-term benefits and side effects
Phase IV trials features
Post-marketing
Designs of clinical trials
Cross-over trials
Factorial design
Cross-over trials
Factorial design
Aim of clinical trials
Explanatory / pragmatic trials
Superiority / equivalence / non-inferiority trials
Explanatory trials
Strict inclusion criteria
Ideal setting of “experimental” conditions
Pragmatic trials
Wide (lax) inclusion criteria
Real conditions
Superiority trial
Refutation of null hypothesis (both treatments are equal) leads to the conclusion that one of them is superior
Equivalence trial
Demonstrate that A and B are equal (efficacies are close enough from a clinical / practical point of view)
Non-inferiority trials
Demonstrate at least same efficacy
Only look to one bound of CI
One-tailed tests
Placebo effect
Sum of non-specific effect of the physician, non-specific effect of the drug and regression to the mean
Distingish placebo effect from pharmacodynamic effect
This makes it possible to substract the placebo effect from the efficacy attributed to the treatment being evaluated.
Randomization
Assigning participants to the different study groups by a random (unpredictable) mechanism
Randomization of a large number of subjects tend to produce
Identical groups with respect to known characteristic as well as unknown characteristics
The only between-group difference in randomized groups will be
Intervention to which subjects have been assigned -> all differences in outcomes will be specifically due to that intervention = strong causal inference
Features of randomization
Gold standard for showing causality
Eliminates biased assignment of treatments
Facilitates blind evaluation of the outcomes
Allows statistical test that assume randomly distributed differences under the null hypothesis
Conditions for randomization
Sufficient large sample size
Truly random assignment method
Masking of the random sequence (MRS)
Ransom sampling vs random allocation (randomization)
Random sampling: small subset (sample) is selected (drawn)
Random allocation: all the study subjects are distributed in groups
Problem in randomization
Conflict emerges between the clinician’s role and the aim of a clinical trial, and as a result, unintentional biases may occur: clinicians may be tempted to cheat (to help patients)
Restricted randomization
Random allocation can be made in blocks -> keep sizes of groups similar
Stratified randomization: 1º stratifying the whole study population into subgroups with the characteristics (strata) then simple random sampling from the stratifies groups
Masking of the random sequence (MRS)
Does not allow to guess to know what group will go the next patient (unpredictability).
This is done at the beginning of the trial.
Blinding of the intervention when assessing outcomes (BIWAO)
Investigators responsible for assessing or adjudicating the outcomes do not know the allocated group.
Particularly important when assessing subjective outcomes (pain or anxiety)
Intention to treat (ITT)
Analysis of the results of the trial that follows the initial random group assignment
Resembles the reality of clinical practice where patients not always comply
Biases the results towards the null value (potential benefit underestimated)
Per protocol (P-P)
Analysis that only includes compliers (advantage of randomization is eliminated
Any difference tends to be exaggerated
Better for equivalence studies
Integrim vs subgroup
Interim: repeated provisional analyses meanwhile the trial is on-going
Subgroup: analyses only including a subset of participants. If the effect is not present in the overall sample, there is not much justification to defend the effect within a subgroup
Both require lower p values (instead of p<.05) for achieving statistical significance.
Important consideration for randomized trials
Despite their high internal validity they have low external validity (low ability to generalize their results)