Randomization

A study's treatment patterns specify the drugs to be administered and the medical procedures to be performed. You assign them to patient positions, creating treatment assignments.

For more information, see:

• Some Concepts
  
• Access
  
• Choosing a Randomization Type
  
• Deleting, Locking/Unlocking, and Verifying Randomizations
  
• Batch Loading a Randomization
  

Some Concepts

Principle 1

The treatment code of a treatment assignment must be unique within a clinical study version, and so randomizations must not overlap. Recipes (instructions creating a randomization, including the treatment patterns involved, their ratio, and the number of related treatment assignments) are overlapping when they contain the same treatment assignment.

For example, Recipe A creates treatment codes 1 to 10; Recipe B creates 1 to 20. Your recipes can overlap, if necessary, while you experiment with alternate recipes. But you cannot use overlapping recipes when creating randomizations. To prevent overlap, when you enter parameters for a randomization the system validates within a recipe to trap for overlapping definitions. However, because the system does not detect overlaps between recipes until you randomize, it is then that you receive an error if you try to create a treatment assignment with a code already in use.

Principle 2

Oracle Clinical assumes that, due to experimental design requirements, your company separates the role of an unblinded randomization analyst, who generates a randomization, from the blinded person who enters the randomization recipe.

Principle 3

The sequence for entering randomization recipes and attributes is flexible; you can continue to modify recipes until you generate a randomization from them. Once it is generated, however, you can only make changes that expand the randomization but that do not change the existing treatment assignments.

Access

Access refers to tasks involved in an interim analysis. An interim analysis is when a quick assessment is made of the study results to decide if it should continue or if it should be stopped because of lack of efficacy (or safety).

For more information, see:

• Interim Analysis
  
• Randomization Privileges
  
• Maintaining Blinding in a Study
  

Interim Analysis

Several functions control who can perform an interim analysis, when, and on what data. In particular, the randomization access status code controls who is allowed to see the randomization in a study. It can have the following values:

Value	Meaning
Open	The study is open label from the start; everybody can see the randomization details.
Closed	The study is currently blinded, so only unblinded users can see the randomization details.
Release	The study was initially blinded (Closed) but is now over and the randomization opened, or released, for analysis.
Access	Allows specified users access to the randomization to perform an interim analysis. (See Creating or Revoking an Access View.)
Multi	Access controlled at the phase level and not at the study level.

A dummy randomization is available to establish the analysis environment. It has no correlation to the real randomization; however, you can use it to test reporting and extracting facilities. The system creates the dummy randomization when you randomize the study. The actual randomization becomes available for analysis when the study's blinding stage is complete.

Randomization Privileges

Most functions within the Randomization subsystem require the privilege RXC_RAND. Typically companies limit RXC_RAND privileges to a small group of specialized users who are not members of the study management teams.

Maintaining Blinding in a Study

While a study is still blind, only privileged users have access to its randomization. If an emergency situation occurs, the system provides ways to disclose a patient's treatment pattern. (See Disclosing Patient Treatment Assignments.) If disclosure occurs, the system generates a record of the disclosure as a blind break.

Choosing a Randomization Type

Oracle Clinical offers four types of randomization: version-level, strata-level, version block-level, and strata block-level—each described in the following sections.

If you create either a version-blocked or strata-level blocked randomization, and you intend to create only one randomization recipe, you can use the easy randomization utility available for these randomization types. From the Design menu, select Version Blocked Randomization (or Strata Blocked Randomization), then select Easy Randomization. Except for allowing only one recipe, the Easy Randomization utilities are identical to the equivalent standard Randomization utilities; if you want different recipes—for example, different ratios of treatment patterns for different patients—use the standard utilities.

For more information, see:

• Version-level Randomization
  
• Strata-Level Randomization
  
• Version Block-Level Randomization
  
• Strata Block-Level Randomization
  

Version-level Randomization

The version type of randomization generates sequential treatment assignments for the same treatment pattern. There is nothing random about this method—the recipe explicitly defines what you produce. For example, kits 1 to 10 on Treatment Pattern A, 11 to 20 on Treatment Pattern B, etc. Designers typically apply version randomization to Phase I open label-type studies.

