Impaired Cardiac Sympathetic Innervation Increases the Risk of Cardiac Events in Heart Failure Patients with Left Ventricular Hypertrophy and Mechanical Dyssynchrony

Background. Left ventricular mechanical dyssynchrony (LVMD), left ventricular hypertrophy, and impaired cardiac sympathetic innervation are closely related to the development of heart failure (HF) and unfavorable outcomes. Methods and Results. A total of 705 consecutive HF patients with reduced left ventricular ejection fraction (EF) < 50% were registered in our hospital HF database. LVMD and left ventricular mass index (LVMI) were evaluated three-dimensionally by gated myocardial perfusion SPECT. LVMD was measured as a heterogeneity index (phase SD) of the regional contraction phase angles calculated by Fourier analysis. Cardiac sympathetic innervation was quantified as a normalized heart-to-mediastinum ratio (HMR) of the 123I-metaiodobenzylguanidine (MIBG) activity. The patients were followed up with a primary end point of lethal cardiac events (CEs) for 42 months. CEs were documented in 246 of the HF patients who had a greater phase SD, greater LVMI, and lower MIBG HMR than those in HF patients without CEs. In the overall multivariate analysis, phase SD, LVMI, and MIBG HMR were identified as significant CE determinants. The three biomarkers were incrementally related to increases in CE risks. Conclusions. Assessment of cardiac sympathetic innervation can further stratify patients with systolic heart failure at increased cardiac risk identified by left ventricular hypertrophy and mechanical dyssynchrony.


Introduction
Management of patients with heart failure (HF) is becoming a critical concern to be resolved due to rapid increases in the number of elderly patients, and the increase in expenditure for life-saving measures and subsequent long-term medical care [1,2]. Despite the recent advances in pharmacological and non-pharmacological treatments, cardiac mortality remains high, with a 5-year survival rate of only 25% after hospitalization for HF with reduced left ventricular ejection fraction (LVEF) [1][2][3][4][5]. Among the known prognostic clinical variables, left ventricular hypertrophy (LVH) is a classical but still important variable, but has been conventionally assessed by electrocardiography or two-dimensional echocardiography. Attention has recently been given to left ventricular mechanical dyssynchrony (LVMD) in relation not only to prognostic implications, but also to the clinical indication of cardiac re-synchronization treatment (CRT). Despite improvements in cardiac prognosis using cardiac implantable electronic devices, there are still limitations in actual prognostic efficacies [6][7][8]. Besides technical issues of device implantation, this is probably because of the limited assessment using an electrical and/or two-dimensional echo data, resulting in an ineffective CRT rate of nearly 30% [9,10]. On the other hand, three-dimensional evaluation of LVMD and LVH can be done routinely in cardiac nuclear imaging, and our previous studies have shown the improvement of prognostic assessment using three-dimensional nuclear techniques, contributing to optimization of the selection of CRT candidates [11][12][13].
The autonomic nervous system plays crucial roles in the maintenance and compensation of cardiac function and, in contrast, in the progression and fatal outcomes of HF [14,15]. Neuroimaging with a norepinephrine analogue, iodinated metaiodobenzylguanidine (MIBG), enables non-invasive delineation and evaluation of pre-synaptic sympathetic innervation in human hearts using a conventional imaging technique [16,17].
A number of investigations have demonstrated independent and incremental prognostic implications of impaired cardiac sympathetic nerve function assessed by cardiac MIBG imaging in HF patients. In particular, multicenter studies established the prognostic values of cardiac MIBG activity in the risk-stratification and the prediction of cardiac outcomes in HF patients, even when conventional prognostic markers such as NYHA functional class, LVEF, and plasma brain natriuretic peptide (BNP) are considered [18][19][20].
In these contexts, a comprehensive understanding and assessment of the pathophysiological conditions such as three-dimensional LVMD and LVH and alterations of cardiac sympathetic innervation are expected to make the prediction and prevention of lethal cardiac outcomes more accurate in HF patients. In this study, we tested the hypothesis that the integration of three-dimensional LVMD and LVH into the impairment of cardiac sympathetic innervation enables more precise evaluation of cardiac risks in HF patients who are conventionally risk-stratified.

