Complete Revascularisation in ST-elevation MI: Do We Really Buy It? Challenging the Current Approach to Bystander Percutaneous Coronary Intervention

The evidence supporting primary percutaneous coronary intervention (PCI) for culprit lesions in ST-elevation MI (STEMI) patients is unequivocal, as reflected in international guidelines that have supported this practice for many years. In contrast, the management of atheromatous lesions in non-infarct-related arteries (NIRAs) has historically been variable, and this is important because such bystander disease is common. For example, in one pooled analysis of 28,282 STEMI patients across eight clinical trials, 52.8% of patients had obstructive disease in NIRAs, defined as ≥50% diameter stenosis in at least one non-culprit vessel.¹ This prevalence was consistent with the findings of two large real-world registries, KAMIR (51.7%) and the Duke Cardiovascular Databank (53.4%).¹

Traditionally, management of bystander lesions in STEMI patients has included a variety of options, including: optimal medical therapy alone, with revascularisation considered only in those with ongoing symptoms; optimal medical therapy plus objective assessment of ischaemia in the NIRA territory using non-invasive tests as the arbiter for further revascularisation; pressure wire assessment of NIRA with targeted PCI of those with flow limitation; and stenting of NIRA based on appearance. All these options are pursued at the discretion of the supervising cardiologist. This is illustrated by a large-scale analysis from the National Cardiovascular Data Registry (NCDR) CathPCI Registry.² Specifically, between July 2009 and March 2018, 359,879 STEMI admissions with multivessel coronary disease across 1,598 US hospitals were analysed. Overall, 38.5% of these patients received multivessel PCI within 45 days of presentation, of whom 76.2% achieved complete revascularisation (CR).² The use of multivessel PCI varied significantly across institutions, with a median rate of 37.9% (interquartile range 30.0–46.5%).²

However, the evidence base has changed substantially over the past decade with the presentation of nine randomised controlled trials (RCTs) (Table 1 ) that have consistently shown significant outcome benefit for CR of bystander lesions over culprit-only PCI.^3–11 This has resulted in both European and US guidelines supporting the CR strategy in STEMI with a Class 1A level of recommendation.^12,13

However, despite guideline endorsement, the uptake of CR in clinical practice remains inconsistent among interventionalists. There are likely to be various reasons for this, such as challenging bystander anatomy, a lack of enthusiasm for more complex procedures out of hours or in sick STEMI patients and pressure on cath lab time. However, an important factor is that although the headline results of these RCTs consistently provide evidence of clinical superiority of CR versus a culprit-only strategy, closer inspection of the details reveals substantial heterogeneity in trial design, inclusion criteria (especially the definition of significant disease) and composite clinical endpoints.^3–11 This heterogeneity raises important questions about the concept that all suitable bystander lesions should undergo PCI, without any personalisation of this strategy. In particular, the suggestion that lesions of a certain diameter stenosis will benefit from stenting demands some greater insight into the mechanism of benefit. Is CR effective by sealing inflammatory lesions? Or by improving the degree of flow limitation? Regardless, it would be expected that prophylactic stenting would reduce the rate of spontaneous MI, which has actually only been seen in some of the trials (e.g. PRAMI and COMPLETE).^3,7

This review critically appraises the current evidence supporting CR in STEMI. We explore whether enthusiasm for the universal adoption of a CR strategy may be misplaced, as well as whether greater efforts should be made to understand which patients, vessels and lesions actually benefit from prophylactic PCI and, conversely, which do not so that we may, in future, develop a personalised approach to CR. The fact that all the RCTs show some outcome advantage to the CR strategy implies that there is definite benefit for some STEMI patients, but does not move us closer to working out whether this benefit applies at the patient, vessel or lesion level, or whether some patients undergo CR completely unnecessarily.^3–11

Current Guidelines and Their Rationale

The most recent European Society of Cardiology guidelines endorse complete revascularisation as a Class 1A recommendation for stable STEMI patients, specifically either during the index procedure or within 45 days.¹² Further, it is also a Class 1B recommendation that PCI of non-culprit lesions be guided by angiographic severity (i.e. visual estimation).¹² The US guidelines also give a Class 1A recommendation to CR in STEMI, as follows:

