Our study showed that SAI QRST quantifies adverse electrical remodeling post-MI, which is characterized by wide QRS, increased magnitudes of spatial QRS and spatial T vectors, J-point deviation, and QT prolongation. We found that in MADIT II post-MI patients with and without BBB, increased SAI QRST independently predicted both VT/VF/SCD and all-cause death, in both the ICD arm and the conventional medical therapy arm. Enlarged SAI QRST was associated with increased risk after adjustment for all composites of the MADIT II risk score (age >70 y, BUN>25 mg/dL, QRS duration >120 ms, NYHA HF class>II, atrial fibrillation). Importantly, increased SAI QRST predicted outcomes in patients with and without BBB. This study showed that enlarged SAI QRST was associated with increased risk of sustained VT/VF, as well as non-sudden cardiac death and HF hospitalizations, but not hospitalizations due to recurrent MI (). Thus, the major finding of our study is that adverse electrical remodeling as quantified by increased SAI QRST carries independent risk of ventricular tachyarrhythmias, HF progression, and non-sudden cardiac death in post-MI patients with systolic dysfunction.
SAI QRST as a Marker of Adverse Electrical Remodeling Post-MI
We proposed SAI QRST as a metric, which quantifies post-MI electrical remodeling. Previous experimental and clinical studies showed development of structural remodeling post-MI 
, characterized by dilatation and distortion of LV cavity shape, apoptosis, thinning, fibrosis and hypertrophy of remote from infarction sites 
, scar formation and gradual remodeling of the scar 
in the central and border zone areas 
, dilatation of LV cavity and/or other heart chambers, activation of the neurohumoral pathways, and left ventricular (LV) systolic and diastolic dysfunction. Larger infarcts tend to show greater increases in cell size in the non-infarcted areas, compared to smaller infarcts 
. Development of regional hypertrophy post-MI is associated with enhanced regional heterogeneity of repolarization and steeper local gradient in spatial dispersion of repolarization, which is a well-known substrate vulnerable to ventricular arrhythmias. Complex structural and electrical remodeling post-MI result in a slowing and fractionation of ventricular conduction, dispersion of repolarization, occurrence of delayed afterdepolarizations 
and create conditions for reentry and arrhythmogenesis 
. LV hypertrophy is a strong, independent predictor of future cardiovascular events and SCD 
In addition, hibernating myocardium is characterized by regional myocyte hypertrophy as well 
. Altered calcium uptake in the sarcoplasmic reticulum, inhomogeneity in sympathetic activation, and increase in interstitial connective tissue might lead to extremely high risk of SCD 
. Despite significant functional improvement of hibernating myocardium after revascularization, inhomogeneity in myocardial sympathetic innervations might persist 
. Thus, increased SAI QRST might be a marker of hibernating myocardium or residual structural remodeling, persisting after revascularization. Further studies are needed to test this hypothesis.
Importantly, as the MADIT II study included post-MI patients only, results of this study should not be simply extrapolated on other populations. Further studies are needed to answer the question if SAI QRST is predicting electrical remodeling in the ventricles with different underlying pathophysiology.
Ultimately post-MI remodeling of cardiac structure is characterized 
by changes in (1) cell geometry (size and shape), (2) gap junctions (distribution and conductivity), and (3) interstitial space (size and distribution). These changes could result in increased curvature of the wavefront 
and significant regional slowing of conduction. Slowing of conduction is a fundamental condition for reentrant circuit: the lower the conduction velocity, the smaller the area in which a reentrant circuit can occur 
In our study, the largest SAI QRST was observed in patients with recent MIs (<3 month, ), when hypertrophy had not been developed yet. We speculate that the Brody effect is responsible for significantly enlarged SAI QRST in early (<3 months) post-MI adverse remodeling. The Brody effect explains that heart dipoles oriented perpendicular to the assumed spherical border zone between high- and low- conductance layers are effectively enhanced, resulting in larger surface potentials 
. On the other hand, those dipoles tangential to the high-low conductivity tissue border are effectively diminished, resulting in lowered surface potentials 
. Progressive enlargement of the infarcted ventricle and thinning and distension of the LV cavity can result in a near-spherical shape of LV and elicit the Brody effect on the blood-tissue border. Future theoretical studies are needed to prove this hypothesis.
Correlation between the Time after Myocardial Infarction and SAI QRST
Obviously, multiple confounding factors (age of patient, size of the scar, preexisting LVH and other significant factors, listed in ) affect the significant nonlinear relationship between the time after MI and SAI QRST. However, all these factors taken together helped to reconcile SAI QRST as possibly being a time-dependent marker of post-MI electrical remodeling. Prospective longitudinal study is needed to prove this hypothesis.
