The Prognostic Value of Echocardiographic Wall Motion Score Index in ST-Segment Elevation Myocardial Infarction

Background When compared to left ventricular ejection fraction (LVEF), previous studies have suggested the superiority of wall motion score index (WMSI) in predicting cardiac events in patients who have suffered acute myocardial infarction. However, there are limited studies assessing WMSI and mortality in ST-segment elevation myocardial infarction (STEMI). We aimed to compare the prognostic value of WMSI in a cohort of STEMI patients treated with primary percutaneous coronary intervention (PCI). Methods A comparison of WMSI, LVEF, and all-cause mortality in STEMI patients treated with primary PCI between January 2008 and December 2020 was performed. The prognostic value of WMSI, LVEF, and traditional risk scores (TIMI, GRACE) were compared using multivariable logistic regression modelling. Results Among 1181 patients, 27 died within 30-days (2.3%) and 49 died within 12 months (4.2%). WMSI ≥1.8 was associated with poorer survival at 12-months (9.2% vs 1.5%; p < 0.001). When used as the only classifier for predicting 12-month mortality, the discriminatory ability of WMSI (area under the curve (AUC): 0.77; 95% CI: 0.68–0.84) was significantly better than LVEF (AUC: 0.71; 95% CI: 0.61–0.79; p=0.034). After multivariable modelling, the AUC was comparable between models with either WMSI (AUC: 0.89; 95% CI: 0.85–0.94) or LVEF (AUC: 0.87; 95% CI: 0.83–0.92; p < 0.08) yet performed significantly better than TIMI (AUC: 0.71; 95% CI: 0.62–0.79; p < 0.001), or GRACE (AUC: 0.63; 95% CI: 0.54–0.71; p < 0.001) risk scores. Conclusions When examined individually, WMSI is a superior predictor of 12-month mortality over LVEF in STEMI patients treated with primary PCI. When examined in multivariable predictive models, WMSI and LVEF perform very well at predicting 12-month mortality, especially when compared to existing STEMI risk scores.


Introduction
Te prognostic value of predischarge echocardiography following an acute myocardial infarction (AMI) has been well established [1]. Currently, the European Society of Cardiology guidelines recommend all patients who sufer an ST-segment elevation myocardial infarction (STEMI) should undergo transthoracic echocardiographic evaluation to determine left ventricular ejection fraction (LVEF) prior to discharge [2]. LVEF is widely recognised as a prognostic marker for mortality outcomes following AMI [3][4][5][6]. After sufering an AMI, detection of left ventricular regional wall motion abnormalities by echocardiography is common, and wall motion scoring is routinely used to defne and quantitate areas of hypokinesia in the afected myocardium [7]. Te wall motion score index (WMSI) numerically sums the average scores for all left ventricular segments into a single parameter. Te prognostic value of WMSI has been investigated in small cohorts of patients with acute myocardial infarction, suggesting superiority to LVEF in predicting mortality [8][9][10].
Furthermore, previous fndings have indicated that in acute myocardial infarction, WMSI may more accurately refect the amount of myocardial damage compared to LVEF when compensatory hyperkinesis of noninfarcted myocardium is present [11]. Tere is also conficting data whether WMSI is a less sensitive predictor in ST-segment elevation myocardial infarction (STEMI) compared to non-ST segment elevation myocardial infarction (NSTEMI) [8,12]. Despite the suggestion of superiority of WMSI over LVEF to predict mortality, there is limited data pertaining to WMSI in STEMI cohorts treated with primary percutaneous coronary intervention (PCI) and on the routine use of regional wall motion scoring in predictive models postmyocardial infarction. Traditional predictive risk scores used to estimate mortality post-STEMI (GRACE and TIMI Risk Scores [13,14]) incorporate variables which often fuctuate during initial presentation such as heart rate and systolic blood pressure, but do not include measures of cardiac function such as LVEF or WMSI, which may provide more accurate predictions of survival. We aimed to evaluate the prognostic value of WMSI compared to LVEF in a large cohort of STEMI patients who underwent primary percutaneous coronary intervention (PCI) and compared the performance with traditional STEMI risk scores.

