Basics of Randomization Purpose of Randomization is intended

Basics of Randomization

Purpose of 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: Ø Blocked vs. Simple Ø Central vs. By Site Ø 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 Block Patien Treatment t ✺ 1 1 1 • Within each unit (block), 1 the treatment ratio is maintained. 2 2 2 Block size (which is 4 in the above case!) 2 is 001 A✺ 002 A✺ ✺ 003 B 004 B✺ ✺ 005 B ✺ 006 A ✺ 007 B confidential! 008 A

Blocked Randomization Why block? 1) Promotes an appropriate treatment ratio 2) 2) Promotes randomness throughout the rand scheme. Block Patien Treatment t 1 1 2 2 001 002 003 004 005 006 007 008 A A B B 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 Ø Weight Ø Age Ø Severity of disease state

Example: Not Stratified Treatment A Treatment B 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 1 A 2 B 3 B 3 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 1 A A B B 1 2 A B 3 4 A B 1 2 3 4 A B B A 2 3 4

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 Site 1 1 A 2 B Site 2 Site 3 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 Site 1 1 A 2 B Site 2 Site 3 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 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

What type of Rand Scheme? 1. By site 2. By site stratified 3. Central 4. Central stratified

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 - 2 mg active kit, Visit 2 - 4 mg 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 101 4 mg Active 332 Visit 1 Active 102 2 mg Placebo 638 Visit 1 Placebo 103 4 mg Placebo 123 Visit 2 Active 104 2 mg Active 875 Visit 2 Placebo Sequential Numbered Random Numbered

Kit List Characteristics Kit # Kit Type 101 4 mg Active 102 2 mg Placebo 103 4 mg Placebo 104 2 mg 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. 2 mg Active vs 4 mg 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 Ø Request that the files are sent in our standard format Ø Discuss this with the customer very early in the process Ø Standard process agreed with Clinical Technologies Ø Clearly document any relationships between data in the file with the treatment assignments Ø E. g. A = placebo, B = active

Generation of Random Lists – Best Practices Ø Ensure lists (electronic and hard copies) are adequately controlled Ø Compare the list uploaded into the computer database with the list provided Ø “Numbers are FREE” ØGenerate more randomization slots than what you think you will need.

Generation of Random Lists – Best Practices ØKit lists ØUse a different number of digits in the kit number vs. patient number ØChoose the largest kit number range as possible ØAllow 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, Web. EZ) Consecutive Kit Numbers Non-consecutive Kit Numbers