<img alt="" src="https://secure.lote1otto.com/219869.png" style="display:none;">
Skip to content

A Gatekeeping Procedure to Test a Group Sequential Design

A recent publication in Biometrics ‘A Gatekeeping Procedure to Test a Primary and a Secondary Endpoint in a Group Sequential Design with Multiple Interim Looks’ greatly extends the results of Glimm et al. ( 2010) and Tamhane et al ( 2010) which studied the problem of testing a primary and secondary endpoint, subject to a gatekeeping constraint, using a group sequential design (GSD) with K = 2 looks. This extends the methodology to provide for multiple (K>2) looks. The methodology is applied to the data from the RALES study (Pitt et al., 1999; Wittes et al., 2001).

The primary author, of the paper, Dr. Tamhane, developed the methodology for the publication, and Cytel’s Cyrus Mehta, President and Co-Founder of Cytel, and Teresa Curto contributed as co-authors with Jiangtao Gou and Christopher Jennison. Dr. Mehta provided the data and discussion for the RALES Study and Teresa Curto, Senior Biostatistician at Cytel performed the analysis on the RALES study data. Using data sets from the RALES study, Teresa obtained log-rank statistics from survival analysis and performed Kaplan-Meier analysis so that survival curves could be examined. This was conducted for each of 5 interim time points and for primary and secondary endpoints (death, cardiovascular death and sudden cardiovascular death).

We asked Dr. Mehta about the objectives, key take-aways and expected impact of the publication.

Cytel: What were the objectives of the publication?
Cyrus Mehta : The objective was to develop a statistically valid and efficient method for testing a secondary endpoint if the primary endpoint showed overwhelming efficacy at an interim analysis of the accruing data resulting in early termination of the trial. In the Randomized Aldactone Evaluation Study (RALES), the treatment arm (spironolactone) was shown to reduce the all-cause mortality (the primary endpoint) at an interim analysis and the trial was stopped. The methods of this paper were used to provide a valid test of the efficacy spironolactone for the secondary endpoint of sudden cardiovascular death.

Cytel: What key learning points could readers of the publication expect to take away?
Cyrus Mehta: There are very subtle statistical issues with testing a secondary endpoint in a study that has interim analyses that can lead the statistician astray, resulting in hypothesis testing procedures that inflate the false positive rate. These subtleties do not arise in trials where there is no hypothesis testing at interim analysis time points and in that case the testing of primary and secondary hypotheses is straightforward.

Cytel: What impact do you expect this publication to have?
Cyrus Mehta :We hope that these methods will lead to a better understanding of multiplicity issues in group sequential designs that can result in incorrect hypothesis testing procedures. As a result more trials will build in at least one interim analysis for testing both the primary and secondary hypotheses so that effective new drugs can be detected earlier and be available to patients who have an unmet medical need.

To read the abstract and access the publication click the button below.


Tamhane, A., Gou, J., Jennison, C., Mehta, C. and Curto, T. (2017). A gatekeeping procedure to test a primary and a secondary endpoint in a group sequential design with multiple interim looks. Biometrics.

contact iconSubscribe back to top