Study Patients
A total of 705 patients were consecutively and prospectively registered in our hospital HF database between April 2011 and December 2016 using the following entry criteria: symptomatic HF established by typical symptoms and signs based on the Framingham criteria including palpitation, dyspnea/orthopnea, neck vein distension, peripheral and/or lung congestion S3/S4 gallop, chest radiographic findings of congestive heart failure, echocardiographic LVEF less than 50% at admission, and a significant increase in BNP. Electrocardiographic and imaging examinations were conducted immediately after admission to establish the diagnosis by excluding non-cardiac diseases with similar symptoms or signs. HF etiologies, such as ischemic versus non-ischemic, were also differentiated using electrocardiographic and cardiac imaging studies with or without a stress approach and coronary angiographic examination depending on the patient's clinical conditions. The following patients were excluded from the analysis: patients aged less than 20 years; patients who refused entry to the study or refused resuscitation treatment; and patients with overt serious non-cardiac disorders such as malignancy, infections, or hemorrhagic diseases. Just before discharge, blood examinations of hemoglobin (Hb), BNP, and creatinine were performed and the estimated glomerular filtration rate (eGFR) was calculated. The plasma BNP level was measured in the initial 310 patients (44%) then and NT-pro BNP level was measured in the next 395 patients (56%). Because of the two different methods, BNP/NT-pro BNP staging was performed on the basis of ESC heart failure guidelines as follows: 0-40 pg/mL and 0-125 pg/mL for stage 1, 41-100 pg/mL and 126-400 pg/mL for stage 2, 101-200 pg/mL and 401-900 pg/mL for stage 3, and 201-pg/mL and 901-pg/mL for stage 4.
All participants provided written informed consent. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki, as reflected in a priori approval by our institution's human research committee.

Two-Dimensional Echocardiographic Assessment
Two-dimensional echocardiography was performed from standard parasternal longaxis and apical four-, three-, and two-chamber views in a left lateral decubitus position using commercially available ultrasound machines equipped with a 2.5-MHz variable frequency transducer by echocardiographic technicians, repeatedly when necessary. The following echocardiographic parameters measured before discharge were used for the analysis: left atrium diameter (LAD, mm); left ventricular end-diastolic diameter (LVDd, mm); septal wall thickness at end-diastole (IVSTd, mm); posterior wall thickness at enddiastole (PTWd, mm); LVEF (%), calculated using the biplane modified Simpson's method; left ventricular volume at end-diastole (EDV, ml); left ventricular volume at end-systole (ESV, mL); E wave velocity (m/sec); and left ventricular deceleration time of E wave (Dct, msec) and septal E/e .

Assessment of LVMD and Left Ventricular Mass Index
LVMD and left ventricular mass index (LVMI) as an index of LVH were evaluated using three-dimensional data derived from resting myocardial perfusion imaging (MPI) with 99m Tc-tetrofosmin of 300 MBq, as shown by our previous study ( Figure 1) [11][12][13]. Briefly, MPI SPECT was performed using an ECG-gated approach with a frame rate of 16 and a single-head gamma camera equipped with a high-resolution, parallel-hole collimator. LVMD and LVMI were measured using three-dimensionally reconstructed ECG-gated data of serial 10-to 14-short-axis SPECT images with the Heart Function View software (HFV version 1.1;Nihon MediPhysics, Tokyo, Japan) [12,13] and Corridor 4DM version 6.0 software (INVIA Medical Imaging Solutions, Ann Arbor, MI, USA) [11], respectively. LVMD was quantified as a standard deviation (SD) of phase angles (degrees) on a phase histogram ( Figure 1). Phase angles at the initiation of systole were calculated mathematically by a high-ordered Fourier analysis of myocardial count changes (i.e., time−activity curve) in each pixel during a cardiac cycle, which is basically identical to a regional time−volume curve. The left ventricular mass was normalized as LVMI (g/m 2 ) using body surface area.

