Periprocedural myocardial infarction: a web of definitions

In 1959, the World Health Organization made the first attempt to formulate criteria for coronary heart disease and myocardial infarction (MI), enabling worldwide epidemiological evaluation of this increasingly important disease [1]. At that time, this lethal disease was defined as an ‘obstruction of coronary circulation causing necrosis of a macroscopic, circumscribed area of the myocardium’, often only encountered during autopsy, with the footnote that indications of this syndrome could be present on an electrocardiogram (ECG). With improved understanding of coronary heart disease, the invention of cardiac imaging modalities and the discovery of cardiospecific biomarkers, definitions of MI evolved, resulting in the universal definition of myocardial infarction (UDMI-1) [2]. This UDMI is now updated to its fourth version (UDMI-4) and endorsed by the various European and American cardiac societies [3]. Still, this definition is also disputed, leading to controversies. The goal of this editorial is to elucidate these concerns, highlight the similarities and differences in applied definitions and suggest future directions for the use of MI definitions in clinical trials and guidelines.

RELEVANCE OF CARDIAC BIOMARKERS

Because myocardial necrosis results in the release of different proteins into the systemic circulation, the detection of increased biomarker concentrations with a rise-and-fall pattern is now a sine qua non for diagnosis of MI. Several of these proteins, such as lactate dehydrogenase, myoglobin and creatine kinase (CK), have been evaluated for their diagnostic accuracy for myocardial necrosis but were non-cardiac specific, potentially leading to false-positive results. However, the myocardial band (MB) fraction of CK (CK-MB) and later the cardiac troponins (cTn) proved to be more specific for MI, with the superior diagnostic accuracy of the latter test [2, 3]. Although the hypothesis is disputed [4], systemic blood cTn concentration increases are thought to be caused by (i) a fast release from the cytosolic pool of the myocardiocytes and (ii) a slow, steady release resulting from intramyocardial protein degradation of the structural pool [5]. This first process especially could facilitate the earlier detection of myocardial damage compared to previously used biomarkers. Currently, 2 proteins of the structural troponin complex are used for these tests, namely the T- (cTnT) and I-variants (cTnI), which have similar diagnostic accuracy. Although there is only one specific assay for cTnT, a variety of assays with different analytical characteristics exist for cTnI, for which different normal values have been formulated [6]. In the early days of cTn, any detected cTn concentration was considered supranormal and subsequently indicative of significant myocardial damage. However, with the emergence of high-sensitivity assays, physiological concentrations can be detected, enabling the identification of pathological cut-off values. These values are expressed as the 99th percentile upper reference limit (URL) of normal, tested in seemingly healthy reference populations, but they are known to be influenced by sex, age and comorbidities such as renal failure and chronic heart disease [7].

PERIPROCEDURAL MYOCARDIAL INFARCTION AS A SEPARATE ENTITY

Although these increased biomarker concentrations almost always indicate the presence of an underlying pathological process that warrants further evaluation, this is not the case for some new classes of MI. Not surprisingly, the advent of surgical and transcatheter revascularization also gave birth to iatrogenic MI as a complication, which is now known as periprocedural myocardial infarction (PMI). The UDMI-1 was the first to categorize this class of MI and to subdivide PMI into 2 types. Post-procedural MI after a percutaneous coronary intervention (PCI) was classified as MI type 4 (MI-4, 4a, 4b, 4c subdivisions), whereas an MI after coronary artery bypass grafting (CABG) was classified as MI type 5 (MI-5). The expansion of the MI classification proved to be a welcome addition, because both procedures in their own way inherently induce some (benign) myocardial injury. For PCI procedures, temporarily reduced coronary blood flow, slow flow or temporary disruption of collateral flow can lead to benign injury, whereas coronary dissection, (side-branch) coronary artery occlusion, coronary perforation or stent thrombosis can lead to significant necrosis, classified as MI-4. During CABG, mechanical compression, cannulation for cardiopulmonary bypass, cardioplegic arrest and coronary preparation are inherent causes of potentially benign injury, whereas insufficient cardioplegic protection, graft-related ischaemia, native coronary plaque rupture or perioperative haemodynamic instability might lead to significant myocardial necrosis, classified as MI-5 [3, 8].

