Randomization

Guide to Randomization on W2H

While there are customizable options for randomization on W2H, certain options are standard and other options may require additional development and associated costs.

The standard options include:

• Basic randomization

• Blocked randomization

• Blocked randomization with Stratification

Basic Randomization:

• This type of randomization procedure divides participants equally across groups regardless of individual characteristics. This process will result in approximately equal group sizes.

• Basic randomization can be accomplished by altering information contained in the settings section for each arm of your particular study.

• Basic randomization should only be used if you have a medium to large “n” and if you expect that all participants will be randomized in a short period of time (less than 2 months).

Blocked Randomization:

• This procedure ensures close balance of participants in each group at any time during the study enrollment period. 

• A “block” is a small group of participants assigned to the arms within a study.  The minimum block size is equal to the number of study arms.

  • If the randomization is balanced (meaning that the same number of participants is assigned to each arm), the block size must be a multiple of the number of arms. For example, a four-arm study could have block sizes of 4, 8 12, etc.  In a block of size 4, 1 participant is assigned to each arm.  In a block of size 8, 2 participants are assigned to each arm.

  • If the randomization is unbalanced (meaning that some arms will have more participants than other arms – see “Arm weights” below), the block size must be a multiple of the sum of the arm weights.  For example, a three-arm study in which two intervention arms each accrue twice as many subjects as the control arm (arm weights of 2, 2, 1) could have blocks sizes of 5, 10, 15, etc.  In a block of size 5, 2 participants are assigned to intervention 1, 2 participants are assigned to intervention 2, and 1 participant is assigned to control.  In a block of size 10, 4 participants are assigned to intervention 1, 4 participants are assigned to intervention 2, and 2 participants are assigned to control. 

• For studies will a small “n”, you are advised to use smaller block sizes (e.g., 2 and 4 for a study with 2 arms and 30 per arm). For larger studies you may use larger block sizes (e.g., 4 and 8 for a study with 4 arms and 200 per arm).

• Blocked randomization should be used in most circumstances and is preferred in order to keep randomization balanced over time. If your anticipated sample size is small or you anticipate a long recruitment period you should use blocked randomization to avoid confounding by temporal trends (e.g., changes in practice patterns over time or a shift in the kinds of patients being referred).

Blocked Randomization with Stratification:

• This approach to randomization is the same as blocked randomization, except that prior to randomization, participants are placed into strata (i.e., categories) on the basis of one or more chosen criteria (e.g., gender, BMI, age); they are then randomized to study arms according to set block sizes as described above. Participants will be allocated proportionally to each intervention while controlling for the chosen criteria so that there is an equal number of participants with each stratification factor in each intervention arm. For example, if you want to use gender as a stratification factor, this process will allocate an equal number of women and men to each intervention arm.

• On the randomization screen select "Block Randomization With Stratification" and click the "Update Configuration" button.

• Once this has been done a new link will appear in the Study Configuration navigation titled "Configure Strata"

• Clicking on this link will take you to a page where you can configure the strata groups for your study. To create a stratum click on the "New" button at the top of the page. When creating strata there are two parameters that can be set: "name" and "block sizes." The name parameter is used across the site as a way to differentiate different strata from each other. The block sizes parameter determines what the potential block sizes are for that stratum.

• On the main strata configuration page each stratum for a study is displayed, as well as the total number of participants that have been randomized into each arm in the current randomization block, the totals for all blocks, and the potential block sizes for each stratum.

• Once all the needed strata have been created the feedback tool can be used to configure how participants are placed in strata. Any encounter during enrollment can be used to assign a participant into a stratum.

Arm Weights:

• All types of randomization procedures can have weights assigned to each arm.

• By default, each arm has a weight of 1. Since all arms have the same weight the arm weights are irrelevant and randomization will happen as expected for the selected randomization option.

• Once an arm has been assigned a weight, all other arms will default to a weight of 0 unless manually assigned. If an arm has a weight of 0 no participants will be randomized in to that arm.

• Arm weights are useful only when you must have arms of different sizes. 

• Example: a two arm, 150 participant study consisting of a control arm and an incentive arm

  • If we want a larger control group than incentive group, we could set the following arm weights

    • Control arm: 2

    • Incentive arm: 1

  • Since the control arm has twice the weight of the incentive arm participants are twice as likely to end up in that arm than the incentive arm. In order to maintain proper randomization we would need the control arm to be twice as large as the incentive arm.

  • If this study were to be blocked, the block sizes would have to be a multiple of 3 (control weight 2 + incentive weight 1).  A block of size three would have 2 control participants and 1 incentive participant.  A block of size 9 would have 6 control participants and 3 incentive participants.

Understanding Caps:

• In W2H you will have to select the size of each of your arms.  However, it is important to set the cap of each arm a bit higher than the target, so that study staff will continue to remain blinded to which arm each participant is being randomized as full enrollment is approaching and up until the last participant is enrolled.  Close attention should be paid to when you are approaching full enrollment so that enrollment can be ceased as soon as you’ve reached your enrollment goal.

Testing Randomization before study launch:

• The most effective way to ensure randomization is working as you would expect is to enroll a reasonable number of participants and observe what arms they are assigned to. Ensure that the proportions assigned to each arm are as expected, and if using blocking, verify that the desired block sizes are appearing.

W2H role permissions and access to randomization details:

• W2H Admins can see the current state of the blocks. 

• W2H Coordinators can’t see current state of the blocks so they will remain blinded to how the blocks are being filled