Transcript Artifact 2

Basics of
Randomization
Purpose of Randomization
Randomization is intended to limit the
occurrence of conscious and unconscious
bias in the conduct and interpretation of a
clinical trial arising from the influence that the
knowledge of the impending treatment
assignment may have on the recruitment and
allocation of subjects.
How Does Randomization
Limit Bias?
If allocation of a patient to a
treatment is done
‘randomly’, then personnel
at a site can not predict the
next treatment assignment
(ie., there is no pattern upon
which to make a prediction).
How is Randomization
Implemented?
• Via a Randomization Scheme (aka, Rand Scheme)
• A Rand Scheme is a list which dictates the order
of treatment assignments (e.g. Active, Placebo)
within a clinical trial.
Randomization Scheme
A list which dictates the order of
treatment assignments.
Characteristics:
 Blocked vs. Simple
 Central vs. By Site
 Stratified vs. Not Stratified
Simple Randomization
Example of Simple
Randomization
• Just like flipping a
coin (heads or tails)
• A simple rand
scheme contains no
blocking.
Patient
Treatment
1
A
2
A
3
B
4
B
5
A
6
A
7
B
8
A
9
B
10
B
Simple Randomization
So, why not always
use a simple rand
scheme?
There are two problems:
Hint 1: Does the list look
random?
Hint 2: What if we want a 1:1
ratio of A to B?
Patient
Treatment
1
A
2
A
3
A
4
A
5
A
6
A
7
A
8
A
9
B
10
B
Blocked Randomization
What is blocking?
• Breaks the rand scheme
into defined units = Blocks
• Within each unit (block),
the treatment ratio is
maintained.
Block Patien Treatment
t
✺
1
001
A✺
1
002
A✺
1
1
2
003
004
005
✺
B
B✺
✺
B
✺
2
006
A
✺
2
007
B
Block size (which is 4 in the above case!)
2 is confidential!
008
A
Blocked Randomization
Why block?
1) Promotes an
appropriate treatment
ratio
2) Promotes randomness
throughout the rand
scheme.
Block Patien Treatment
t
1
001
A
1
002
A
1
1
2
003
004
005
B
B
B
2
2
2
006
007
008
A
B
A
Stratified vs. Not Stratified
Use a stratified randomization when
patient characteristics greatly influence
the effectiveness of the treatment.
Promotes an equal distribution of
treatments across patient populations.
Examples
 Weight
 Age
 Severity of disease state
Example: Not Stratified
Treatment A
Treatment B
50 %
50 %
1
2
Which treatment is likely to have the fewest
heart attacks? 1. Treatment A
2.
Treatment B
Stratified Randomization
If stratified on Age (Patients <55 and Patients >=55 ),
the result is essentially two rand schemes:
For Patients < 55
For Patients >=55
Patient
Treatment
Patient
Treatment
1
A
1
A
2
B
2
B
3
B
3
A
4
A
4
B
Stratified Randomization
Terminology Note:
Stratification Factor: The characteristic
of interest (e.g. Age)
Stratification Level (Strata): The groups
within the stratification factor (e.g.
Age<55 vs. Age>=55)
Stratified Randomization
Another Example:
Let’s stratify patients on hair color like:
• Blonde
• Brown
• Red
Stratified Randomization
• So, what is the stratification factor?
Hair Color!
• And, what are the stratification levels?
• Blonde
• Brown
• Red
Stratified Randomization
Now, how many separate rand schemes are created?
Blonde
Brown
Red
Patient
Trt
Patient
Trt
Patient
Trt
1
2
3
A
A
B
1
2
3
A
B
A
4
B
4
B
1
2
3
4
A
B
B
A
Stratified Randomization
• For convenience,
these 3 schemes are
combined into one
list as such:
Seq Order
Hair Color
Trt
1
Blonde
A
2
Blonde
A
3
Blonde
B
4
Blonde
B
5
Brown
A
6
Brown
B
7
Brown
A
8
Brown
B
9
Red
A
10
Red
B
11
Red
B
12
Red
A
Central vs. By Site
Randomization
Central
By Site (site stratified)
Typically smaller trials
Larger trials (>100 pts)
(25-100 pts)
All patients are
randomized from the
same rand scheme.
A portion of the rand scheme
is allocated to each site and
patients are randomized
based upon the site at which
they are enrolled.
Central Randomization
Rand Scheme
Patient
Trt
1
A
2
B
3
B
4
A
5
B
6
A
Note: Order of assignments does not vary
based upon where patient was
randomized.
For example, the third patient will always
be assigned to ‘B’ regardless of which site
recruits the third patient.
Site 1
2 B
5 B
Site 2
Site 3
1
A
3
B
4
A
6
A
Central Randomization
Ask yourself:
Site 1
Would you have known to only
send ‘B’ kits to Site 1?
5 B
OR only send ‘A’ kits to Site 3?
Site 2
No….there is no way to predict
which drug will be used where
because it depends on when
patients arrive for treatment.
2 B
Site 3
1
A
3
B
4 A
6
A
By Site Randomization
Site 1 Rand
Scheme
Site 2 Rand
Scheme
Site 3 Rand
Scheme
Patient
Trt
1
A
2
B
Patient
Trt
1
B
2
A
Patient
Trt
1
B
2
A
Site 1
1 A
2 B
Site 2
Site 3
1
B
2
A
1
B
2
A
By Site Randomization
Note: The order of
the drug
assignment at the
site is known…it
follows the site’s
rand scheme.