“In selected, hemodynamically stable patients with STEMI and multivessel disease, after successful PCI of the infarct-related artery, PCI of significantly stenosed non-infarct related arteries is recommended to reduce the risk of death or MI and improve the angina-related quality of life.”¹³

These international guidelines are based on the consistent superiority of the CR versus culprit-only strategy in STEMI patients in a series of nine RCTs, whose design and outcomes are summarised in Table 1.^3–11 In every case, the RCTs show benefit for CR, and yet, given the discrepant and discordant inclusion and drivers for advantage, there remain some important concerns and uncertainties about this strategy and, in particular, the appropriateness for its universal application.

Some studies have also now reported long-term follow-up, including CvLPRIT (5.6-year follow-up) and DANAMI-3-PRIMULTI (10-year follow-up), and these demonstrate sustained benefit over the long term.^14,15 Despite these data, detailed forensic analysis raises important questions about the strength of the case for the universal application of the CR strategy in STEMI.

Evidence from Randomised Controlled Trials: The Case for Clinical Uncertainty

The RCTs that have been published in this field show extensive variability in inclusion criteria, definition of bystander disease, clinical composite endpoints and drivers for advantage in a CR strategy. Our intention here is to undertake a forensic examination of these differences and, following a comprehensive description of them, to analyse how the flaws in our current data could be addressed by research looking at mechanistic questions.

Patient Selection

First, as is always the case for RCTs, the generalisability of the results to consecutive patients in front-line clinical practice is unclear, given the highly selective nature of the trial inclusion criteria. For example, the PRAMI trial only enrolled 19% of screened patients.³ Furthermore, a secondary publication by the DANAMI-3-PRIMULTI authors, examining the top recruiting site, identified evidence of unreported exclusions and potential sampling bias.¹⁶ Their main findings were that: non-screened patients had similar characteristics and all-cause mortality compared with the excluded patients, from which it could be inferred that unreported screening and exclusion of the target population had occurred; and randomised study participants had a lower mortality rate than eligible but excluded individuals (HR 3.41; 95% CI [2.69–4.32]; p<0.001) and eligible but non-screened individuals (HR 3.37; 95% CI [2.36–4.82]; p<0.001).¹⁶ These findings suggest that trial participants are not representative of the broader real-life STEMI population. It is particularly notable that 44% of the non-screened cohort (131 patients) did not meet any exclusion criteria and were omitted from the screening log for unspecified reasons.¹⁶

The second important consideration regarding patient selection is the substantial variability in the criteria by which the trials defined bystander disease. As indicated in Table 1, some trials relied on purely visual angiographic assessment of the NIRA lesions, but the degree of stenosis varied significantly (e.g. >50% stenosis in PRAMI; >70% stenosis in a single view or >50% in two views in CvLPRIT).^3,4 In other RCTs, a combination of visual estimation of a varying degree of diameter stenosis and functional significance by pressure wire assessment was used to define target bystander lesions (e.g. DANAMI-3-PRIMULTI used angiographic stenosis >90% or >50% with fractional flow reserve [FFR] ≤0.80, whereas COMPLETE used visual stenosis >70%, or 50–69% with FFR ≤0.80).^5,7 Inevitably, this range of definition must mean that some bystander disease would be a candidate for CR in one trial but not in another. This introduces uncertainty about the optimal population on which to focus the CR strategy in real-world clinical practice, and certainly raises a question about the likelihood of universal benefit.

