Adaptive Randomization

Overview

Randomization serves as our causal mechanism to draw conclusions about the effect of an intervention in a clinical trial. However, there are many approaches to randomization including both static (i.e., fixed) and dynamic (i.e., changing) ratios. In this module we first do a brief review of static randomization approaches before diving into three unique adaptive randomization (AR) approaches to modify a study’s allocation ratio: baseline covariate AR, outcome/response AR, and information balance AR.

Slide Deck

 

You can also download the original PowerPoint file.

Code Examples in R

Various packages exist to assist in implementing randomization approaches in R:

• randomizeR: implements static randomization schema, with a Journal of Statistical Software article to follow along for more information
• carat: implements covariate adaptive randomization designs (six different strategies included as of package version 2.2.1), with support to implement the appropriate statistical analysis after data has been collected
• RARfreq: implements response-adaptive randomization procedures
• Information balance AR approaches generally have to be custom coded with the choice of information borrowing software.

References

Below are some references to highlight based on the slides and code:

• FDA Adaptive Design Clinical Trials for Drugs and Biologics Guidance for Industry Guidance Document: FDA guidance document on adaptive trial elements

• Recent innovations in adaptive trial designs: A review of design opportunities in translational research: 2023 review paper examining adaptive and novel trial elements with included case studies