Cardiac 123 I-MIBG Imaging
In a stabilized patient's condition, cardiac imaging with 123 I-MIBG of 111 MBq was performed using a gamma camera equipped with a low-energy, general-purpose collimator in a fasting and resting condition for 15-30 min (early image) and 4 h (late image) after an intravenous tracer injection, as described previously [11][12][13][14]. Cardiac 123 I-MIBG activity was measured as a heart/mediastinum ratio (HMR) by a region of interest being automatically set on the upper mediastinum and the whole cardiac region on a planar anterior image with dedicated MIBG software (smart MIBG software, Tokyo, Japan) operated by an experienced nuclear medicine technician without knowledge of any clinical data. 123 I-MIBG washout kinetics from the heart was calculated as the washout rate (WR) from early and late cardiac 123 I-MIBG activities without a decay correction. The cardiac MIBG HMR was standardized to the medium-energy collimator condition by the mathematical method established by cross-calibration phantom experiments [18]. This method can minimize variations in MIBG HMR originating from different collimators such as low-and medium-energy ones between institutions or study populations. sion 6.0 software (INVIA Medical Imaging Solutions, Ann Arbor, MI, USA) [11], respectively. LVMD was quantified as a standard deviation (SD) of phase angles (degrees) on a phase histogram (Figure 1). Phase angles at the initiation of systole were calculated mathematically by a high-ordered Fourier analysis of myocardial count changes (i.e., time−activity curve) in each pixel during a cardiac cycle, which is basically identical to a regional time−volume curve. The left ventricular mass was normalized as LVMI (g/m 2 ) using body surface area.

Cardiac 123 I-MIBG Imaging
In a stabilized patient's condition, cardiac imaging with 123 I-MIBG of 111 MBq was performed using a gamma camera equipped with a low-energy, general-purpose collimator in a fasting and resting condition for 15-30 min (early image) and 4 h (late image) after an intravenous tracer injection, as described previously [11][12][13][14]. Cardiac 123 I-MIBG activity was measured as a heart/mediastinum ratio (HMR) by a region of interest being automatically set on the upper mediastinum and the whole cardiac region on a planar anterior image with dedicated MIBG software (smart MIBG software, Tokyo, Japan) operated by an experienced nuclear medicine technician without knowledge of any clinical data. 123 I-MIBG washout kinetics from the heart was calculated as the washout rate (WR) from early and late cardiac 123 I-MIBG activities without a decay correction. The cardiac MIBG HMR was standardized to the medium-energy collimator condition by the mathematical method established by cross-calibration phantom experiments [18]. This method can minimize variations in MIBG HMR originating from different collimators such as low-and medium-energy ones between institutions or study populations.

Follow-Up Protocol
Patients were regularly followed up at an out-patient care unit by cardiologists for one year or more when the patients survived. The primary endpoint was lethal cardiac events including sudden cardiac death, death due to progression of pump failure, lifethreatening lethal ventricular tachyarrhythmias, and appropriate ICD therapy. Clinical outcomes were confirmed retrospectively by reviewing medical records. Sudden cardiac death was defined as witnessed cardiac arrest and death within 1 h after acute onset of Phase histograms for measurements of phase SD (left panels) and heart/mediastinum ratio (HMR) of MIBG activity (right panels). Case 1: a 72-year-old female with ischemic cardiomyopathy who had reduced LVEF (32%), but had nearly normal values of LV mass index (97 g/m 2 ), phase SD (19), and MIBG HMR (2.66). No cardiac event occurred during the follow-up period. Case 2: a 68-yearold male undergoing cardiac re-synchronization therapy who had non-ischemic cardiomyopathy with reduced LVEF (28%), increased LV mass index (178 g/m 2 ), increased phase SD (40), and decreased MIBG HMR (1.48). He died of progressive pump failure during the follow-up period.

Follow-Up Protocol
Patients were regularly followed up at an out-patient care unit by cardiologists for one year or more when the patients survived. The primary endpoint was lethal cardiac events including sudden cardiac death, death due to progression of pump failure, lifethreatening lethal ventricular tachyarrhythmias, and appropriate ICD therapy. Clinical outcomes were confirmed retrospectively by reviewing medical records. Sudden cardiac death was defined as witnessed cardiac arrest and death within 1 h after acute onset of symptoms or unexpected death in patients being well within the previous 24 h. This study was based on the principles outlined in the Declaration of Helsinki, and informed consent for enrollment in our database and usage for clinical study was obtained according to the guidelines of the ethics committee of our hospital.

Statistical Analysis
A statistical value is shown as mean ± 1 SD. Mean values were compared between the two groups using the unpaired t-test, and categorical variables were compared using the Chi-square test. The Kaplan−Meier analysis using the key parameters identified in this study was used to create a time-dependent, cumulative event-free curve, and the log-rank test was also used for comparison of the curves, if necessary. Following univariate analysis, multivariate analysis with a Cox hazard proportional model was performed using the statistically appropriate number of significant variables identified to calculate the hazard ratios and 95% confidence intervals (CIs). A receiver operating characteristic (ROC) analysis was performed to determine an optimal cutoff value of an independent significant parameter for differentiating a risk category. Global Chi-square values were calculated when incremental prognostic values were needed to be clarified, using several significant variables identified above. A computer software program, SAS for Windows, version 9.4 (SAS Institute, Cary, NC, USA), was used for these analyses. A p-value less than 0.05 was considered significant.

