In a new episode of the NIH Collaboratory Podcast, Drs. Jonathan Moyer and David Murray discussed their recent publication, “Evaluating Analytic Models for Individually Randomized Group Treatment Trials With Complex Clustering in Nested and Crossed Designs.” The episode was moderated by Patrick Heagerty, co-chair of the Biostatistics and Study Design core working group.

Listen to the podcast. For alerts about new episodes, subscribe for free on Spotify, Amazon Music, Apple Podcasts, or SoundCloud.

Individually randomized group treatment trials, or IRGT trials, are those in which participants are randomized to conditions individually but receive the intervention in a group format, delivered by shared agents. Prior to randomization, the individual outcome measures are all independent; after randomization, individual outcomes become correlated over time due to interactions with shared agents or groups.

Moyer noted that the IRGT design is likely more common than investigators realize.

“John and I worked on a project reviewing all of the clinical trials newly supported or approved by NIH in [Fiscal Year] 2023. And of all of the trials that involved individual randomization, about half were IRGTs and about half were RCTs… Only a couple of those IRGT trails were recognized,” said Moyer.

Even if an IRGT trial is recognized as such, identifying the proper analytic model can be tricky. There are multiple dimensions to consider. The trials may be fully nested, with  agents in both arms, or partially nested, with agents in only one arm. In a crossed design, the agents interact with both arms.

The investigation found substantial type I error rate inflation in nested designs when analytic models did not account for multiple membership and when analytic model weights characterizing the association with multiple agents did not match the data generating mechanism. In the podcast, Moyer and Murray translated their findings into guidance for investigators.

“ If investigators have a choice between a nested or crossed design, and it’s not expected that there’s going to be a lot of cross-arm contamination, then a crossed design might be a good choice over the nested design,” said Moyer.

Moyer is a statistician in the NIH Office of Disease Prevention. He is a longtime member of the NIH Pragmatic Trials Collaboratory’s Biostatistics and Study Design Core. Murray is the NIH associate director for prevention and the director of the Office of Disease Prevention. He is a longtime member of the NIH Pragmatic Trials Collaboratory’s Biostatistics and Study Design Core.