In the so-called composite/combined endpoints or outcomes (CCEs/CCOs) the number of patients that suffer at least one of predefined events (which are components of the outcome variable), are encompassed in one single estimator.CCEs/CCOs are particularly used in clinical trials in the cardiovascular field. (1) For instance, the number of patients who die or suffer a non-fatal myocardial infarction during the followup period illustrates what a composite/combined endpoint, composite/combined outcome is. Even though these endpoints/outcomes are a frequently used methodologic means, the clinical and biological significance of the effect of an intervention assessed by means of CCEs/CCOs, as well as its interpretation, are concepts that have been poorly addressed in the medical literature. Both these aspects are related to the motivation to use CCEs/CCOs.

A recent systematic review attempted to synthesise the advantages and disadvantages that authors of medical literature attributed to the use of combined/composite endpoints, combined/ composite outcomes in clinical trials, as well as the recommendations for their use. (2) Tables 1, 2 and 3 synthesise the findings from this review.

Table 1. Advantages attributed to combined/composite endpoints/outcomes. These can:
1	Reduce the magnitude of the sample needed
2	Estimate the net clinical benefit of a treatment
3	Improve the understanding of the effect of the interventions and avoid competitive risks
4	Avoid the need to choose a single primary endpoint variable in cases where many others could be of equal importance
5	Avoid adjustment by means of multiple comparisons

Table 2. Disadvantages attributed to combined/composite endpoints/outcome
1	Practical interpretation could be problematic when the components of the CCE/CCO are of very different clinical importance (“qualitative heterogeneity”)
2	Interpretation can be problematic if the event rates or relative risk reduction vary notably among the components of the CCE/CCO (“quantitative heterogeneity”)
3	Potential masking of an increase in a damaging associated with the experimental intervention
4	Possibility of secondary bias in competitive risks
5	The greater the number of components, the more complex it will be to carefully assess the CCE/CCO
6	Excessive influence of the subjective components of the CCE/ CCO (“clinician-induced”)
7	Need to adjust the alpha error to draw confirmatory conclusions from the components

Table 3. Suggested recommendations for the use of combined/composite primary endpoints/outcomes
A. To construct primary combined/composite endpoints-outcomes
1	The CCE/CCO must be pre-specified and all its components must be well defined prior to initiation of the clinical trial
2	Avoid components that are unlikely to exert an effect on the intervention
3	Avoid components of no clinical importance
4	Avoid significant heterogeneity among the components in terms of clinical importance as well as the effect on the intervention
5	“Weigh” the component to ascertain their relative importance
B. To report the results form clinical trials using CCEs/ CCOs
1	All the components of the CCE/CCO also have to be reported separately (that is, as secondary outcome variables/ endpoints) so as to determine whether the CCE/CCO has been dominated by any of its components, or whether the findings on the effects of the intervention over the components are not consistent
2	The authors and directors of journals should ensure that CCEs/CCOs have been clearly reported and should avoid suggesting that an effect on their individual components has been demonstrated (the effect on the components should be demonstrated jointly rather than as demonstration of efficacy on the individual components)
3	Report the “worse” event that the patients have had at the completion of the study, based on a priori listing of the considered events, from “worse” to “better”
4	Report all the possible combinations among the components, grouping them according to the first event and according to the different components

Irrespectively of the advantages attributed to the use of CCEs/CCOs, the majority of authors coincided in that the interpretation of the trial's results could be problematic in the presence of the following circumstances: 1) heterogeneity of the components of the CCEs/CCOs with respect to their importance/ clinical relevance; 2) heterogeneity of the components in the frequency of the events; and 3) heterogeneity of the effect of the intervention on such components. The more evident these circumstances are, the more complicated the assessment of the intervention will be. Perhaps this is the reason why all authors agreed on the importance of reporting the effect of the intervention on the components of the CCEs/CCOs in order to assess whether its effect is consistent (Table 3).