Differences in Trial Primary Outcomes and Variability in Drivers of Positive Outcomes

The composite endpoints for the RCTs are highly variable, as indicated in Table 2. It is notable that the observed benefit was driven by mortality in only one trial, FIRE, although it is surely important that only 35% of the enrolled cohort in that RCT were STEMI patients, with the rest being non-STEMI patients.¹⁰ With the exception of FIRE, mortality is not a driver for the clinical advantage of CR in these trials. In fact, the components of the composite endpoints that drove the difference between the strategies are highly variable across the trials. For example, in the 10-year DANAMI-3-PRIMULTI outcome data that have recently been published, the main driver was revascularisation, with the rate of MI and death being very similar between the groups.¹⁵ In contrast, in COMPLETE, the component that dominated the difference between the groups was MI.⁷

In several trials, the main driver for the significant difference in the composite endpoint was revascularisation, rather than MI or death. The COMPARE-ACUTE trial evaluated FFR-guided revascularisation in STEMI patients with multivessel disease.⁶ At 1 year, the primary outcome occurred in 7.8% of the CR group, compared with 20.5% of the infarct-artery-only group, with revascularisation accounting for most of the observed benefit.⁶ However, only one-third of these revascularisations were performed for acute coronary syndrome (ACS). The remainder were triggered by stable angina (29.1%), silent ischaemia (3.9%) or angiographic findings alone in the absence of symptoms or demonstrable ischaemia (33%).⁶ Overall, therefore, 72.8% of these procedures were classified as non-urgent. This raises an important question: if the need for revascularisation differs between groups over the follow-up period, but the incidence of MI and death does not, can a routine early strategy of CR truly be justified? There is a strong argument that a more efficient approach would be to defer bystander intervention until it is ischaemia-driven. This interpretation of the data seems to be reinforced by the recently presented 10-year follow-up of DANAMI-3-PRIMULTI, in which there was a significantly lower rate of the composite endpoint among the CR patients, but this was due to fewer revascularisations in the CR group, with no numerical difference in MI or death.¹⁵ The obvious advantages of deferring CR are that, without exposing the patients to a higher rate of MI or death, it would save cath lab time and operator stress in performing routine CR at the index primary PCI or soon after, and that the overall number of patients receiving CR would be substantially fewer than via the currently recommended pathway. In addition, this would be achieved without any cost to the patients in terms of hard events like MI or death.

Is the Benefit of Complete Revascularisation Really Large Enough to Justify Applying it as a One-size-fits-all Strategy?

Across the RCTs, a substantial proportion of patients in the culprit-only group experienced no adverse events during follow-up, whereas many patients in the CR group still experienced events, despite having undergone intervention on all angiographically or physiologically significant lesions. For example, in PRAMI, 23% of patients in the culprit-only arm had an adverse event, meaning that 77% remained event-free; in the CR arm, 9% still experienced an event.³ COMPARE-ACUTE reported similar findings: 79.5% of patients in the culprit-only group remained event-free, whereas 7.8% of those in the CR group still experienced an adverse event.⁶ These observations should logically raise questions about the mechanism of benefit of the CR strategy. Specifically, if the concept is that bystander lesions are at risk of precipitating future events, perhaps by virtue of pan-coronary inflammation, and that stenting these lesions can prophylactically interfere with this risk, then would we not expect a much higher proportion of subsequent events to be avoided? Clearly, to test this concept, the inclusion of any revascularisation event that is not driven by ACS in the composite endpoint is flawed.

Given that all nine RCTs show benefit for CR, it would be foolhardy to suggest that this approach is not beneficial, but it is logical to question whether it benefits all patients and all their bystander lesions. Specifically, the variability of the size of the benefit for CR and the driver for that difference allows for the hypothesis that some patients/lesions benefit from prophylactic stenting and that some do not. If we were able to identify more precisely which patients/lesions do obtain outcome benefit from this intervention and which do not, it would be possible to devise a more targeted approach to CR that would yield larger differences between the CR group and the culprit-only group despite less prophylactic PCI. This surely must be the focus for future research in this field.