Design.
A cohort consisting of consecutive STEMI patients treated with primary PCI within 12 hours of symptom onset during a study period between January 2008 and December 2020 at the Prince Charles Hospital, a quaternary referral centre in Queensland, Australia, was analysed. Patients were included for analysis if they survived the index PCI procedure, underwent transthoracic echocardiography within the index admission, and had both LVEF and WMSI calculated on the echocardiogram. Patients were excluded if they did not undergo transthoracic echocardiography during the index admission, did not have either LVEF or WMSI calculated, or were deemed salvage PCI (out-of-hospital cardiac arrest with emergency intubation prior to PCI). Tis study was conducted according to the World Medical Association Declaration of Helsinki and was approved by the Human Research Ethics Committee of the Prince Charles Hospital (LNR/2018/QPCH/47412) and the Australian Institute of Health and Welfare Ethics Committee (EO2020/2/1147).

Patient
Characteristics. Patient demographics, cardiac risk factors, and procedural data were collected at the time of the index admission and recorded in the cardiac catheterisation database. First medical contact was defned as either the time from paramedic arrival for emergency medical service transfer or primary PCI centre arrival for patients who self-presented.

Echocardiographic Measurements.
A transthoracic echocardiogram was performed by cardiac sonographers as part of the routine standard of care during the index admission post-PCI. Te left ventricular (LV) regional wall motion scores were examined using a 16-segment model as per American Society of Echocardiography (ASE) guidelines ( Figure 1) [15,16]. Individual segments were scored as follows: (1) normal or hyperkinetic, (2) hypokinetic (reduced thickening), (3) akinetic (absent or negligible thickening), and (4) dyskinetic (systolic thinning or stretching). WMSI was calculated by adding the scores of individual segments and dividing this total by the number of segments assessed. LVEF was obtained by Simpson's bi-plane method [17] and is calculated as the percent ratio of the stroke volume (LV end-diastolic volume minus LV end-systolic volume) and LV end-diastolic volume.

Study Outcome.
Te primary outcome was all-cause mortality. Patients were followed up after primary PCI with the angioplasty nursing service and routine hospital clinic appointments. Outcomes were assessed at 30-days and 12months post-PCI. Mortality and cause of death data were confrmed with data linkage to the Australian Institute of Health and Welfare's National Death Index registry, which records the death information from all states and territories in Australia.

Statistical Analysis.
Patient characteristics and procedural variables were compared by 12-month survival. Continuous variables were summarised as mean (and standard deviation (SD)) and tested between groups using a Student's t test if approximately normally distributed, or summarised as median (and interquartile range (IQR)) and tested using Wilcoxon's rank-sum test otherwise. Categorical variables were summarised as frequency (%) and tested between groups using Pearson's chi-square test or Fisher's exact test as appropriate. Where possible, WMSI and LVEF were analysed as continuous variables. Assumptions of linearity were assessed by using the link test and testing fractional polynomial terms to determine the best functional form for continuous variables to be included in logistic regression models. Where cut-points were used, these were chosen based on clinical relevance or prior literature. Associations between WMSI and LVEF and mortality (30-days and 12 months) were analysed using logistic regression. To facilitate comparison of WMSI and LVEF efect sizes, LVEF was transformed to its complement (by subtracting the value from 100) and both variables were standardised. Confounders of interest were identifed a priori based on a literature review. Tese included patient demographics and past medical history, measures of disease severity, and periprocedural variables. Eligible variables with univariable p values <0.20 were entered into a multivariable model. Stepwise backward selection was used to determine the base model. Excluded variables were re-entered and tested in the fnal model which was selected based on Akaike's Information Criteria. Modelling was repeated for each outcome time point (30-day and 12-months) and for each alternative measure of cardiac disease severity (LVEF and WMSI) and compared with existing STEMI risk scores (GRACE and TIMI). Te GRACE and TIMI risk scores were chosen as the comparators as they are the most widely validated STEMI risk scores.
Model ft was assessed by comparing the number of observed and predicted events across probability deciles and tested using Hosmer-Lemeshow's goodness of ft test. Internal validation was assessed further by a graphical comparison of calibration belts, which compare predicted and observed values. Parametric receiver operator curve (ROC) analysis was used to compare the discriminatory ability of WMSI to LVEF, both alone and in the presence of covariates, and to compare WMSI to the GRACE and TIMI risk scores. WMSI cut-of values that maximise the sensitivity or specifcity were examined according to the Youden or Liu criteria. Nelson-Aaelen methods were used to compare survival over time among groups defned using the optimal cut-points. Analyses were performed using the Stata statistical software package (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LLC).