The system assigns the following type codes to version-level randomizations:

• SVTP: Study Version Level, Treatment Pattern

• SVTPX: Study Version Level, Treatment Pattern, Extra Assignments Only

The system maintains the recipe within the treatment pattern records. For each treatment pattern, you can specify how many treatment assignments you require and the code from which to start coding them.

To create a version-level randomization, from the Design menu, select Randomization, choose Version Randomization, then select Treatment Patterns. Choose a study, then click the Treatment Patterns button. Specify the parameters of your recipe and use the Regimens in Treatment Pattern and Drugs buttons to access windows to further specify the randomization. Click the Copy Treatment Pattern button to copy a pre-existing treatment pattern within your study.

To create a version-level randomization from a specific treatment assignment, from the Design menu, select Randomization, choose Version Randomization, then select Treatment Assignments.

Strata-Level Randomization

The strata-level type of randomization is similar to the version-level randomization described above, except that each stratum in the study can have its own set of treatment assignments. For example, males can use kits 1 to 10 on Treatment Pattern A, 11 to 20 on Treatment Pattern B; females can use kits 100 to 110 on Pattern A, etc. This is also designed to apply to Phase I open label-type studies.

The following type codes are assigned to stratum-level randomizations:

• STFP: Stratum Level, Treatment Pattern

• SFTPX: Stratum Level, Treatment Pattern, Extra Only

From the Design menu, select Randomization, choose Strata Randomization, and then select Treatment Patterns. Choose a study, then click the Treatment Patterns button. Specify the treatment pattern information, then use the Regimens in Treatment Pattern and Drugs buttons to access windows to further specify the treatment pattern. Click the Copy Treatment Pattern button to copy a treatment pattern already existing in your study.

To maintain the strata assigned to a study, from the Design menu, select Randomization, choose Strata Randomization, and then Stratification. Choose a study, then click the Stratum Assigned to the Study button to view the descriptions of the strata assigned to the study. Click the Factors button to associate the strata to a factor; click the buttons, Details of the Stratum, and Treatment Patterns to refine the randomization.

To create a strata-level randomization from a specific strata, from the Design menu, select Randomization, choose Strata Randomization, and then Treatment Assignments. Select the strata and click the Randomize button.

Version Block-Level Randomization

The version block-level type of randomization generates a random set of treatment assignments. Within the randomization, each block of treatment assignments contains the same combination of treatment patterns but in different sequences. For example, two patterns A and B and a block size of four could produce ABBA BABA ABAB BBAA ABBA, etc.

The following type codes are assigned to study version block definition-level randomization:

• SVBD: Study Version Level, Block Definition

• SVBDX: Study Version Level, Block Definition, Extra Only

To create a version-blocked level randomization, from the Design menu, select Randomization, choose Version Blocked Randomizations, and then Treatment Patterns. Choose a study, then click the Treatment Patterns button. Specify the parameters of your recipe and use the Regimens in Treatment Pattern and Drugs buttons to access windows to further specify the randomization. Click the Copy Treatment Pattern button to copy a pre-existing treatment pattern within your study.

To create a version block-level randomization, from the Design menu, select Randomization, choose Version Randomization, then select Block Definition. Choose a study, then click the Block Definitions button. Specify the parameters of your recipe, then click the Treatment Pattern button to further specify the randomization. Click the Copy Treatment Pattern button to copy an existing treatment pattern in your study.

To create a version block-level randomization from a specific treatment assignment, navigate to Design, Randomization, Version Blocked Randomization, and then Treatment Assignments.

Strata Block-Level Randomization

The strata block-level type of randomization is similar to the version block definition-level described above, except that the randomization belongs to a stratum of the study.

