Instructor:
Kelley Kidwell, PhD (University of Michigan)

Workshop Dates and Times:
Recorded on August 13-14, 2024
Now available for self-paced learning

Workshop Format:
Two-Day Online Workshop Recording

In the United States, the Food and Drug Administration’s Orphan Drug Act defines a rare disease as one that affects less than 200,000 people. Although a given rare disease may affect only a small number, cumulatively, an estimated 7,000-10,000 rare diseases impact 25-30 million Americans and 236-446 million people globally. Only 11-15% of these diseases have had at least one product show potential for diagnosis, prevention or treatment under the orphan drug designation in the US, and only 4-6% have a product approved. A large gap must be bridged to adequately address the significant burden of rare diseases, but challenges in designing clinical trials create obstacles to meeting this urgent need.

Randomized controlled trials, and in particular blinded, parallel-group randomized trials, are considered the gold-standard for producing evidence-based conclusions about treatment efficacy. However, in a rare disease context, such trials are concomitant with practical and ethical concerns. Inherently small population sizes make it difficult or impossible to recruit a large enough sample to adequately power a randomized trial, and including a placebo arm may further complicate recruitment, as patients may try to avoid the possibility of not receiving treatment. Ultimately, these and other challenges result in rare disease trials being more likely than non-rare disease trials to be non-randomized, single-arm, and open-label.

In this workshop, we’ll discuss the current rare disease or small sample clinical trial landscape including cross-over, N-of-1 and adaptive designs. We’ll present designs, examples in the fields and analytic approaches.

Then, we’ll introduce small sample SMART (snSMART) designs. Sequential, multiple assignment, randomized trial (SMART) designs are often motivated to identify tailored sequences of treatments or adaptive interventions or dynamic treatment regimens (DTRs) in larger samples. SMARTs employ at least two randomizations in sequence where only some groups may be re-randomized based on response or other characteristics related to previous treatment. We have turned standard SMART designs and analyses on their head, and instead of focusing on DTRs, we apply the design to small samples to obtain more information from a small sample of individuals. This short course will provide an overview of small sample SMART (snSMART) designs with corresponding Bayesian and frequentist methods for analyses. The differences between snSMART and SMART designs will be highlighted and methods to analyze snSMART data, calculate sample size, add adaptive components, incorporate external data, and dose-find and confirm will be presented. Many of our methods are motivated by the current snSMART ARAMIS which seeks to find an effective treatment for individuals with isolated skin vasculitis, but the methods apply broadly to chronic, rare diseases that remain relatively stable over the trial period.