Transcript randomization - CardioGroup.org

RANDOMIZED
CONTROLLED TRIAL
Instructor: Fabrizio D’Ascenzo
[email protected]
www.emounito.org
www.metcardio.org
Role MD
CONFLICT OF INTEREST
None
AIM OF THE COURSE
A critical appraisal
- Theorical
- Practical
of RCT
SOME HISTORY
- 600 B.C.:Daniel of Judah compared the health
effects of the vegetarian diet with those of a royal Babylonian diet
over a 10-day period. (Book of Daniel 1:1–21)
-1952 The Medical Research Council trials on streptomycin for
pulmonary tuberculosis are rightly regarded as a landmark that
ushered in a new era of medicine. (Hill AB. The clinical trial. N
Engl J Med 1952; 247:113–119)
RANDOMIZED

It prevents selection bias and insures against accidental
bias.
 It produces comparable groups, and eliminates the
source of bias in treatment assignments.
RANDOMIZATION
 It permits the use of probability theory to express the likelihood of chance as a
source for the difference between outcomes.
 It facilitates blinding (masking) of the identity of
treatments from investigators, participants, and
assessors, including the possible use of a placebo
RANDOMIZATION
 „Produces groups that are not systematically different
with regard to known and unknown prognostic factors
 „Permits a valid analysis
 Permutation test is justified by randomization
 Standard analyses are valid approximations of the
correct permutation test
CRUCIAL CONCEPTS
PHASE
STRUCTURE
SUPERIORITY AND INFERIORITY
RANDOMIZATION
BLINDING
SAMPLE SIZE
AD INTERIM ANALYSIS
ITT VS AT
SUBGROUP ANALYSIS
PHASE
PHASE
Phase I trials
Objective to determine a safe drug dose
Design usually dose escalation/de-escalation
Subjects healthy volunteers or patients with disease
Phase II trials
Objective to determine a safe drug dose
Design often single arm
Subjects patients with disease
Phase III trials
Objective to compare efficacy of the new treatment with the standard
regimen
Design usually randomized control
Subjects patients with disease
STRUCTURE
STRUCTURE
Parallel group
Cluster randomized
Crossover
Factorial
PARALLEL
Most randomized controlled trials have
parallel designs in
which each group of participants is exposed to
only one of the
study interventions.
CLUSTER RANDOMIZED
A cluster randomized trial is a trial in which
individuals are randomized in groups (i.e.
the group is randomized, not the
individual).
CLUSTER RANDOMIZED
CROSSOVER
This design, obviously, is appropriate only
for chronic conditions that are fairly stable
over time and for interventions that last
a short time within the patient and that do
not interfere with one another.
CROSSOVER
CROSSOVER
Removing patient variation in this way makes crossover
trials potentially more efficient than similar sized, parallel
group trials in which each subject is exposed to only one
treatment
In theory treatment effects can be estimated with greater
precision given the same number of subjects.
that the effects
CROSSOVER
of one treatment may “carry over” and alter
the response to
subsequent treatments.
The usual approach to preventing this is
to introduce a washout (no treatment) period
between
consecutive treatments which is long enough
to allow the
effects of a treatment to wear off.
FACTORIAL DESIGN
two or more experimental interventions are not only
evaluated separately but also in combination and against
a control
FACTORIAL DESIGN
It allows evaluation of the interaction that
may exist between two treatments.
FACTORIAL DESIGN
two or more experimental interventions are not only
evaluated separately but also in combination and against
a control
SUPERIORITY AND
INFERIORITY
SUPERIORITY AND INFERIORITY
• FDA’s regulations on adequate and wellcontrolled studies (21 CFR 314.126)
describe four kinds of concurrently
controlled trials that provide evidence of
effectiveness.
• Three are superiority controlled trials:
 placebo
 no treatment
 dose-response controlled trials
SUPERIORITY
SUPERIORITY
A properly designed and conducted superiority
trial,
is entirely interpretable without further
assumptions
(other than lack of bias or poor study conduct)
INFERIORITY
The difference between the new and active
control treatment is enough to support the
conclusion that the new test drug is also
effective
INFERIORITY LIMIT
M 1 = the largest clinically acceptable
difference (degree of inferiority) of the test
drug compared to the active control
INFERIORITY LIMIT
 The critical problem, and the major focus of this
guidance, is determining M 1 , which is not measured in
the NI study (there is no concurrent placebo group).
