Master's degree in Statistics and Operations Research

Adaptive Designs for Confirmatory Clinical Trials

Date:

June 25 to 28. MORNING: 9 to 13h (June 25, 26 and 27) and 9 to 12h (June 28).

Instructor

Werner Brannath

I am Professor of Applied Statistics and Biometry at the Faculty of Mathematics and Computer Science at the University of Bremen and Head of the Department of Biometry at the Competence Centre for Clinical Studies Bremen (KKSB) and member of the Board of Directors of the KKSB.

I studied mathematics and physics in Karlsruhe and Vienna and was a research assistant at the Institute of Statistics at the University of Vienna and the Institute of Medical Statistics at the Medical Faculty.

Before coming to Bremen, I was an Associate Professor at the Medical University of Vienna. I also spent a year on a research fellowship at the Statistics Department of Stanford University.

I am an active member of the German Region (IBS-DR) of the International Biometric Society, have chaired its Working Group on Adaptive Design and Multiple Testing Procedures and served as Region President from 2019 to 2021, and am currently chairing a Working Group again.

I have also been an Associate Editor for Biometrics and am currently an Associate Editor for the Biometrical Journal. I have been (and still am) a member of several independent data and safety monitoring boards for clinical trials.

My research interests include the development and investigation of statistical methods for multiple testing, complex clinical trials, in particular adaptive designs and trials for personalised medicine.

My current research also involves the development and evaluation of machine learning methods and digital health applications.

I am active in serval research networks, e.g. the DFG Research Training Group π3, the Research Unit "LifeSpan AI" and the Leibniz ScienceCampus "Digital Public Health". I am also involved as the responsible statistician in a number of clinical trials and medical studies at the KKSB.

In Bremen I teach and supervise bachelor and master students in Mathematics and Biostatistics and am responsible for the master programme in Medical Biometry / Biostatistics.

Language

English

Description

The objective of confirming the efficacy and safety of new treatments in a clinical trial usually requires the full pre-specification of the trial design, particularly to meet the anticipated and pre-specified error rates.

This includes the definition of the study population and any subpopulation studied, the primary and secondary endpoints, all study treatments, randomisation and treatment allocation schemes, the sample sizes and the statistical analysis strategies.

Since the appropriate specification of these elements requires reliable a priori information, which is not always available before the start of the trial, adaptive designs have been invented that allow some design elements to be specified during the trial, based on the available study data and other internal and external information gathered during the trial.

Examples of such adaptively chosen study design elements are the final sample sizes, the final treatments to be studied, and the final (sub)populations for which the effect of the selected treatments shall be confirmed. Moreover, it may be difficult to fully specify in advance the way of how these design elements will be chosen based on the interim data. This is the case, for example, for the selection of treatments and/or study populations with promising treatment effects and safety profiles.

New statistical principles and methods are required to cope with the control of error rates with such data-driven design adjustments. This course will provide a solid introduction to these principles and corresponding non-canonical statistical methods, and will show how adaptive designs can be planned and analysed in R using the rpact package.

The course will be based on a book on confirmatory adaptive designs co-authored by the lecturer. Main Literature Wassmer, G., Brannath, W. (2016). Group Sequential and Confirmatory Adaptive Designs in Clinical Trials. Springer Series in Pharmaceutical Statistics. Springer, Cham. https://doi.org/10.1007/978-3-319-32562-0_6

Course goals

The main aim of this course is to give the participants an understanding of why and how adaptive designs work from a mathematical, statistical and practical point of view, as well as the limitations of such designs and the corresponding statistical methods.

Participants will also learn to plan and analyse group sequential and adaptive designs using the R package rpact.

Course contents

After reviewing the main statistical principles for clinical trials, I will introduce classical group sequential designs, which can be considered as a starting point for confirmatory adaptive designs. I will then present the basic statistical inference principles for adaptive designs. I will thereby focus on single hypothesis designs and sample size reassessment methods. I will also give an introduction to multiple testing, including the closed testing principle, and will discuss adaptive designs with multiple hypotheses as they arise when treatments or populations are selected (or newly introduced) at an interim analysis. All methods will be illustrated by examples and the course participants will have the opportunity to plan and analyse example studies using the R package rpact.

Provisional schedule

Day 1 – Group Sequential Clinical Designs

  • Statistical Principles for Clinical Trials
  • Group Sequential Designs
  • Practical Part: Planning Group Sequential Trials in R with rpact

Day 2 – Basic Principles and Methods for Adaptive Designs

  • Adaptive Combination Tests
  • Conditional Error Function Principle
  • Conditional Power and Sample Size Reassessment
  • Practical Part: Planning and Re-planning Adaptive Designs in R with rpact

Day 3 – Optimal Adaptive Designs and Estimation Methods

  • Optimal Adaptive Designs
  • Estimation Methods for Adaptive Designs
  • Practical Part: Analysing Adaptive Designs in R with rpact

Day 4 – Adaptive Designs with multiple Hypotheses

  • Introduction to Multiple Testing
  • Adaptive Designs with Multiple Hypotheses
  • Practical Part: Planning & Analysing Adaptive Designs with multiple Hypotheses in rpact

Prerequisites

Basic knowledge in probability theory and statistical inference (hypothesis testing and confidence intervals)

Targeted at

MSc and PhD students (as well as researcher) in operations research, applied mathematics, statistics and data science programs

Evaluation

Small projects in which participants are asked to design and compare adaptive study designs and analyse corresponding sample data

Software requirements

The free statistical software R (with package rpact installed) and R-studio