Statistical Analysis Plan

aim is to detect eventual trends with regard to the safety and/or efficacy of the tested treatment before the primary analysis. If the early data show a clear benefit with the tested treatment, it is unethical to continue exposing patients to less effective standard therapy. Similarly, if there is clearly no benefit with the tested treatment, the protocol should be modified or the trial stopped.

The SAP should describe:

• The planned interim analysis, including when it will take place.
  
• How the unblinding of data will be handled (in blinded trials).
  
• The scope of decisions on the trial conduct and on which criteria they will be based.
  

Repeated looks at the efficacy data during the course of a trial suffer from the same issue of multiplicity as explained previously. They can inflate the overall type I error rate of the trial (i.e. the probability of having a false positive outcome). In the case of interim analyses, specific methodology has been developed to establish thresholds for the significance of statistical tests supporting the decision to stop or not to stop a clinical trial. Examples of some of the methods are the Pocock, Haybittle-Peto and O'Brien Fleming.