2023/04: Some challenges with implementing estimands in real life.

Session within the first Effective statistician conference.

This shortcourse will discuss the application of the ICH E9(R1) estimand addendum to oncology clinical trials. Using examples we will illustrate how the estimand framework supports transparent formulation of clinical trial objective(s) for following challenging drug development questions:

  • Complexity of treatment options and existence of potential curative procedures (such as e.g. CAR-T) or treatment sequence across different phases in hematological trials.
  • Application of the hypothetical strategy.
  • Treatment switching.
  • A very gentle introduction into the potential outcomes framework from causal inference. This is essential to understand why we actually randomize in clinical trials.
  • The use of a principal stratum estimand to assess the impact of post-randomization events such as development of antidrug antibodies on efficacy or safety of a new drug. Often, such type of questions are answered using simplified analyses that are only valid under strong assumptions. To make these assumptions transparent we will discuss estimands, estimation methods, and impact on data collection.

More on the oncology estimand WG: oncoestimand.org.

This shortcourse was initially presented at the 76th Annual Deming Conference on Applied Statistics by Evgeny Degtyarev (Novartis) and Kaspar Rufibach (Roche). The slidedeck that will form the basis for this shortcourse is available online here.