Randomization

Randomization in Greenlight Guru Clinical 

The randomization in Greenlight Guru Clinical uses block randomization with random block sizes.

The pseudorandom number generator is from https://www.npmjs.com/package/seedrandom using ARC4 with RC4-drop(256). The blocks sizes are 1-5 * NumberOfTrialGroups.

Enable Randomization

You need to enable and activate the randomization add-on function in the study builder, to use the randomization function. This can either be done when creating the study but also later by clicking on the “Edit Study” button in the study overview.

To define randomization for your study, go to processes and click on “+ New process” or edit an already existing process. You can find randomization under the Advanced Options in your process.

Click on the "randomization" button and specify your target inclusion.

Randomization Settings

Target inclusion

Target inclusion (TI) is the TOTAL number of subjects you wish to randomize in your study and is always split evenly between the number of trial groups,

e.g if you have two groups and want 50 in each group, the target inclusion is 100.

The randomization IDs generated are numbers between 1 and Target Inclusion, currently, it is not possible to customize randomization ID.

The system will aim to generate as close to your target inclusion as possible, but it requires your target inclusion is divisible with your number of trial groups, e.g. TI = 150, Number of groups = 3.

Site as a factor

The randomization must either stratify by sites or ignore sites.

If sites are used as a factor (stratifying by sites), it will create a randomization list PER site, from where the site pulls randomization records, no sharing of blocks.

The target inclusion number will be split evenly between all sites using the randomization process at time of study start.

If sites are ignored, it will generate a central randomization list, from where all sites then pulls randomization records as needed, they share blocks.

Stratification by question

You can add stratification to your randomization, thereby segmenting the study population into strata. For information on stratification within clinical trials in general, please see:https://www.sciencedirect.com/science/article/pii/S0895435698001383?casa_token=ivePL7zL3g0AAAAA:a5JDzPuqInl7Gherwp-TYiTZpHQHOvZ1LC6PZdg4U_dd3477SQCKOWtB8QhroDDN82ZcAncWPsI 

In Greenlight Guru Clinical, you can select possibility questions (multiple-choice, Yes/No) and use them for stratification in your study. 

All of the question possibilities will be used to generate strata, and if multiple questions are used, the strata will be equal to the product of possibilities.

For example, if you have set up a stratification with a Yes/No question (smoker?), there will be two strata: Yes &  No

If you then add a multiple-choice question with three possibilities ( Blood pressure: low, middle, high), it will result in six (2 x 3) combinations:

Yes | Low

Yes | Middle

Yes | High

No | Low

No | Middle

No | High

The chosen target inclusion will be split evenly between strata. Please see examples at the bottom.

Randomization groups

Subsequently, add your randomization groups. You can add as many groups as you want, but can not have less than two.

Select if you want the stratification to be based on a site level or if you want to ignore the sites

Add "Stratify by question" if needed. Select the data event and question of interest. This could be relevant if you need to stratify by e.g. gender.

Note: You can only stratify by questions that have pre-defined answers, e.g. multiple choice and yes/no questions.

 

Enabling randomization adds a "randomization event" in your data event list, and you can drag it to where in the process the subject should be randomized.

It will show up as a regular data event in the site overview. To randomize the subject, click on the data event and then randomize.

Other

Blinding

Study personnel can be blinded in Greenlight Guru Clinical by removing "View Randomize" permission in the collaborator section, which will remove their ability to see which group the subjects are allocated to.

Export

It is possible to export a list with an overview of all randomization IDs and the corresponding trial group. This requires the appropriate collaborator permission “Export Randomization list” which is a study permission. 

Upload Randomization List

It is possible to upload a randomization list into your process, e.g. if the study should not use random block size, if the distribution to the groups should not be 50/50, or for other reasons. Read the help article Upload Randomization List for more info.

Template text to use in articles or protocols:

Participants were randomized to either <Control> or <Intervention> by the randomization tool in SMART-TRIAL, which uses permuted block randomization with random block sizes of the size 1-5*TrialGroups. The pseudorandom number generator is from https://www.npmjs.com/package/seedrandom using ARC4 with RC4-drop(256).

Examples of how randomization records are distributed:

Central randomization with stratification
Target Inclusion	100
Sites	Aalborg	Copenhagen
Groups	Intervention	Control
Use site as factor?	NO
Stratification(Smoker?)	Yes	No
	CENTRAL LIST
Minimum rando. records assigned:	100
Intervention (I)	50
Control (C)	50
I: Smoke = Yes	25
I: Smoke = No	25
C: Smoke = Yes	25
C: Smoke = No	25
Site as factor, no stratification
Target Inclusion	100
Sites	Aalborg	Copenhagen
Groups	Intervention	Control
Use site as factor?	Yes
NO Stratification
	Aalborg	Copenhagen
Minimum rando. records assigned:	50	50
Intervention	25	25
Control	25	25
Central randomization, no stratification
Target Inclusion	100
Sites	Aalborg	Copenhagen
Groups	Intervention	Control
Use site as factor?	NO
NO Stratification
	CENTRAL LIST
Minimum rando. records assigned:	100
Intervention	50
Control	50

Contact [e-mail] for more information on how to use randomization in your study. 

 

Amendments to Randomization in Ongoing Study

Adding new sites to an ongoing study

Not using site as a factor

The new site will use the central list.

Site as a factor

The new site will have a randomization list created for it, with an amount of records corresponding to:

Target Inclusion / number of sites

Example:

Target inclusion of 120, with 2 sites. Adding a third site will create a list with the following amount of records: 120 /3 = 40

Adding a fourth: 120/4 = 30

Adding a fifth:  120/5 = 24

Adding a sixth: 120/6 = 20

and so forth. This results in the latter sites having relatively fewer records, and patients they can randomize. 

Adding more randomization records (blocks)

If you wish to add more randomization records (blocks) to the study, e.g., to allow for sites to continue recruiting, it can be done by increasing the Target Inclusion number.

Note that if stratify by question is used, the records allocated to each site is then split evenly between the strata.

Note that the system will always round UP when generating additional records.

No sites added after study start, central randomization lists

Increasing the target inclusion from e.g. 100 to 200 adds 100 records to the central randomization list.

No sites added after study start, site as a factor

Increasing the target inclusion from e.g. 100 to 200, will split the difference (100) between the existing sites.

I.e. if there are two sites, each site will get an additional 50 records.

Sites added after study start, central randomization list

Increasing the target inclusion from e.g. 100 to 200 adds 100 records to the central randomization list.

All sites pull records from the same list, using the same blocks, always finishing a block before starting a new one.

Sites added after study start, with site as a factor

The sites added after study start, will have fewer records available, as described above.

Increasing the target inclusion here will therefore behave a bit differently than the other cases.

The system will seek to balance the records between each site, such that sites with many records already, will have none or few records generated, and sites with few records will have many generated.

Example:

Target Inclusion: 100

Site A: 50 records

Site B: 50 records

(added after study start) Site C: 33 records

(added after study start) Site D: 25 records.

In amendment mode, the target inclusion is increased from 100 to 200. Upon finishing the amendment, the system checks how many records each site should have using the calculation TargetInclusion/NumberSites ->  200/4  = 50.

New records are generated for sites with LESS than 50 records.

Site A: 50 records (No new records)

Site B: 50 records (No new records

Site C: 50 records (17 records generated)

Site D: 50 records. (25 records generated).

So, to create additional records for Site A and Site B, the target inclusion would have to be raised by more than 200, which would also generate more blocks for sites C and D.

Help Video

The video below demonstrates how to enable randomization, add randomization to a process, and how to randomize a subject.