UNIVERSAL DEFINITION OF MYOCARDIAL INFARCTION

In the UDMI, biomarker concentration increases are the cornerstone for diagnosis of MI-4 and MI-5, for which cTn was identified as the only accepted test, abandoning CK-MB measurements [8]. Given the difference in invasiveness of PCI and CABG, different cut-off concentrations that warrant further evaluation were proposed in the UDMI-4. For PCI-related MI-4, cTn concentrations that exceed 5 × URL indicate further diagnostic evaluation. Of note, in the UDMI-4, cTn increases always warrant supporting evidence, in terms of angiographic complications, regional wall motion abnormalities on imaging or electrocardiographic changes (i.e. all new ischaemic changes, including all ST-segment deviations and T-wave inversions). For CABG-related MI-5, the cTn cut-off was set arbitrarily at 10 × URL, also warranting supporting evidence. However, because ST-segment deviation and T-wave changes are common after CABG and not per se indicative of prognosis-impairing damage, only the development of ST-segment elevation with reciprocal depression, new left bundle branch block and new pathological Q-waves are considered diagnostic for MI-5 in UDMI-4 [3]. Although supported by the various cardiac societies, the UDMI consensus statements have also been criticized for their conservative and arbitrarily chosen cTn cut-off values, because 90% of patients having CABG would exceed the proposed cut-off concentration of 10 × URL [8].

DEFINITION OF THE SOCIETY OF CARDIOVASCULAR ANGIOGRAPHY AND INTERVENTIONS

In response to the uncertain prognostic importance of the proposed cut-off concentrations in the UDMI, an alternative definition of PMI (MI-4 and -5) was proposed by the Society for Cardiovascular Angiography and Interventions (SCAI) in 2013 [9]. The main incentive of this working group was to introduce a new definition for ‘clinically relevant’ PMI, applicable for use in clinical trials. Interestingly, the SCAI definition identified CK-MB as the biomarker of choice for detection of PMI, based on several studies that designated diagnostic superiority to CK-MB over cTn for detection of PMI. Still, according to the SCAI definition, cTn can be used in the absence of CK-MB, with specific cut-offs. Although the SCAI working group initially questioned the prognosis-influencing validity of biomarker elevations without angiographic complications or imaging evidence, it still proposed a solitary large biomarker concentration to be diagnostic of PMI. Additionally, the advocates of the SCAI definition prefer the use of the local laboratory’s upper limit of normal (ULN), as opposed to the URL, which is based on a non-local reference population. Of note, the SCAI definition does not discriminate between the procedural invasiveness of PCI and CABG, suggesting similar cut-off values for PMI after both procedures, resulting in 1 uniform definition of PMI following PCI and CABG. As such, if CK-MB is used, solitary CK-MB increases of >10 × ULN are indicative of PMI (without supporting evidence), whereas CK-MB elevations of >5 × ULN warrant additional ECG abnormalities (new left bundle branch block, new pathological Q-waves). If CK-MB measurements are unavailable, cTn concentrations can be determined using a 7:1 cTn:CK-MB ratio (resulting in a solitary cut-off of cTn >70 × ULN or >35 × ULN in the presence of ECG abnormalities). The similarities and differences between UDMI-4 and the SCAI definition are summarized in Table 1.

SCIENTIFIC BASIS OF THE DEFINITIONS

In contrast to the extensive literature supporting the definition of primary MI (non-procedure-related MI, spontaneous MI, also known as MI type I/II), the scientific basis of the PMI definitions is less firm. For MI-1, specific biomarker (cTn) cut-offs have been studied extensively and proved to be related to survival and MI during long-term follow-up [10]. This adverse prognosis reflects the impact of immediate myocardial damage, as well as the consequence of potential plaque rupture in a later setting [9]. However, because PCI and CABG have the potential of inducing some (non-significant) myocardial injury, formulation of specific cut-offs for procedure-related MI is more challenging. Following an elegant meta-analysis by Domanski et al. [11] that demonstrated that even minor increases in biomarkers (CK-MB and cTn) are related to 30-day and 1-year survival after CABG, the proposed cut-off value for cTn of 10 × URL was even more solidly established in the UDMI. Although long-term survival might be the most important outcome in clinical research, it does not necessarily reflect the rate of PMI in this setting. Importantly, these minor biomarker elevations could also be the expression of a more pronounced cardiovascular risk profile, because (baseline) cTn values are known to be increased in these patients. As such, this risk profile, which negatively influences prognosis, could be a potential confounder of long-term survival.

