Too Easy to Misconstrue: Is It Time to Ditch Periprocedural MI?
There are many definitions and some may even be useful, but there’s little consensus on MI endpoints in clinical trials.
If there’s one thing cardiologists and cardiovascular surgeons can agree upon, it’s that serious, high-risk coronary artery disease needs fixing—it’s the how and by whom that are contentious.
But measuring the risks and benefits of surgical versus interventional approaches has itself become an increasingly quarrelsome debate, with one clinical outcome measure proving particularly polarizing: periprocedural myocardial infarction.
Revascularizing an occluded artery, either with a stent or bypass graft, can injure myocardium and produce a biomarker spike that, to varying degrees, gets picked up by this endpoint. With the introduction of high-sensitivity biomarker assays, there’s an ever-growing chorus of cardiologists who believe the assessment of periprocedural myocardial infarction—no matter how it’s defined—muddies the water, adding nothing but “noise.” Others argue that MIs during a procedure or shortly thereafter are an important indicator of future adverse events, including mortality. What matters, they say, is choosing the right definition of MI.
For Kenneth Mahaffey, MD (Stanford University, CA), the assessment and adjudication of these early MIs is a historical artifact, harkening back to an era when physicians were treating patients with acute IV antiplatelet and anticoagulant therapy where the focus was on short-term clinical outcomes.
“If you go way back to the [glycoprotein] IIb/IIIa inhibitor days in the 90s and early 2000s, we were looking at 30-day endpoints,” Mahaffey told TCTMD. “So you wanted to include periprocedural events. People thought they were important and still think they are important. However, you were actually trying to mitigate acute MIs in the hospital setting with acute therapies. PURSUIT, GUSTO, and all of those trials that studied the IIb/IIIa inhibitors, that's why these events were used—acute interventions in the acute phase of the disease with short-term follow-up.”
Now, however, the adjudication of MI has become a head-scratching complication for clinical trialists. Not because these events are hard to identify, but rather the opposite. The use of cardiac troponin (cTn) as a biomarker, including when measured by high-sensitivity troponin assays, has allowed physicians to diagnose periprocedural MI in many of their patients undergoing revascularization.
“The noise from these events of dubious significance has become ‘deafening,’” said Ajay Kirtane, MD (NewYork-Presbyterian/Columbia University Irving Medical Center, New York, NY). He thinks it’s high time to end the routine ascertainment of periprocedural biomarkers in coronary revascularization clinical trials, particularly since this isn’t done in routine practice.
PCI has evolved to same-day discharge for a large proportion of patients, he said, which means physicians/researchers don’t even have the patient in the hospital long enough to document, based on biomarkers, if they had periprocedural MI.
We don't want somebody not collecting [cTn] because they don't want to know, or they want to suppress it. On the other hand, to make people routinely collect it, it's practically impossible to do. Ajay Kirtane
“Combine that with the fact that most hospitals have now gone to troponin assays entirely, and even with troponin assays many hospitals are adopting high-sensitivity troponin assays, which means that even if you did happen to collect it, it's going to be positive, in some cases, in more than 50% of cases,” he told TCTMD. Instead of routinely collecting cTn, or even creatine kinase-myocardial band (CK-MB), Kirtane thinks a “hybrid approach” might be more feasible. If a patient had angiographic complications during the procedure or continues to have chest pain afterwards, situations where they are going to stay in the hospital, then a blood draw for cTn would be justified.
“That type of approach codifies what good clinical practice would be,” he said. “We don't want somebody not collecting [biomarkers] because they don't want to know, or they want to suppress it. On the other hand, to make people routinely collect it, it's practically impossible to do.”
Taking things one step further, John Spertus, MD (Saint Luke’s Mid America Heart Institute, Kansas City, MO), said he believes that if researchers want to drop periprocedural MI from the primary endpoint then all MIs should be dropped. In fact, Spertus told TCTMD he simply doesn’t buy that MIs are as clinically relevant today as they were in the past. MIs used to be catastrophic events, he said, but that’s no longer the case, especially with events detected by high-sensitivity assays.