Two examples representing antagonistic situations on the difficulty entailed in the interpretation of the results when CCEs/CCOs are used are the HOPE (3) study (Table 4) and the DREAM (4) study (Table 5). Whilst in the former, heterogeneity in the three previously mentioned domains is small, in the latter it is completely the opposite. So then whilst it is very plausible that the effect of the intervention on the HOPE study can be applied to its three components, it might be quite difficult to do the same in the DREAM study. And, in contrast, their authors concluded:' this large, prospective, blinded international clinical trial shows that 8 mg of rosiglitazone daily, together with lifestyle recommendations, substantially reduces the risk of diabetes or death by 60% in individuals at high risk for diabetes''. Even though this asseveration is literally correct, it is quite clear that stating that the mentioned intervention reduces the risk of diabetes or death by 60% induces the reader to get the impression of the benefit that the intervention has on both components; an spurious, meaning misleading, conclusion that, on the other hand, tends to magnify the effect of the treatment assessed in the trial.

Table 4. Results from the HOPE study (ramipril in cardiovascular prevention)
	HOPE study
	Intervention	Control	Relative risk (IC 95%)
AMI or CVA or cardiovascular death	651 (14%)	826 (17,8%)	0,78 (0,7-0,86)
· Cardiovascular death	282 (6,1%)	377 (8,1%)	0,74 (0,64-0,87)
· AMI	459 (9,9%)	570 (12,3%)	0,8 (0,7-0,9)
· CVA	156 (3,4%)	226 (4,9%)	0,68 (0,56-0,82)

Table 5. Results from the DREAM study (rosiglitazone in subjects at high risk for diabetes)
	DREAM study
	Intervention (n=2.635)	Control (n=2.634)	Risk rate (IC 95%)
New-onset diabetes or death	306 (11,6%)	686 (26%)	0,4 (0,35-0,46)
· New-onset diabetes	280 (10,6%)	658 (25%)	0,38 (0,33-0,44)
· Death	30 (1,1%)	33 (1,3%)	0,9 (0,55-1,5)

Aimed at analysing the real frequency with which potentially problematic CCEs/CCOs are used in clinical trials in real life and how results are reported, a second systematic review of clinical trials in the cardiovascular field published in journals with a high impact factor was carried out. (5) This review showed that only 68% of all clinical trials (n=114) reported the effect of the intervention on all the components and up to 14% did not report it at all. In trials which did report these results, it was observed that just only 14% used CCEs/CCOs that presented a situation similar to that seen in the HOPE study; in other words, little heterogeneity of the components in the three domains. It was also shown that, whereas the effect of the interventions on the most important components (such as mortality) remained small (RR 0.92; CI 95% 0.87-0.96), the effect increased as the relative importance of the components decreased, in such a way that in the least relevant components (e.g. revascularisation), the effect was significantly greater (RR 0.67; CI 95% 0.49-0.93). A similar situation was observed when the frequency of events was analysed. That is, both the weight of the “effect” of the intervention on the CCE/CCO as well as the frequency of the events of the CCE/CCO corresponded to variables of a lesser clinical importance.

In current literature, the frequent use of CCE/CCO with a marked gradient of clinical importance among its components and in those in which the magnitude of the effect of the intervention is predominant in the components of lesser importance suggests that the magnification of the real benefit derived from the intervention could be a motivation to use CCEs/ CCOs. The reader of medical literature should bear these considerations in mind with the aim of assessing the results of clinical trials employing CCEs/CCOs in a sensible way.

References

1. Freemantle N, Calvert M, Wood J, Eastaugh J, Griffin C. Composite outcomes in randomized trials: greater precision but with greater uncertainty? JAMA. 2003; 289(19):2554-9.
2. Ferreira-González I, Permanyer-Miralda G, Busse JW, Bryant DM, Montori WM, Alonso-Coello P, et al.. Methodologic discussions for using and interpreting composite endpoints are limited, but still identify major concerns. J Clin Epidemiol 2007; 60:651-7.
3. Yusuf S, Sleight P, Pogue J, Bosch J, Davies R, Dagenais G. Effects of an angiotensin-converting-enzyme inhibitor, ramipril, on cardiovascular events in high-risk patients. The Heart Outcomes Prevention Evaluation Study Investigators. N Engl J Med. 2000; 342:145-53.
4. DREAM (Diabetes REduction Assessment with ramipril and rosiglitazone Medication) Trial Investigators. Effect of rosiglitazone on the frequency of diabetes in patients with impaired glucose tolerance or impaired fasting glucose: a randomised controlled trial. Lancet. 2006; 368:1096-105.
5. Ferreira-González I, Busse JW, Heels-Ansdell D et al. Problems with use of composite end points in cardiovascular trials: systematic review of randomised controlled trials. BMJ. 2007;334(7597):786. Epub 2007 Apr 2.