The following type codes are assigned to the stratum version block definition-level randomization:

• SFBD: Stratum Level, Block Definition

• SFBDX: Stratum Level, Block Definition Extra Only

To create a strata block-level randomization, from the Design menu, select Randomization, choose Strata Blocked Randomization, and then Treatment Patterns. Choose a study, then click the Treatment Patterns button. Specify your treatment patterns, and use the Regimens in Treatment Pattern and Drugs buttons to access windows to further specify the treatment pattern. Click the Copy Treatment Pattern button to copy a treatment pattern that exists in your study.

To create a version block-level randomization, from the Design menu, select Randomization, choose Strata Blocked Randomization, and then Block Definition. Choose a study, then click the Block Definitions button. Specify the parameters of your recipe, then click the Treatment Pattern button to further specify the randomization. Click the Copy Treatment Pattern button to copy a treatment pattern existing in your study.

To create a strata block defined-level randomization from a specific set of treatment assignments, from the Design menu, select Randomization, choose Strata Blocked Randomization, and then Treatment Assignments. Select a strata and block definition, then click the Randomization button.

Deleting, Locking/Unlocking, and Verifying Randomizations

Oracle Clinical randomization includes utilities to remove randomizations, to lock—that is, to prevent changes to randomization treatment assignments—and to unlock locked randomizations. Open the Randomizations window by navigating to Design, Randomization, Randomization Type, then Treatment Assignments, and clicking the Show Randomizations button.

To test for changes to a randomization, click the Verify button in this window. Oracle Clinical regenerates the randomization using the original seed number, and creates a list of differences between the original randomization and the verification randomization. The report is named randval.out.

Note:

Randomization reports are formatted in landscape mode, so be sure to configure your page setup to this mode before printing a hard copy of the report.

Batch Loading a Randomization

Oracle Clinical provides utilities to transfer the details of a randomization in or out of Oracle Clinical. These utilities handle the following situations:

• Downloading a copy of a study's randomization details from Oracle Clinical for use in another application
• Loading legacy randomization details into Oracle Clinical
• Unusual randomization requirements that Oracle Clinical's standard functionality cannot emulate

  Note:

  If you batch load a legacy 'Block' Randomization into Oracle Clinical, it will not result in a block randomization. The batch load process will never populate TREATMENT_ASSIGNMENTS.RANDOM_BLOCK_ID or create blocks. If you want a block randomization, you should use the Oracle Clinical functionality.

  In addition, if you download an Oracle Clinical system-generated randomization, edit it, and batch load it back into Oracle Clinical, the previously populated RANDOM_BLOCK_ID is no longer populated.

Before you can batch load a randomization into Oracle Clinical, you must define the treatment patterns and strata. Create strata by navigating to, Strata, then Stratification. Define treatment patterns by navigating to Design, Treatments, then Treatment Patterns. Batch loading a stratification randomization is the same as manual creation, except that you cannot specify the number of treatment kits.

See Maintaining a Study Stratification Factor for a description of stratification factors.

Load the file into the directory specified in the RAND_ROOT setting of the OCL_STATE local reference codelist. If RAND_USERDIRS in the same codelist is set to Y, load the file into the user-specific subdirectory for the user who will run the job. See the Oracle Clinical Administrator's Guide for more information.

In this section:

• Downloading a Randomization
  
• Validating and Loading a Randomization
  
• Handling a Legacy Randomization
  

Downloading a Randomization

You can create an ASCII file of all the randomization details in a study. Navigate to Design, Randomization, Randomization Batch Load, then Download Randomization.

This PSUB submission function requires one parameter, Description, which the system applies as the name of the ASCII output file. The formats for the file are shown in this section.

• The name of the Randomization File: must include, and defaults to, the directory specified in the RAND_ROOT setting of the OCL_STATE local reference codelist. If RAND_USERDIRS in the same codelist is set to Y, it is stored in the user-specific subdirectory for the user who runs the job. See the Oracle Clinical Administrator's Guide for more information.

• The Study Code of a Study

Only the output file name is passed as input. This file is stored in

Editing the Randomization Download

If you intend to force in a randomization (download, edit, and reload a randomization), the system limits you to the following changes:

• Altering the treatment pattern assigned to a treatment assignment code

• Creating new treatment assignments

• Deleting treatment assignments

• Changing the replacement and/or latest flags

File Format

The tables in this section specify the format of the data for the different randomization types.

