Quetelet seminar
- When
- 07-10-2024 from 09:00 to 11:30
- Where
- Leslokaal 0.3 Victor Van Straelen, Building S8, Campus Sterre
- Language
- English
- Organizer
- Jan De Neve
- Contact
- Jan.DeNeve@ugent.be
adaptive designs
First, introductory talk
Title: Statistical Principles and Methods for Confirmatory Adaptive Designs
Abstract:
Confirmatory adaptive designs are sequential designs for clinical trials that allow adjustments to the trial design at interim analyses based on all information collected to date, without compromising the type I error rate. Examples of design adaptions include changes in the sample size, the selection or addition of study treatments, changes in the study population, and changes in the primary endpoints. An important feature of adaptive designs is that there is no need to pre-specify the adaptions rules to control the type I error rate. Controlling the Type I error rate under this flexibility requires non-standard statistical principles and methods. In this talk I will review the basic principles of confirmatory adaptive designs. This will include the basic invariance principle and its implementation via combination tests and conditional error functions. I will also review tools such as the conditional power and sample size recalculation based on conditional power. Finally, I will indicate how to deal with multiple hypotheses, that are either considered at the beginning of the study or come along with the interim design adaptations.
Literature
Wassmer, G. and Brannath, W. (2016). Group Sequential and Confirmatory Adaptive Designs in Clinical Trials. Springer Series in Pharmaceutical Statistics. Springer International Publishing, Switzerland
Second, scientific talk
Title: Theory and application of optimal adaptive designs
Abstract:
For ethical and economic reasons, sample size calculations are required when designing a clinical trial. These calculations must be based on assumptions about the treatment effect, which are sometimes difficult to make. A pilot study is often conducted to estimate the treatment effect in advance. However, with traditional testing methods, the data from the pilot study cannot be used for the final hypothesis test at the end of the clinical trial. At least in cases where it is difficult to recruit participants, it seems reasonable to have study designs that allow the pilot study to be used for the final hypothesis test, while ensuring control of the type I error rate. This is possible with adaptive designs, where, for example, all data from an interim evaluation can be used to adjust sample sizes mid-trial. However, the efficacy of adaptive designs has been questioned from the outset and is still an issue today. The development of optimal adaptive designs is therefore an important task.
In this talk, I will present a 20-year-old optimality theory, together with recent extensions, that provides an analytical expression for the most important statistical component of an adaptive design, namely the so-called "conditional error function". The resulting adaptive design is optimal in the sense that it minimises the expected sample size for a given parameter value or a given mixture of parameter constellations. Furthermore, the theory provides a kind of maximum likelihood approach and allows optimisation under constraints such as minimum and maximum sample size. We will present examples of such optimal adaptive designs and some lessons that can be learned from the theory and the examples.
Literature
Brannath, W., Bauer P. (2004). Optimal Conditional Error Functions for the Control of Conditional Power. Biometrics 60, 715–723.
Brannath, W., Dreher, M., Scharpenberg, M. (2024). Optimal monotone conditional error functions. arXiv preprint arXiv:2402.00814.