Multiplicity Issues in Clinical Trials ICSA 2010

From MultXpert

Jump to: navigation, search

Session at the 2010 International Chinese Statistics Association meeting in Indianapolis, IN on June 23, 2010 organized by Alex Dmitrienko (Eli Lilly and Company) and chaired by Yan D. Zhao (UT Southwestern Medical Center).


Multiple testing problems with general logical restrictions in clinical trials

Presented by Alex Dmitrienko, Research Advisor, Global Statistical Sciences, Eli Lilly and Company.

This presentation discusses multiple testing problems with general logical restrictions arising in clinical trials. The mixture-based framework can be used to construct gatekeeping procedures that account for complex logical relationships among multiple null hypotheses. These gatekeeping procedures control the familywise error rate in the strong sense and enable trial sponsors to set up powerful procedures based on p-value-based tests (eg, Hochberg-type or Hommel-type tests) or parametric tests (eg, regular and step-down Dunnett tests).

Download slides.

Statistical give and take: Power implications on strong control of the Type I error rate in a clinical trial setting

Presented by Chris Holland, Director, Biostatistics, MacroGenics.

Much progress has been made over the past decade with the development of gatekeeping statistical procedures that provide strong control of the Type I error rate in increasingly complex clinical trial study designs that require, for example, multiple dose groups and coprimary endpoints to undergo formal hypothesis testing. The advantage that such procedures provide is greater assurance that positive results will make it into an approved product's labeling and can therefore be used for marketing claims. The disadvantage, of course, is the need to increase the sample size in order to provide adequate power for meeting a study's key objectives. However, the true power implications between competing procedures or approaches are not always well understood. This is a particularly relevant problem in situations where strong Type I error control is more of a "nice to have" than a regulatory requirement, such as when dealing with key secondary endpoints. In this presentation we will look at one such procedure, the truncated Holm test, applied to a situation that involves multiple treatment groups and co-primary endpoints. The properties of this "separable" test will be examined with respect to varying truncation fractions and Type II error rates compared to a standard Holm procedure (which is not separable and therefore does not provide strong control of the Type I error).

Download slides.

Hommel-based gatekeeping multiple comparison procedure: an example

Presented by Jane Xu, Executive Director, Biostatistics and Data Management, Sepracor (jointly with Thomas Brechenmacher, Dainippon Sumitomo Pharma).

In recent years increasingly more and more clinical trials are conducted where the objectives are hierarchically ordered. In such cases it is important that appropriate multiple comparison adjustment methods are applied to ensure the family-wise error rate (FWER) is controlled in the strong sense. In this talk, a motivating example will be given to illustrate a Hommel-based gatekeeping multiple comparison procedure that controls the FWER. This method is derived from the work done by Dmitrienko Tamhane 2010. Details of the procedure will be described and how it is applied to the motivating clinical trial will be illustrated. The results from this procedure will be compared with results using the Bonferroni-based tree gate-keeping procedure described in Dmitrienko Wiens Tamhane Wang 2007.

Download slides.

Subgroup analysis in clinical trials: Enrichment design and multiple testing

Presented by Mohamed Alosh, Expert Mathematical Statistician and Team Leader, Division of Dermatology and Dental Drugs, FDA (jointly with Mohammad Huque, FDA).

Subgroup analyses are commonly used in clinical trials with the objective of learning about differential treatment effect across subgroups. Due to power consideration among other factors, clinical trials are seldom considered for establishing an efficacy claim for a subgroup in case the trial fails to establish an efficacy claim for the total population. However, through proper study design and analysis the clinical trial can be designed to establish efficacy claim for the total population as well as for the subgroup, thus increasing the chance of a positive trial. The concern that clinical trials are underpowered for subgroups can be relaxed somewhat through enrichment of the patient population for a priori identified subgroup and by using statistical testing strategies which spend the overall Type I error rate more efficiently than many traditional methods. In this presentation we consider a multiple testing strategy for the total population and the subgroup with the following features: (i) ensuring consistency of efficacy findings of the total population and that of the subgroup so that the results of the study overall are interpretable and (ii) allowing the significance level for testing the subgroup to adapt to the efficacy findings of the total population in a general form. We consider application of the proposed methodology to clinical trial data.

Download slides.