Results
The patients included 517 males (73.3%) and had a mean age of 67.1 ± 12.6 years and a mean LVEF of 35.7 ± 10.6%. During a mean follow-up period of 42.2 months, lethal cardiac events were documented in 246 (35%) of the patients: 190 died due to HF progression, 22 had lethal ventricular arrhythmias, 22 had sudden cardiac death, and 12 had appropriate ICD shock treatment against lethal ventricular arrhythmias. Patients with cardiac events were older and leaner, had a greater NYHA functional class, more frequently had a history of lethal ventricular tachyarrhythmias, and had lower levels of eGFR and hemoglobin when compared with those without cardiac events (Table 1). Figure 1 shows the measurements of the phase SD and MIBG HMR in two typical cases. Case 1 with LVEF of 32% had nearly normal LVMI (97 g/m 2 ) and HMR (2.66), and a relatively small phase SD (19.0 degrees), and had no cardiac event during the follow-up period ( Figure 1, Case 1). In contrast, Case 2 had markedly reduced LVEF (28%) and HMR (1.48), had increased LVMI (178 g/m 2 ) and phase SD (40 degrees), and the patient died of progression of HF ( Figure 1, Case 2). In addition to echocardiographic functional abnormalities, patients with cardiac events had both greater phase SD and LVMI and lower HMR than did those without cardiac events: phase SD, 37.4 ± 10.8 vs. 33.1 ± 9.9 degrees, p < 0.0001; LVMI, 129.3 ± 35.7 vs. 117.9 ± 28.7 g/m 2 , p < 0.0001; MIBG HMR, 1.71 ± 0.39 vs. 1.99 ± 0.48, p < 0.0001 (Table 2). In the multivariate Cox analysis, phase SD, LVMI, and MIBG HMR were identified as significant independent variables, as well as age, NYHA functional class, BNP/NT-pro BNP, eGFR, amiodarone use, and LV ESV (Table 3).
Using cutoff values determined by the ROC analysis (Supplementary Figure S1), highrisk HF categories were clearly discriminated from low-risk categories as follows. Patients with a phase SD greater than 37 degrees, MIBG HMR less than 1.89, or LVMI more than 122.9 g/m 2 had significantly lower event-free rates than the other patients ( Figure 2). The combined use of LVMI or phase SD with MIBG HMR more clearly discriminated patients with greater CV risks from others ( Figure 3). LVMI > 122.9 g/m 2 , Phase SD > 37 degrees, and MIBG HMR < 1.89 were incrementally related to increases in CV rates, leading to the lowest CV-free curve when all of the three abnormalities were accumulated ( Figure 4). Values are shown as means ± one standard deviation, MI-myocardial infarction; PCI-percutaneous coronary intervention; CABG-coronary artery bypass grafting ; ICD-implantable cardioverter-defibrillator ; CRT-cardiac resynchronization therapy ; eGFR-estimated glomerular filtration rate ; NYHA-New York Heart Association Classification; ACE-I-angiotensin-converting enzyme-inhibitors; ARB-angiotensin-receptor blockers; ns-no significance.       (c) Phase SD and LVMI   Furthermore, when analyzed separately in heart failure groups with male and female, age (65 years old older, or younger), LVEF (reduced EF (EF < 40%) or mid-range EF (EF ≥ 40%)), the similar Kaplan-Meier event-free curves were obtained, respectively. (Supplementary Figure S2A-H).
When analyzed separately in heart failure groups with and without ischemic etiology, similar Kaplan−Meier event-free curves were obtained, strongly indicating the critical values of the accumulation of the three prognostic variables ( Figure 5).
When the significant multivariate variables, including clinical prognostic biomarkers (Table 3), were combined, LVMI, phase SD, and MIBG HMR showed synergistically increased prognostic values maximally, with a Chi-square value of 188.8 ( Figure 4).
When analyzed separately in heart failure groups with and without ischemic etiology, similar Kaplan−Meier event-free curves were obtained, strongly indicating the critical values of the accumulation of the three prognostic variables ( Figure 5).  Furthermore, when analyzed separately in heart failure groups with male and female, age (65 years old older, or younger), LVEF (reduced EF (EF < 40%) or mid-range EF (EF ≥ 40%)), the similar Kaplan-Meier event-free curves were obtained, respectively. (Supplementary Figure S2A-H).
When analyzed separately in heart failure groups with and without ischemic etiology, similar Kaplan−Meier event-free curves were obtained, strongly indicating the critical values of the accumulation of the three prognostic variables ( Figure 5).
When the significant multivariate variables, including clinical prognostic biomarkers (Table 3), were combined, LVMI, phase SD, and MIBG HMR showed synergistically increased prognostic values maximally, with a Chi-square value of 188.8 (Figure 4).
When the significant multivariate variables, including clinical prognostic biomarkers (Table 3), were combined, LVMI, phase SD, and MIBG HMR showed synergistically increased prognostic values maximally, with a Chi-square value of 188.8 (Figure 4).

