This paper introduced Simes-based parallel gatekeeping procedures and Bonferroni-based parallel gatekeeping procedures with logical restrictions (with applications to clinical trials with multiple treatment arms).

This paper described testing procedures for clinical trials in which the primary family is a parallel gatekeeper, that is, at least one significant result needs to be found in the primary family before one can begin testing secondary hypotheses. The paper discusses parallel gatekeeping procedures based on the Bonferroni and resampling tests. Software implementation of these procedures is discussed in Chapter 2 of Dmitrienko, Molenberghs, Chuang-Stein, and Offen (2005).

Chapter 2 gave a detailed overview of gatekeeping strategies in clinical trials. Clinical trial examples and SAS code are included. A good way to get started in this area.

The paper introduced parametric gatekeeping procedures based on the Dunnett test. This approach was applied to dose-finding clinical trials with multiple endpoints. Dunnett-based gatekeeping procedures account for the correlation among the dose-placebo comparisons and multiple endpoints. They are more powerful than Bonferroni-based gatekeeping procedures described in Dmitrienko, Offen and Westfall (2003).

This paper extended the fallback testing method (Wiens, 2003) to the multi-stage case. Multiple testing procedures proposed in the paper can be used to carry over a predetermined fraction of the Type I error rate to the next family even if no significant results are found in the previous family. Applications included dose-finding studies with multiple endpoints.

It was shown in this paper that parallel gatekeeping procedures (Dmitrienko, Offen and Westfall, 2003) have a simple stepwise structure that considerably simplifies their implementation and interpretation of the results.

Tree gatekeeping procedures serve as an extension of serial and parallel gatekeeping procedures and can be used in clinical trials with multi-dimensional objectives (multiple endpoints, doses, tests, etc) and clinical trials with logical restrictions. Note that the algorithm introduced in this paper may fail to meet the monotonicity condition, which may cause it to violate the gatekeeping property. It is shown in Dmitrienko, Tamhane, Liu and Wiens (2008) how the algorithm can be modified to rectify this problem.

This paper gave a review of recent developments in the world of gatekeeping procedures, inclucing serial, parallel and tree gatekeeping procedures, with clinical trial examples.

This working paper introduced a class of general gatekeeping procedures that have an attractive multistage form (it extends the method proposed by Guilbaud (2007) who focused only on Bonferroni-based gatekeeping procedures). These procedures do not not require the application of the closed testing principle and are easy to implement. A modified version of the working paper will be published in Biometrical Journal (Dmitrienko, Tamhane and Wiens, 2008).

It was shown in this paper that the algorithm for constructing tree gatekeeping procedures used in Dmitrienko, Wiens, Tamhane and Wang (2007) may fail to satisfy the gatekeeping property. A general algorithm based on the monotonicity condition introduced by Hommel, Bretz and Maurer (2007) was proposed to rectify this problem.

This paper proposed a straightforward approach for setting up multistage/stepwise gatekeeping procedures based on the Bonferroni test. This method was later extended in Dmitrienko, Tamhane and Wiens (2007) to construct a general class of gatekeeping procedures based on arbitrary separable tests.

This paper introduced a class of Bonferroni-based gatekeeping procedures that includes popular tests (e.g., Holm and fallback tests) and gatekeeping procedures (e.g., Bonferroni-based parallel gatekeeping procedure, Dmitrienko, Offen and Westfall, 2003).

The paper discussed serial gatekeeping strategies in clinical trials.

The paper described the fixed-sequence test which serves as an example of a serial gatekeeping procedure (all tests must be significant in the primary family before one can perform secondary analyses).