ABSTRACT
Background In May 2021, the US Food and Drug Administration (FDA) released a revised draft guidance for industry on ''Adjustment for Covariates in Randomized Clinical Trials for Drugs and Biological Products.'' This guidance discusses adjustment for covariates in the statistical analysis of randomized clinical trials in drug development programs. It specifically focuses on the use of prognostic baseline factors to improve precision for estimating treatment effects. The impact depends on the specifics of the trial, but typical sample size reductions range from 5-25% (at no cost). Despite regulators such as the FDA and the European Medicines Agency recommending covariate adjustment, it remains highly underutilized leading to inefficient trials in many disease areas. This is especially true for binary, ordinal, and time-to-event outcomes, which are quite common in COVID-19 trials and are, moreover, prevalent as primary outcomes in many disease areas (e.g. Alzheimer's disease and stroke). Research and guidance on this topic could therefore not be more timely. In response to the FDA draft guidance on covariate adjustment, this session invites experts who represent a variety of viewpoints, coming from academia and Pharmaceutical industry. The aim of this session is to provide insight into the state-of-the-art methods at a high level and from a practical perspective. We moreover want to discuss the main obstacles that lead to the underutilization of covariate adjustment, all of which we aim to surmount in this session. Finally, we want to discuss the connections of the different talks to the FDA draft guidance and provide options for better practice. Talk by Min Zhang ''Covariate adjustment for randomized clinical trials when covariates are subject to missingness.'' One practical issue that may have limited the use of covariate adjustment is that covariates are often subject to missingness. Existing statistical methodologies often ignore this issue and assume covariates are completely observed. We discuss conditions under which robust covariate adjustment can be achieved when the missingness of covariates is present. We study various methods for handling missing data and compare their performances in terms of robustness and efficiency through comprehensive simulation studies. Recommendations on strategies for handling missing covariates to achieve robust covariate adjustment are provided. Talk by Mark van der Laan on ''Targeted Learning of causal effects in randomized Trials with continuous time-to-event outcomes.'' Targeted maximum likelihood estimation (TMLE) provides a general methodology for estimation of causal parameters in the presence of high-dimensional nuisance parameters. Generally, TMLE consists of a twostep procedure that combines data-adaptive nuisance parameter estimation with semi-parametric efficiency and rigorous statistical inference obtained via a targeted update step. In this talk, we demonstrate the practical applicability of TMLE for standard survival and competing risks settings where event times are not confined to take place on a discrete and finite grid. We demonstrate TMLE updates that simultaneously target point-treatment-specific survival curves and treatmentcause- specific subdistributions in the competing risk setting, across treatment and time points. We consider the case that we only observe baseline covariates as well as the case that we also track time-dependent covariates that potentially inform censoring/drop-out. This results in estimates that are not only fully efficient, but also respect the natural monotonicity of survival functions and cause-specific subdistributions. It moreover makes sure that the sum of subdistributions and survival equals 1. We propose a super-learner for the causespecific conditional hazards that incorporate many possible Cox models as well as a variety of highly adaptive Lasso estimators. Asymptotic theoretical guarantees are given and finite-sample robust performance is demonstrated with simulations. We illustrate the usage of the considered methods for a ovo Nordisk Leader study as well as for publicly available data from a trial on adjuvant chemotherapy for colon cancer. Talk by Kelly Van Lancker on ''Combining Covariate Adjustment with Information Monitoring and Group Sequential Designs to Improve Randomized Trial Efficiency'' In this talk, we focus on the knowledge gap in statistical methodology that leads to the underutilization of covariate adjustment. A first obstacle is the uncertainty of its efficiency gain and corresponding sample size reduction at the design stage;an incorrect projection of a covariate's prognostic value risks an over- or underpowered future trial. A second open problem is the incompatibility of many covariate-adjusted estimators with the commonly used group sequential, information-based designs (GSDs). To overcome these challenges, we suggest combining covariate adjustment with information monitoring and continuing the trial until the required information level is surpassed. Since adjusted estimators typically have smaller variance than standard estimators, the information accrues faster leading to faster trials. Building on this, we propose a new statistical method that orthogonalizes estimators in order to (1) have the independent increments property needed to apply GSDs and (2) simultaneously improve (or leave unchanged) the variance at each analysis. Such a method is needed in order to fully leverage prognostic baseline variables to speed up clinical trials without sacrificing validity or power. We prove that this method has properties such as the independent increments, consistency, asymptotic normality, and correct type I error and power, and evaluate its performance in simulations and data analyses. Discussion by Frank Bretz This talk will discuss connections between the three previous presentations in the session and recommendations in the May 2021 FDA revised draft guidance for industry document on ''Adjustment for Covariates in Randomized Clinical Trials for Drugs and Biological Products.'' It will moreover touch on the broad impact of covariate adjustment for the pharmaceutical industry and provide advice on better practice.
ABSTRACT
The COVID-19 pandemic continues to affect the conduct of clinical trials globally. Complications may arise from pandemic-related operational challenges such as site closures, travel limitations and interruptions to the supply chain for the investigational product, or from health-related challenges such as COVID-19 infections. Some of these complications lead to unforeseen intercurrent events in the sense that they affect either the interpretation or the existence of the measurements associated with the clinical question of interest. In this article, we demonstrate how the ICH E9(R1) Addendum on estimands and sensitivity analyses provides a rigorous basis to discuss potential pandemic-related trial disruptions and to embed these disruptions in the context of study objectives and design elements. We introduce several hypothetical estimand strategies and review various causal inference and missing data methods, as well as a statistical method that combines unbiased and possibly biased estimators for estimation. To illustrate, we describe the features of a stylized trial, and how it may have been impacted by the pandemic. This stylized trial will then be revisited by discussing the changes to the estimand and the estimator to account for pandemic disruptions. Finally, we outline considerations for designing future trials in the context of unforeseen disruptions.Copyright © 2022 American Statistical Association.