Patient
Trt
1
A
2
B
Patient
Trt
1
B
2
A
Patient
Trt
1
B
2
A
Site 1
1 A
2 B
Site 2
Site 3
1
B
2
A
1
B
2
A
By Site Randomization
Rand Scheme for Site 5
Patient
Trt
1
A
2
B
3
B
4
A
5
A
6
B
Site 5
A
B
B
If we planned to ship 3 kits to Site 5, what kit types would we ship?
Types of Rand Schemes
Each of the Rand Scheme characteristics can be combined to
produce different types of Rand Schemes.
Examples are:
• Central
• Central and Stratified
• By Site and Stratified
Note: All of the above Rand Schemes are blocked
and a block size must be designated. Simple
randomization is rarely used.
What type of Rand Scheme?
1.
By site
2.
By site stratified
3.
Central
4.
Central stratified
25 %
1
25 %
2
25 %
3
25 %
4
What type of Rand Scheme?
1.
By site
2.
By site stratified
3.
Central
4.
Central stratified
25 %
1
25 %
2
25 %
3
25 %
4
Dynamic (Adaptive) Randomization
• Used when randomization needs to be stratified
on various levels and the sample size is very
small.
• Special feature is that the study drug
assignment is NOT fixed at the beginning of the
trial (i.e.., it’s dynamic!).
• The assignment is determined at the time of
randomization based upon the type of patients
currently enrolled, the characteristics of the
current patient and need of the trial to ‘fill all
the cells.’
Random Lists
Previous discussion revolved around types and
characteristics of rand schemes. However, there are two
types of ‘random’ lists used in most trials:
1. Randomization Scheme – List used to associate
patients to treatments (e.g. active, placebo).
2. Kit List – List used to associate kit numbers to kit
types (e.g. Visit 1- 2mg active kit, Visit 2- 4mg active kit)
Random Lists
• Randomization Scheme: Links patients to treatments
Patient
Trt
1
A
2
B
Note: The patient and treatment
assigned are present, but NO Kit #
is listed.
• Kit List: Links kit numbers to kit types
Kit #
Kit Type
8432
A
4492
B
Note: The Kit # and Kit Type are
present, but which patient is NOT
listed.
Kit Lists
• Definition: A list of kit numbers associated
with the content of the kit.
• In a randomized clinical trial, the association
between the kit number and the treatment is
random (in other words…you can’t guess the contents of
the kit based upon the kit number)
Kit Lists
Why ‘random’?
•
We want the number on the kit to, in no way,
indicate what the kit contains (assists with
blinding).
Note: You may also here the word ‘scrambled’ in
reference to kit lists.
Kit Lists
• Like Rand Schemes, Kit Lists have different
characteristics.
• The type of kit list generated for a trial will
depend upon the method used to assign the
kit to a patient.
• We’ll discuss ‘Method of Randomization’
later today.
Kit Lists
Examples: Sequential vs. Random Numbers
Kit # Kit Type
Kit #
Kit Type
101
4mg Active
332
Visit 1 Active
102
2mg Placebo
638
Visit 1 Placebo
103
4mg Placebo
123
Visit 2 Active
104
2mg Active
875
Visit 2 Placebo
Sequential Numbered
Random Numbered
Kit List Characteristics
Kit #
Kit Type
101
4mg Active
102
2mg Placebo
103
4mg Placebo
104
2mg Active
Kit numbers are
consecutive (101 – 104)
Kit Type is
designated…there are 4
types of kits.
Remember: A Kit List associates a kit number with a
kit type (e.g. 2mg Active vs 4mg Active), NOT just a
treatment (Active vs. Placebo).
Kit List Characteristics
Kit Numbers are nonconsecutive and
randomly ordered
There are 4 types of
kits, 2 for Visit 1 and 2
for Visit 2
Kit #
Kit Type
332
Visit 1 Active
638
Visit 1 Placebo
123
Visit 2 Active
875
Visit 2 Placebo
Again: Kit Type is indicated, NOT treatment!
Kit Lists
Summary:
•Kit Numbers may be consecutive or
random.
•Kit Lists establish the relationship between
the TYPE of kit and the kit number.
Generation of Random
Lists – Best Practices
 Request that the files are sent in our standard format
 Discuss this with the customer very early in the
process
 Standard process agreed with Clinical Technologies
 Clearly document any relationships between data in the
file with the treatment assignments
 E.g. A = placebo, B = active
Generation of Random
Lists – Best Practices
Ensure lists (electronic and hard copies) are
adequately controlled
Compare the list uploaded into the
computer database with the list provided
“Numbers are FREE”
Generate more randomization slots than what
you think you will need.
Generation of Random
Lists – Best Practices
Kit lists
Use a different number of digits in the kit
number vs. patient number
Choose the largest kit number range as
possible
Allow for hyphens on the kit label, to help
with reading long digits (e.g 100-456)
Random Lists at Clinical Services
• Random Lists include:
• Randomization Schemes
• Kit Lists
• Random List SOP (“Procedure for the
Control of Random Lists” - GQA.005)
governs the process of requesting, receipt
and storage.
Randomizing Patients
???
Let’s say, the clinician has received the
study drug and has a patient ready to
receive drug.
How does the clinician know which kit to
give the patient?
There are two methods used to assign a specific drug
kit to a patient:
1) Single Randomization
2) Double Randomization
Method of Randomization
Single
vs.
Double
One Random List
Two Random Lists
Manual process
(No automation)
Automation required
(IVRS, IWRS, WebEZ)
Consecutive Kit
Numbers
Non-consecutive
Kit Numbers