Objective Validation of methods to assess the area at risk (AAR) in patients with ST elevation myocardial infarction is limited. A study was undertaken to test different AAR methods using established physiological concepts to provide a reference standard.
Main outcome measured In 78 reperfused patients with first ST elevation myocardial infarction, AAR was measured by electrocardiographic (Aldrich), angiographic (Bypass Angioplasty Revascularization Investigation (BARI), APPROACH) and cardiovascular magnetic resonance methods (T2-weighted hyperintensity and delayed enhanced endocardial surface area (ESA)). The following established physiological concepts were used to evaluate the AAR methods: (1) AAR size is always ≥ infarct size (IS); (2) in transmural infarcts AAR size=IS; (3) correlation between AAR size and IS increases as infarct transmurality increases; and (4) myocardial salvage ((AAR-IS)/AAR×100) is inversely related to infarct transmurality.
Results Overall, 65%, 87%, 76%, 87% and 97% of patients using the Aldrich, BARI, APPROACH, T2-weighted hyperintensity and ESA methods obeyed the concept that AAR size is ≥IS. In patients with transmural infarcts (n=22), Bland–Altman analysis showed poor agreement (wide 95% limits of agreement) between AAR size and IS for the BARI, Aldrich and APPROACH methods (95% CI −22.9 to 29.6, 95% CI −28.3 to 21.3 and 95% CI −16.9 to 20.0, respectively) and better agreement for T2-weighted hyperintensity and ESA (95% CI −6.9 to 16.6 and 95% CI −4.3 to 18.0, respectively). Increasing correlation between AAR size and IS with increasing infarct transmurality was observed for the APPROACH, T2-weighted hyperintensity and ESA methods, with ESA having the highest correlation (r=0.93, p<0.001). The percentage of patients within a narrow margin (±30%) of the inverse line of identity between salvage extent and infarct transmurality was 56%, 76%, 65%, 77% and 92% for the Aldrich, BARI, APPROACH, T2-weighted hyperintensity and ESA methods, respectively, where higher percentages represent better concordance with the concept that the extent of salvage should be inversely related to infarct transmurality.
Conclusions For measuring AAR, cardiovascular magnetic resonance methods are better than angiographic methods, which are better than electrocardiographic methods. Overall, ESA performed best for measuring AAR in vivo.
- Myocardial infarction
- coronary artery disease
- CT scanning
- cardiac remodelling
- cardiac function
- imaging and diagnostics
- stable angina
- 12 lead ECG
- myocardial viability
Statistics from Altmetric.com
- Myocardial infarction
- coronary artery disease
- CT scanning
- cardiac remodelling
- cardiac function
- imaging and diagnostics
- stable angina
- 12 lead ECG
- myocardial viability
Clinical outcome after acute myocardial infarction (MI) largely depends on final infarct size (IS) and is improved by early reperfusion.1 IS is determined by the extent of collateral flow, myocardial metabolic demand during occlusion, duration of coronary occlusion and size of the initial area at risk (AAR).2 3 Determining both AAR and IS allows the assessment of myocardial salvage and evaluation of the efficacy of reperfusion strategies.
Delayed enhanced cardiovascular magnetic resonance imaging (DE-CMR) reliably detects IS and accurately distinguishes between viable and non-viable myocardium.4 However, determination of the AAR in the clinical setting remains challenging. Although single-photon emission computed tomography (SPECT) is an accepted technique,5 6 radiation exposure, low spatial resolution and difficult logistics requiring tracer injection before reperfusion complicate this method.