Discussion
Impairment of cardiac sympathetic innervation was shown to be independently and incrementally related to lethal cardiac outcomes in combination with three-dimensional LVMI and LVMD, contributing to a further improvement in the risk-stratification of patients with HFrEF evaluated conventionally.

Assessment of LVMI and LVMD
The presented findings strengthened the prognostic values of LVMI and LVMD, which are already known. Methodologically, there are several rationales for the use of threedimensional data for LVMI and LVMD. Morphological and functional left ventricular remodeling occurs globally and is responsible for the manifestation and exacerbation of HF, leading to lethal outcomes. Electrical remodeling assessed by QRS complex duration and two-dimensional assessment are surrogated but limited, as in this point of view. In addition to conventional cardiac functional parameters, the recent widely available technology that was used in this study enabled semi-automatic quantification of LVMI and LVMD at a routine nuclear cardiology practice, without any additional cost and imaging time. These advances in three-dimensional analysis enable more precise risk-stratification of HF patients and overcome the technical limitations of conventional methods [9][10][11], possibly contributing to a more cost-effective diagnostic strategy in combination with an assessment of myocardial ischemia and viability for the differentiation of HF etiology [4].

Assessment of Cardiac Sympathetic Innervation
The prognostic values of cardiac neuronal remodeling assessed by the cardiac 123 I-MIBG activity have been consistently shown in the past two decades [16,17,20]. It has, however, been desirable to standardize the quantitative method for cardiac 123 I-MIBG activity as HMR to cancel the effects due to methodological variations such as collimators used in facilities [18][19][20]. It is important to compare MIBG data among different facilities or among different investigations. The method presented for the standardization of MIBG HMR to a medium-energy collimator condition [21] can be easily performed by using widely available computer software (smart MIBG software, Tokyo, Japan). When using standardized MIBG HMR data in the previous multicenter studies [18][19][20], this method identified identical prognostic values of HMR, contributing to a more extensive use of the MIBG index for the prediction of long-term outcomes. This study used a standardized HMR of MIBG activity. The standardization technique was established by the cross-calibration phantom experiments with various collimators available clinically in Japan and Europe [21,22]. The standardization can minimize differences in MIBG HMR originated in variations of low-energy, low-medium-energy, and medium-energy collimators. Furthermore, our recent study shows that this standardization method can be applied in cardiac event prediction models [23]. Prognostic values of cardiac autonomic innervation assessed by standardized HMR are also shown in HF patients with preserved or mid-range LVEF, as well as in HF patients with reduced LVEF [24]. Thus, the standardization technique in cardiac MIBG imaging can contribute to facilitating the wide-spread use of the neuroimaging in a clinical setting and in a multicenter study for precise identification and monitoring of life-threatening cardiac conditions.