Study Population.
During the study period, 1712 STEMI patients were treated with primary PCI. Of these patients, 1441 (84%) had echocardiographic data available during the inpatient admissions. After exclusions, 1181 patients with both LVEF and WMSI calculated were included in the fnal cohort. Exclusions are listed in Supplementary Figure 1 Table 1. Patients who died were on average older had a lower mean GFR.
Excellent internal calibration was observed for all multivariable models (Models 1-4) in predicting 12-month Critical Care Research and Practice mortality, with close agreement between observed proportions and predicted probabilities and confdence intervals symmetrically distributed around the diagonal line.

Discussion
Tis study provides an evaluation of the prognostic value of WMSI in comparison to LVEF and traditional STEMI risk models in a large cohort of STEMI patients treated with primary PCI. To our knowledge, this is the largest study examining WMSI, LVEF, and outcomes in the STEMI population. In this study, individually, WMSI, LVEF, and TIMI risk scores were very good predictors of mortality at 30 days post-STEMI and performed signifcantly better than the GRACE risk score. Consistent with other literature [8,12,18], this study has demonstrated the superiority of WMSI when compared individually to LVEF at predicting 12-month mortality outcomes in STEMI patients. When compared to traditional STEMI risk scores (GRACE and TIMI), the multivariate risk models in this study, adjusting for confounders and mediators by incorporating WMSI and LVEF, performed signifcantly better (Table 2). Te incremental beneft of WMSI over LVEF in predicting mortality when incorporated into a multivariate model, however, was only small. Te authors believe that this may be due to inclusion of more relevant confounders such as ischaemic time and renal function, which have previously not been studied when examining WMSI or not traditionally  incorporated into the same STEMI mortality risk modelling. Te validity of traditional risk scores, in particular the GRACE risk score, may not be as relevant in contemporary STEMI treatment with primary PCI [19], and similar to the fndings of this study, displayed weak predictive value of the GRACE score. Other risk scores such as the PAMI and CADILLAC risk scores, although less validated, may have provided a higher predictive value than the traditional risk scores used in our study. Tis study also demonstrated a WMSI cut-of of ≥1.8 was a signifcant and strong predictor of 12-month mortality. Utilising this cut-of may help identify patients who traditionally may have only been identifed as having mild LV dysfunction or normal LV function measured by LVEF and may be at increased risk of mortality. Te superiority of WMSI compared to LVEF may be explained by several factors, including the nongeometric assumptions with WMSI calculations which are present with LVEF calculations as well as over-compensation of LVEF calculations when incorporating hyperkinetic segments. Additionally, the inferolateral wall is not represented when calculating the LVEF by Simpson's bi-plane method and in patients with isolated inferolateral MI, LVEF will likely be overestimated.
Early studies investigating regional wall motion scoring were positive in demonstrating a correlation with LVEF [20]; however, there were variations in the WMSI calculation models used and were in smaller cohorts of acute myocardial infarction patients. Later studies [21,22] investigated cohorts of thrombolysed STEMI patients, demonstrating that the use of a 16-segment model to calculate WMSI was better correlated to LVEF than the existing 11-and 14-segment models. Carluccio et al. [22] showed that a WMSI >1.5 was a powerful predictor of subsequent cardiac events which performed better than LVEF <40%. Moller et al. [12] examined a large cohort of 767 acute myocardial infarction patients (376 STEMI) of which 146 patients underwent primary PCI over a median 40-month follow-up. Similar to the results of our study, the authors used a 16-segment model and showed the superiority of WMSI in predicting all-cause mortality (1-year mortality of 18%) over LVEF when incorporated into a Cox regression model.
Several more recent studies have examined WMSI in comparison to other echocardiographic measurements including global longitudinal strain (GLS) and demonstrated good correlation with LVEF, with results of both WMSI and GLS demonstrating superiority to LVEF in predicting combined endpoints of heart failure and mortality in acute myocardial infarction patients with regional wall motion abnormalities [23]. Mistry et al. [24] examined 163 STEMI patients and compared GLS and LVEF and WMSI using a 16-segment model. Whilst GLS best correlated with WMSI (r 2 � 0.55.), WMSI was best correlated with relative infarct  We believe, despite the fndings of superiority of WMSI over other measures of systolic function including LVEF, issues surrounding the use of WMSI in risk stratifcation postmyocardial infarction may pertain to the lack of consistency in application of wall motion scoring models and variations in cut-of values in the existing literature. Additionally, when compared to existing WMSI literature, there have been much larger studies examining LVEF and mortality across broader indications than acute myocardial infarction [26,27]. Similar to our study, other studies [8,12] have suggested a WMSI cut-of value of ≥1.8 has validity in predicting outcomes at 12-months. A standardised cut-of value for WMSI may provide clinicians with another tool to predict mortality and assist with identifying patients at higher risk than previously identifed when stratifed by traditional LVEF measurements or using existing STEMI risk scores. Te clinical utility of WMSI over other echocardiographic methods of assessment such as strain evaluation is that, despite poor acoustic windows, WMSI can usually be assessed, and it does not require postprocessing using additional software or additional expertise to analyse. In line with the ASE guidelines, we recommend the use of the ASE 16-segment model [15,16] over the use of the 17segment model which includes the apex segment, for routine assessment of WMSI scoring, because thickening of the tip of the apex and endocardial excursion are often not well visualized.

