E/e' in relation to outcomes in ST‐elevation myocardial infarction

Myocardial infarction (MI) is a high‐risk condition especially when filling pressure is raised, and earlier reports have suggested that E/e’ is associated with poor outcome. However, whether E/e’ predicts risk better than LVEF, which is the current standard of practice, is not known. We investigated this question in the largest and most rigorous study of MI patients so far.


| INTRODUC TI ON
ST-segment elevation myocardial infarction (STEMI) remains a high-risk cardiovascular event despite therapeutic advances such as primary percutaneous coronary intervention (PCI). 1 Mortality in the acute phase is largely secondary to loss of pumping strength, which leads to congestive heart failure (CHF), shock, and arrhythmias originating from the left ventricle. Cardiovascular mortality postdischarge, however, comprises a number of different components. Real-world data have shown these to primarily be re-infarction, stroke, CHF, and sudden cardiac death. 2 While scoring systems such as TIMI and GRACE have been validated for risk prediction, echocardiography is also performed routinely as depressed left ventricular (LV) ejection fraction (LVEF) during the index admission identifies patients at risk of late mortality (≥40 days postdischarge). 3 Elevated LV filling pressure is also an important risk indicator in STEMI, 4 and CHF indeed accounts for a significant proportion of both short-term and long-term deaths. 2 This begs the question of whether echocardiographic measures of LV filling pressure may be valuable risk predictors in STEMI, especially E/e' which seems especially attractive as it is both directly influenced by myocardial function 5,6 and predictive of adverse LV remodeling post-MI. 7 Unfortunately, contemporary reports associating E/e' with outcomes post-MI have yielded conflicting results. Firstly, a link between E/e' and in-hospital death remains unproven. 8 Secondly, the relationship between E/e' and long-term outcomes is controversial: While initial data showed that E/e' was a very strong predictor of poor outcome 13 months post-MI in n = 250 patients, 9 a larger study performed by the same authors, which controlled better for residual confounding, showed E/e' to be nonsignificant, once key covariates such as age and LVEF were entered into the model (n = 388). 10 On this background, we performed the largest study so far relating E/e' to outcomes in STEMI. We compared utility of adding E/e' to clinical covariates and LVEF in sequential testing. In order to achieve a satisfactory sample size, we analyzed 2-year follow-up data after STEMI admission.

| Patients and data sources
The study was based on prospectively acquired clinical registries combined with administrative data as follows. Patients were originally identified in an Interventional Cardiology Procedural Registry as having successfully undergone percutaneous coronary intervention (PCI) for ST-elevation MI at a single tertiary center (National Heart Centre Singapore). Angiographic and echocardiographic data were extracted from digital sources. Linkage to outcome events was performed using 2 main data sources: firstly, registries at Singapore National Registry of Diseases Office (NRDO) were used to analyze myocardial infarction, ischemic stroke, and death events. Secondly, claims data at Ministry of Health were used to analyze admissions with heart failure. Data were subsequently anonymized before being made available by NRDO to our biostatistician (FG). Detailed information about databases including their audited accuracy is available as an online supplement (Appendix S1).
All cases admitted between 1 August 2007 and 31 December 2010 were considered for inclusion. We excluded patients that (a) had not undergone echocardiography within a week of PCI and (b) had neither Singapore citizenship nor permanent residency.
We counted only the first recorded admission meeting these criteria if more than 1 existed. The study had ethics approval from SingHealth Centralized Institutional Review Board which was granted with waiver of requirement for patient consent. It was nonexperimental in design and did not violate the Declaration of Helsinki.

| Echocardiography
Measurements were derived according to guidelines as recommended by the American Society of Echocardiography. 11,12 In brief, LVEF was derived from end-diastolic volume (EDV) and end-systolic an appropriate assumption given that the prevalence of elevated LA pressure is high in a population of STEMI cases. 15

| Clinical outcomes
The primary short-term outcome was all-cause death during the index admission. The primary long-term outcome was time postdischarge to a composite outcome comprising (i) myocardial infarction, (ii) ischemic stroke (iii) admission with CHF, and (i) all-cause death.
The secondary long-term outcome was time postdischarge to allcause death alone. Long-term outcome analyses were based on events up to 24 months postdischarge. Short-and long-term models were nonoverlapping such that outcome events were counted only once: in-hospital deaths were not included in long-term survival analyses, which began at the time of discharge. This was due to the fact that ongoing medical treatments were felt to be important confounders in long-term analyses, and robust data for medications were only available at the time of discharge.

| Statistical analysis
Short-term outcomes were analyzed using logistic regression while adjusting for age, sex, and LVEF. Long-term outcomes were analyzed in Cox regression models adjusted for age, sex, diabetes, revascularization procedure, history of myocardial infarction, hypertension, renal function, treatment with beta blockers, ACE inhibitors or angiotensin receptor blockers, statin, clopidogrel, and for LVEF.
Information about drugs was taken at the time of discharge for all drugs studied. All analyses were performed using Stata v. 13 (College Station, Texas, USA). Risk is reported as proportions for short-term events and as cumulative incidence (based on analysis of personyears) for long-term events.

