Home > Research > Publications & Outputs > Multiplicity
View graph of relations

Multiplicity: discussion points from the PSI multiplicity expert group

Research output: Contribution to journalJournal articlepeer-review

  • Alan Phillips
  • Chrissie Fletcher
  • Gary Atkinson
  • Eddie Channon
  • Abdel Douiri
  • Thomas Jaki
  • Jeff Maca
  • David Morgan
  • James Henry Roger
  • Paul Terrill
<mark>Journal publication date</mark>09/2013
<mark>Journal</mark>Pharmaceutical Statistics
Issue number5
Number of pages5
Pages (from-to)255-259
Publication StatusPublished
Early online date26/07/13
<mark>Original language</mark>English


In May 2012, the Committee of Health and Medicinal Products issued a concept paper on the need to review the points to consider document on multiplicity issues in clinical trials. In preparation for the release of the updated guidance document, Statisticians in the Pharmaceutical Industry held a one-day expert group meeting in January 2013. Topics debated included multiplicity and the drug development process, the usefulness and limitations of newly developed strategies to deal with multiplicity, multiplicity issues arising from interim decisions and multiregional development, and the need for simultaneous confidence intervals (CIs) corresponding to multiple test procedures. A clear message from the meeting was that multiplicity adjustments need to be considered when the intention is to make a formal statement about efficacy or safety based on hypothesis tests. Statisticians have a key role when designing studies to assess what adjustment really means in the context of the research being conducted. More thought during the planning phase needs to be given to multiplicity adjustments for secondary endpoints given these are increasing in importance in differentiating products in the market place. No consensus was reached on the role of simultaneous CIs in the context of superiority trials. It was argued that unadjusted intervals should be employed as the primary purpose of the intervals is estimation, while the purpose of hypothesis testing is to formally establish an effect. The opposing view was that CIs should correspond to the test decision whenever possible.