Beyond a Binary Approach

Given the data presented so far, it seems implausible that all bystander lesions of a certain severity in all STEMI patients require stenting. Given the consistent advantage for CR over culprit-only strategies in current RCTs, despite variability in what drives that benefit, it seems increasingly likely that identifying the specific features of bystander lesions that confer benefit from stenting could pave the way for more selective approaches: targeting only high-risk lesions for PCI while managing low-risk lesions with optimal medical therapy alone. To this end, there are now imaging systems available that show great promise in the identification of the risk of future ischaemic events, and even mortality, in patients with coronary disease. It is, therefore, possible that application of these technologies to patients with STEMI and bystander disease may be able to shed light on which of these patients and their lesions are most likely to cause events and conversely, which lesions are very low risk for this natural history.

Tools for Personalised Risk Stratification of Future Acute Coronary Syndrome Events

CT Coronary Angiography-based Predictive Tools

It has become widely recognised that there are high-risk plaque features associated with future ACS events. These adverse plaque characteristics (APCs) on CT coronary angiography (CTCA) include positive remodelling, low-attenuation plaque, napkin-ring sign and spotty calcification.¹⁷ These APCs are associated with established intravascular ultrasound and histological features of advanced atherosclerotic plaque and thin-cap fibroatheroma, both of which are risk factors for future ACS.¹⁷ Based on a wide range of evidence from studies, including ROMICAT II, SCOT-HEART and PROMISE, lesions can be labelled as high-risk plaque by the CAD-RADS 2.0 system when two or more features are present.^18–20 This stratification tool facilitates the application of aggressive primary prevention strategies in stable patients and the diagnosis and management of ACS patients.¹⁸ The true role of CTCA-derived high-risk plaque labelling is probably still unresolved.

Predictive Models Based on CT-derived Fractional Flow Reserve

FFR derived from CT (FFR_CT; HeartFlow) is an imaging tool that uses raw data from standard CTCA plus other clinical data to model coronary blood flow using computational fluid dynamics within an anatomical construction of the coronary anatomy and myocardial mass. This allows for modelling of FFR in all the major coronary vessels, providing an estimate of flow limitation down these vessels in a lesion-specific manner.²¹

FFR_CT has been progressively validated with reference to invasive coronary angiography and intracoronary pressure wire-derived FFR in the DISCOVER-FLOW, DeFACTO and NXT studies.^22–24 The diagnostic performance of FFR_CT is robust, with a pooled sensitivity of 90% and specificity of 72% relative to invasive FFR in one meta-analysis that included 18 studies (1,535 patients).²⁵ The clinical value of FFR_CT has now been tested in a series of studies (including PLATFORM, ADVANCE, FISH&CHIPS) and two randomised trials (FORECAST and PRECISE), as summarised in Table 3.^26–31 These studies have presented some consistent outcomes for the use of FFR_CT in clinical practice that can be summarised as follows: the use of FFR_CT is associated with a lower rate of other tests, including invasive coronary angiography, as well as a lower rate of invasive coronary angiography that demonstrates no significant atheroma, is cost neutral overall (or cost saving in some subgroups) and this is all achieved with no excess in adverse events, such as MI or death.

There are limited data available that use FFR_CT in the context of STEMI. In XPECT-MI, baseline FFR_CT was performed in 38 STEMI patients within 10 days of the index event, and intracoronary FFR was measured in NIRAs at baseline and at follow-up (45–60 days after STEMI).³² FFR_CT showed a closer correlation with the follow-up (versus baseline) invasive FFR (sensitivity 100.0%, specificity 90.0%).³² These data imply that FFR_CT soon after an index STEMI may be a better assessment of the functional significance of NIRA lesions than early invasive FFR.

