Understand the concepts used in minimization

Minimization can be particularly useful for studies that aim to balance subjects across treatment groups. A key aspect of working this this type of randomization design is understanding the main concepts that it operates with.

As a non-random method of treatment allocation in clinical studies, minimization aims to balance treatment groups by subject factors (ensuring that common subject characteristics have a balanced weight across stratum factors). In recent history, it was proven that minimization provides better-balanced treatment groups given its flexibility to integrate more demographic or prognostic factors into its algorithm.

Let's go through the key concepts that define minimization's algorithm. Each main concept is tied to a product area where one of the three involved user roles perform a different task for creating and setting up minimization.

For the purpose of this exercise, we will describe the concepts in the context of an oncology study that uses minimization with five stratum factors: Country, Age, Gender, Race, and Cancer Stage.

Weight

In Oracle Clinical One Platform, weight represents an even number (from 0 to 99) that you must use to define the importance of each minimization stratum factor. How you define the weight of each stratum factor determines how each subject is assigned to a treatment arm and what drug they ultimately receive during the course of a study. In other words, the higher the weight number of a stratum factor is, the higher the priority it has when balancing subjects across treatment groups.

Using a sequential number for defining the weight is not mandatory. Moreover, a study designer typically defines the weight number based on the study protocol and a statistician's recommendations.

For example, as a study designer, you define the Weight number of a Cancer Stage minimization stratum factor as the highest among a list of other stratum factors, such as Country, Age, Gender, and Race.

Two new subjects of similar age, gender, and country are enrolled into the study. To balance out treatment groups, the two subjects will be assigned to two different treatment arms based on the severity of their illness (the Cancer Stage stratum factor).

For step-by-step instructions on how to define a minimization design, see Define the minimization.

Probability factor

When it comes to minimization, the probability factor is a measure or an estimate of the possibility one subject may have to be assigned to a certain treatment arm and receive a certain type of drug during the course of a study. This probability is measured on a scale from zero (impossibility) to one (certainty).

In Oracle Clinical One Platform, the probability factor is a decimal number included in a randomization list. The probability factor must be between 0 and 1 (exclusively) and can only have up to a maximum of six decimal places.

For example, in an oncology study, as a clinical supply manager, you can define the probability factor of a first subject being assigned to Treatment Arm A as 0.728, based on the associated stratum factors. Then, as you go further down the randomization list, you can alternate between a number closer to 1 (certainty) and a number closer to 0 (impossibility) for the probability factor associated with each subject and randomization number.

Note:

In general (and especially if your study contains three treatment arms), we recommend a randomly assigned probability factor, not an alternating one.

For step-by-step instructions on how to upload a randomization list, see Upload a randomization list for minimization.

Minimization cohorts

Minimization cohorts represent the groups of subjects that are created in the system based on stratum factors defined by a study designer.

In Oracle Clinical One Platform, minimization cohorts are associated with each minimization design they belong to. Minimization cohorts can be enabled for enrollment and randomization of subjects in the study.

For example, in an oncology study, as a study manager, you can enable all minimization cohorts for enrollment and randomization (Country, Age, Gender, Race, and Cancer Stage) and then define a Randomization Limit of 50 and a Notification Limit of 50%. That way, every time the limit of 25 subjects is reached in each minimization cohort, the study manager (or any other assigned study team member) receives a notification about this reaching this limit.

For step-by-step instructions on how to configure settings for minimization cohorts, see Specify enrollment settings for minimization cohorts.