Thank you for joining us for another episode of the Guidance Recap Podcast. My name is Kylie Haskins, and I am today’s host. In this episode, I am excited to be talking with Dr. Dan Rubin, who is a statistician in the Division of Biometrics IV in CDER’s Office of Biostatistics. Dr. Rubin will be sharing some thoughts with us on the newly published final guidance titled, “Adjusting for Covariates in Randomized Clinical Trials for Drugs and Biological Products.” Welcome, Dr. Rubin! Thank you for speaking with us today.
Can you explain to the audience why adjusting for covariates is important and provide some of the reasons that FDA issued this document?
Sure – for listeners less familiar with this area it is common in randomized controlled trials to measure information on participants before randomization occurs, such as demographic information and disease characteristics. These measurements are called baseline covariates. A covariate adjusted analysis is a comparison of outcomes between treatment groups that tries to account for this information when estimating and performing statistical inferences for treatment effects. For example, this could involve controlling for random differences in ages between treatment groups. When adjusting for baseline covariates that are prognostic for the outcome, an adjusted analysis usually leads to more precise estimation and greater statistical power than an unadjusted analysis. Additionally, covariate adjustment does not require the collection of new data, as it makes use of information already collected during the study. The FDA issued the document to encourage use of covariate adjustment and provide more recommendations on how it can be implemented.
When should sponsors adjust for covariates during drug development?
Sponsors should decide to use covariate adjustment ideally before the clinical trial starts, but definitely by finalization of the statistical analysis plan and before unblinding results. We recommend that sponsors look for factors or characteristics that will be measured in the trial that might be prognostic for treatment outcomes. For example, in a trial for a type 2 diabetes these factors could include body mass index, hemoglobin A1c measurements, or other known comorbidities. Randomization by chance can lead to an imbalance of subjects with these variables between the treatment groups in the trial. Covariate adjustment attempts to account for this to improve the estimator of the overall treatment effect.
Can you give us a brief overview of the evolution of this guidance?
Yes, the first draft version of this guidance published in April 2019 and was titled “Adjusting for Covariates in Randomized Clinical Trials for Drugs and Biologics with Continuous Outcomes.” This initial version considered covariate adjustment for trials using endpoints measured on a continuous scale such, as blood pressure, and provided recommendations for covariate adjustment using only linear models. We received positive comments on the guidance and also received requests for additional information.
In response to these comments we revised the draft guidance in 2021 to elaborate on previous recommendations. We also provided additional recommendations on adjusting for covariates in more general circumstances such as trials where the primary endpoint is a binary measure of success or failure (such as survival) or trials in which the primary endpoint is an ordinal scale or a time to event outcome. The revised draft guidance also expanded the recommendations for covariate adjustment beyond linear models to include nonlinear statistical models as well. The FDA encouraged stakeholders to submit comments on the revised draft guidance to the Federal Register.
We considered the feedback to the revised draft guidance to be largely positive, and the final guidance does not make large substantive changes. We have made several clarifications, corrections, and updates requested by stakeholders. We have also given other examples of methods for covariate-adjusted estimation, inference, and testing so that the final guidance will be less prescriptive about recommended techniques. Finally, we have added material on selecting covariates for adjustment, alignment with the estimand framework discussed in other guidances, and several additional topics.
How do you anticipate this guidance will affect external and internal stakeholders?
We hope that this guidance will be well received by both internal and external stakeholders. The intention of the guidance is to promote the use of covariate adjustment and correct a few misconceptions around its use. External stakeholders will welcome this information. While the FDA encourages covariate adjustment for external stakeholders, it is certainly not required, and a simple unadjusted analysis is still accepted to support a new drug application. For internal stakeholders we anticipate that FDA staff will be grateful to have the recommendations to provide sponsors when questions on adjusting for covariates arise.
For our final question, what are a couple of key items that you especially want listeners to remember?
I really want listeners to remember that the FDA encourages covariate adjustment because we believe that it is a low hanging fruit that can be used to improve the efficiency of a clinical trial analysis without creating additional burdens for sponsors. We encourage sponsors to discuss covariate adjustment with the FDA during the development of the protocol, particularly for situations not explicitly covered in the guidance.
Dr. Rubin, thank you for taking the time to share your thoughts on the final guidance on adjusting for covariates. We have learned so much from your experience and insights on this document. We would also like to thank the guidance working group for writing and publishing this final guidance.
To the listeners, we hope you found this podcast useful. We encourage you to take a look at the snapshot and to read the guidance.