Power of a study

Sample size and power

• clinical studies such as randomised controlled trials are designed to test whether a difference exists between two or more interventions in terms of specific outcomes or endpoints
• in order for a study to detect a statistically significant difference between treatments, the study must be large enough (i.e. have enough subjects participating in the study) for a sufficient number of endpoints of interest to occur
  
  
• the 'power' of the study is the ability of the study to reliably detect a difference between interventions
  • number of subjects required to be included in the study in order to have sufficient power must be made before a study begins
  • ideally clinical study reports should indicate that a power calculation has been made - it is common for studies to stipulate a power of 80–90%
  • the power of a study refers to its ability to detect a difference only in that endpoint on which the power calculation is based, i.e. the primary endpoint. The power used in the study may not be sufficient to reliably detect differences in other (secondary) endpoints or in subgroups
    
    
  • the power of a study can be calculated via knowledge of the type II error (a type II error arises when the null hypothesis is accepted when it is false) (1)
    • power (the chance that the study will detect the minimum difference as statistically different) = 1 - type II error
    • in effect the power of a study is
      • the likelihood that a study will detect an effect when there is an effect there to be detected

Reference:

1. Evidence Based Medicine 2006;11:69-70.
2. MeReC Briefing (2005);30:1-7.
3. Wiebe S. The principles of evidence-based medicine.Cephalalgia. 2000;20 Suppl 2:10-3.
4. Best Pract Res Clin Obstet Gynaecol. 2005;19(1):15-26.