Diagnosis of ST elevation myocardial infarction (STEMI) is supported by the ECG that is routinely performed on admission and most patients will undergo coronary angiography followed by percutaneous coronary intervention (PCI). Both electrocardiographic and coronary angiographic measures have previously been used to measure the AAR. The Aldrich score is based on the total amount of ST segment deviation on admission,7 whereas the Bypass Angioplasty Revascularization Investigation (BARI) and APPROACH methods are based on vessel anatomy and location of coronary obstruction.8 9 Myocardial ischaemia leads to inflammation and tissue oedema in the infarct-related artery (IRA) territory.10 This oedema is shown as an area of increased signal intensity (SI) on T2-weighted cardiovascular magnetic resonance (T2w-CMR), and animal studies suggest that this hyperintense area delineates the AAR in STEMI.11 12 More recently, endocardial surface area (ESA) on DE-CMR has been used to measure AAR.13 14 Although promising, for all the proposed methods the number of studies are small and more validation in humans with appropriate reference standards is needed.15 16
In this study electrocardiographic, angiographic and CMR techniques for measuring AAR were evaluated using established and straightforward physiological concepts based on the ‘wavefront phenomenon’ of infarct progression from endocardium to epicardium with only minimal lateral extension.2
Between August 2006 and March 2008, 78 consecutive patients with a first STEMI referred to our institution for primary PCI and presentation <12 h after symptom onset were prospectively studied. The definition of MI was based on the recent consensus document including appropriate rise and fall in cardiac biomarkers.17 Patients with claustrophobia, contraindications to CMR (cerebral clips, pacemakers, defibrillators, neurostimulators), severe congestive heart failure or cardiogenic shock (Killip class III and IV) at the time of CMR, atrial fibrillation, age <18 years, severe renal failure (stage 4 or 5) and pregnancy were excluded.
After pretreatment with aspirin, heparin and clopidogrel, patients underwent standard catheterisation with multiple selective contrast injections in the left and right coronary arteries prior to PCI. Thrombosuction and downstream administration of intravenous antiplatelet agents (abciximab), intracoronary nitroglycerin and adenosine were left to the discretion of the interventional cardiologist. PCI was performed in all patients and every patient received at least one stent.
Two observers, blinded to the electrocardiographic and CMR data, independently reviewed all angiograms. Discrepancies were solved by discussion. Collaterals were defined by visual assessment using the Rentrop classification.18 Antegrade flow in the IRA before and after PCI was characterised using the TIMI scoring system.19
Angiographic AAR assessment
Angiographic AAR measurement for the IRA was performed using the following methods.
AAR was calculated by grading all terminating arteries.8 All branches were scored 3, 2, 1 or 0 points, corresponding to large, medium, small or absent. The ventricular base to apex distance, approximated from the coronary angiogram, provided the basis for assessing the relative distribution of coronary branches. Branches were considered large if their length exceeded two-thirds of the distance from base to apex, medium if one-third to two-thirds the distance, small if less than one-third of the distance and absent if less than one-fifth of the distance from base to apex. The AAR was calculated as a percentage of the left ventricle (LV) by dividing summed scores of a jeopardised area by the total score of the entire LV.
Modified APPROACH score
AAR was calculated by dividing the LV into regions that represent the proportion of myocardium perfused by each coronary artery based on post-mortem pathology studies.20 In this study the modified APPROACH score was used.14 Scoring was performed using a template in which the culprit artery is selected, as well as proximity of the lesion and size of side branches (small/absent, medium or large). The template then provided one of 39 possible AAR scores as a percentage of the LV.
Electrocardiographic AAR assessment (Aldrich method)
Two observers, blinded to the coronary angiography and CMR data, independently determined the Aldrich scores on the standard 12-lead admission ECG.7 Using the TP segment as the isoelectric line, ST elevation was measured manually, 60 ms after the J-point to the nearest 0.5 mm in each lead. Any ST elevation ≥1 mm was measured in leads II, III, avF for inferior STEMI and in all leads for anterior STEMI. Electrocardiographic AAR was then calculated using the Aldrich score (anterior MI: 3×(1.5(number of leads with ST↑) − 0.4); inferior MI: 3×(0.6(∑ST↑ II, III, AVF) + 2.0)) and computed as a percentage of LV.
