Composite outcomes.

Clinical trials are very expensive, particularly in the US. Many patients are needed to show statistical significance if the number of events we are looking at is infrequent. A way of avoiding these problems is to combine multiple events into a composite outcome to enroll fewer patients and reduce overall costs.

A heart failure medication study may seek to combine the outcomes of "cardiovascular death" and "hospitalization due to heart failure" into a single composite outcome termed "cardiovascular death or heart failure hospitalization." Combining these events would successfully increase the number of events, as patients who experience either cardiovascular death or heart failure hospitalization will now contribute to the primary analysis. This makes it easier to show statistical significance by increasing the number of events with the same number of patients.

Composite outcomes do have a significant limitation or trade-off. All clinical events included within a composite outcome are treated as equally clinically important. Going back to our previously stipulated example, it is evident that death carries greater clinical significance than hospitalization, a statement that virtually any layperson or clinician would agree with. Some experts would argue that due to its importance, death should never be combined with anything else in a composite outcome. Beyond that, another drawback of the traditional approach is the focus on the first event, irrespective of the events that may follow. Referring to our previous example once more, if a patient is hospitalized for heart failure and subsequently dies a month later, their death will be excluded from the primary analysis. This can lead to potentially skewed results since the initial event, even if not the most clinically significant, will be prioritized. (1)

Literature examples of potentially problematic composites:

*Studies that assign the same value to outcomes of clearly different value:

TIME trial:

From the abstract:

"Results:  After 1 year, improvements in angina and quality of life persisted for both therapies compared with baseline, but the early difference favoring invasive therapy disappeared. Among invasive therapy patients, later hospitalization with revascularization was much less likely (10% vs 46%; hazard ratio [HR], 0.19; 95% confidence interval [CI], 0.11-0.32; P<.001). However, 1-year mortality (11.1% for invasive; 8.1% for medical; HR, 1.51; 95% CI, 0.72-3.16; P = .28) and death or nonfatal myocardial infarction rates (17.0% for invasive; 19.6% for medical; HR, 0.90; 95% CI, 0.53-1.53; P = .71) were not significantly different. Overall major adverse cardiac event rates were higher for medical patients after 6 months (49.3% vs 19.0% for invasive; P<.001), a difference which increased to 64.2% vs 25.5% after 12 months (P<.001).

Conclusions:  In contrast with differences in early results, 1-year outcomes in elderly patients with chronic angina are similar with regard to symptoms, quality of life, and death or nonfatal infarction with invasive vs optimized medical strategies based on this intention-to-treat analysis. The invasive approach carries an early intervention risk, while medical management poses an almost 50% chance of later hospitalization and revascularization." (2)

-How does that look like visually?

*Some studies have also used them inappropriately when outcomes or harms show a benefit or harm with opposite directions

SYNTAX trial:

From the study's abstract:

"Conclusion: CABG remains the standard of care for patients with three-vessel or left main coronary artery disease, since the use of CABG, as compared with PCI, resulted in lower rates of the combined end point of major adverse cardiac or cerebrovascular events at 1 year."

Their conclusion is interesting, and potentially inaccurate since it is based on the following results:

"Most of the preoperative characteristics were similar in the two groups. Rates of major adverse cardiac or cerebrovascular events at 12 months were significantly higher in the PCI group (17.8%, vs. 12.4% for CABG; P = 0.002), in large part because of an increased rate of repeat revascularization (13.5% vs. 5.9%, P<0.001); as a result, the criterion for noninferiority was not met. At 12 months, the rates of death and myocardial infarction were similar between the two groups; stroke was significantly more likely to occur with CABG (2.2%, vs. 0.6% with PCI; P = 0.003)." (3)

The composite and the conclusion seem to imply PCI is much worse strategy than CABG. However, as they correctly point out in the results section, the difference was mostly driven by repeat revascularization need. Mortality and MI outcomes were not different between groups. Stroke was worse in the CABG group, and it was statistically significant.

References:

(1) Pocock SJ, Ariti CA, Collier TJ, Wang D. The win ratio: a new approach to the analysis of composite endpoints in clinical trials based on clinical priorities. European Heart Journal. 2012;33(2):176-182.

(2) Pfisterer M, Buser P, Osswald S, et al. Outcome of elderly patients with chronic symptomatic coronary artery disease with an invasive vs optimized medical treatment strategy: one-year results of the randomized time trial. JAMA. 2003;289(9):1117.

(3) Serruys PW, Morice MC, Kappetein AP, et al. Percutaneous coronary intervention versus coronary-artery bypass grafting for severe coronary artery disease. N Engl J Med. 2009;360(10):961-972.