Nonfatal MI Not a Death Surrogate in Cardiology: Meta-analysis

A meta-analysis spanning more than 100 cardiology RCTs over three decades in more than a million patients offers a stern rebuke to the notion that nonfatal MI can serve as a surrogate for mortality, researchers say.

Whether the trials evaluated primary or secondary prevention, or even revascularization, the data’s message was the same: reducing the rate of nonfatal MI does not, in turn, reduce the rate of either all-cause or cardiovascular death.

Senior author David L. Brown, MD (Washington University School of Medicine in St Louis, MO), told TCTMD that the impetus for their analysis was encountering cardiologists who were convinced that the opposite was true—namely, that PCI not only reduces nonfatal MI but also “if you just wait long enough” improves survival. Similar debates have been going on in oncology, related to whether the term “progression-free survival” is a valid surrogate to overall death, he pointed out.

Brown drew a distinction between association—where endpoints simply go in the same direction—and surrogacy. “MI is associated with death, because in some people they have an MI and they die. That’s not really what surrogacy is. Surrogacy is: if you can reduce that risk of the surrogate, the true outcome should be reduced,” he explained. “That’s what we don’t see.”

This flies in the face of what every cardiologist has encountered when treating a patient who presents to the emergency department with acute MI only to die in the cath lab or hospital. “We all agree that death is on the natural pathway of MI in some people. This is asking a different question, which is if you can reduce the occurrence of MI—however you define it—do you alter the mortality? The answer is really no,” said Brown.

The researchers’ conclusion will perhaps disappoint the 86% of respondents to a Twitter poll, posted earlier this month by Brown, who said they expected a reduction in nonfatal MI would translate into a reduction in either all-cause or CV death, or both.

For Robert W. Yeh, MD (Harvard Medical School, Boston, MA), the message that lowering nonfatal MI rates won’t by default decrease death corroborates what some have long suspected.

“It’s not altogether surprising that if one holds MI to a standard as a surrogate endpoint that it won’t pass that test, even more so in recent years,” he commented to TCTMD. Yeh gave the example of PCSK9 inhibitor trials FOURIER and ODYSSEY, which showed MI reductions but no effect on mortality: “When those papers came out, it was clear at that time that ‘MIs aren’t what they used to be.’”

Ajay Kirtane, MD (NewYork-Presbyterian/Columbia University Irving Medical Center, New York, NY), who previously explored MI definitions in a blog for TCTMD, described the new meta-analysis as an interesting, exhaustive effort. As cardiologists, it makes sense to note that “not everything we do is going to directly impact a patient’s mortality, certainly over the relatively short term that is the time span of many clinical trials,” Kirtane agreed.

“The challenge is what you make of it at the end of the day,” he commented. “I don’t think we can say it’s unimportant to prevent nonfatal myocardial infarction. That would obviously be a stretch way too far. But I do think it’s reasonable to think that not all things that are adjudicated as a nonfatal MI are necessarily the same.”

More granularity in defining what constitutes an MI and verifying which MI types have occurred would improve trial design, Kirtane urged, and Brown himself made a similar point.

“It’s important to be circumspect when you’re looking at things that might be categorized as nonfatal MIs to know exactly what the circumstances are,” Kirtane said, referencing the discussions over cardiac biomarker elevations after PCI.

The prognostic significance of various definitions for periprocedural MI have been intensely debated in recent years, with some arguing it may be time to ditch the endpoint. This study, published online in JAMA Internal Medicine, may serve to broaden the scope.

MI Put to the Test

Led by Kevin O’Fee, MD (Washington University School of Medicine in St Louis), the researchers searched PubMed through the end of 2020 for trials of at least 1,000 patients and 24 months of follow-up.

They ultimately analyzed 144 trials with 1,211,897 patients, setting the threshold for surrogacy at a coefficient of determination (R₂) value of 0.8 and 95% CI excluding 0.6. Values for R₂ fall between 0 and 1.00, ranging from no to perfect surrogacy.

Their calculations showed nonfatal MI didn’t meet this bar for all-cause mortality (R₂ 0.02; 95% CI 0.00-0.08) or cardiovascular mortality (R₂ = 0.11; 95% CI 0.02-0.27). Nor did it serve as a stand-in for all-cause death when looking separately at primary prevention, secondary prevention, mixed prevention, or revascularization trials. It didn’t matter whether the trials were conducted before 2000, in the years 2000-2009, or thereafter, or whether their follow-up duration was below 4 years, up to 6 years, or beyond 6 years.

The R₂ cutoff of 0.8, Brown acknowledged, is “arbitrary and stringent,” though based on oncology’s threshold for surrogacy.

“Definitely, I confess to both of those sins if they are sins—but it should be stringent if you’re going to ask the FDA to approve a treatment based on the effect on a surrogate without doing a large enough or long enough trial to see what the effect is on overall mortality, which is what most people are most concerned about,” he said.

Even with thresholds of 0.6 or 0.4, none of the calculations came close, Brown observed. “If it had been a lot more borderline, I think this research would have died with this trial.” Instead, it’s sparked conversations.

Why Didn’t Mortality Track?

One explanation for their results is that there are many different kinds of MIs and that the field’s understanding of them isn’t complete, said Brown, pointing to myocardial infarction with nonobstructive coronary arteries (MINOCA) as an example. “I think it’s all mixed together. I think you can have MINOCA and still have an 80% LAD lesion. I think you can have an 80% LAD lesion and still have ‘INOCA’—you can still be ischemic from other mechanisms. But our paradigm has been to focus on the severe obstructive epicardial lesion for so long,” he commented.

