Abstract
This issue of Clinical Trials includes proceedings of the 17th annual conference on statistical issues in clinical trials, a day focused on the ubiquitous problem of covariate adjustment in randomized clinical trials. Publication of the proceedings of our conference began in 2010 and includes invited papers from our speakers along with edited morning and afternoon panel discussions, with audience members, both in-person and online.1–16 Table 1 lists our speakers and panelists.
2025 University of Pennsylvania conference on statistical issues in clinical trials: covariate adjustment in randomized clinical trials: new methods and applications.
Michael Rosenblum (Johns Hopkins University) contributed but was unable to join due to illness.
Dr Mehrotra is primarily associated with Merck & Co; Dr Shaw is primarily associated with Kaiser Permanente Washington Health Research Institute.
I will start by highlighting a memorable comment from one of our speakers. Dr Laura Balzer spoke of how her work can impact people living with HIV, both internationally and here at home, offering a quote from one of her high-school teachers: The world should be better because you’re in it. It’s that simple. Well, simple to say, but oh so difficult to do because it means you also need to be clear-sighted enough to challenge those things that may have been handed down but aren’t right, those things that don’t fit the sense of justice and community we espouse. It means being bold enough to change the things that need changing just because it is right and because you care about leaving the world better than you found it.
Dr Balzer reminds us of the opportunities we have to embrace a broader purpose—to promote statistical innovation and rigor, while always keeping in mind the patients and communities we serve.
With this perspective in mind, I turn to themes of the day. Our keynote speaker, Dr Stuart Pocock, began with a thoughtful overview based on his extensive experience with covariate adjustment in clinical trials and the potential to improve efficiency. Dr Pocock discussed commonly used methods and the rationale behind their use and demonstrated varied levels of efficiency gains in real-world case studies. He touched on regulatory issues which were subsequently expanded upon by Dr Daniel Rubin, who presented current key US Food and Drug Administration guidance on covariate adjustment in randomized clinical trials for drugs and biologics, including the importance of pre-specifying covariates, the number of covariates relative to sample size, and settings in which adapted standard errors should be computed, with a discussion of both linear and non-linear models. He demonstrated results regarding the efficiency, benefit, and robustness of analysis of covariance (ANCOVA) and commented on areas of interest with respect to further research. Dr Kelly Van Lancker demonstrated how covariate adjustment can improve efficiency in specialized trial designs combining principles from group sequential and information-adaptive designs. Our final speaker of the morning, Dr Anqi Zhao, addressed covariate adjustment in randomized experiments with missing data, pointing out that while covariate adjustment can yield efficiency gains, missing data often hinder its use in practice. She found that, at least asymptotically, propensity score weighting offers advantages over covariate adjustment when dealing with the missingness.
In the afternoon session, Dr Ting Ye discussed key principles for covariate adjustment, including estimand-focused analyses, assumption-lean robustness, and fit-for-purpose variance estimation. She highlighted several specific methods of interest including ANHECOVA (Analysis of Heterogeneous Covariance Effects) and the doubly robust augmented inverse probability weighted (AIPW) estimators. Dr Bingkai Wang followed with a talk about covariate adjustment when rerandomization, equivalently covariate-constrained randomization, was used to ensure balance and discussed asymptotic properties of several estimators, showing explicitly that covariate adjustment can provide efficiency gains beyond those achieved with rerandomization. Our final speaker, Dr Laura Balzer, described the process of working with the study team of a community-level cluster-randomized HIV trial in Africa to validly implement an adaptive pre-specification method for covariate adjustment using targeted minimum loss estimation (TMLE). The method yielded a reduced standard error of the estimator and with the utility of this machine-learning approach demonstrated for multiple trial examples.
Drs Courtney Schiffman, Frank Harrell, and Dylan Small led the morning and afternoon panel and audience discussions. This year’s discussions were spirited; controversial topics included the use of marginal versus conditional estimands and the role in practice of newer machine-learning and causal-inference–based approaches versus more traditional model-based approaches. Dr Jeffrey Morris (Director, Biostatistics Division, Department of Biostatistics, Epidemiology & Informatics, UPenn) delivered closing remarks noting that covariate adjustment, while long recognized as a way of improving trial efficiency, has not always been optimized in practice and that our conference showed how theory and practice advance hand-in-hand, where real-world trial complexities drive the development of new methods, while theoretical insights offer the structure and justification needed for their effective application.
I hope you enjoy these papers and edited transcripts of the panelists’ comments and audience discussions.
Footnotes
Acknowledgements
My thanks to the program committee: Drs Susan Ellenberg, Wei-Ting Hwang, Yimei Li, Devan Mehrotra, Michael Proschan, Nicholas Seewald, and Pamela Shaw and our Administrative Director, Ms Marissa Fox, for outstanding support. The Department of Biostatistics, Epidemiology and Informatics at the Perelman School of Medicine, University of Pennsylvania, provided the conference facility and administrative support, and the American Statistical Association the Society for Clinical Trials provided in-kind support for the conference. Additional financial support was provided by Merck & Co. I am grateful to colleagues at Memorial Sloan Kettering Cancer Center, Dr Colin Begg (Editor in Chief) for guidance, and Ms Katherine Cheung for expert editorial assistance. And finally, I thank the wonderful reviewers who volunteered to provide thoughtful and important comments to our authors.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.