Making Sense of MI: EXCEL and SYNTAX Tackle Event Definitions

The rates of periprocedural myocardial infarction following PCI or CABG surgery vary significantly depending on the definition—some of these are associated with an increased long-term risk of death after revascularization while others are not, according to two new analyses from the EXCEL and SYNTAX-Extended Study (SYNTAXES) investigators.

In the EXCEL trial, a study that has been criticized for failing to initially present MIs based on the Universal Definition (UDMI), which was a prespecified secondary endpoint, investigators reported that the rates of periprocedural MI after PCI and CABG varied substantially—and consequently so could the primary endpoint—depending on how the infarctions were defined.

While the rate of periprocedural MI favored PCI as part of the protocol-defined primary endpoint, the advantage flipped to CABG surgery when investigators tallied up MIs ascertained according to the Universal Definition. Importantly, the prognostic significance of those MIs varied, too, with protocol-defined periprocedural MIs after CABG and PCI appearing to be more prognostically relevant, say investigators.

“Periprocedural MIs differ by their definition but so does their absolute and relative importance with respect to either all-cause mortality or cardiovascular mortality,” Gregg Stone, MD (Icahn School of Medicine at Mount Sinai, New York), EXCEL lead investigator, told TCTMD.

Periprocedural MIs identified with the UDMI were only linked to mortality after surgery, not after PCI, suggesting it is unfair to compare these rates because the events have different long-term implications after PCI and surgery, said Stone. Instead, the protocol definition that captured periprocedural MIs was associated with a similar risk of death after PCI and CABG, making it a much better endpoint to compare the two revascularization strategies, he added. 

I’m very sensitive to the idea that [periprocedural MIs] are somewhat artificial and that [these events] can modulate the Kaplan-Meier event curves. Patrick Serruys

In SYNTAXES, investigators also turned their attention to the prognostic significance of periprocedural MIs using different definitions.

Like the EXCEL researchers, they found that event rates were highly dependent on how they were defined, with the lowest rates observed when using the SYNTAX and Fourth Universal Definition of MI, both of which required accompanying electrocardiographic evidence of myocardial damage. They also found that periprocedural MIs identified with SYNTAX criteria or the Fourth UDMI were significantly linked with long-term all-cause mortality, deeming these events more “clinically relevant.”

“Periprocedural MI is not really an event when you think about it,” Patrick Serruys, MD, PhD (National University of Ireland, Galway and Imperial College London, England), senior SYNTAXES investigator, told TCTMD. “It’s not like death or reintervention or myocardial infarction at follow-up. You take a blood sample and decide, based on the definition you’ve decided on, if you have a periprocedural event. The issue is that there are multiple definitions.”

Michael Reardon, MD (Houston Methodist Hospital, TX), who wasn’t involved in EXCEL or SYNTAX, said the new analyses show just how hard it is to assess periprocedural MIs with two revascularization strategies as distinct as PCI and CABG. “This is really hard to do right,” he said. “If it were easy to do right, we’d have a way that everybody agreed on, but it’s hard to do right because you have such vastly different procedures to which people react very differently.”

For Reardon, EXCEL showed that CABG surgery was associated with a significantly lower risk of revascularization, as well as a much-debated reduction in all-cause mortality, and those endpoints, rather than MI, are “the big things.” The fact that different definitions of MI “can move you from either side of a P value . . . tells us the P values are so fragile that it’s probably not as important in [clinical] decision-making.”

EXCEL Tackles the UDMI Question

In EXCEL, investigators used a modified version of the Society for Cardiovascular Angiography and Interventions (SCAI) definition for periprocedural MI as part of their primary composite endpoint of death, stroke, and MI. The EXCEL definition required a CK-MB rise more than 10 times the upper reference limit (URL) as a standalone measure or a rise of more than 5 times the URL with supporting electrocardiographic, angiographic, or imaging evidence of myocardial ischemia. Stone has argued that if periprocedural MI is included as part of a primary or secondary endpoint, it should be clinically meaningful and this is why they used the modified SCAI definition to identify periprocedural MIs.

“By clinically meaningful, we mean prognostically linked with other adverse outcomes,” he said. “And it should have similar prognostic relevance for the two different test therapies.” 

As noted, outcomes using the Third UDMI, a prespecified secondary endpoint, weren’t initially reported at 5 years, leading to a deluge of criticism from the surgical community and contributing to the decision by the European Association for Cardio-Thoracic Surgery to withdraw their support for the current treatment recommendations for left main CAD. The EXCEL investigators have previously explained that cardiac troponin levels, the preferred biomarker for the Universal Definition, weren’t routinely collected as part of the trial. As such, they said, it was unclear how to report the data. In this new analysis, the researchers present the procedural UDMI rates using assessments of peak CK-MB, which was collected in all patients, troponin levels in 919 patients in whom it was collected, and a mixed collection of CK-MB and troponin. The Universal Definition uses different for biomarker elevations with PCI and CABG surgery, with the Third UDMI requiring more-extensive myonecrosis after CABG surgery than after PCI, said Stone.