| Population characteristics
We identified n = 660 patients that met inclusion criteria, comprising 555 males (84%) and 105 females (16%), with a mean age of 59 ± 12 years. Median length of stay was 4 days (inter-quartile range: 3-5). There were 184 patients (28%) with elevated E/e'; this subgroup was significantly older, with a higher prevalence of smoking, diabetes, hypertension, dyslipidemia, chronic kidney disease, and peripheral vascular disease (Table 1). On the contrary, a similar proportion of patients with raised E/e' had a prior history of myocardial infarction and revascularization, and prescription of cardioprotective therapeutics did not differ in this subgroup (antiplatelet drugs, angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, beta blockers, and diuretics) ( Table 1). Patients with elevated E/e' differed in terms of echocardiographic findings: LV dimensions were larger and LVEF was lower ( Table 2).

| Predictors of outcome
In short-term models, there were 19 in-hospital death events.

| D ISCUSS I ON
While considerable progress has been made in STEMI care, mortality and morbidity remain high, 1 and risk stratification is important. 3,16 While filling pressure has been shown to be an important predictor of poor outcome in MI, earlier reports seeking to connect E/e' to clinical outcomes have yielded mixed results. The present study found an association between E/e' and in-hospital death in patients with STEMI. There was also an association with long-term risk of composite outcome events, as well as with death, which was noted in unadjusted analyses. However, after rigorous adjustment for confounding, E/e' no longer predicted long-term outcome events independently. In contrast, LVEF was related to long-term outcomes while retaining a very high degree of statistical significance in the fully adjusted model.

| Short-term risk
The present study is the first to find a relationship between shortterm risk of death post-STEMI and E/e' done shortly after admission. DINAMIT and IRIS. Therefore, the fact that E/e' appears to identify high-risk patients over and above LVEF in the acute phase of STEMI remains valid, and one may argue that E/e' may still be "operationally interesting" for trials and registries seeking to enroll high-risk STEMI cases for the purpose of evaluating the efficacy of drugs or devices.

| Long-term risk
The number of long-term outcome events in the present study was larger than in any previous publication in this field, which enabled us to adjust coefficients for a range of potential confounders. This is TA B L E 1 Basic characteristics of study population the correctness of estimates in models based on simulated data and found that the EPV ratio is in fact less important than the structure of the data itself. While Peduzzi et al found more stable estimates when the EPV ratio was at least 10, 27 Vittinghoff & McCulloch found that observational research can relax that requirement when conducted to analyze causal influences, as models remain accurate when EDV ratio is below 10. Instead, they argued that models built for prediction need a higher EPV ratio of ≥20. 28 The present study found that LVEF remained very highly significant in the final model, but not E/e'. While patients who develop CHF post-STEMI often have raised filling pressures, this is not always the case. One way to interpret our findings is that CHF deaths may be a relatively more important problem during the admission-at which time at-risk patients can be identified by raised E/e'-whereas death events may occur in poor LVEF patients over the first 2 years postdischarge irrespective of whether they had elevated filling pressure or not at the time of the initial STEMI. Future work in this area will benefit from more granular data on mortality, which will show whether the death events identified by LVEF are different to those associated with raised E/e'.

| LI M ITATI O N S
In addition to limitations related to sample size discussed in detail above, an important consideration is the fact that the population under study constitutes a subset of STEMI admissions where pri- Myocardial Infarction Registry does not collect data on NT-proBNP, which is therefore not included in the present report.

| CON CLUS ION
In conclusion, in the largest study population and most rigorous analysis to date, we found that E/e' is a stronger predictor of in-hospital death than LVEF, but that the relationship between E/e' and adverse outcome events at 2 years was rendered nonsignificant once LVEF was entered into the model. While these results suggest that E/e' may help to identify STEMI patients at highest risk of death in the acute phase, our data do not support an independent role of E/e' in predicting long-term outcome in STEMI.

ACK N OWLED G M ENTS
The authors gratefully acknowledge invaluable help from Ministry of Health and National Registry of Disease Office, Singapore, with data handling and merging of registries.

CO N FLI C T O F I NTE R E S T S
Authors report no relationships that could be construed as a conflict of interest.

AUTH O R CO NTR I B UTI O N S
All authors made important contributions in terms of concept and design of the study, and associated analysis. All authors helped to  .091 1.9 (0.9-3.9) .093 Note: Composite outcome events include myocardial infarction, ischemic stroke, heart failure admission, and all-cause death. Model 1 provides an unadjusted estimate of risk.
Estimates in model 2 were adjusted for age, sex, diabetes, revascularization procedure, history of myocardial infarction, hypertension, renal function, beta blockers on discharge, ACE inhibitors or angiotensin receptor blockers on discharge, statin on discharge, and clopidogrel on discharge.
Estimates in model 3 were adjusted for the same covariates as model 2, but also for LVEF.
Models were nonoverlapping: in-hospital deaths were not considered in long-term analyses.
analyses were run by FG. Work done with government agencies in Singapore was facilitated by MYC and KKY.