Under this model, PCI “works, it makes the angiogram look better, it frequently makes the patient feel better for whatever reason,” said Brown. “But I think that did a disservice to the field, because I think [MI’s] biology is much more complicated than we give it credit for.”

Moreover, trials may be mixing some type 2 MIs in with type 1 MIs, further muddying the data, Brown noted. Even in clinical practice, “we need to be much more open-minded about what an MI is, what causes an MI, and try to make a very precise diagnosis every time someone comes in with an MI, [so we can better] understand what’s going on,” he said. “Same thing in clinical trials: I think that the adjudication needs to be much more thorough and encompassing of all the mechanisms that we currently know about, with an open mind to the fact that there may be mechanisms we haven’t stumbled upon yet.”

In an accompanying editorial, Associate Editor Tracy Y. Wang, MD (Duke Clinic Research Institute, Durham, NC), and Editorial Fellows Mithi del Rosario, MD, and Vinay Guduguntla, MD (both University of California, San Francisco), also offer a few explanations as to why surrogacy didn’t pan out.

For one thing, only three of the RCTs specified MI type, making it impossible to know how the disparate pathophysiologic characteristics and management of type 1 and 2 MIs might have affected outcomes, they write. Moreover, “few cardiac trials have durations of follow-up long enough for deaths to accrue; with longer follow-up, it is possible that the correlation between nonfatal MI and mortality will be found to be stronger,” the editors write.

Yeh mentioned other possibilities, as well. “The MIs we diagnose now are just less significant in many ways than the ones we used to diagnose, because of changes in biomarker sensitivity [and] definitions,” he noted. “In addition, we just do a better job of treating MIs than we did historically, but even historically the fatality rate associated with MI is not 100%—in fact, it’s a lot less than that.”

Nonfatal MI Still Matters

This doesn’t mean nonfatal MIs aren’t worthy of attention, all agreed.

“Like many people have said, the results of this [study] do not mean nonfatal MI shouldn’t be an endpoint,” stressed Brown, “because it costs money to take care of these people and there’s a quality-of-life impact when somebody gets admitted with chest pain that turns out to be an MI.”

The knowledge that nonfatal MI isn’t a valid surrogate for death can help put its relevance in perspective, he said. “To me, it puts MI into the basket of adverse events that you want to avoid but may not kill people if you don’t.” Groin bleeds, contrast nephropathy, and the like, Brown continued, “are all adverse outcomes. MI shouldn’t be given a loftier perch because people think ultimately we’ll be saving lives by preventing them. . . . It’s still a perfectly appropriate outcome along with all those other things that patients don’t like.”

Of course, patients would rather not have an MI and it’s worth preventing one—the question is how to weigh it against other events, Kirtane said, calling out bleeding as potentially “more important than a small nonfatal MI. . . . The reality is [its relevance] depends on the type of MI.” Another variable is the timeliness of treatment, which can affect prognosis, he added.

Yeh, like the others, said that it’s necessary to frame discussion of the meta-analysis in the right way. Nonfatal MI, surrogate or not, isn’t on the same scale as LDL cholesterol, which he called “the quintessential surrogate in cardiovascular medicine.”

“Nobody cares about their LDL number. You don’t know, walking around, what your LDL [level] is. It only matters in as much as it is a surrogate for your future cardiovascular risk,” Yeh observed, adding, “It serves one function.”

Nonfatal MI, on the other hand, is more on the scale of stroke, he suggested. “It’s an unexpected event, it’s scary, it requires hospitalization, it might require a procedure [that’s] not zero risk, it might lead to some long-term consequence. Just because it doesn’t per se lead to cardiovascular death, or all-cause mortality, doesn’t mean that it’s not a worthy thing to reduce in and of itself.”

When measuring MI, however, it’s important to capture events “that are consequential to patients,” whether those are prognostically meaningful or have an otherwise noticeable impact, Yeh asserted. “For example, one thing that’s not that is periprocedural MI based on a minimal troponin elevation. . . . Those should not drive whether devices or therapies are approved or not approved.”

Kirtane, on the topic of whether a carefully designed trial could pick up on an MI-related mortality difference after PCI, said it would be unlikely to work in an all-comers MI population but perhaps would be successful in high-risk patients if the follow-up were long enough.

“If you look at, for instance, non-STEMI ACS and take the right patients with risk factors and a presentation that’s acute, . . . I don’t think there’s any question that you’re going to be preventing adverse outcomes such as recurrent MI and death, and if you had a large enough study of death alone, by taking those patients and revascularizing them,” he said. “The challenge is that most of the studies include patients that are low risk, patients that might not have a true ACS, so that’s where some of the noise comes in. And they don’t follow the patients long enough.”

Until there’s greater precision in clinical adjudication of type 1 versus type 2 MIs, it could be hard to detect a survival benefit—if one does exist—with revascularization, said Brown. Type 2 accounts for anywhere from 2% to 58% of all MIs, the researchers note in their paper, and its treatment focuses on addressing the underlying comorbidity causing a “supply and demand mismatch.” Not differentiating between type 1 and 2, they say, runs the risk of blunting any potential surrogacy for type 1 MI.

“That’s an area where we definitely to get better and more precise, and maybe somebody should develop a biomarker [for type 2 MI] or something like that. I think that would shine a lot of light on this issue generally,” Brown said.