October 21, 2016: Demographics, Subgroup Analyses, and Statistical Considerations in Cluster-Randomized Trials
Monique L. Anderson, MD, MHS, Division of Cardiology, Duke Clinical Research Institute, Duke University School of Medicine
Demographics, Subgroup Analyses, and Statistical Considerations in Cluster-Randomized Trials
Pragmatic clinical trials; PCT; Cluster-randomized trials; CRT; Heterogeneity of treatment effect; HTE; Race/ethnicity reporting; Demographic subgroup analysis; Systematic review; FDA; NIH
- Definitions for this topic:
– Cluster-randomized trials (CRTs) are experiments in which intact social units, rather than independent individuals, are randomly assigned to intervention groups.
– Treatment effect is the comparison between treatment groups in a trial and is usually measured by relative risk, odds ratio, or arithmetic difference.
– Subgroup analysis is an evaluation of treatment effects for a specific endpoint in subgroups of patients defined by multiple baseline characteristics. It is undertaken to investigate the consistency of trial conclusions across different subpopulations.
– Heterogeneity of treatment effect (HTE) refers to circumstances in which the treatment effects vary across levels of baseline characteristics (e.g., men/women, black/white). HTE is expressed in a statistical model as an interaction between the treatment group and baseline variable.
- As a trial approach, PCTs may be optimal for understanding how a treatment impacts demographic subgroups. But it is unclear how to address demographic subgroups in CRTs, if they should be addressed at all. Also, it is unclear how often CRTs in healthcare settings address baseline demographics in design and analysis of the trial.
HTE analyses in general are uncommon but are encouraged, and in some cases are mandatory.
When HTE analyses by demographic subgroups are expected, then imbalance issues may need to be addressed in the CRT.
It is important to consistently collect and report trial demographics and to have strategies to ensure balanced design and attention to optimal detection of HTE.
While data collection has improved, relying on administrative data in the EHR for race and ethnicity is not sufficient. Training of frontline personnel is necessary.
For More Information
For more information on NIH policies related to this topic, visit Inclusion of Women and Minorities as Participants in Research.
@PCTGrandRounds, @Collaboratory1, @PCORnetwork