Contradictory Findings in SAI QRST
Surprisingly, in our previous preliminary analysis of an ongoing prospective study of patients with primary prevention ICD diminished, but not increased SAI QRST was associated with increased risk of VT/VF with appropriate ICD therapies 
. Our data were obtained in somewhat different patients populations. In our earlier study patients with both ischemic and non-ischemic cardiomyopathy were enrolled, whereas MADIT II enrolled post-MI patients only. More females and African-Americans were included in our prior study. Differences with current data can be ascribed to these differences in studied populations, but importantly, SAI QRST was a strong, independent predictor of outcomes in both, independently performed, studies. We currently can only speculate on the underlying mechanism. One possible explanation came from modeling studies, which showed that LV hypertrophy could be presented by both diminished and increased magnitudes of spatial QRS and T vector 
. It is possible that different underlying pathophysiology might result in a different pattern of SAI QRST changes. Specifically, differences in the activation distribution and in the heart and body surface geometry can be due to underlying pathophysiology. This hypothesis is indirectly supported by the experimental studies, which identified signaling pathways, affecting some, but not all features of remodeling in different disease models 
. Further theoretical and clinical studies are needed to clarify this intriguing discrepancy. Future large population studies might help to detect U-shaped risk associated with SAI QRST.
Risk Stratification in Post-MI Patients with BBB
Analysis of ECG in BBB is challenging and frequently inconclusive. Therefore, our finding of strong, independent predictive value of SAI QRST in BBB patients is clinically important and promising. In this study, SAI QRST predicted outcomes after adjustment for BBB status and QRS duration. Interestingly, in BBB patients, increased SAI QRST more strongly predicted sustained VT/VF events with appropriate ICD therapies, rather than mortality. At the same time in patients without BBB, SAI QRST predicted mortality rather than appropriate ICD therapies.
ECG Predictors of Outcomes in MADIT II
Multiple ECG markers of SCD risk have been tested in MADIT II data analysis, but only a few demonstrated independent predictive value after adjustment for clinical predictors of outcomes. Increased repolarization lability (QT variability and T wave variability) has been shown to be significant, independent predictor of appropriate ICD therapies in the ICD arm 
, but did not predict SCD in the conventional medical therapy arm. Frequent ventricular premature beats (>3/10 min) were associated with death in the conventional medical therapy arm and with appropriate ICD therapy in the ICD arm 
, whereas heart rate turbulence was not predictive. In post-MI patients with resolved Q waves, fragmented QRS was associated with increased risk of cardiac events 
, but was not associated with cardiac death independently of Q waves. Prolonged QRS duration predicted SCD in the conventional medical therapy arm 
, but did not predict VT/VF/SCD in the ICD arm. Thus far, SAI QRST seems to be the only ECG predictor of both VT/VF/SCD and total mortality in both MADIT II arms, which underscores the importance of post-MI adverse electrical remodeling in the pathogenesis of arrhythmia and heart failure progression.
This was a retrospective analysis of prospectively collected data of randomized controlled trial MADIT II, and, therefore, statistical power of subgroups analysis was limited. Patients without an available ECG have been excluded from this analysis. However, absence of statistically significant interaction between SAI QRST and BBB status in the Cox model suggests a similar effect of SAI QRST on survival in patients with and without BBB. Validation of predictive value of SAI QRST in a prospective study is needed before its implementation into clinical practice.
As previously shown, appropriate ICD therapies may overestimate the frequency of SCD 
. However, study of the predictors of ventricular arrhythmia is critical for understanding SCD mechanisms and developing mechanistically- sound risk markers of SCD.
As MADIT II study population is predominantly presented by white males, we did not apply race- and gender-specific thresholds for SAI QRST in this study. At the same time, voltage ECG parameters are known to be gender-specific 
and further development of gender- and possibly race-specific thresholds of SAI QRST will be needed in the future.
SAI QRST is significantly different in patients with BBB as compared to those without BBB 
. Therefore, threshold of SAI QRST that is specific for ventricular conduction status might demonstrate better precision in prediction of outcomes. However, in this study we utilized one uniform threshold of SAI QRST for patients with and without BBB to simplify its implementation.
In this study only baseline ECGs were analyzed. Longitudinal changes in SAI QRST over post-MI time course in correlation with longitudinal changes in LV morphology, assessed by imaging, are unknown. Future theoretical and prospective longitudinal clinical studies are needed to confirm whether or not SAI QRST is indeed a marker of electrical remodeling.
Adverse electrical remodeling as quantified by increased SAI QRST carries independent risk of ventricular tachyarrhythmias, HF progression and non-sudden cardiac death in post-MI patients with systolic dysfunction. In post-MI MADIT II patients with and without BBB, increased SAI QRST independently predicted both VT/VF/SCD and all-cause death, both in the ICD arm and in the conventional medical therapy arm. Enlarged SAI QRST was associated with increased risk after adjustment for all composites of the well-known MADIT II risk score (age >70 y, BUN>25 mg/dL, QRS duration >120 ms, NYHA HF class>II, atrial fibrillation).
The development of our novel ECG marker of adverse electrical remodeling, SAI QRST, and demonstration of its predictive value for SCD/VT/VF and all-cause mortality in a blinded analysis of landmark ICD trial MADIT II opens up new opportunities for SCD risk stratification. Future studies of SAI QRST are warranted and might uncover novel mechanisms of electrical remodeling in the post-infarction period.