PARADISE-MI Results Obscured As Post Hoc Analysis Finds Flaws

Ted Bosworth

September 14, 2022

A post hoc analysis of the PARADISE-MI trial, although not intended to alter the conclusions generated by the published data, suggests that clinically relevant benefits were obscured, providing the basis for recommending different analyses for future studies that are more suited to capture the most clinically significant endpoints.

"What these data show us is that we need clinical trial designs moving towards more pragmatic information that better reflect clinical practice," reported Otavio Berwanger, MD, PhD, director of the Academic Research Organization at Hospital Israelita Albert Einstein, São Paulo, Brazil.

The reevaluation of the PARADISE-MI data, presented at the annual congress of the European Society of Cardiology in Barcelona, was based on a win ratio analysis and on the inclusion of investigator-reported endpoints, not just adjudicated events. Both appear to reveal clinically meaningful benefits not reflected in the published study, according to Berwanger.

In PARADISE-MI, which was published in the New England Journal of Medicine last year, more than 5,500 patients were randomized to the angiotensin receptor neprilysin inhibitor (ARNI) sacubitril/valsartan or the ACE inhibitor ramipril after a myocardial infarction. A reduced left ventricular ejection fraction (LVEF), pulmonary congestion, or both were required for enrollment.

For the primary composite outcomes of death from cardiovascular (CV) causes or incident heart failure, the ARNI had a 10% numerical advantage, but it did not reach statistical significance (hazard ratio [HR], 0.90; P = .17).

"PARADISE-MI was a neutral trial. This post hoc analysis will not change that result," Berwanger emphasized. However, the post hoc analysis does provide a basis for exploring why conventional trial designs might not be providing answers that are relevant and helpful for clinical practice.

New Analysis Provides Positive Trial Result

When the data from PARADISE-MI are reevaluated in a hierarchical win ratio analysis with CV death serving as the most severe and important outcome, the principal conclusion changes. Whether events are reevaluated in this format by the clinical events committee (CEC) or by investigators, there is a greater number of total wins than total losses for the ARNI. Combined, sacubitril/valsartan was associated with a win ratio of 1.17 (95% confidence interval, 1.03-1.33; P = 0.015) over ramipril.

Using a sports analogy, Berwanger explained that the win ratio analysis divides the total number of wins to the total number of losses to provide a much more clinically relevant approach to keeping score. It also used a hierarchical analysis so that the most serious and important events are considered first.

In addition to CV death, this analysis included first hospitalization for heart failure and first outpatient heart failure events. CEC-defined events and events reported by investigators were evaluated separately.

The ARNI had more wins than losses in every category for all outcomes, whether CEC adjudicated or investigator reported, but most of this benefit was generated by the endpoint of CEC-adjudicated CV deaths. This accounted for 36.9% of all events (investigator-documented CV death accounted for 0.7%). This is important because PARADISE-MI, like many standard trials, was conducted on a time-to-primary event basis.

"In this type of analysis, the first event is what counts. Usually time-to-first-event analyses are dominated by nonfatal events," Berwanger explained. He believes that placing more weight on the most serious events results in an emphasis on what outcomes are of greatest clinical interest.

In addition, Berwanger argued that it is important to consider investigator-reported events, not just CEC-adjudicated events. While adjudicated events improve the rigor of the data, Berwanger suggested it omits outcomes with which clinicians are most concerned.

Investigator, Adjudicated Outcomes Differ

Again, using PARADISE-MI as an example, he reevaluated the primary outcome based on investigator reports. When investigator-reported events are included, the number of events increased in both the ARNI (443 vs. 338) and ramipril (516 vs. 373) arms, but the advantage of the ARNI over the ACE inhibitors now reached statistical significance (HR, 0.85; P = .01).

"The data suggest that maybe we should find definitions for adjudication that are closer to clinical judgment in the real world and clinical practice," Berwanger said.

One possible explanation for the neutral result in PARADISE-MI is that benefit of an ARNI over an ACE inhibitor would only be expected in those at risk for progressive left ventricular dysfunction, and it is likely that a substantial proportion of patients enrolled in this trial recovered, according to Johann Bauersachs, MD, PhD, professor and head of cardiology at Hannover (Germany) Medical School.

"You cannot predict which patients with reduced LV function following an MI will go on to chronic remodeling and which will recover," said Bauersachs, who was an ESC-invited discussant of Berwanger's post hoc analysis.

He agreed that Berwanger has raised several important issues in standard trial design that might have prevented PARADISE-MI from showing a benefit from an ARNI, but he pointed out that there are other potential issues, such as the low use of mineralocorticoid antagonists in PARADISE-MI, that may have skewed results.

However, he agreed generally with the premise that there is a need for trial design likely to generate more clinically useful information.

"We have now seen the win-ratio approach used in several studies," said Bauersachs, citing in particular the EMPULSE trial presented at the 2022 meeting of the American College of Cardiology. "It is a very useful tool, and I think we will be seeing it used more in the future."

However, he indicated that the issues raised by Berwanger are not necessarily easily resolved. Bauersachs endorsed the effort to consider trial designs that generate data that are more immediately clinically applicable but suggested that different types of designs may be required for different types of clinical questions.

Berwanger reports financial relationships with Amgen, AstraZeneca, Bayer, Bristol-Myers Squibb, Pfizer, Servier, and Novartis, which provided funding for the PARADISE-MI trial. Bauersachs reports financial relationships with Amgen, AstraZeneca, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Cardior, Corvia, CVRx, Novartis, Pfizer, Vifor, and Zoll.

This article originally appeared on, part of the Medscape Professional Network.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.