CMR was performed 5±2 days after admission. Images were acquired during multiple breath holds on a 1.5 T MRI system (Intera R11.3, Philips Medical Systems, Best, The Netherlands) with a dedicated five-element phased array surface coil. To image myocardial oedema, a double inversion T2-weighted turbo spin echo sequence with spectrally selective inversion recovery (SPIR) fat suppression was used in multiple contiguous short axis slices covering the entire LV (TR/TE: 2 R–R intervals/100 ms, flip angle 90°, FOV 350 mm, matrix 236×186, voxel size 1.48×1.89×8 mm). Delayed enhancement (DE) was performed 10–15 min after administration of 0.2 mmol/kg body weight gadolinium-diethylenetriaminepentaacetic acid (Magnevist, Bayer Schering Pharma, Berlin, Germany) using a 3D segmented inversion recovery gradient echo sequence (average TR/TE 3.9/2.4 ms, flip angle 15°, FOV 400 mm, matrix 256×256, reconstructed voxel size 1.56×1.56×6 mm). A Look–Locker sequence was used to find the inversion time that optimally suppressed the signal of non-infarcted myocardium (typical range 200–280 ms).
CMR image analysis
CMR images were analysed offline using commercially available software (CAAS MRV 3.0, Pie Medical Imaging, Maastricht, The Netherlands) and blinded to the electrocardiography and angiography results. Endocardial and epicardial contours were manually traced on the T2 SPIR and DE images.
The AAR was quantified on the T2 SPIR images by delineating myocardium with a SI threshold >2SD above the mean of a remote region outside the IRA territory.16 Areas of central hypoenhancement were included and endocardial ‘slow flow’ artifacts were excluded by manual adjustment of contours.
IS was quantified on the DE images by delineating areas of hyperenhancement using a SI threshold of >5SD above the mean of a remote non-infarcted region, including areas of microvascular obstruction (MVO).21
AAR and IS were expressed as a percentage of LV mass. The transmural extent of infarction was calculated by dividing the hyperenhanced volume by the total volume of the infarcted segments. The infarct ESA was calculated as a percentage of the hyperenhanced ESA divided by total LV ESA (including MVO).
For each method, the data of 25 randomly selected patients were analysed twice by one observer on two different occasions and a second observer independently analysed all data to calculate intra- and interobserver variability.
Patients were classified into three groups: (1) subendocardial infarction (transmurality <50%, n=13); (2) near transmural infarction (transmurality >75%, n=22); and (3) intermediate transmurality (transmurality 50–75%, n=43). Each method was evaluated by using the following established physiological concepts: (1) AAR should always be larger than or equal to IS; (2) in nearly transmural infarcts, AAR should approach IS; (3) as infarct transmurality increases, the correlation between AAR and IS should increase; (4) myocardial salvage (defined as AAR-IS)/AAR ×100) should be inversely correlated with transmurality, closely following the inverse line of identity (salvage =100% − transmurality). The deviation from this ‘line of identity’ was expressed as the percentage of patients within a ±30% margin and as the SE of estimate (SEE). Individual AAR methods were ranked from 1 (worst) to 5 (best) based on how well they agreed to each of these physiological concepts.
Continuous variables were presented as mean±SDs and proportions (%) were used for categorical variables. Intra- and interobserver variability were assessed by calculating intra- or interclass correlation coefficients (ICC). In near transmural infarcts, agreement between AAR and IS was tested using Bland–Altman analysis, with reporting of bias and 95% CIs. Correlations were computed using the Pearson correlation coefficient. The predictive accuracy of the AAR methods concerning the inverse relation between myocardial salvage and infarct transmurality was measured using the SEE. A two-tailed p value ≤0.05 was considered statistically significant. The data were analysed using SPSS V.17.0 for Windows (SPSS Inc).
Baseline characteristics are shown in table 1. The majority of patients were men (71%), approximately half of the patients had single vessel disease (51%) and in 53% the right coronary artery (RCA) was the IRA. Final TIMI 3 flow was established in 91%. Mean IS was 16±11% of LV mass. A trend in decreasing pre-PCI TIMI flow was observed with increasing transmurality (p=0.07). The mean Rentrop scores and symptom to balloon times in patients with <50%, 50–75% and >75% infarct transmurality were 0.46, 0.60 and 0.59 (p=0.90) and 251±82, 204±75 and 228±94 min, respectively (p=0.17).
