The British Medical Journal has published a systematic review that examined the use of composite end points in cardiovascular trials. The aim was to assess the importance to patients of each component in the composite end points and to examine the frequency of these events in the studies.
The study examined cardiovascular randomised controlled trials published in the 18 months from 1st January 2002 in six medical journals (Lancet, Annals of Internal Medicine, Circulation, European Heart Journal, JAMA and New England Journal of Medicine). The advantage of using composite endpoints is that trials can be smaller in terms of numbers of patients and duration. 114 trials were identified that used composite endpoints.
Only 68% of the studies reported full data for each of the composite end point components. Over half of the studies (56%) showed a high degree of variability of importance to patients in the composites. This is due to inclusion of outcomes that are very important to patients, for example cardiac death, in combination with outcomes that are insignificant to patients, for example doubling of serum creatinine concentration.
This analysis also found that the measures that were of least importance to patients contributed the most to the event rates and treatment effects. Removing these less important outcomes from the composite would mean fewer trials reached statistical significance.
It is clear from this review that when interpreting randomised controlled trials using composite endpoints, each of the components of a composite endpoint should be viewed separately. This will ensure that claims relating to outcomes of greatest importance to patients are not exaggerated.
For example, if a study was highly statistically significant on a composite endpoint of cardiovascular death, non-fatal MI and re-admission to hospital but had similar event rates for cardiovascular death and non-fatal MI with the significance being driven by a large difference in the rate of re-admission to hospital it would be important to ensure that the intervention was not interpreted as reducing deaths or non-fatal heart attacks.