Artificial Intelligence-enabled Quantitative Coronary Plaque and Haemodynamic Analysis

Invasive intracoronary imaging studies, most notably PROSPECT I and PROSPECT II, have demonstrated that vessel- and lesion-specific characteristics predict future coronary events.^33,34 In these studies, a plaque burden ≥70%, minimum luminal area ≤4 mm² and the presence of thin-cap fibroatheroma were independent predictors of subsequent events.^33,34 These invasive findings provide a foundation for contemporary non-invasive approaches, such as artificial intelligence (AI)-enabled quantitative plaque and haemodynamic analysis, which seek to identify similar high-risk features without the need for intracoronary imaging. The original EMERALD study retrospectively assessed the CTCA characteristics of 72 patients who were admitted with an ACS event and who had undergone CTCA up to 2 years before that admission.³⁵ The study compared APCs and newly defined adverse haemodynamic characteristics (AHCs; including FFR_CT, change in FFR_CT across the lesion (ΔFFR_CT), wall shear stress and axial plaque stress) in both culprit and non-culprit lesions to determine whether there was a dose–response relationship between these factors and high-risk lesions. Not only did culprit lesions have a greater degree of stenosis and longer lesion length, but they also exhibited more APCs and AHCs.³⁵

Subsequently, and based on the findings of EMERALD, the EMERALD II study enrolled 351 patients with ACS who had also had a recent CTCA and used a newly designed AI model to further refine the risk stratification of lesions according to plaque characteristics, including APC and AHC and other clinical data.³⁶ The best-performing AI-enabled quantitative coronary plaque and haemodynamic analysis (AI-QCPHA) features identified in the derivation cohort were ΔFFR_CT, total plaque volume, low-attenuation plaque volume and averaged percentage total myocardial blood flow.³⁶ ΔFFR_CT was the single most impactful predictor. The area under the receiver operator characteristic curve increased from 0.78 using CTCA characteristics alone to 0.84 with the addition of AI-QCPHA features, and, again, this was associated with the number of characteristics in the model.

EMERALD II provides a persuasive proof of concept for future research strategies that aim to identify lesion-specific risk based on plaque characteristics. In the context of STEMI, where non-culprit lesions are often judged solely based on their visual appearance, such AI-guided assessment could potentially shift CR from a largely empirical approach to a more risk-guided, precision strategy.

Coronary Artery-specific Inflammation as a Biomarker for Risk: The Fat Attenuation Index

The development of coronary atheroma is the product of a complex vascular inflammatory pathological pathway.³⁷ Epicardial adipose tissue around coronary arteries is involved in a dynamic interaction with the underlying vessel by releasing proinflammatory adipocytokines that influence the adjacent vascular wall. High levels of vascular inflammation have been shown to be associated with suppressed differentiation of perivascular fat cells and reduced lipid accumulation; these factors modify the composition of perivascular adipose tissue. The result is a higher water content and lower lipid content in perivascular adipose tissue; this can be detected on CT imaging and has recently been described as the perivascular fat attenuation index (FAI). FAI represents a novel CTCA-derived biomarker that can quantify metabolic changes in perivascular adipose tissue, hence representing a surrogate for coronary inflammation.³⁷

The initial promising validation experiment demonstrated that FAI values were significantly lower in individuals without coronary artery disease than in those with atherosclerosis, and this was independent of conventional risk factors as well as plaque burden and calcification.³⁸ Follow-on experiments demonstrated an association between higher FAI and plaque with high metabolic activity and inflammation, as indicated by the uptake of ¹⁸F-sodium fluoride on PET imaging, and intravascular ultrasound features of high-risk plaque.^39,40 An association between FAI and significant vessel flow limitation as assessed by FFR has also been demonstrated.⁴¹

Based on the extensive validation data, FAI was then tested in clinical studies as a potential imaging biomarker for cardiovascular risk. In the CRISP-CT study, FAI was a strong and independent predictor of both cardiac and all-cause mortality in two large cohorts of patients (derivation and validation) undergoing CTCA.⁴² In both populations, high FAI values in all three coronary vessels were significantly associated with the risk of all-cause mortality. Specifically, in the derivation cohort (n=1,872), a high perivascular FAI around the right coronary artery (RCA) was independently associated at a median of 6 years with a significantly higher risk of both all-cause mortality (adjusted HR 2.55; 95% CI [1.65–3.92]; p<0.0001) and cardiac mortality (adjusted HR 9.04; 95% CI [3.35–24.40]; p<0.0001).⁴² The validation cohort included 2,040 patients and a median follow-up of 4.5 years, and reported similar associations between FAI and both all-cause mortality (HR 3.69; 95% CI [2.26–6.02]; p<0.0001) and cardiac mortality (HR 5.62; 95% CI [2.90–10.88]; p<0.0001).⁴² However, the association with cardiac mortality was statistically significant only for the RCA and left anterior descending artery, and not for the left circumflex artery. The FAI was restricted to the RCA, which was used as the reference vessel for subsequent analysis.⁴² In both populations, FAI improved the area under the curve for prognostic discrimination. This RCA FAI represents a potentially powerful tool with which to risk stratify for mortality.⁴²

