December 2, 2016: Lessons Learned from the NIH Collaboratory Biostatistics and Design Core
Andrea J. Cook, PhD, Senior Investigator, Biostatistics Unit, Group Health Research Institute, and Affiliate Associate Professor, Department of Biostatistics, University of Washington
Lessons Learned from the NIH Collaboratory Biostatistics and Design Core
Biostatistics; Study design; Pragmatic trial; Cluster randomization; Stepped-wedge design; Statistical analysis; Sample size calculation; Constrained randomization; Outcome ascertainment
- It’s important to conduct pragmatic trials to move research quickly into practice—yet pragmatic trials add complication to the design and analysis. The first design question is, Can the study question be answered using a pragmatic trial approach?
- The choice of which quantity to estimate should be based on the scientific question of interest, but statistical tradeoffs, including power, must also be considered.
- Sample size calculations need to take into account variable cluster size; variability has potentially major implications for the trial’s power and analysis.
A constrained randomization method can be used to balance a large number of characteristics:
– First, simulate a large number of cluster randomization assignments (A or B but not actual treatment); remove duplicates.
– Across these simulated randomization assignments, assess characteristic balance; restrict to those assignments with balance.
– Randomly choose a randomization scheme from the “constrained” pool, and randomly assign treatments to A or B.
What are your recommendations on reevaluating/updating your approach to sample size calculations during the trial?
For More Information
Read more about biostatistics for pragmatic trials with links to related publications at the Biostatistics Core’s webpage.
@PCTGrandRounds, @Collaboratory1, @PCORnetwork