Clinical Implications
The synergistic prognostic values of the neurohumoral, structural, and electromechanical remodeling's demonstrated here are conducive to deepening understanding of HF pathophysiology and to the development of more effective management strategy of heart failure. Improvement in HF risk assessment by the presented method can greatly reduce under-and/or over-estimations of cardiac mortality risks, both of which might have hindered clinically appropriate and cost-effective selection of HF management [25]. This includes therapies using devices such as ICD, CRT, and CRTD and transcatheter mitral valve repair, which have been shown to incrementally improve prognosis beyond standard neurohormonal treatments in a certain subgroup of HF patients. The integration of LVMD and cardiac sympathetic function assessment into conventional prognostic parameters might be helpful for selecting HF patients at increased risk who can benefit more definitively from advanced device therapy. Recently, we developed a machine learning (ML)-based risk model using 123 I-MIBG HMR to differentially predict cardiac death mode (pump failure death versus sudden/arrhythmic death) in HF patients [23]. The ML method can easily analyze various sorts of clinical, cardiac functional, and autonomic functional variables to identify the best model for risk-stratification and subsequent selection of the therapeutic strategy. Thus, a better HF risk-based approach has the potential for establishing a clinically reasonable health-care management strategy for treatment and prevention of unfavorable clinical outcomes, while corresponding appropriately to growing medical costs.

Study Limitations
Despite the entry and exclusion criteria in this study, the possibility of selection bias cannot be completely ruled out. This is because of the features such as an observational cohort study. Patients with systolic dysfunction who had undergone imaging studies for assessment of LVMI, LVMD, and cardiac sympathetic innervation were analyzed. It is necessary to establish better risk-based prophylactic and therapeutic strategies for patients who were identified to be at more increased risk for cardiac mortality by our method. It is also necessary to clarify effects of novel drugs recently available on LVH, LVMD, and cardiac sympathetic innervation in HF patients at greater risks. This study did not include any patients who had major valvular heat diseases responsible for advanced heart failure such as aortic stenosis (AS) or mitral regurgitation (MR), who may had been indicated for transcatheter aortic valve implantation (TAVI) or Mitra Clip ® treatment. Because of the possibilities of cardiac MIBG imaging for identifying the therapeutic effects of TAVI or Mitra Clip ® [26,27], a future study will clarify the therapeutic effects on cardiac sympathetic innervation and function in correlation with long-term clinical outcomes. Despite the possible prognostic values of right heart information in left-sided heart failure [28,29], there was no right-heart data available in all patients of this study. Besides technical difficulties in the assessment particularly in patients with advanced heart failure and left ventricular dilatation, the presented study aimed to correlate cardiac MIBG acidity quantified as standardized HMR with left ventricular functional parameters. Cardiac MIBG tomographic imaging enables identifying regionally denervated but viable myocardium. However, as patients with advance heart failure often have profound and global impairment of cardiac sympathetic innervation, it is actually impossible not only to visually identify cardiac MIBG activity but also to reconstruct cardiac MIBG activity using tomographic data. Therefore, MIBG tomographic data were not obtained in all patients enrolled for this study. Finally, cost-effective analysis is necessary when both myocardial gated perfusion SPECT and cardiac MIBG studies are performed in HF patients. In addition to the assessment of myocardial ischemia, viability, and left ventricular function, this study clearly showed that conventional myocardial perfusion gated SPECT study can be utilized to provide novel biomarkers that improve the risk-assessment of HF patients. Importantly, the quantitative information on LVH and LVMD presented here was obtained without any additional cost or time in a standard gated SPECT protocol. These findings indicate the possibilities that this method can not only improve the cost-effectiveness of a standard gated SPECT study per se, but also rationalize a total management cost in HF patients by reducing unnecessary or ineffective interventional approaches by more precise risk-stratification in HF patients. In this context, further study is needed to establish a more appropriate risk-based strategy using the imaging modalities from a prognostic and economic perspective.

Conclusions
Impairment of cardiac sympathetic innervation as well as three-dimensionally assessed left ventricular hypertrophy and mechanical dyssynchrony have critical prognostic values. The independent and synergistic interactions of these biomarkers on long-term outcomes indicate not only pathophysiological implications in the development and progression of heart failure, but also the clinical significance to help physicians more appropriately risk-stratify and select better management strategy in HF patients.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/jcm10215047/s1, Figure S1: Phase SD and LVMI, standardized late HMR in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events, Figure S2: (A). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the ischemic etiology group. (B). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the non-ischemic etiology group. (C). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the group of males. (D). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the group of females. (E). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the group < 65 years old. (F). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the group ≥ 65 years old. (G). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the group with EF < 40%. (H). Phase SD and LVMI in three dimensions cut-off value determined by ROC analysis for the prediction of cardiac events combination with three parameters in the group with EF ≥ 40%. Table S1: CONSORT 2010 checklist of information to include when reporting a randomised trial.