Intra- and interobserver variability
The ICC for all methods was moderate to excellent (for Aldrich 0.95 and 0.75, for BARI 0.89 and 0.86, for APPROACH 0.89 and 0.79, for T2w-CMR 0.92 and 0.96, for IS by DE 0.91 and 0.96, for ESA by DE 0.90 and 0.87, respectively).
Table 2 shows that, overall, 51 (65%), 68 (87%), 59 (76%), 68 (87%) and 76 (97%) of patients using the Aldrich, BARI, APPROACH, T2w and ESA methods, respectively, had an AAR that was ≥IS. Thus, ESA performed best and Aldrich worst. T2w and BARI performed equally well and were given a similar score of 3 (table 4). In the subgroup of subendocardial infarcts, AAR was ≥IS in all patients, irrespective of the method used. The results of the other subgroups are also shown in table 2.
AAR approximation of IS in near transmural infarcts
Bland–Altman analysis was used to assess the agreement between AAR and IS in a subcohort of 22 patients with near transmural infarcts (figure 1). Because near transmurality was defined as >75% infarction and not 100% (mean transmurality was 82±4%), a small positive bias was expected. This was true for all methods except the Aldrich method (bias −3.5), which was thus considered the worst technique. The 95% limits of agreement from worst to best were −22.9 to 29.6 (Δ52.6), −28.3 to 21.3 (Δ49.6), −16.9 to 20.0 (Δ36.8), −6.9 to 16.6 (Δ23.6) and −4.3 to 18.0 (Δ22.4) for the BARI, Aldrich, APPROACH, T2w and ESA methods, respectively. Accordingly, ESA received the highest score, followed by T2w, APPROACH, BARI and Aldrich (table 4).
Correlation of AAR and IS with respect to transmurality
Figure 2 shows the correlation of AAR with IS, separated by transmurality subgroups. Physiology dictates that this correlation should improve with increasing infarct transmurality. For the Aldrich method, a reverse pattern was observed. For the BARI method, although the correlation went up from subendocardial to intermediate transmurality infarcts, it went back down in near transmural infarcts. For the remaining techniques, the correlations increased with increasing amount of transmurality. The correlation in near transmural infarcts was highest for ESA (r=0.93, p≤0.001). Accordingly, ESA was given the highest relative score, followed by T2w, APPROACH, BARI and Aldrich (table 4).
Inverse relationship of myocardial salvage with transmurality
The relationship of myocardial salvage with infarct transmurality is expected to be inverse (ie, myocardial salvage is large in subendocardial infarcts and minimal in near transmural infarcts) and, ideally, observations should be close to the inverse ‘line of identity’ (defined as 100% minus transmurality). The percentage of patients within a ±30% margin around the inverse ‘line of identity’ (high = better) were 56%, 76%, 65%, 77% and 92% for the Aldrich, BARI, APPROACH, T2w and ESA methods, respectively (table 3) and the SEE values were 64.7, 44.8, 33.1, 26.6 and 21.7 for the Aldrich, BARI, APPROACH, T2w and ESA methods, respectively (figure 3). Based on the latter, the ESA method was given the highest score and the Aldrich method the lowest (table 4).
Table 4 shows the final scores assigned to each AAR method according to how well they agreed with the specified physiological concepts. From worst to best, the methods were as follows: Aldrich, BARI, APPROACH, T2w-CMR and ESA.
The findings of our study can be summarised as follows. Using established physiological concepts to test the performance of different AAR methods in patients with STEMI, CMR methods were better than angiographic methods (BARI and APPROACH), which were better than electrocardiographic methods (Aldrich). Concerning CMR, the ESA method performed consistently better than the hyperintense area on T2w-CMR.