“If you go into any heart failure clinic, 20% of those patients are going to have positive troponin even though they’re not having a heart attack,” he said. “We’re allowing a biomarker to define something that we’re interpreting as clinicians to be something very different. We’re interpreting spontaneous MI as a patient coming in with a STEMI, with crushing chest pain, and developing heart failure. We associate spontaneous MI as something different from a periprocedural MI, but in reality, it might not be so different.”
So Many Definitions, Little Consensus
Both the Third and Fourth Universal Definition of Myocardial Infarction (UDMI) differentiate procedural events shortly after PCI (type 4a) and CABG surgery (type 5) by using different thresholds for elevations in cTn values, with surgical events requiring a larger increase in baseline values (> 5 versus > 10 times the 99th percentile upper reference limit [URL] for PCI and CABG surgery, respectively). Importantly, both definitions require additional clinical evidence, such as new ECG changes, the development of new pathological Q waves, or angiographic or imaging evidence supporting MI.
Despite the name, the Universal Definition is far from the only MI definition out there, which only adds to the confusion.
The Academic Research Consortium (ARC), a collaboration between several research organizations that aim to develop consensus definitions and standard nomenclature for clinical trials, has its own criteria, recommending that both CABG and PCI use the same standard for periprocedural MI: an absolute increase of cTn more than 35 times the URL plus additional measures, such as new Q waves or angiographic/imaging evidence. The Society for Cardiovascular Angiography and Interventions (SCAI) also uses the same threshold for surgery and PCI but prefers CK-MB over cTn as the biomarker of choice, defining periprocedural MI as an event within 48 hours of the procedure with CK-MB increases 10 times the upper limit of normal (ULN). If new pathological Q waves are present, the SCAI threshold is lowered (CK-MB > 5 times ULN).
In the EXCEL trial—where squabbling over periprocedural MI has reached a fever pitch, investigators went one step further and used a modified SCAI definition for MI. The EXCEL definition required a CK-MB rise more than 10 times the URL as a stand-alone measure or an increase of more than five times the URL with supporting electrocardiographic, angiographic, or imaging evidence of myocardial ischemia.
Given all the different definitions, biomarkers, and thresholds for events, some clinical trialists have dispensed with periprocedural MI altogether. That’s the route the NOBLE investigators took in their trial of PCI versus CABG surgery for left main CAD. While EXCEL showed PCI was noninferior to surgery at 3 years in patients with a low-to-intermediate SYNTAX score, subsequent follow-up data showed the trial results changed if different MI definitions were used. In NOBLE, only nonprocedural MIs were counted. At 5 years, CABG was superior to PCI for reducing the risk of death, stroke, and MI.
We’re allowing a biomarker to define something that we’re interpreting as clinicians to be something very different. John Spertus
Despite the confusion periprocedural MIs bring to the table, not everybody thinks the endpoint should be abandoned.
For Hector Garcia-Garcia, MD, PhD (Medstar Washington Hospital Center, DC), periprocedural MI remains a valid endpoint that should continue to be part of clinical trials testing different revascularization strategies. There are data, he said, linking these events to worse clinical outcomes.
For example, in a study-level meta-analysis published in 2019, Garcia-Garcia and colleagues reported that 30% of all periprocedural MIs—regardless of the definition or thresholds used—were associated with a higher risk of mortality. A similar patient-level meta-analysis of five contemporary stent trials showed that periprocedural MIs correlated with a higher risk of death.
Garcia-Garcia, who led the writing committee standardizing the ARC-2 clinical endpoint definitions and who continues to serve on the clinical event committees of several trials, argued that the ARC-2 definition takes into account the technical nature of PCI and CABG surgery by applying a meaningful threshold for periprocedural MI that largely excludes “noise.” In their review of the data, an increase of cTn > 35 times ULN has been shown to be the most suitable for detecting a large, prognostically relevant MI for both PCI and CABG surgery. The SCAI definition also uses a troponin T increase 35 times the ULN as criteria for periprocedural MI if CK-MB values are not available.
“It’s a very high threshold, and not many patients achieve that,” said Garcia-Garcia. “At the end of the day, what we’re trying to capture is whether these procedures are inducing some injury that would affect outcomes.”