First Card Type - All Randomization Types

From:	To:	Parameter
1	- 10	"STUDY"
11	- 25	Study Code
26	- 35	Version Code

Stratum Card Type - For Stratified Randomizations Only

One card for each stratum involved in the randomization.

From:	To:	Parameter
1	- 10	"STRATUM"
11	- 21	Stratum Id
22	- 51	Stratum Description
53	- 61	Starting Treatment Assignment Code
62	- 73	Ending Treatment Assignment Code

Non-stratum Card Type -For Non-stratified Randomizations Only

One card for the entire randomization.

From:	To:	Parameter
1	- 10	"NOSTRATUM"
11	- 51	Null
52	- 61	Starting Treatment Assignment Code
62	- 71	Ending Treatment Assignment Code

Treatment Pattern Card Type - For All Randomizations

One card for each treatment pattern involved in the study.

From:	To:	Parameter
1	- 10	"PATTERN"
11	- 25	Pattern Code
28	- 80	Treatment Pattern Description

Treatment Assignment Card Type - For All Randomizations

One card for each treatment assignment in the randomization.

From:	To:	Parameter
1	- 10	"TRTASG"
11	- 20	Treatment Assignment Code
21	- 30	Patient Code
-	31	Latest Assignment Flag
-	32	Replacement Patient Flag
33	- 47	Owning Location Code

Match the structure of the force-in file as one of the following:

• STUDY
  • STRATUM
    • PATTERN
    • TRTASG
    • TRTASG
    • etc.
    • PATTERN
    • TRTASG
    • PATTERN
    • TRTASG
  • STRATUM
    • PATTERN
    • TRTASG
    • PATTERN
    • TRTASG
    • PATTERN
    • TRTASG

OR

• STUDY
  • NOSTRATUM
    • PATTERN
    • TRTASG
    • TRTASG
    • PATTERN
    • TRTASG
    • TRTASG
    • PATTERN
    • TRTASG
    • TRTASG

Validating and Loading a Randomization

This form reloads an ASCII file containing the randomizations for a study, checking that it meets all randomization requirements. It is a standard PSUB submission form. From the Randomization menu, select Randomization Batch Load, and choose Validate and Load Randomization. The system checks that the treatment patterns exist within the clinical study and that there are no existing locked treatment assignments. It then deletes all the existing randomization information and re-creates it from the file.

Copy the input randomization file into the directory specified in the RAND_ROOT setting of the OCL_STATE local reference codelist. If RAND_USERDIRS in the same codelist is set to Y, copy the file into the user-specific subdirectory for the user who will run the job. See the Oracle Clinical Administrator's Guide for more information.

Handling a Legacy Randomization

When you have a randomization created outside the system—a converted randomization—Oracle Clinical reads the randomization from the file and loads it into the database, translating external codes into the corresponding design codes for strata and treatment patterns. The system takes this input and creates treatment assignments and randomization records. Oracle Clinical also validates that the load process ran successfully.

Before you can execute the load, the clinical study, clinical study version, treatment patterns, and, possibly, strata, have to exist.

To load and validate a converted randomization:

1. Navigate to Design, Randomization, Randomization Batch Load, and Load a Converted Randomization.
2. Reformat a flat file from the existing legacy randomization system into Oracle Clinical's format. See "Converted File Format."
3. Copy the input randomization file into the directory specified in the RAND_ROOT setting of the OCL_STATE local reference codelist. If RAND_USERDIRS in the same codelist is set to Y, copy the file into the user-specific subdirectory for the user who will run the job. See the Oracle Clinical Administrator's Guide for more information.
4. Use the Maintain Conversion Specifications module to inform Oracle Clinical how to translate the treatment pattern and strata codes supplied by your legacy system into Oracle Clinical codes. See "Entering Conversion Parameters." Some randomization systems treat replacement patients as different strata. Oracle Clinical does not require this, so if a stratum exists only to separate replacement patients from normal patients, you can flag it as a replacement stratum and map it to the same stratum in Oracle Clinical as the normal stratum from the legacy system.
5. Load the converted randomization into Oracle Clinical.
6. Check the output report for any errors encountered in loading the data.
7. Optionally, run a report to read the original data file and compare it with the Oracle Clinical database to confirm that the data loaded correctly. See "Validating a Loaded Randomization."

