U.S. flag An official website of the United States government
  1. Home
  2. Science & Research
  3. Science and Research Special Topics
  4. Advancing Regulatory Science
  5. Statistical methods to improve precision and reduce the required sample size in many phase 2 and 3 clinical trials, including COVID-19 trials, by covariate adjustment
  1. Advancing Regulatory Science

Statistical methods to improve precision and reduce the required sample size in many phase 2 and 3 clinical trials, including COVID-19 trials, by covariate adjustment

CERSI Collaborators: Michael Rosenblum, PhD; Joshua Betz, MS; Kelly Van Lancker, PhD; Bingkai Wang, PhD

FDA Collaborators: Daniel Rubin, PhD, CDER; Greg Levin, PhD, CDER; Boguang Zhen, PhD, CBER; Gene Pennello, PhD, CDRH

Project Start Date: March 1, 2020

Regulatory Science Challenge

Investigators are addressing the following FDA research priority: “Developing methods and tools to improve and streamline clinical and post-market evaluation of FDA-regulated products.”

Project Description and Goals

Clinical trials are often conducted to learn whether new medical treatments are safe and effective. Data collected when participants first enter a trial are called baseline variables. Examples of baseline variables include age, sex, and disease severity. Sometimes, due to chance, there are imbalances in these variables between those assigned to the experimental treatment arm (or group) and those assigned to the control arm (or group). For example, in some trials participants in the treatment arm may have higher or lower baseline disease severity compared with participants in the control arm.

When baseline variables (e.g., older age as a risk factor for worse outcome among those with COVID-19 infection) are related to the outcome, (e.g., one intended to support effectiveness of a treatment), taking the baseline variables into account in the data analysis of a trial’s results can lead to more precise estimates of treatment effectiveness (i.e., smaller standard errors for the estimates). More precise estimation means that at the planning stage, trials could be conducted with fewer participants than when baseline variables are not considered in the data analysis. Unfortunately, in many clinical trials information in baseline variables is not considered, leading to a greater chance that the trial will fail due to greater uncertainty regarding conclusions that can be drawn, potentially wasting resources. At the planning stage, not accounting for baseline information may lead to the planning of a larger trial size and longer trial duration than is necessary.

A major barrier to using baseline variables (called covariate adjustment) is that for many common types of clinical trial outcomes, such as binary, ordinal, and time-to-event outcomes, confusion remains as to what statistical approach is appropriate. The results of this project will help to overcome this barrier by demonstrating how to appropriately adjust for baseline variables to improve precision of treatment effect estimates for outcomes of these types. The statistical adjustments will be demonstrated on case studies in several disease areas as examples of best practices for the use of baseline variables in clinical trial data analysis. Investigators plan to disseminate these case studies to the public through a free, online tutorial on the FDA public website.


Williams, N., Rosenblum, M. & Díaz, I. (2022) Optimising precision and power by machine learning in randomised trials with ordinal and time-to-event outcomes with an application to COVID-19. Journal of the Royal Statistical Society: Series A (Statistics in Society), 1– 23. https://doi.org/10.1111/rssa.12915

Wang, B., Susukida, R., Mojtabai, R., Amin-Esmaeili, M., and Rosenblum, M. (2021) Model-Robust Inference for Clinical Trials that Improve Precision by Stratified Randomization and Adjustment for Additional Baseline Variables. Journal of the American Statistical Association, Theory and Methods Section. https://www.tandfonline.com/doi/full/10.1080/01621459.2021.1981338

Kelly Van Lancker, Joshua Betz, Michael Rosenblum. Combining Covariate Adjustment with Group Sequential, Information Adaptive Designs to Improve Randomized Trial Efficiency. Under review: https://arxiv.org/abs/2201.12921

Back to Top