Tutorium 1: Interim Monitoring of Clinical Trials

Interim Monitoring of Clinical Trials

Referent: Christopher Jennison, University of Bath
Kosten: 80 Euro 
Zeit: 20.3.2005, 15-20 Uhr
Ort: Melanchthonianum, Universitätsplatz 8-9, Hörsaal XVII

The tutorial will survey group sequential methods for the interim monitoring of clinical trials. For many studies, such methods are required under international guidelines. They meet ethical needs, aid administration and contribute to efficiency in the development of new treatments.

We shall start with an introduction to interim monitoring, a little history on its origins and evolution in the medical context, and discussion of the scope of application including types of testing problems and forms of early stopping. We shall note the theoretical foundations of group sequential methods (stating results, rather than deriving them) and see the benefits that arise from a common theory that covers most response types and statistical models.

The tutorial will then move on to a specific example illustrating the process of design, monitoring and analysis. Implementation of group sequential methods really needs purpose-built software and several packages are currently available. We shall illustrate examples using the East package.

Survival studies play a key role in medical research. They have special features of variable observation length for different subjects due to staggered entry, right censoring at interim analyses, and censoring because of loss to follow-up. The general group sequential theory seen earlier extends to survival data but the amount of information (in the formal statistical sense) that will be available at a given calendar time cannot be predicted exactly and special methods are needed to cope with this. The "error-spending" tests that have emerged in response to such problems have other useful attributes and now form a key part of the toolbox for interim monitoring and testing.

In our final session, we shall discuss some of the general issues of interim monitoring including blinded and unblinded methods for estimating a response variance with a view to modifying sample size. We shall look at flexible approaches based on repeated confidence intervals and stochastic curtailment and contrast these methods with optimised procedures based on rigorously enforced stopping rules.

The book "Group Sequential Methods with Applications to Clinical Trials" by Jennison & Turnbull (Chapman & Hall/CRC, 2000) is the natural companion to this tutorial. It covers the above topics in depth and offers material on further related topics.