Lead EXCEL investigator Gregg Stone, MD (Icahn School of Medicine at Mount Sinai, New York, NY), has also said he’s not ready to jettison periprocedural MI as an endpoint, particularly for clinical studies comparing different strategies, such as a DES versus guideline-directed medical therapy in ISCHEMIA or DES versus CABG.
INSERT TWEET: https://twitter.com/GreggWStone/status/1317521122007617536
In September, the EXCEL researchers made the case that periprocedural MIs captured using their study definition were clinically and prognostically meaningful. In their analysis, periprocedural MI was associated with similar risk of death after PCI and CABG surgery, while MIs using the Third Universal Definition were only linked with mortality in the setting of surgery.
For studies comparing two stents, Stone said on Twitter that the routine measurement of biomarkers might be unnecessary.
The Stent Didn’t Fail, the MI Definition Did
Back in October at TCT Connect 2020, a new low-profile, fixed-wire DES (Svelte Medical) came up short against a contemporary everolimus-eluting stent in the OPTIMIZE trial. The investigational device missed achieving noninferiority based on a primary composite endpoint that included target-vessel MI. But 90% of these MIs, which proved to be the primary driver of the composite endpoint, were periprocedural events; of those, 96% were not supported by ECG changes.
In total, just 25% of trial patients had cTn levels measured and these patients accounted for 80% of the target-vessel MIs, said lead investigator Dean Kereiakes, MD (The Christ Hospital, Cincinnati, OH). The trial failed, Kereiakes said, not because of the stent, but because of the biomarker definition.
At the end of the day, what we’re trying to capture is whether these procedures are inducing some injury that would affect outcomes. Hector Garcia-Garcia
Mahaffey, who was part of the Standardized Data Collection for Clinical Trials Initiative that worked with the US Food and Drug Administration to define cardiovascular and stroke endpoints for clinical trials, stressed periprocedural MIs can no longer be defined by biomarker elevations alone.
“That’s been baked into the definitions—you have to have something else happen, either clinically or with imaging,” he said. “Was there vessel closure? Dissection? Was there a loss of a graft? Is there a new regional wall motion abnormality? The bar to meet the endpoint can no longer just be an enzyme elevation to whatever level you arbitrarily decide is going to define your endpoint. You need to have the elevation and something else.”
When deciding on a clinical endpoint as part of a randomized trial, Mahaffey said one critical question is whether it’s important to the study in terms of the treatments evaluated. “And if it is, then you need to figure out a way to define it,” he said. While there are now several consensus documents defining MIs, what trialists don’t spend enough time thinking about is how they will systematically ascertain and adjudicate those events.
“That’s where you can get into trouble,” said Mahaffey.
In fact, the EXCEL researchers found themselves in hot water for this very reason. According to the trial protocol, they had selected the modified SCAI definition for MI as part of the primary endpoint, but they also included the Third UDMI as a secondary endpoint. However, collection of cTn was optional in the trial, and values were collected in just 919 of the 1,905 patients. When the 5-year results were presented and published late last year, the UDMI data were not included in the New England Journal of Medicine paper. Facing a barrage of criticism for this omission, investigators said the low numbers of patients with cTn measures was the reason the data was not immediately published.
But the UDMI-based results favored CABG surgery. That, along with controversy surrounding the higher risk of death at 5 years with PCI, led surgeon David Taggart, MD, PhD (University of Oxford, England), to withdraw his name as one of the authors. Many in the surgical community called foul, prompting the European Association for Cardio-Thoracic Surgery to withdraw their support for the current treatment recommendations for left main CAD.
To TCTMD, Kirtane said that several ongoing studies he’s been involved with over the years have had to troubleshoot how to avoid protocol violations because researchers failed to collect biomarkers at the required time interval after the procedure. It’s not deliberate malfeasance or negligence, he asserted, but simply because it’s not part of standard clinical practice.
