In a study supported by the NIH Collaboratory, researchers found that imbalance in individual-level baseline covariates influences bias in the observed treatment effect in cluster randomized trials. Using race as an example, the study highlights the importance of reducing covariate imbalance in the design stage of cluster randomized trials and of using statistical analysis techniques to minimize the resulting bias.
The innovative study, published in Contemporary Clinical Trials, used computer simulation models validated by real-data simulations from a large clinical trial to examine the influence of baseline covariate imbalance on treatment effect bias. They found that bias was proportional to the degree of baseline covariate imbalance and the covariate effect size. In the simulations, trials with larger numbers of clusters had less covariate imbalance. Statistical models that adjusted for important baseline confounders were more effective than unadjusted models in minimizing bias.
The authors recommend several design approaches and statistical analysis techniques for both reducing covariate imbalance and minimizing bias. Using the results of available prior data can help researchers identify important baseline confounders when designing cluster randomized trials.
This work was supported within the NIH Collaboratory by the NIH Common Fund through a cooperative agreement from the Office of Strategic Coordination within the Office of the NIH Director, and by a research supplement from the NIH Common Fund to promote diversity in health-related research.