Using the protocol definition, the rate of periprocedural MI was 6.1% with cardiac surgery and 3.6% with PCI (P = 0.015). When MI rates were assessed using the Universal Definition, surgery had the advantage regardless of which biomarkers were used. In the “mixed” cohort that assessed events with CK-MB and troponin, the rate of periprocedural UDMI was 2.2% in the surgical arm and 4.0% in the PCI group (P = 0.025). When the analysis was restricted to patients with CK-MB measurements only, which can be used as part of the UDMI if troponin isn’t available, the rate of UDMI was 1.4% in the surgical arm and 3.3% among those treated with PCI (P = 0.007). In patients with available cardiac troponin measures, the rate of UDMI was 4.6% with PCI and 3.5% with surgery (P = 0.38).  

As we showed in EXCEL, only large biomarker elevations, consistent with large, extensive myonecrosis, are prognostically related to subsequent death. Gregg Stone

In terms of their prognostic importance, periprocedural MIs using the protocol definition were associated with a more than twofold increased risk of cardiovascular mortality at 5 years in the entire cohort (adjusted HR 2.18; 95% CI 1.13 to 4.23), with a similar risk after PCI and CABG surgery (P = 0.86 for interaction). However, when assessing the UDMI after CABG—based on either CK-MB or troponin, or a mix of both—increased rates of this endpoint wwere associated with a higher risk of all-cause and cardiovascular death at 5 years. No such association was seen between UDMI and mortality after PCI.

When analyzed by biomarker levels alone, MI based on peak CK-MB ≥ 10x URL or peak troponin ≥ 70x URL was associated with an increased risk of all-cause and cardiovascular mortality at 5 years.

“As we showed in EXCEL, only large biomarker elevations, consistent with large, extensive myonecrosis, are prognostically related to subsequent death,” said Stone. “That would mean that essentially all procedural myocardial infarctions after CABG would include elevations that high, but many after PCI would have lesser biomarker elevations.”

To TCTMD Stone said procedural MI is one of the most potentially biased endpoints, and “it’s one of the reasons why there’s no widespread agreement as to what the optimal definition should be.”

SYNTAXES and Multiple MI Definitions

In SYNTAX, patients with three-vessel disease and/or left main CAD who were eligible for both PCI and CABG based on the heart team’s assessment were randomized to receive a DES (Taxus Express; Boston Scientific) or surgery. In the trial, the primary endpoint was a composite that included all-cause mortality, stroke, MI, or repeat revascularization. Periprocedural MI, which was included as part of the MI endpoint, was defined by peak CK-MB/peak total CK > 10% (or CK-MB ≥ 5 ULN) and ECG criteria, such as new Q waves on ECG.

To TCTMD, Serruys said with so many definitions of periprocedural MI now circulating, including the Universal Definitions that preference cardiac troponin, they sought to assess the clinical relevance of periprocedural events based on the SYNTAX, ISCHEMIA, EXCEL, SCAI, and Fourth Universal definitions. The Fourth UDMI and ISCHEMIA definitions both use different thresholds for CK-MB increases after PCI and CABG surgery and rely on ECG and imaging criteria to adjudicate periprocedural MI. The SCAI definition relies on ECG criteria if CK-MK is ≥ 5x ULN (no ECG criteria is needed if CK is ≥ 10x ULN) but uses the same criteria for PCI and CABG surgery.

“We wanted to apply the five definitions to see what happens,” said Serruys. “Obviously, they have a major impact on the time to events or the composite endpoint. The clinical relevance is also quite different.”

Using SYNTAX and the Fourth UDMI, the rates of periprocedural MI were similar. With SYNTAX, periprocedural MI occurred in 2.7% and 2.4% of the PCI and CABG-treated patients (P = 0.756). With the UDMI, the rates were 3.0% and 2.1%, respectively (P =0.281). Using the ISCHEMIA definitions, the rates were higher and favored PCI, with 6.0% periprocedural MIs in the PCI arm and 8.8% in the CABG group (P = 0.03). When using either the SCAI or EXCEL definitions, the rates of periprocedural MI were the same: 5.7% in the PCI group and 16.5% in the surgical group (P < 0.001).

