Randomization Configuration

Owned by John Paliwoda
Last updated: Dec 08, 2015 by Mary Lukowski

Overview

StudyTRAX supports two randomization methods: Simple and Permuted Block.

If the Simple randomization method is employed, when a subject is randomized one of the Treatment Groups is randomly selected based on the weight of each Treatment Group.

If the Permuted Block randomization method is employed, the subject is first placed in a subgroup based on their values for the stratification factors and optionally their site. If there are no stratification factors and subjects are not stratified based on site, then there will only be one subgroup that all subjects are assigned to. When a subject is randomized, if there is no active Block Size, one is randomly selected from the list of Block Sizes (if there is more than one specified). During randomization a treatment group is randomly selected based on each treatment's weight, the treatments of other subjects in the same subgroup, and the remaining slots available within the current block size. The result being that when all the slots in a block size are filled for a particular subgroup of subjects, each treatment will be represented in proportion to its weight.

Data Elements

Blinding methods
  1. None: No blinding is done and the Treatment Name is shown for randomized subjects on the Subject Overview page. No special role is required to export the Treatment Name or Treatment Code.
  2. Double: Only the Treatment Display Name is shown for randomized subjects on the Subject Overview page. A special role is required to export the Treatment Name or Treatment Code or to break the blinding, in an audited action. The Treatment Display Name can be the same for every Treatment so users will not be able to discern whether two subjects have the same Treatment.

  3. Double (with Randomization ID): A Randomization Id will be displayed and not any information about the treatment. A special role is required to export the Treatment Name or Treatment Code or to break the blinding, in an audited action.

    The Treatment Display Name can be exported by users with the Data Set role, so it can be determined if two subjects had the same treatment. If this is not desired, use the same Treatment Display Name for each Treatment.

Randomization ID Format

Available to be set if Blinding Method Double (with Randomization ID) is selected. This dictates the format of the Randomization ID that will be displayed for a randomized subject. There are two special variables available for use in the Randomization ID Format

  1. {SiteCode} : The subject's actual Site Code replaces this text.

  2. {Seq:0000}: The number representing the subject's sequence in being randomized replaces this code. For example, if the subject is the 20th subject to be randomized, this string will be replaced with "0020".

    If a subject is un-randomized, their sequence number is not reused.

All other text is interpreted as literals. When a subject is randomized, the Randomization ID is created based on the format specified.

Randomization Seed

The seed to use for the random number generation function. If a seed is used, random number generator will generate the same sequence of random numbers and block sizes across each strata.  Thus, only applicable if stratification factor(s) used.

  • If Seed Used - Fixes the treatment asignment and block size sequence order to be the SAME within each stratification group.
Randomization Method
  1. Simple: Subjects are randomly assigned a treatment.
  2. Permuted Block: The subjects within a subgroup are randomly assigned a treatment but the assignment probability is constrained by the number of slots allocated to the treatment (based on weight) available within the current block size.
Block Size

If the Randomization Method is Permuted Block, one or more Block Sizes must be provided. The Block Size ensures that all treatments are represented in proportion to their weight after a certain number (the value of the current Block Size) of subjects within a subgroup are randomized. If more than one Block Size is entered, a Block Size is selected at random each time a new Block Size is needed (that is, when the current Block Size is filled). Each Block Size entered must be a multiple of the sum of Treatment Weights. For example, if there are two Treatments, one with a weight of 2 and the other with a weight of 1 (sum of weights is 3), then the Block Sizes entered must be a multiple of 3 (e.g., 3, 6, 9, etc.).

Is Site Specific

If the Randomization Method is Permuted Block, selecting this option means subjects will be placed in subgroups by Site. If Stratification Factors are defined, this essentially adds Site as a Stratification Factor.

Treatment Groups

The definition of the treatments available for a subject to be assigned to during randomization. See the section above on Blinding Methods for details on how the fields are used. The Weight is only used if the Randomization Method is Permuted Block, in which case it defines the Weight a Treatment is given within a Block Size. For example, if the current Block Size is 3 and there are two Treatments: A with a Weight of 2 and B with a Weight of 1, then after the first three subjects within the same subgroup are randomized, two will be assigned Treatment A and one will be assigned Treatment B.

Stratification Factors

Stratification Factors are the variables whose values are used to create subgroups of subjects. For example, if Gender and Race are selected as Stratification Factors, then subjects are placed in the same subgroup if they have the same value for the Gender and Race variables. Stratification Factors are added using the Variable Finder tab. Note: if "Is Site Specific" is selected, site is essentially added as a Stratification Factor.

Randomization class

StudyTRAX utilizes the Microsoft.Net Random class to create the random numbers used to select the treatment code. The current implementation of the Random class is based on a modified version of Donald E. Knuth's subtractive random number generator algorithm. For more information, see D. E. Knuth. The Art of Computer Programming, Volume 2: Seminumerical Algorithms. Addison-Wesley, Reading, MA, third edition, 1997.