Pitfalls of Adaptive Designs - A practical view
Thomas Zwingers - Managing Director - Estimate GmbH

Interim analyses of clinical trials according to an adaptive design offer many advantages over conventional analyses, but they are more complicated than conventional types of analyses.

The major practical and theoretical problems we can face in a trial with an adaptive design are :

·         „Comparability of patients“ – in an adaptive design we not only have make sure that treatment groups are comparable with respect to patients characteristics, but also we want the patients of the different stages of the design to be comparable. The first issue we usually handle by randomising, but what about the second issue?

·          „Overrunning“ – depending on the time when the last patients has been included until the results of the interim analysis are available, we usually recruited a variable number of addittional patients. If it is the intention of the interim analysis to stop one or more treatment groups we face the problem of how to deal with patients in treatment groups which we want to stop on the basis of the interim results.

·          „Choice of Hypotheses“ – the adaptive design give much flexibility with respect to the trial hypotheses we want to test, especially if we have a Phase II/III design, where we usually want to continue to the second stage only with the most promising treatments. How do we deal with mis-specified hypotheses?

We will give examples for the above mentioned problems from real trials and try to give advice to overcome such problems.