There is a major interest in measuring the AAR and myocardial salvage in STEMI for use as surrogate endpoints in clinical trials evaluating the effectiveness of new treatment strategies in acute MI. Although it is an accepted technique for the clinical assessment of AAR, SPECT has several limitations such as impractical logistics, suboptimal spatial resolution and radiation burden.6 To validate the effectiveness of other more clinically feasible methods we used established physiological concepts of infarct progression that can be accurately measured by DE-CMR as a reference standard.4 In essence, the subendocardial lateral boundaries of the infarct are established within the first 40 min after coronary occlusion and myocardial necrosis progresses in a transmural wavefront over a period of 3–6 h.2
The ST segment deviation score developed by Aldrich was calibrated to predict the final IS estimated by the Selvester QRS score in a population that did not receive reperfusion therapy.7 Similar to our results, the poor performance of the Aldrich score in estimating AAR was previously demonstrated in a study using SPECT.22 This poor performance might be explained by the fact that the ECG only momentarily captures the ST segment, which is known to behave dynamically during the acute phase of STEMI. In addition, the ST segment changes that are prominent during coronary occlusion represent the ischaemic myocardium, which will resolve when the ischaemic myocardium is either infarcted or reperfused. With progressive myocardial necrosis, the ST segment deviation that is initially directed towards the ischaemic region will be replaced by QRS deviation away from the infarcted region.23 The Aldrich score will therefore progressively underestimate the initial AAR when infarction proceeds.
In the angiographic scores BARI and APPROACH, the jeopardised myocardium is based on the territory supplied by all significantly stenotic vessels. Of interest, neither angiographic method has been validated for its use to quantify the AAR in patients with STEMI nor against a proper reference standard (ie, SPECT). Moreover, angiographic methods other than BARI and APPROACH were previously shown to correlate poorly with the AAR assessed by technetium-99m sestamibi.24 Ortiz-Pérez et al found an excellent correlation between both the BARI and APPROACH methods and IS in a subgroup of patients with near transmural infarcts, suggesting indirectly that these techniques might be useful to measure the AAR (r=0.90, p<0.001 for both).14 Despite the fact that our definition for near transmurality was more strict (>75% vs >50% for Ortiz-Pérez et al), our correlations for the BARI and APPROACH methods were only moderate at best (r=0.43 and r=0.67, respectively) and thus we could not confirm their results. In addition, our results showed that up to 50% of patients (for APPROACH) had an AAR that was smaller than IS in the subcohort of near transmural infarcts.
Dark blood T2w-CMR is increasingly being used for measuring the AAR.25 26 Myocardial injury predisposes to oedema formation and the increase in tissue water can be detected by T2w-CMR as an area of hyperintense signal. However, its validation and reproducibility is still limited.11 12 15 Furthermore, many technical issues remain and T2w imaging using conventional dark blood turbo spin-echo techniques is hampered by artifacts such as inferolateral wall signal loss due to cardiac motion and bright subendocardial rims due to stagnant blood. These artifacts generally lead to overestimation of the size of the T2w abnormality.27 Newer methods such as bright blood T2w imaging and T2w mapping may overcome these technical limitations, but these methods have even less validation.28 Furthermore, the SI thresholds used to define abnormal myocardium for both DE-CMR and T2w-CMR are arbitrary, which could also affect IS, AAR and salvage calculations.
To date, the results of studies using dark blood T2w-CMR for measuring the AAR are confusing. Some studies have found that the AAR ≥IS in almost 100% of patients, while we and others have observed this in only 87% and 65% of patients, respectively.29 In a recent study in 38 patients, Berry et al also compared electrocardiographic and angiographic methods with T2w-CMR imaging to assess the AAR.26 They found a good correlation (r=0.78, p<0.0001) and agreement between T2w-CMR and the APPROACH method for measuring the AAR. This finding led to their conclusion that T2w-CMR is a good measure for the AAR. In their study they only used one physiological concept to check for internal consistency, which was that AAR ≥IS. For both T2w-CMR and APPROACH, higher percentages of patients with AAR ≥IS were found than in our study (92% and 88% vs 87% and 76%, respectively). They did not test the other complementary physiological concepts as was performed in our study. Based on our findings, it does not seem appropriate to validate T2w-CMR against the angiographic techniques (ie, APPROACH) since (1) we did not find a good relationship between T2w and angiographic measures of AAR and (2) APPROACH performed poorly against our reference standard.
