Dynamic Randomization — Striking a Balance

Case Studies

The Problem:

Randomization — that is, the truly random assignment of patients to a treatment group in a clinical trial — is the gold standard approach to producing statistically valid results in drug studies. Randomization is intended to eliminate bias — theoretically, the myriad of variable and potential variable differences between patients, both known and unknown, should even out over time in infinitely large samples. But study sample sizes in the real world are not infinite. In many cases, they’re not even what one would call ‘large,’ mathematically speaking. In real life scenarios, such as new drug trials, a study can wind up having a serious imbalance in prognostic factors and assignment to study treatment groups. Serious enough, in some cases, that imbalances can call study results into question.

The Solution:

Dynamic Randomization is an alternate strategy in the biostatistician’s toolbox. In a dynamic approach, there is no pre-existing randomization schedule as there is for studies using static randomization designs. Rather, the randomization table is dynamically built as the study proceeds. For each new patient to be randomized into the study, the IRT algorithm takes into account the balance/imbalance that will result from the assignment of the new patient to each treatment group, and assigns that patient to the group that will best serve the goal of balancing group size and, as study may require, key variables. Factors that might be considered include gender, age, stage of disease, or certain aspects of an individual’s past medical history. Dynamic randomization is sometimes referred to as a ‘minimization’ algorithm because it minimizes the imbalances in patient assignments. ICH has accepted the use of these dynamic techniques as a valid approach to randomizing subjects for drug trials.

Veracity Logic’s dynamic randomization algorithm is available as one of several randomization tools that are part of our standard CORE platform. When applied to a project, dynamic treatment assignment is tested as part of the standard project startup validation process. Three sample randomizations (with 150 subjects in each sample) are produced and provided to the client’s biostatisticians for review and approval prior to releasing the system for customer testing.

Forced Randomization …

Case Studies

Uh oh, we don’t have that drug type on the shelf…!

Clinical site A has been slow but steady enrolling subjects for your study, and the IRT resupply schedule has been working just fine, provisioning the site with just the right amounts of the two types of study drug used in the study. But an unexpected bolus of three eligible subjects has enabled the site to exceed its usual goals. Unfortunately, the randomization schedule has just assigned three kits of the same treatment group to the new subjects. Problem is, site A only has two kits of that group left in its inventory! You don’t want to lose the third subject. What to do?!

Veracity Logic’s IRT accommodates a variety of randomization schemas including “Forced Randomization”. In the Forced Randomization scenario, the inventory management module of the IRT ‘knows’ that site A is one kit short on the assigned treatment group. The system automatically skips ahead in the randomization schedule to the first available kit in a treatment group that is currently available in site A’s inventory. It then assigns that kit/treatment group to the subject. This ‘forcing’ of assignment is balanced by the system in subsequent randomizations in which the ‘skipped’ kit/treatment group is assigned at the next study opportunity.

A special notification sent to specified users documenting when a Forced Randomization assignment has occurred.

In other words, Veracity Logic’s IRT gives you the power and flexibility you need to meet real-life scenarios with real-life solutions.