The MI–Mortality Mismatch: When Lowering MI Doesn't Extend Life

John M. Mandrola, MD


November 04, 2021

One of the many great things about practicing medicine is watching it change.

Take myocardial infarction: not that long ago, it largely meant an occluded coronary artery, resulting in cardiac injury then scar. This often led to heart failure and ventricular arrhythmias. Any treatment that reduced MI events would seem a sure bet to reduce death rates.

And that is why scientists who ran clinical trials felt comfortable using MI as a surrogate marker for survival: You reduce MI, you lengthen life.

Times have changed. A study published last week led by David Brown, MD, and Kevin O'Fee, MD, from Washington University in St. Louis has challenged the notion that MI is a reliable surrogate for mortality.

JAMA Internal Medicine published the meta-analysis of 144 randomized clinical trials (RCTs) that included more than a million patients studied over a 40-year period.

The findings have major implications for evidence-based practice. First the study then its many messages.

The MI-Mortality Meta-Analysis  

The authors searched PubMed from its inception to December 2020 for RCTs of treatments to reduce MI, cardiovascular (CV) death and all-cause death in patients with coronary artery disease. They limited the search to the New England Journal of Medicine, the Journal of the American Medical Association and the Lancet because the most robust trials tend to be published in these journals.

The authors aimed to correlate the effect of a treatment on MI and both CV and all-cause death. In other words: if a treatment reduced MI, then did it also reduce the odds of death?

Each trial acted as a data point. The authors then plotted odds ratios of MI reduction on the X-axis and either all-cause- or CV-death reduction on the y-axis and drew a correlation line.

They quantified the relationship with a coefficient of determination called R2, which falls between 0 and 1. Zero indicates absence of any correlation and 1 indicates perfect correlation. A good surrogate should have an R2  ≥ 0.8.

The results were not subtle:

Nonfatal MI did not come close to the threshold of surrogacy for either CV death (R2 = 0.11) or all-cause death (R2 = 0.02).

The type of trial (primary or secondary prevention), the date of the trial (pre- or post-2000) and the length of trial did not matter: nonfatal MI still did not come close to meeting the criteria for surrogacy for CV or all-cause death.

Surrogate Markers Warrant Caution

The most obvious take-home message is the caution we should have for surrogate markers of outcomes.

In essence, clinicians have only two goals: help patients live better and longer. Much of the cardiology evidence base focuses on survival. The problem is that in a typical CV trial not that many patients die, which means you need to enroll a lot of patients or run the trial for many years to test for a survival benefit — and both add costs.

The compromise has been to include surrogate markers that occur at a higher rate than death. More events mean more statistical power to discern differences between treatments.

MI seems an obvious choice. It fulfills two of the three criteria for a good surrogate: it is associated with death in observational studies and has biological plausibility. But the most important criteria for surrogacy comes from trials. If MI is a good surrogate for mortality, treatments that reduce it should also reduce death.  

This meta-analysis strongly suggests that this is not the case.

This has immense implications for translation of clinical trials to practice. Consider the FOURIER trial of the lipid-lowering PCSK9 inhibitor drug, evolocumab. The drug was associated with a reduction in MI of 27% compared with placebo (P < .001), but there were no differences in CV death or overall death. In fact, numerically more patients died in the evolocumab arm (444) than in the placebo arm (426). 

Why a reduction of MI may fail to correlate with survival also teaches three important lessons about evidence translation today.  

Modern cardiology now detects smaller MIs because of high-sensitivity troponin tests and uses better therapies (eg, drug-eluting stents and P2Y12 inhibitors). In other words: as our ability to test and treat improves, MI becomes less reliable as an endpoint.

The second reason MI may have failed as a surrogate is multimorbidity and competing causes of death. As patients live longer and with more comorbidity, ischemia represents a smaller proportion of their causes of death.

Consider the ISCHEMIA-CKD trial, which compared an invasive vs conservative strategy in patients with CAD and advanced kidney disease. These were high-risk patients; cardiac event rates occurred in more than one third of patients. The invasive strategy was associated with a 15% reduction in total MI and a near 30% reduction in nonprocedural MI, but neither CV death nor all-cause death differed between the two treatment arms.

Multimorbidity also plays a role in the third possible reason for lack of surrogacy. We can call this the target problem.

Troponin assays have increased the recognition of type 2 MI — a mismatch of supply and demand without thrombosis. The approach to type 2 MI is to treat the underlying cause of the mismatch. Patients with multimorbidity (say anemia, sepsis, atrial fibrillation) are apt to have type 2 MI. Cardiac interventions directed at plaque rupture in these patients will have minimal to no impact.

Implications for Science and Its Consumers:

This paper will influence scientists who design future trials. If a drug or device is intended to improve survival, then MI, as it is currently defined, is inadequate as a surrogate. This will make it harder to bring new drugs and devices to market. But that's the price of success.

This paper also informs the consumers of medical evidence — clinicians and patients.

When we are told a drug or device has reduced a "composite primary endpoint," we must look deeper into the components. If the driver of "significance" is something other than CV death or all-cause death, say nonfatal MI, we should wonder how much value this new therapy will actually deliver.

You might push back and say…prolonging life isn't the only the job of a clinician. We are also tasked with improving quality of life. And avoiding an MI may improve quality of life. The key word in that sentence is "may." We should not assume a therapy that reduces a nonfatal surrogate improves quality of life (QOL).

Take the EMPEROR-Preserved trial of empagliflozin (Jardiance) vs placebo in patients with heart failure and a preserved ejection fraction. Empagliflozin significantly reduced the surrogate marker of heart failure hospitalizations but did not substantially improve QOL measured by the Kansas City Cardiomyopathy Questionnaire (KCCQ) summary score.

The other problem with QOL as an endpoint in CV trials is the need for proper placebo controls. This can be done with drug trials, but it is much harder with procedure trials. Patients in procedure vs medical therapy trials will know their treatment assignment — a huge bias when measuring quality of life.


I have increasingly come to believe that evidence needs an expiration date. There may have been a time when MI was a reasonable surrogate for mortality. But the extremely low correlations reported in this study suggest that MI should no longer be considered a surrogate for mortality.

Times change and so should our evidence translation.

This study deserves attention because it reminds us to question prevailing dogma, changes the way we assign value to our therapies, and helps plan future trials.

These are the hallmarks of meaningful science. I congratulate the authors for conceiving such a study and the journal editors for publishing it.

John Mandrola, MD, practices cardiac electrophysiology in Louisville, Kentucky and is a writer and podcaster for Medscape. He espouses a conservative approach to medical practice. He participates in clinical research and writes often about the state of medical evidence. 

Follow John Mandrola on Twitter

Follow | Medscape Cardiology on Twitter


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.