 It must be estimated (really assumed) based on the
past performance of the active control and by
comparison of prior test conditions to the current test
environment
INFERIORITY LIMIT
One approach is to specify the equivalence
margin on the basis of a clinical notion of a
minimally important effect.
BUT
clearly subjective
The equivalence margin is often chosen with
reference to the effect of the active control
in historical placebo-controlled trials.
INFERIORITY LIMIT
Someones claims that a positive noninferiority trial
implies that the new treatment is superior to placebo.
However, this claim requires an assumption that the effect
of the active control in the current trial is similar to its
effect in the historical trials.
INFERIORITY LIMIT
Differences with respect to design features or by
an inconsistency in the effect of the active controls
among the historical placebo-controlled trials
(beyond that expected by random chance)
>
is often based on the lower bound of a confidence
interval for that effect(accounting for within-trial and
trial-to-trial variability)
LIMIT OF INFERIORITY
• Non-inferiority studies are not conservative in nature
since limits in the design and conduct of the study will
tend to bias the results towards a conclusion of similarity.
•
Poor compliance with the study medication, poor
diagnostic criteria, excessive variability of
measurements, and biased end-point assessment.
RANDOMIZATION
RANDOMIZATION
1- To conceal
2- To generate
TO CONCEAL
Allocation concealment prevents investigators
from influencing which participants are
assigned to a given intervention group
>
Increasing risk of selection bias
Evidence shows that reports of trials reporting inadequate
allocation concealment are associated with exaggerated treatment
effects
TO GENERATE
Use of computer or random number table
http://www.randomization.com/
TO GENERATE
Balanced randomisation involves selecting
certain baseline covariates (called
balancing variables) and incorporating them
into the randomisation scheme in a way
SIMPLE (UNRESTRICTED) RANDOMISATION
No other allocation generation approach, irrespective
of its
complexity and sophistication, surpasses the
unpredictability
and bias prevention of simple randomisation.
SIMPLE (UNRESTRICTED)
RANDOMIZATION
With small sample sizes, simple randomisation
(one-to-one allocation ratio) can yield highly
disparate sample sizes in the groups by
chance, although becoming negligible with trial
sizes greater than 200.
BUT
However, interim analyses with sample sizes of
less than 200 might result in disparate group
sizes.
RESTRICTED RANDOMISATION
It controls the
probability of obtaining an allocation sequence
with an
undesirable sample size imbalance in the
intervention
groups
BLOCKING METHODS
Blocks may be fixed or variable
If the block size is fixed, especially if small
(six participants or less), the block size could
be deciphered in a not double-blinded trial.
Longer block sizes—eg, ten or 20—rather
than smaller block sizes—four or six—and
random variation of block sizes help
preserve unpredictability.
RANDOM ALLOCATION RULE
For example, for a total study size of 200, placing 100
group
A balls and 100 group B balls in a hat and drawing them
randomly without replacement symbolises the random
allocation rule.
It is usually reported as use of envelopes
LIMITS OF RANDOMIZATION
Balanced randomisation introduces
correlation between
treatment groups, which violates the
statistical assumption that
all patients are independent
Balanced (simple)
randomisation forces
the outcomes between
treatment arms to
be similar
(apart from
any treatment effect)
LIMITS OF RANDOMIZATION
Variables used in the randomisation process should
subsequently be adjusted for in the analysis?
STRATIFIED
RANDOMIZATION
For example, with 6 diabetics, there is 22%
chance of 5-1 or 6-0 split by block
randomization only.
Stratified randomization is the solution to
achieve balance within
subgroups: use block randomization
separately for diabetics and non-diabetics.
STRATIFIED RANDOMIZATIOn
 The block size should be relative small to maintain balance in small
strata. Increased number of stratification variables or increased
number of levels within strata leads to fewer patients per stratum.
 Subjects should have baseline measurements taken before
randomization.
 Large clinical trials don’t use stratification. It is unlikely to get imbalance
in subject characteristics in a large randomized trial.
BLINDING
BLINDING (MASKING)
Keeping the trial participants, care providers,
data collectors, and some times those
analysing the data, unaware of which
intervention is being administered to which
participant, so that they will not be
influenced by that knowledge.
BLINDING (MASKING)
Do not use single, double…
But symply report who is blinded
1) Patients
2) Those assessing the outcome
3) Those administering the intervention
BLINDING (MASKING)
The success of blinding could be assessed
early in the first days of the study if possible
before the evidence of efficacy
Subjects could be asked to guess treatment
assignment, but they should be allowed to
express uncertainty and answer ‘‘do not
know.’’
