Bleeding as Mortality Surrogate Questioned for Antithrombotic Trials

Bleeding does not appear to be a good proxy for mortality in the setting of antithrombotic therapy for coronary artery disease, a meta-analysis of RCTs shows.

In pooled results from four dozen trials, there were weak correlations between the effects of treatment on nonfatal major or minor bleeding and the effects on either all-cause or cardiovascular death, Toshiki Kuno, MD, PhD (Montefiore Medical Center, Albert Einstein College of Medicine, New York, NY), and colleagues report in a study published online recently in JACC: Advances.

“Our results were consistent regardless of bleeding definitions, study period, follow-up period, and in patients with acute coronary syndrome,” they write.

The findings, which Kuno also presented in a poster at the recent American College of Cardiology/World Congress of Cardiology (ACC/WCC) 2023 meeting, “should be informative for researchers who are considering using bleeding outcomes as a study endpoint to be a surrogate for death [or combining] bleeding outcomes and death as NACE,” or net adverse clinical events, they say. “Preferably, alternative analyses such as the win ratio analysis can be used to take into account for the priorities of the composite outcomes.”

Speaking to TCTMD, Kuno noted that a similar study published in October 2021 in JAMA Internal Medicine showed that MI was not a good surrogate for mortality in more than 100 RCTs evaluating treatment or prevention of CAD, leading him and his colleagues to explore the question for bleeding, which is included as a major endpoint in trials of antithrombotic therapies.

Sorting Out Surrogates

Kuno et al’s analysis included 48 RCTs of antithrombotic therapies with a total of 181,951 participants. Investigators defined major and minor bleeding events according to what was used in the trials, mostly BARC or TIMI criteria. They assessed trial-level correlations between nonfatal bleeding and mortality using the coefficient of determination (R-square), ranging from 0 (absence of surrogacy) to 1 (perfect surrogacy). If the lower limit of the 95% CI is above 0.72, it’s a strong correlation, and if the upper limit of the 95% CI is below 0.50, it’s a weak correlation.

The R-squares for the correlations between major/minor bleeding and either all-cause death or CV death were both 0.09, with confidence intervals indicating weak relationships. Correlations remained weak when confining the results to major bleeding.

The findings indicate not only that bleeding is not a good surrogate for mortality, but also that combining bleeding with MACE or death in a NACE endpoint is not recommended because it places bleeding and mortality on the same level, Kuno said. “NACE is not a good outcome as a primary, and it is hard to interpret for a randomized trial with dual antiplatelet therapy strategies,” he said. “A win ratio may be a preferred analysis.”

Kuno and his colleagues say bleeding is an important endpoint in and of itself. “In general, physicians should select antithrombotic agents based on individualized ischemic and bleeding risks; however, NACE as the primary endpoint may be inconclusive regarding which antithrombotic agents should be used for each patient,” they write. “In addition, NACE has a risk of bias toward the null since thrombotic and bleeding outcomes often tend to go in opposite directions. Moreover, a composite outcome with a different magnitude of components may result in misleading impressions of the impact of treatments.”

Marco Valgimigli, MD, PhD (Cardiocentro Ticino, Lugano, Switzerland), however, disagreed. Commenting on the analysis for TCTMD, Valgimigli said NACE is a valid endpoint for trials in the right situations. “I think a NACE endpoint is justified as long as the bleeding events which are included in the composite endpoints have a similar association with mortality than other nonfatal endpoints included,” he said in an email.

He pointed to a prior study his group performed using data from the TRACER trial, which showed that BARC 2 bleeding was independently associated with mortality but that this association was not as strong as the relationship between MI and mortality. BARC 3 bleeding had an association with mortality similar to that seen for MI. “Therefore, a NACE endpoint which includes death, MI, stroke, and major (and not minor) bleeding makes sense to me,” Valgimigli said. “At the same time, if the NACE endpoint includes also BARC 2 bleeding, [it] does not seem to be appropriate and might be misleading.”

He supported inclusion of MI and major bleeding in composite endpoints for trials because of their associations with mortality. “Of course, having a study powered for mortality alone is better off than having it powered on a composite endpoint,” he said. But he noted that the GUSTO trial published in 1993 was powered for 30-day mortality and included more than 40,000 patients. “Who can afford randomizing more than 50,000 patients these days, taking into account that mortality has substantially decreased since the conduct of GUSTO?” Valgimigli said.

As to the larger issue of bleeding as a surrogate for mortality, he noted that multiple studies have established an association between bleeding and mortality, adding that “mortality is a very multifactorial endpoint and [it] would be surprising to expect that the vast majority of fatal events only depend on bleeding. Bleeding is most likely one of many causes.”

In addition, the approach used in the current analysis “does not take into account the benefit of the antithrombotic treatment that was at the same time associated with bleeding as side effects,” meaning that the ischemic benefits and the increase in bleeding may have canceled each other out when it came to the association with mortality, Valgimigli said.