‘Arbitrary’ Hodgepodge of Composite Endpoints Used Across ASCVD Trials

Less than half of the recent trials testing therapies for atherosclerotic cardiovascular disease (ASCVD) use the same endpoint, and many alter those endpoints after enrollment begins, according to a new analysis.

Among 50 randomized, controlled trials published in prominent medical journals, a little more than one-quarter used the standard composite three-point endpoint of nonfatal MI, nonfatal stroke, and cardiovascular death. The remainder use a seemingly capricious mix of different combinations.

The results, which were published this week in JAMA Internal Medicine, are surprising because cardiovascular trials are not pulled together in a slapdash way: they are large, expensive, highly organized studies considered the gold standard of evidence-based medicine, said lead investigator Daniel Shepshelovich, MD (Columbia University Irving Medical Center, New York, NY).

“It’s not like these are small trials of rare disorders where you might say, ‘OK, this is what we looked at,’” he told TCTMD. “These are very planned out studies.”

Shepshelovich said they focused on interventions in ASCVD to have a homogeneous patient population, and the choice of endpoint was “more chaotic and random” than expected. When trialists deviated from the three-point MACE endpoint, almost no other study used the same primary endpoint as others. This makes it difficult, as a hospitalist who treats patients with ASCVD, to compare outcomes of different interventions, said Shepshelovich.

It may be important for investigators to look at a different endpoint in a specific trial, but “they should rationalize it,” he said. “If they do choose to deviate from it, it would be good to explain why because otherwise it seems a bit arbitrary, which is what I feel we have right now.”

Paul W. Armstrong, MD (University of Alberta/Canadian VIGOUR Centre, Edmonton, Canada), who has previously called for a reappraisal of the use of composite endpoints in clinical research, said the addition of a softer clinical event, which is usually done to maintain the study’s power, can complicate a clinical trial where treatment decisions vary around the globe.

“How you define something, whether it’s rehospitalization or need for revascularization, for example, these are the sorts of things that have some discretion to them and could potentially lead to bias,” he told TCTMD. Such decisions, said Armstrong, are also culturally sensitive because they’re affected by healthcare system resources, such as the number of hospital beds and third-party payments.

Even MI, which is considered a hard endpoint, can be biased depending on the definition used by trialists, said Armstrong. In a 2017 white paper on the use of composite endpoints, he noted that composite endpoints can be unreasonably combined, poorly defined, and inadequately reported.

The issue, said Shepshelovich, is important because “we want to have the best science for our patients.”

Shifting Endpoints

The analysis spans 2019 to 2023 and includes studies published in the New England Journal of Medicine, The Lancet, JAMA, the European Heart Journal, Circulation, and the Journal of the American College of Cardiology. The 50 randomized trials were a mix of primary and secondary prevention studies and included interventions such as lipid-lowering therapies, antidiabetic agents, nutritional supplements, antihypertensive drugs, polypills, and anti-inflammatory medications, among others. Two-thirds of the trials were positive. 

It’s not like these are small trials of rare disorders where you might say, ‘OK, this is what we looked at.’ These are very planned out studies. Daniel Shepshelovich

While 14 trials used the three-point endpoint of nonfatal MI, nonfatal stroke, and cardiovascular death, the remaining 36 studies used 29 different endpoint combinations. Of the 32 studies that were positive, 22 reported endpoints other than the three-point MACE endpoint.

Eight of the trials didn’t even define the primary endpoint before the launch of the study, listing “cardiovascular disease” or “MACE” on the initial ClinicalTrials.gov posting. The endpoints were later edited to provide more detailed descriptions of the primary endpoint.

Concerningly, the original and published primary endpoints differed for 23 randomized trials, with components of the original primary endpoint changed after enrollment began in 15 studies. The most commonly cited reason for changing the primary endpoint—as happened in the ISCHEMIA trial when it expanded from cardiovascular death and MI to a five-point composite endpoint—was a need to boost statistical power. 

In 2005, the International Committee of Medical Journal Editors began requiring trial registration as a prerequisite for publication. That change, said Armstrong, was a transformative moment, as it laid out the intentions of the study investigators along with their preselected clinical endpoints. NEJM currently requires investigators to publish the study protocol and any amendments made during the trial.   

“There are times in the evolution of a clinical trial where there’s a legitimate need to modify the endpoint,” said Armstrong.

Today, there are additional considerations when it comes to choosing an endpoint, said Armstrong, highlighting the rising prominence of patient-reported outcomes, such as quality of life. Additionally, while mortality is unambiguous, the US Food and Drug Administration has surrogate endpoints, such as the use of intravenous diuretics as a proxy for hospitalization, in some clinical trials.

“We all recognize that surrogate endpoints are sometimes misleading, but if your cholesterol is lower and natriuretic peptides are lower, those align pretty well with prognosis,” he said. “And of course, we’re into quality of life and what matters to patients.”

Given the complexity of the issue, Armstrong said it would be beneficial if journal editors, along with clinical trialists, could reach a consensus on agreed-upon composite endpoints and systematic definitions, making sure to account for differences between countries. Such a consensus would provide greater homogeneity of composite endpoints across randomized trials and would help physicians and researchers make comparisons of the evidence. He would also like to see consensus-derived differential weighting of clinical events.