If subjects are asked to guess treatment
assignment, subjects’ answers (for example
placebo/treatment/Do not know) should be
reported for each group.
DOUBLE DUMMY
SAMPLE SIZE COMPUTATION
SAMPLE SIZE COMPUTATION
The aim of an a priori sample size calculation
is mainly to determinate the number of
participants needed to detect a clinically
relevant treatment effect.
 Type 1 error and power are usually fixed at conventional
levels (5% for type I error, 80% or 90% for power).
 Assumptions related to the control group are often prespecified on the basis of previously observed data or
published results, and the expected treatment effect is
expected to be hypothesised as a clinically meaningful
effect.
HOW TO MINIMIZE SAMPLE SIZE
• Use Continuous Measurements Instead of
Categories
• Use More Precise Measurements
• Use Paired Measurements
• Expand the Minimum Expected Difference
• Use Unequal Group Sizes
SAMPLE SIZE FOR NON INFERIORITY
For NI trial
A small sample size is needed
To evaluate sample size for a new drug with 20% of
failure compared to 25% in the standard group:
- 1461 for group for superiority trial
- 298 for group (limit of inferiority 5)
- 133 for group (limit of inferiority 10)
AD INTERIM ANALYSIS
AD INTERIM ANALYSIS
The interests of participants should be best
served
if recruitment is closed as soon as a clear answer
is
available
Vs
The interests of society
should be best met if recruitment continues until
there is a clear answer (such that the results are
sufficiently conclusive to lead to changes in the
clinical management of future patients).
INDEPENDENT DATA
MONITORING COMMITTEE
 Is the inclusion rate of patients acceptable and as expected?
 Is there an unexpectedly high rate of severe or life-threatening
adverse events, which may indicate the premature closure of
the trial?
 Is the outcome of the trial treatment comparable with that of
the previous experience upon which the specific trial is based?
INDEPENDENT DATA
MONITORING COMMITTEE
If the interim analysis demonstrated a statistical significant
differences between the trial treatments that exceed the
differences defined by the statistical guidelines of the
trial
then
this would warrant closure of the study.
INTENTION TO TREAT
vs
AS TREATED
INTENTION TO TREAT
• Use every subject who was randomized according
to randomized treatment assignment.
• „Ignore noncompliance, protocol deviations,
withdrawal, and anything that happens after
randomization
• The ITT analysis holds the randomization as of
paramount importance
• ŠDeviation from the original randomized groups can
contaminate the treatment comparison
WHY INCLUDE NONCOMPLIANT SUBJECTS
IN ITT ANALYSIS?
 „Compliance or noncompliance occurs after randomization
 „Attempting to account for noncompliance by excluding
noncompliant subjects can bias the treatment evaluation
 In clinical practice, some patients are not fully compliant
 „Compliant subjects usually have better outcomes than
noncompliant subjects, regardless of treatment
AS PROTOCOL/AS TREATED
All participants are analyzed according to the
treatment they actually received, regardless
of what treatment they were originally
allocated.
While this may have some initial appeal,
once again the effect of random allocation
is compromised, making the interpretation of
the results difficult.
Intention to treat analysis
How can we decide on 5 events?
As treated analysis
SUBGROUP ANALYSIS
SUBGROUP ANALYSIS
If many are performed, it becomes likely that one or more will
spuriously be statistically significant.
In fact, if the subjects in a trial randomized between treatment
groups A and B are partitioned into G mutually exclusive subgroups
and a statistical significance test at α=0.05 is conducted
within each subgroup, then even if there is no true effect,
the probability of at least one significant result is 1 – (1 – α )G
.
SUBGROUP ANALYSIS
For α 0.05 and G 5, this probability is 23
percent;
for α 0.05 and G 10, the probability is 40
percent
Subgroup analyses also produce misleading
reversals of
effects, especially if the overall result is barely
significant.
SUBGROUP ANALYSIS
A commonly used method for adjusting is
dividing the overall significance level by the
total number of subgroup analyses, also called
the Bonferroni method.
For example, in a study with a significance
level of 0.05 and 10 subgroup analyses, the
significance level for each subgroup analysis
would be 0.005.
However, some statisticians state that
significant results are rarely observed after
adjustment with the Bonferroni method
SUBGROUP ANALYSIS
The pre-specificed ones
that is according to stratified randomization
are the most reliables ones
TAKE HOME MESSAGES
- Check how randomization is performed
- Check how blinding is performed
- Check about superiority and inferiority
structure
THANKS A LOT!!!!