Unlike downloading, editing, and loading Oracle Clinical data—processes that identify everything with Oracle Clinical codes and structure—converting and loading legacy data allows for code translation and for merging strata. It is more powerful but also more complex. Oracle recommends using the download format, if possible.

Entering Conversion Parameters

When converting randomization data from legacy systems into Oracle Clinical, you need to create conversion specifications to match the external data to Oracle Clinical parameters.

If the treatment pattern in the legacy system is based on the same code as the treatment pattern in Oracle Clinical, there is no need to enter a conversion parameter.

From the Randomization menu, select Randomization Batch Load, choose Enter Conversion Parameters, and then select the study. The Version window of Maintain Conversion Specifications has two buttons:

• Strata opens the Strata window, where you map the strata from the legacy system to the corresponding target Oracle Clinical codes. You can map each target stratum to more than one legacy stratum. Since some legacy systems use strata differently from Oracle Clinical's normal and replacement assignments, you can set a flag to indicate that Oracle Clinical merges the strata into Oracle Clinical's replacement assignments.

• Patterns opens the Treatment Pattern Conversions window of the Maintain Conversions window, where you map the legacy treatment pattern code (Old Pattern field) to the corresponding Oracle Clinical code. You can have only a one-to-one correspondence.

Converted File Format

The file format must conform to the following requirements:

• Left-align all character fields, with no leading zeros.

• Right-align and zero-fill all numeric fields.

• Left-align the version code, adding trailing spaces to fill the 10-character code.

• Treatment and patient codes are character fields but are limited to a single, leading, non-numeric character. Right-align these fields, and do not add leading zeros; for example, S1001.

Clinical Study Card Type

From:	To:	Parameter	Type	Length
1	-10	"STUDY"	CHAR	9
11	- 25	Study Code	CHAR	15
25	- 34	Null	CHAR	10
35	- 45	Starting Seed number	NUMBER	11
46	- 56	Ending Seed number	NUMBER	11
57	- 59	Block size	NUMBER	3

Stratum Card Type

From:	To:	Parameter	Type	Length
1	- 9	"STRATUM"	CHAR	9
11	- 14	Investigator code	NUMBER	5
15	- 16	Stratum number	NUMBER	2
17	- 31	Stratum abbreviation	CHAR	15
32	- 41	Starting patient number	CHAR	10
42	- 51	Ending patient number	CHAR	10

Treatment Pattern Card Type

From:	To:	Parameter	Type	Length
1	- 9	"PATTERN"	CHAR	9
10	- 24	Pattern Code	CHAR	15
18	- 80	Pattern Description	CHAR	63

Patient Card

From:	To:	Parameter	Type	Length
1	- 9	"PATIENT"	CHAR	9
10	- 19	Patient Code	CHAR	10
20	-34	Owning Location	CHAR	15

Converted file structure

• STUDY
  • STRATUM
  • PATTERN
    • PATIENT
    • PATIENT
    • etc.
  • PATTERN
    • PATIENT
  • PATTERN
    • PATIENT
  • STRATUM
  • PATTERN
    • PATIENT
  • PATTERN
    • PATIENT
  • PATTERN
    • PATIENT
  • PATTERN
    • PATIENT

Validating a Loaded Randomization

If required, you can run a report to read the original data file and compare it with the Oracle Clinical database to confirm that the load was correct. From the Design menu, select Randomization, choose Randomization Batch Load, and then Validate Loaded Randomization. The Randomization Take-on display of the PSUB submission form appears. Enter the name of the loaded randomization file in the Description field.