Diagnostic value of the electrocardiogram in the assessment of prior myocardial infarction

This is a single-centre, retrospective study. All patients who underwent CMR with LGE between January 2014 and December 2017 at the Deventer Hospital were screened for eligibility. Patients were excluded if no ECG was available within 6 months before the CMR or if the electronic medical record (EMR) was not accessible. Finally, a patient was excluded if the outcome of the CMR was inconclusive or if the ECG showed an acute or hyperacute MI.

Data were collected from the EMR by reviewing reports on CMR with LGE, ECGs, letters and reports of clinicians, reports and letters of clinicians in other hospitals, measurements of vital signs, and laboratory results.

CMR was performed using a 1.5 T magnetic resonance imaging (MRI) system (Signa HDxt, GE Healthcare, Milwaukee, WI, USA). Scans were carried out using an eight-channel HD cardiac array coil, following the CMR protocol of the hospital. Ten minutes after the administration of 0.2 mmol/kg intravenous gadolinium, LGE images were made, including the two-chamber, four-chamber and short-axis views.

All CMR with LGE images were analysed in a clinical setting (not for research purposes) by an experienced radiologist and discussed by a multidisciplinary team, consisting of a radiologist and at least one cardiologist. CMR images with inconclusive or unclear descriptions in the EMR were reassessed by an experienced radiologist. Patients were divided in two groups: (1) patients with a pattern of LGE typical for MI and (2) those without an LGE pattern typical for MI. A distinction was made between subendocardial hyperenhancement and transmural hyperenhancement (>50% of the myocardium). Images with a pattern of LGE typical for MI were divided in different segments based on the standardised 17-segment model of the American Heart Association [12]. The different segments were merged into two categories: an anterior and inferior category. The anterior category consisted of the segments anterior, anterolateral, anteroseptal, septal and apical; the inferior category consisted of the segments inferior, inferolateral, inferoseptal and lateral.

Twelve-lead ECGs were extracted from the EMR and blinded before assessment. ECG assessment was performed by two independent cardiologists separately and analysed as to the presence or absence of signs associated with prior MI without knowledge of the results of CMR with LGE. Both cardiologists used the ECG criteria according to the expert consensus document of the European Society of Cardiology [6]. ECGs with signs of a left or right bundle branch block or signs of ischaemia were marked as negative (absence of MI). Inter-observer disagreement was resolved by the judgement of a third cardiologist.

Statistical analysis was performed using SPSS (version 25, IBM SPSS Statistics for Windows). Continuous data are presented as means with standard deviation or as median with first and third quartile if the data had a non-normal distribution. Analyses were performed using the Student’s t-test and Mann-Whitney U test. Categorical data were presented as numbers (n) and percentages (%) and compared with the chi-squared test. The diagnostic value of the ECG (sensitivity, specificity, positive predictive value (PPV) and negative predicted value (NPV)) were calculated and shown in a 2 × 2 table. The relationship between the independent variables age, gender, patient history, hypertension, diabetes mellitus (DM), hypercholesterolaemia, (history of) smoking and outcome of the CMR was determined by multivariate logistic regression. Determinants which affected the regression coefficient of the association between ECG classification and MRI outcome by more than 10% were added to the model. Inter-rater variability between the classification of ECGs by the two cardiologists was analysed using Cohen’s kappa [13]. A p-value <0.05 was considered to be statistically significant.

The Medical Ethical Committee of the Isala Hospital Zwolle in The Netherlands reviewed the research protocol and concluded that the rules laid down in the Medical Research Involving Human Subjects Act did not apply to this research proposal.

Of 1033 patients that underwent CMR with LGE between January 2014 and December 2017, 974 were included in the analysis (Fig. 1), 566 men and 408 women. Their median age was 63.5 years. In 205 patients (21.0%) MI was present on CMR with LGE. In this group, the percentage of women (21.0% vs 47.5%, p < 0.001) was statistically significantly lower and the median age (67.1 vs 62.3 years, p < 0.001) was statistically significantly higher compared to the patient group without MI. An overview of patient characteristics and CMR data is shown in Tab. 1.

In Tab. 2, the results of ECG assessment and those of CMR with LGE as the gold standard with regard to a history of prior MI are shown. Signs of previous MI based on CMR were seen in 205 patients (21.0%). In 179 out of 974 patients (18.4%) the ECG showed electrocardiographic changes that were associated with a prior MI. The ECG correctly showed signs of a prior MI in 38.0% of the patients with MI on CMR. Of the assessed ECGs, 56.4% were false positive (CMR showed no MI) and 16.0% were false negative. In 84.0% of the patients without MI on CMR, no ECG changes compatible with prior MI were found (Tab. 2).

The sensitivity of an ECG in detecting an MI was 38.0% (95% CI 31.6–44.8%). The specificity was 86.9% (95% CI 84.4–89.1%). The PPV and NPV of the ECG were 43.6% (95% CI 36.4–50.9%) and 84.0% (95% CI 81.4–86.5%) respectively (Tab. 3).

After categorisation, the ECG showed a sensitivity of 63.3% (95% CI 50.8–74.8%) for an MI in the anterior category, compared to a sensitivity of 20.2% (95% CI 13.1–28.8%) in the inferior category. If MI was seen in both categories, the ECG showed a sensitivity of 43.5% (95% CI 29.8–57.9%). A comparable PPV of the ECG divided into the anterior and inferior category was found (43.0% (95% CI 34.4–51.9%) vs 42.9% (95% CI 30.0–55.9%) respectively). An increase in size and transmurality showed non-significant improvement of the sensitivity (Fig. 2).

The odds ratio of a positive versus a negative ECG for MI on CMR with LGE was 4.062 (95% CI 2.859–5.771). In multivariate logistic regression, a patient history of CVD appeared to be a confounder in the association between ECG findings and MI on CMR with LGE. After correction for this confounder, the odds ratio of a positive versus a negative ECG was 3.394 (95% CI 2.283–5.046). Other patient characteristics (age, gender, presence of DM, hypercholesterolaemia, hypertension and (history of) smoking) were not found to influence the association between ECG findings and MI on CMR with LGE.

The retrospective study design leads to several limitations, including the problem of missing data in the EMR and the inability to check the correctness of certain classifications like family history or other risk factors. Besides, ECGs were classified into two groups: presence or absence of MI on the ECG. However, in clinical practice, there is a wide range of doubtful cases in which the application of predefined criteria can be discussed. This may have influenced the reliability of the results. In clinical practice the physician always has more information, which helps him/her in making a decision, like the presence of risk factors, use of medication, blood tests and previous ECGs. Finally, the external validity of this study is limited. In this study all patients who underwent CMR with LGE between January 2014 and December 2017 were selected. This resulted in a heterogeneous patient group, in which the CMR with LGE was performed for a variety of indications. The results of our study might be different from those in other populations such as patients with a suspicion of CVD.

In future research the above-mentioned limitations could be minimised by using a prospective study design, whereby all patients in whom an ECG is performed also have to undergo CMR with LGE. However, such a study design is not feasible in clinical practice because of the high costs and a lack of availability of CMR systems.

We recommend to at least consider performing CMR with LGE in patients with signs of prior MI on the ECG if the outcome has clinical consequences for secondary prevention.