THE GOLD STANDARD FOR PERIPROCEDURAL MYOCARDIAL INFARCTION

In a definition in which biomarkers are the cornerstone of a diagnosis, it is difficult to assess and improve the accuracy of the diagnostic compared to a gold standard. Especially in the perioperative phase, ECG and echocardiographic findings have limited significance in detection of PMI. Furthermore, chest pain is the rule rather than the exception following cardiac surgery and therefore unreliable as well. For (quantification of) myocardial necrosis, cardiac magnetic resonance can be considered a contemporary gold standard, possibly enabling the evaluation of biomarker cut-offs for accurate detection of significant PMI. An important limitation of this modality is the required absence of arrhythmia and the necessary fixed position of patients, which can be cumbersome in the immediate postoperative period. Still, some studies have reported the results of such an approach, describing the incidence of new myocardial necrosis to vary between 20% and 78%, with a median infarct size of 2.5% of the left ventricle [12]. In these studies, cTn proved to be of superior diagnostic accuracy compared to other markers such as CK-MB. Whether this limited amount of myocardial necrosis, compared to the benefit of revascularization, negatively influences prognosis, remains questionable.

DEFINING PERIPROCEDURAL MYOCARDIAL INFARCTION IN CONTEMPORARY CLINICAL TRIALS

PCIs are well established and are the preferred revascularization strategy in ongoing acute coronary syndromes. Additionally, they are performed in patients with more stable coronary artery disease, especially in the case of one- and two-vessel disease. Several studies and registries have also suggested a potential benefit of PCI over CABG for treatment of three-vessel or left-main disease, which warranted confirmation in randomized trials with long-term follow-up. Subsequently, various trials have addressed these clinical questions, most notably the Synergy between Percutaneous Coronary Intervention with TAXUS and Cardiac Surgery (SYNTAX) trial [13], the Evaluation of XIENCE versus Coronary Artery Bypass Surgery for Effectiveness of Left Main Revascularization (EXCEL) trial [14] and the Nordic-Baltic-British Left Main Revascularization (NOBLE) study [15]. Although patient characteristics, coronary anatomy, study design and eventual short- and long-term outcomes differed on the various time points within and between studies, the study findings have led to the incorporation of a class I indication (level of evidence A) for PCI and CABG for treatment of three-vessel and left-main disease (in case of low or intermediate SYNTAX scores) [16]. Interestingly, these trials handled PMI differently.

The first striking difference can be found in the importance of PMI as a complication and its incorporation in the study’s end points. The EXCEL and SYNTAX trials considered PMI as a relevant perioperative complication, which was also included in the primary outcome measure (EXCEL: composite of death, stroke and MI; SYNTAX: death, stroke, MI and repeat revascularization). In contrast, in the NOBLE trial, PMI was registered [no difference between CABG (7%) and PCI (5%; P = 0.53)], but not incorporated in the study’s primary outcome, which included all-cause mortality, non-procedural MI, repeat revascularization and stroke. Of note, the Academic Research Consortium-2 consensus document, which serves as a guideline for cardiovascular clinical trial design, supports the inclusion of PMI as a complication (using a combination of the UDMI and SCAI definitions) [17]. Still, the prognostic relevance of PMI is not well established because several studies failed to demonstrate a relation between the rate of PMI and survival [18, 19].

Secondly, different definitions of (P)MI were applied in these studies. The SYNTAX trial applied its own protocol definition based on absolute and relative CK(-MB) values and ECG criteria, whereas the EXCEL trial based its definition of PMI mainly on the SCAI criteria; the NOBLE trial applied the UDMI. Moreover, NOBLE only deemed patients with normal baseline biomarker concentrations assessable for PMI; in contrast, EXCEL only required CK-MB to be returned to normal levels, regardless of cTn concentrations (potentially affecting the postoperative concentration).

INFLUENCE OF DEFINITIONS ON PERIPROCEDURAL MYOCARDIAL INFARCTION AS A COMPLICATION

In the aftermath of the publication of the 5-year results of the EXCEL trial, controversies arose in the cardiovascular community, especially with regard to the applied definition of MI and the inclusion of PMI in the primary composite outcome. Already in 2017, an interesting retrospective study by Cho et al. [20] demonstrated substantial differences in incidence and clinical relevance of PMI according to the various definitions in 7697 patients undergoing PCI and CABG for whom serial biomarker measurements were available. Rates of PMI after CABG ranged from 2.9% to 18.3%, depending on the definition, and 3.2% to 18.7% for PCI. A worrying finding was the apparent predilection of the definitions for 1 of the procedures while disadvantaging the other.