Limitations
Tis is a single-centre study with a large cohort of consecutive STEMI patients treated with primary PCI. Patient inclusion in the study was reliant on the availability of echocardiographic data at the index presentation. Tere were patients who were excluded from analysis as WMSI and/or LVEF were either not assessed or not calculated, which may introduce bias. Patients who were intubated prior to primary PCI with unknown neurological status were also excluded to avoid introducing mortality bias (noncardiac related death). Mortality outcomes were lower than previously reported studies examining WMSI in acute myocardial infarction, yet consistent with current primary PCI literature, we presume this was due to the treatment with contemporary primary PCI strategies. Tis study did not examine confounders such as medication compliance or cardiac rehabilitation, which may also infuence mortality post-STEMI. Further investigations incorporating echocardiographic parameters such as WMSI and LVEF into predictive risk models for STEMI are warranted.

Conclusion
In STEMI patients who underwent primary PCI, both WMSI and LVEF derived from transthoracic echocardiography are highly prognostic for short term mortality. When compared individually, WMSI using the 16-segment model is superior to LVEF measurements at predicting mortality at 12 months. A WMSI cut-of value ≥ 1.8 was correlated with poorer all-cause mortality at 12 months. When compared in multivariable predictive models, the wall motion score index and left ventricular ejection fraction perform very well at predicting 12-month mortality, especially when compared to traditional STEMI risk models.

Data Availability
Te clinical data used to support the fndings of this study are restricted by the Prince Charles Hospital Human Research and Ethics Committee in order to protect patient privacy and confdentiality. Data are available from the relevant Prince Charles data custodian/s for researchers who meet the criteria for access to confdential data.