To avoid inflation in the rate of type 1 error, or false positives, in individually randomized group treatment (IRGT) trials, researchers should choose an analytic model that accounts for the correlations in outcome measures that arise when study participants receive an intervention from the same source, according to a report from the NIH Pragmatic Trials Collaboratory’s Biostatistics and Study Design Core.
The report was published online ahead of print in Statistics in Medicine.
Many IRGT trials randomly assign individuals to study arms but deliver the study intervention through shared “agents,” such as clinicians, therapists, or trainers. After randomization, interactions between participants who share the same agent can lead to correlations in study outcomes. The delivery agents may be nested in or crossed with study arm, and participants may interact with a single agent or multiple agents. There has been no systematic effort to identify the appropriate analytic models for these complex study designs.
To address this knowledge gap, members of the NIH Collaboratory’s Biostatistics and Study Design Core conducted a simulation study to examine the performance of a variety of analytic models for IRGT trials in which complex clustering arises from participants interacting with multiple agents or single agents in both nested and crossed designs. They found substantial inflation in the type I error rate in studies with nested designs when the analytic model did not account for participants interacting with multiple agents.
This article is the latest in a series of reports completed this year by members of the Biostatistics and Study Design Core to explore analytic approaches to clinical trials with complex clustering and other novel design features:
- Designing Individually Randomized Group Treatment Trials With Repeated Outcome Measurements Using Generalized Estimating Equations
- Robust Analysis of Stepped Wedge Trials Using Composite Likelihood Models
- Planning Stepped Wedge Cluster Randomized Trials to Detect Treatment Effect Heterogeneity
Lead author Jonathan Moyer, a statistician in the NIH Office of Disease Prevention, led a discussion of complex clustering in pragmatic trials in a session of the NIH Collaboratory’s weekly Rethinking Clinical Trials webinar series: “The Perils and Pitfalls of Complex Clustering in Pragmatic Trials.”
Learn more about the NIH Collaboratory’s Biostatistics and Study Design Core.