Since the publication of the results of the EXCEL trial and the accompanying controversy, 2 post hoc analyses have been published that aimed to evaluate the influence of various definitions on the study outcome. The Extended Survival SYNTAX trial results were re-evaluated by Hara et al. [21]; they found the SYNTAX protocol definition and the UDMI-4 to be more strongly associated with long-term survival than the EXCEL/SCAI definition. Of note, rates of PMI differed significantly between the various definitions applied to the Extended Survival SYNTAX population. At the same time, Gregson et al. [22] also presented a re-appraisal of the EXCEL results (3- and 5-year), comparing the protocol definition to the UDMI-3. Although the authors stand by the protocol definition, because it was associated with a similar 5-year hazard after PCI and CABG, a notable difference between the PMI rates was found when UDMI-3 was applied (decreasing the PMI rate in patients who had CABG from 6.1% to 2.2%, an absolute decrease of 36 MIs) [23]. Figure 1 shows the influence of the definition of (P)MI used on the diagnosis of MI as a complication in post hoc analyses of both trials.

INFLUENCE OF PERIPROCEDURAL MYOCARDIAL INFARCTION DEFINITIONS ON COMPOSITE END POINTS

Improvements in the treatment of cardiovascular disease have led to a decline in morbidity and mortality, resulting in a reduction of clinically relevant end points in trials. In response to this trend, composite end points have been proposed to increase the event rate and lower the required sample size, facilitating trial feasibility in terms of costs and follow-up duration. An important disadvantage of such an approach is the challenging interpretation of its results because the end point can be composed of outcomes that vary in their clinical meaningfulness. In contemporary revascularization trials comparing PCI and CABG, mortality (which is an unambiguous dead or alive) is included in the composite outcome, whereas stroke is incorporated in this end point as well and valued equally. Still, the severity of stroke can vary immensely among patients, ranging from temporary complaints to life-long disability. Moreover, an even more disputable component of this composite end point is the MI, which also includes PMI in the EXCEL and SYNTAX trials. The question is whether it is justified to value PMI in the same manner as death, when the influence of PMI on prognosis is debatable and its diagnosis is so definition-dependent.

The EXCEL trial found no difference between PCI and CABG after 5 years in terms of the composite outcome [PCI: 203 events; CABG: 176 events, odds ratio 1.18 (0.94–1.48)] and concluded that both PCI and CABG are feasible and safe for patients with left-main disease with low or intermediate anatomical complexity (a recommendation that is also incorporated in recent guidelines). Still, an absolute decrease of 36 negative events (as described above [22]) in the control group would hypothetically lead to a significant difference between the 2 groups for the composed primary outcome, in favour of CABG (Table 2, simple 2 × 2 contingency table, authors’ own calculations). Of note, this table does not suggest that the UDMI is the justified definition of PMI. However, it illustrates the definition dilemma, given that the mere adoption of a different definition of PMI apparently leads to an important change in the interpretation of the trial’s results, namely, from ‘no difference’ to ‘inferiority’ for PCI after 5 years, potentially even influencing guideline making.

FUTURE DIRECTIONS

In conclusion, based on the differences in definitions of PMI, the variety in biomarkers used and the corresponding cut-off concentrations, the incorporation of PMI in the end points of clinical trials and the potential influence of the definitions used on PMI rate and composite outcomes, we suggest the following future directions:

• Although cTn is superior for diagnosis of primary MI, these results might not be extrapolatable to PMI. Future research is warranted to identify the most appropriate biomarker for identification of significant periprocedural myocardial necrosis.

• Because cTnT and cTnI are measured using different assays with differing normal values, these values should be normalized and corrected for.

• With the advent and routine use of high-sensitivity new generation biomarker assays that enable the measurement of physiological concentrations, multiples of the URL/ULN for diagnosis of PMI should be re-evaluated in the definitions because these were based on assays from previous generations.

• Current biomarker cut-off concentrations are based on long-term survival and not necessarily on the actual event of the PMI and are therefore potentially confounded by the patient’s cardiovascular risk profile.

• Increases in the concentrations of solitary biomarkers without supporting evidence in terms of imaging, angiographic or ECG abnormalities seem to be of insufficient value.

• Trials using biomarker-based PMI definitions should report on baseline concentrations and adapt the cut-off value for PMI accordingly.

• Biomarker cut-off concentrations for diagnosis of PMI are rarely weighted against a gold standard. Given that ECG changes are common after CABG and echocardiographic evaluation is operator- and interpreter-dependent, cardiac magnetic resonance imaging might be the preferred standard.

• It seems counterintuitive to capture PMI in 1 definition following 2 procedures that are inherently different in their cardiac invasiveness. It might be more justified to define PMI distinctly following CABG or PCI.

• Because the influence of PMI on prognosis remains questionable and its diagnosis is extremely definition-dependent, the incorporation of PMI into trial composite end points seems unjustified.

Conflict of interest: none declared.

REFERENCES