The ESA method is measured on DE-CMR images14 and is based on the fact that the lateral boundaries of the AAR equal the lateral boundaries of infarction after 40 min.2 Variation in the percentage of the AAR that becomes necrotic is primarily caused by variation in the transmural extent of the infarction.20 Of all the methods tested for measuring the AAR, we found that infarct ESA consistently performed best. Using T2w-CMR as the reference standard, the results from previous studies are conflicting. One study found a good correlation between T2w-CMR and ESA (r=0.77, p<0.001)13 while, in another study, this correlation was poor (r=0.41, p=0.002).30 Apparently, this discrepancy can be caused by variation in either the ESA or T2w-CMR measurements. Therefore, instead of a head-to-head comparison, we provided new evidence by comparing both ESA and T2w-CMR against a reference standard.
For the determination of myocardial salvage, the use of different techniques to assess the individual components of AAR and IS may cause an offset problem—in other words, two different techniques may be highly correlated but one may always be consistently underestimating or overestimating the other. The combined use therefore appears problematic, especially in combining electrocardiographic and angiographic measurements with CMR techniques. For instance, if the AAR is smaller than the IS, the measurement of salvage will be useless.
Limitations of the study
Our results are based on a limited number of patients and should therefore be interpreted with care. However, to date, this is a larger study than others to compare different techniques for measuring the AAR against a reference standard. Several limitations of our approach of using physiological criteria as the reference method need to be addressed. First, we only provided a relative ranking of methods, and absolute values regarding the level of precision and accuracy—such as could be obtained with a direct comparison with a true reference standard of the AAR—are not available. Another limitation is the assumption that infarct size and transmurality, which are fundamental components of the physiological criteria, are perfectly measured by DE-CMR. Although this is an established method to assess infarct size and transmurality, image artifacts or reduced image quality on DE-CMR could have introduced variability in the application of the physiological criteria. Finally, an important related limitation is the possibility that CMR methods of measuring the AAR (ie, ESA and T2w) may have been favoured compared with non-CMR methods, given that the reference physiological criteria were assessed using CMR. Although they are different measurements, less variation may occur when comparing one CMR measurement with another CMR measurement than with a non-CMR measurement. Although SPECT imaging was not available to provide precise measurements of the AAR, we are confident that our prespecified physiological concepts provided a good reference and that our conclusions are valid. For example, when using the APPROACH method, the AAR was smaller than IS by DE-CMR in 24% of patients overall and the correlation with near transmural infarcts was only moderate, indicating this method is poor. Concerning our CMR protocol, we obtained DE-CMR images 10–15 min after gadolinium administration. Although we achieved good image quality using this protocol, which was reflected in our high inter- and intra-observer variability, a longer delay could have resulted in improved visualisation of subendocardial infarcts. We used specific cut-offs to quantify the AAR and IS by T2w-CMR and DE-CMR (>5SD and >2SD), which was based on the literature. There are certain limitations as we have discussed already with these cut-offs and, although this may affect absolute size measurements, the correlation between methods remains unchanged.
For measuring AAR, CMR methods performed better than angiographic methods, which performed better than electrocardiographic methods. Concerning CMR, ESA performed consistently better than dark blood T2w-CMR. Our results also show that combining electrocardiographic or angiographic methods of measuring AAR with CMR determined infarct size, the determination of salvage may be problematic.
Competing interests None.
Ethics approval The study was approved by the Institutional Review Board of Maastricht University Medical Center and patients gave written informed consent.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.