Subsequently, the ORFAN study has added to the evidence that FAI may have a future role in screening for cardiovascular mortality.⁴³ ORFAN was a multicentre UK cohort study that evaluated the prognostic value of FAI, using an AI-enabled, CTCA-derived risk stratification tool in patients undergoing CTCA. An initial observational cohort (Cohort A) included 40,091 patients and was used to quantify major adverse cardiovascular events (MACE), including cardiac death at a median of 2.7 years. In ORFAN, although 81.1% of patients (n=32,533) had no obstructive coronary artery disease, they nevertheless accounted for 63.7% of cardiac deaths (1,118/1,754) and 66.3% of total MACE (2,857 of 4,307).⁴³ This illustrates the potential flaw associated with assessment of risk based solely on anatomical parameters such as atheroma burden.

In a nested cohort of 3,393 patients, elevated FAI scores across the three major coronary arteries (left anterior descending artery, left circumflex artery, RCA) independently predicted cardiac death and MACE at a median of 7.7 years, even among patients with no or minimal plaque.⁴³ Specifically, the presence of one or more ‘inflamed’ artery(ies), defined as FAI >75th percentile, carried a nearly 30-fold increased risk of cardiac mortality and a 13-fold higher risk of MACE compared with low FAI vessels. Furthermore, the availability of AI-risk FAI resulted in a change in management in 45% of 744 real-world patients, manifested as statin initiation, dose escalation or the addition of anti-inflammatory therapies.

These properties make both AI-QCPHA and AI-risk FAI attractive tools with which to risk assess STEMI patients and vessels in those with disease in NIRA.

Potential Future Application of AI-driven Biomarkers for Identification of High-risk Bystander Disease in Patients with STEMI

The all-or-nothing approach to CR used in previous RCTs has clearly demonstrated benefit. However, the current guideline-directed strategy for universal CR places a significant burden on cath lab time, resources and operator workload, without equipping interventionists with the tools to identify which patients or lesions stand to benefit most. Emerging technologies that enable detailed imaging and risk profiling may allow us to personalise bystander revascularisation. After all, how do we know that bystander revascularisation must be complete in every individual? A more selective patient- or lesion-specific approach could ultimately achieve equivalent or even superior clinical outcomes with less intervention.

The PICNIC study (NCT06506448) is investigating whether we can identify markers for patient-, vessel- and lesion-level risk in a STEMI population with bystander coronary disease after culprit primary PCI. During the index admission, blood tests will be taken for prespecified inflammatory markers and a CTCA will be performed, which will be used to apply AI-QCPHA and FAI analyses. Patients will be followed for 24 months for a composite endpoint (all-cause mortality, cardiac arrest, ACS, additional revascularisation by coronary artery bypass grafting or PCI, rehospitalisation for angina, heart failure, stroke and ventricular tachyarrhythmia at 24 months) and an association sought between clinical and AI-CTCA parameters and clinical events. We hypothesise that the risk associated with non-culprit lesions is highly variable and that an AI model incorporating clinical and AI-QCPHA and FAI parameters will be able to risk stratify patients and lesions. The concept of AI-guided CTCA for all STEMI patients with multivessel disease is appealing but may inevitably raise logistical and capacity challenges for cardiac CT services. However, PICNIC is designed to establish this proof of concept, which, if successful, may lay the foundation for a new generation of trials focused on precision targeting of bystander revascularisation.