Donald Cutlip, MD (Beth Israel Deaconess Medical Center, Boston, MA), as well as others, said the controversy surrounding the diagnosis of clinically relevant procedural MI after PCI has existed for decades, but has recently again become a focus following the trials of PCI versus surgery for left main CAD. On the whole, Cutlip urged that periprocedural MI should remain part of a clinical endpoint, but said it’s important to select the “right” definition.
In interventional cardiology trials, some of the MI definitions are challenging to use, he added.
“The Universal Definition requires supporting evidence for MI diagnosis, but I think most investigators who try to apply that in a clinical trial have run into issues where either it doesn't really add any specificity to the diagnosis or it's too difficult to obtain the high-quality supporting evidence to allow the definition to be applied,” Cutlip told TCTMD. “Most would agree that having a biomarker-only definition is better. And then it depends on how high should it be? I think, for the most part, most interventionalists have accepted the SCAI definition, which is pretty high but also seems to be clinically relevant.”
In the debate surrounding EXCEL, the surgical community emphasized that the 5-year results showed surgery was associated with a significantly lower risk of death compared with PCI. Taggart, the surgeon who withdrew his name from EXCEL, told TCTMD at that time that “death is arguably the most important endpoint of this trial.”
If death is the most important endpoint of a study—and it could be argued it’s the most relevant for patients and their families—then why not make all-cause mortality the primary endpoint? Size and money, say the experts.
“If you’re doing a strategy trial like PCI versus CABG rather than a device trial where you're comparing one device to another, I think it makes sense to have mortality as the most important endpoint,” said Cutlip. “The problem, of course, is the trials would have to be fairly large or the mortality difference would have to be fairly large to be able to show [a difference between treatments]. With 5-year follow-up, you might be able to show a difference, and here I think there is some evidence in the trials showing that maybe there is a mortality difference at 5 years. But to design it that way upfront may be a little more challenging.”
Dropping periprocedural events while keeping spontaneous MIs as part of a composite endpoint is not an unreasonable strategy, added Cutlip. “Maybe mortality itself is not enough, but if you have mortality, stroke, and spontaneous MI, and maybe something like new heart failure, that might be enough as an endpoint to tease out differences between treatments and it doesn’t rely on something that is clearly problematic to define.”
The only reason researchers include MI as part of a clinical composite endpoint is because to not do so would require massive, unaffordable trials with mortality as the primary outcome measure, said Spertus. “We want to create more outcomes, more endpoints, and so we add MI to make the trial more feasible. But my concern is that we’re focusing on something that isn’t as important as we wish it were and as we interpret it to be.”
Instead of focusing on the definition of MI, Spertus believes researchers should shift toward patient-reported outcomes, such as symptoms, function, and quality of life.
“That’s what we’re really trying to improve,” he said. “We want our patients to feel better and live longer. We should measure mortality, and we should measure the patient’s health status really carefully. Because the health status comes from the patient and we don’t believe our patients, we think of it as a soft, squishy, or unimportant endpoint, but I think that’s wrong.”
As the creator of the Seattle Angina Questionnaire (SAQ), Spertus acknowledged his personal bias but stressed the SAQ is reproducible, sensitive to change, and an important tool to capture symptoms, function, and quality of life. In ISCHEMIA, the invasive strategy on top of optimal medical therapy didn’t prevent a range of major cardiovascular events compared with medical therapy alone; however, it did provide significant angina relief in those who reported daily or weekly angina.
In other words, it was a “huge win” for invasive management with PCI/CABG if the patient had symptoms, though that benefit would have been missed if researchers didn’t track patient-reported outcomes, said Spertus.
Garcia-Garcia noted the ARC-2 collaborators, which included Spertus, Stone, and Taggart, as well as officials from the FDA, do provide important considerations if trialists use patient-reported outcomes as part of a clinical endpoint. Still, Garcia-Garcia worries about their use and doesn’t imagine these “softer,” more subjective outcomes replacing mortality or MI anytime soon, particularly given some of the ascertainment challenges during clinical trials.
Patient-oriented outcomes can and should be part of clinical trials, he said. Still, the first order of business is do no harm, and that’s measured with hard clinical endpoints, such as mortality and MI, including periprocedural MI. “One of the issues with quality of life is that human beings have very short memories,” said Garcia-Garcia.