In terms of the prognostic implications, periprocedural MIs identified with the SYNTAX and Fourth UDMI were more strongly associated with mortality than the EXCEL- or SCAI-defined periprocedural events. With SYNTAX, periprocedural MI after PCI was an independent predictor of death at 1 and 10 years, but periprocedural MI after CABG was only an independent predictor for mortality at 1 year. Similar findings were observed with the Fourth UDMI and ISCHEMIA definitions. With the SCAI and EXCEL definitions, periprocedural MI after PCI was associated with an increased risk of death at 1 and 10 years, but these events had no impact on mortality in the CABG arm.

With enzymatic elevations but no accompanying ECG changes, increases in CK-MB after PCI had an impact on the risk of death at 1 and 10 years, but they were not associated with higher mortality in the surgical arm. “Basically, CK-MB 10x [ULN] for percutaneous intervention has a long-term impact, even after 10 years,” said Serruys. “The fact that you had a lot of enzyme release during the procedure is never a good thing. It seems to be a little bit different for the surgeon.”

With CABG surgery, elevations in CK-MB with ECG changes, such as new Q waves or persistent left bundle branch block, were associated with a significant risk of death at 1 year but only linked to a trend toward higher mortality at 10 years. In contrast, periprocedural MI based on the same criteria was associated with early and later mortality in the PCI arm.

On the whole, Serruys said these new data show that increased biomarker elevations plus evidence of permanent, irreversible signs of myocardial damage, such as new Q waves, loss of viable myocardium on imaging, or permanent wall motion abnormalities, portend poor long-term outcomes.

Clinical Implications

With respect to the unpublished UDMI data, which triggered a BBC Newsnight investigation, Reardon said he doesn’t believe there was any intention on the part of the EXCEL clinical researchers to obscure or hide these data. Many clinical researchers, in hindsight, wish they had designed their trials a little bit better given how difficult it is to come up with relevant clinical endpoints that will have a meaningful impact on practice.

“These definitions show us just how hard it is to adjudicate important myocardial infarctions between PCI and coronary artery bypass surgery that may have a bearing on your survival,” said Reardon. “We continue to refine the definitions to get better at it, but [even] as we look back at the original trial there is still important [information] there to use.”   

In an editorial, Donald Cutlip, MD (Beth Israel Deaconess Medical Center, Boston, MA), calls the different definitions a dilemma for researchers, particularly for comparisons of two distinctly different procedures like PCI and CABG surgery. In both studies, supporting criteria—including new Q waves, documented vessel occlusion, or imaging evidence of myocardial damage—had a big impact on MI diagnosis after CABG surgery.

The P values are so fragile that it’s probably not as important in [clinical] decision-making. Michael Reardon

“Biomarker elevation alone [after CABG], even at high levels, without these supporting criteria increased the frequency of procedural MI but appeared to add little, if any, to the value of MI as a correlate with subsequent mortality,” he writes. “In contrast, supporting criteria seemed to detract from biomarker thresholds for defining procedural MI after PCI.”

Going forward with future research, Cutlip believes it may be necessary to use different criteria for periprocedural MIs with CABG and PCI and that supporting ECG or imaging criteria should be a requirement after CABG surgery. With the UDMI, however, the supporting criteria don’t appear to add much in the PCI setting and “should be removed for a standalone biomarker threshold that is meaningful.” High-level biomarker increases, such as those used as part of the SCAI definition, are clearly associated with increased mortality, he said.

To TCTMD, Serruys took a step back, saying that the ultimate goal of any lifestyle intervention, drug, or device therapy is to prolong and improve quality of life. Periprocedural MI, he said, is a somewhat “artificial” endpoint, especially if there are no late consequences in terms of clinical events from isolated cardiac biomarker release. Nonetheless, these events are included as part of time-to-event composite endpoints to adequately power studies. 

“What I have learned with surgery in my career is that you need to talk about long-term follow-up when you talk about revascularization,” he said. “At 5 or 10 years, that’s really where you see a difference. I almost have the temptation to say let’s forget about periprocedural MI. Either you will pay the bill at the end with heart failure or death, or with poor quality of life, but it’s almost a technicality that you have these periprocedural MIs. Of course, we must try to eliminate that, and we’re making progress every day in that direction, but I’m very sensitive to the idea that [periprocedural MIs] are somewhat artificial, and that [these events] can modulate the Kaplan-Meier event curves.”

For Reardon, if the risk of CABG surgery for left main CAD was equivalent to the risk of PCI, the patient experience was the same, and the surgeon planned to use arterial grafts, then everybody would choose surgery, even cardiologists. “The problem is the patient experience is not the same, the risk is not the same, and not all surgeons use arterial grafts,” he said. 

Note: Stone and several co-authors of the EXCEL analysis are faculty or employees of the Cardiovascular Research Foundation, the publisher of TCTMD.