primary versus secondary endpoints

Last reviewed 01/2018

Primary vs. secondary endpoints and sub-group analyses

• clinical studies such as randomised controlled trials are designed to answer specific questions, e.g. whether one drug is more effective than another at preventing a particular event occurring
  • primary endpoint
    • the specific event that the study is designed to assess the effect of the drugs upon
    • the primary endpoint is important to study design because the sample size required in order to adequately power a study is dependent on the number of primary events that are expected to occur over a given time period
      • the incidence of some events of interest (e.g. death) in a particular study population being studied may actually be quite low in the period over which the study can be conducted - thus, in order to reliably study the effectiveness of two different interventions in preventing the event of interest, many study subjects would need to be enrolled in the study
      • however if there is a large study population then this may make the study very expensive to conduct and recruitment of sufficient numbers of subjects into the study may be difficult
    • composite primary endpoints
      • composite primary endpoints are frequently used to overcome the difficulties associated with designing a study around a single event of interest (e.g. death)
      • an example of a composite primary endpoint in a cardiovascular study might be cardiovascular death and/or non-fatal myocardial infarction and/or any kind of stroke - thus the study would be designed with such a composite primary endpoint (e.g. any major cardiovascular event) and would require fewer subjects, as the total number of events to be detected is in effect increased
      • however when appraising the results of studies that combine several event types into one composite endpoint, it can be difficult to determine the true effect of the intervention on each of the event types
  • secondary endpoints
    • these are additional events of interest, but which the study is not specifically powered to assess
    • because the design of the study was not based around the secondary endpoint (s), analyses of secondary endpoints need to be viewed with caution
    • the caution in interpreting results is also true with respect to sub-group analyses, in which primary or secondary endpoints are considered in a smaller sub-group of the whole study population (e.g. only those subjects with a specific baseline characteristic) - this is because the power of the study is based on the entire study population
    • secondary endpoints and any sub-group analyses should be pre-specified in the studyprotocol and methods section of the study report
      • if the secondary endpoints and sub-group analyses have not been pre-specified then there is a possibility that the data from the study has been analysed in many ways to find any kind of positive result in favour of the intervention ( 'data dredging'). If this is the case then the study results cannot be relied upon alone as providing reliable information - because if a set of data is analysed and re-analysed enough times, a positive result may occur purely by chance
    • secondary endpoints do not have the same statistical authority as the primary endpoint, and it is more likely that positive changes in secondary endpoints are due to chance. It has been stated that secondary endpoint results should only be used to help interpret the primary result of the trial or to provide information, or prompts, for future research (4)

     

Reference:

1. MeReC Briefing (2005);30:1-7.
2. Best Pract Res Clin Obstet Gynaecol. 2005;19(1):15-26.
3. Wiebe S. The principles of evidence-based medicine.Cephalalgia. 2000;20 Suppl 2:10-3.
4. Drug and Therapeutics Bulletin 2006; 44(3):21.