Sudden cardiac death after myocardial infarction.

Abstract

This article refers to ‘Predictors of sudden cardiac death in high-risk patients following a myocardial infarction’ by K.F. Docherty et al., published in this issue on pages 848–855. Sudden cardiac death (SCD) accounts for 15–20% of global mortality and coronary heart disease is the most common underlying substrate.1 The weeks to months following myocardial infarction (MI) have been recognized as a particularly vulnerable period, where the absolute rate of SCD is acutely elevated.2 Despite this observation, randomized trials of implantable cardioverter-defibrillator (ICD) therapy in the immediate post-MI period have not demonstrated an overall survival benefit in high-risk patients with left ventricular systolic dysfunction (LVSD).3 Whether additional strategies for SCD risk enrichment could identify post-MI patients who may benefit from sudden death prevention interventions remains unknown. In this issue of the Journal, Docherty and colleagues4 sought to develop a risk score for the prediction of SCD in patients with acute MI associated with heart failure and/or LVSD. The analytic cohort included 13 202 patients from three randomized trials of pharmacotherapy interventions (β-blocker, angiotensin receptor blocker, or aldosterone receptor blocker) in this population. Employing a competing risk framework with established candidate predictors of SCD, the authors derive a risk score using pooled data from the Effect of Carvedilol on Outcome After Myocardial Infarction in Patients with Left Ventricular Dysfunction (CAPRICORN)5 and Valsartan in Acute Myocardial Infarction (VALIANT)6 trials. Significant predictors of SCD were then scaled to an integer-based risk score, which was validated in the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (EPHESUS).7 The authors identify a range of SCD risk factors that have been previously reported including age, elevated heart rate, active smoking, left ventricular ejection fraction (LVEF) ≤30%, atrial fibrillation, prior MI, heart failure, diabetes mellitus, chronic kidney disease, and absence of revascularization at index MI.1, 8 The SCD risk model demonstrated reasonable discrimination for death subtype in both derivation and validation cohorts (C-index 0.70–0.72) and was well calibrated when assessed over 2 years of follow-up. Likewise, the SCD risk model improved continuous reclassification of SCD risk at 2 years when compared to LVEF ≤35% alone. In the validation cohort, the gradient of 2-year SCD risk ranged from 3.4% to 13.4% in the lowest and highest risk subgroups, respectively. Shorter-term SCD risk gradients in the derivation cohorts ranged from 0.5–3.4% (at 40 days) and 0.7–5.1% (at 90 days). While the authors should be congratulated for presenting this large, collaborative analysis with rigorous evaluation of SCD risk prediction, there are several caveats to consider when evaluating the clinical implications of these findings. First, in the cohorts assessed,5-7 less than 50% underwent revascularization and the prevalence of contemporary pharmacotherapy for MI and LVSD was low. Indeed, all of the pharmacotherapy interventions employed in the considered trials (β-blockade, angiotensin receptor blockade, mineralocorticoid receptor blockade) have each been associated with a reduction in SCD.9 Second, in the contemporary era of secondary prevention pharmacotherapy and percutaneous intervention, less than 10% of patients with acute MI develop an LVEF <35%.10 Taken together, the SCD risk score developed in this study applies only to a significant minority of contemporary post-MI patients and the identified event rates would likely need to be scaled downward across all subgroups before contemporary use. Second, it is important to consider the diversity of mechanisms of SCD in the post-MI population. As highlighted in an autopsy-based study of VALIANT,11 the majority of adjudicated sudden death in the first month after MI was related to recurrent MI and/or myocardial rupture, whereas the majority of adjudicated sudden death after 90 days was related to presumed ventricular arrhythmia. To the extent that the mechanism of SCD is relevant to its prevention (e.g. ICD therapy for prevention of arrhythmic SCD), it will be important to further understand the specificity of the reported SCD score for arrhythmic vs. non-arrhythmic SCD. While the definition of SCD was relatively harmonized across the trials evaluated in this study,5-7 there was no distinction in the SCD endpoint between arrhythmic vs. non-arrhythmic mechanism. Validation of the score in a post-MI cohort with continuous rhythm surveillance [e.g. Defibrillator in Acute Myocardial Infarction Trial (DINAMIT)],3 would better discern the potential utility of the risk score to identify ICD-remediable SCD. Third, while the authors should be congratulated for employing a competing-risk framework in the derivation of their risk score, the use of a competing risk model does not exclude the association of SCD risk factors with non-SCD mortality. Put plainly, patients at increased risk for SCD are also likely to be at increased (and numerically excessive) risk of non-SCD mortality. As a consequence, patients at increased absolute risk for SCD are not necessarily at increased proportional risk of SCD. This discordance was present for the SCD subgroups in this study wherein the proportional risk of SCD in the ‘highest risk’ subgroups (33–35% across all time points) was numerically lower than the proportional risk of SCD in the ‘lowest risk’ subgroups (40–42% across all time points). To that end, while the derived SCD model identifies patients at increased absolute risk for SCD, it does not follow that these patients necessarily would benefit from ICD therapy. In fact, when evaluating the implications of absolute and proportional SCD risk, we and others have found that projected ICD benefit is relatively insensitive to absolute SCD risk, whereas it is exquisitely sensitive to proportional risk.12, 13 In this formulation, the proportional risk of SCD serves as a ‘rheostat’ for ICD benefit at any given absolute SCD rate (Figure 1).13, 14 For example, for patients deemed ‘highest-risk’ by the study's SCD risk score (i.e. 15% SCD risk at 2 years), the projected impact of ICD therapy on all-cause mortality could range from null to a 50% relative reduction depending on the prevalence of non-SCD competing risk. After integrating the observed proportional SCD risk at 2 years in this ‘highest risk’ subgroup (33%), we would project a limited survival benefit to ICD therapy in this subgroup (Figure 1). In keeping with this projection, several SCD risk factors in the derived risk score (e.g. older age, chronic kidney disease, diabetes mellitus) have also been associated with non-SCD mortality and further associated with attenuated ICD benefit in previous primary prevention ICD trials.15 Finally, it is worth noting that risk scores derived using relatively short-term follow-up, as was present in this study, are likely to underestimate the potential ‘life years saved’ from ICD therapy in patients at low absolute but high proportional risk of SCD over a longer time horizon (e.g. younger patients).12 Taken together, this study from Docherty and colleagues4 represents an important first step in enhancing the paradigm of SCD prevention in patients with acute MI. Looking ahead, there remain several avenues of future investigation. For example, during the immediate post-MI period where SCD risk is transient and highest, can further refinement of mechanism (arrhythmic vs. non-arrhythmic) identify candidates for short-term protective strategies, like the wearable cardioverter-defibrillator, as the salutary effects of intensive pharmacotherapy and revascularization accrue? In the longer-term, given emerging evidence that the arrhythmogenic substrate following MI evolves over the course of months to years, it is likely that robust SCD models will need to incorporate more dynamic assessment of SCD risk over time. Indeed, previous analysis of the VALIANT study has shown that clinical predictors of SCD in the near-term following MI differ from predictors of SCD in the long-term.16 Furthermore, as the authors demonstrate,4 the presence of a low LVEF (<35%) in patients with a history of MI is a relatively poor discriminator of death subtype. To that end, the incremental value of markers specific to the pathophysiology of ventricular arrhythmias (e.g. ventricular scar phenotyping, repolarization heterogeneity) warrant further examination. Other strategies including the use of programmed ventricular stimulation as a means to risk stratify patients with acute MI are currently being evaluated.17 Finally, as we consider a broader population strategy to reduce the global burden of sudden death, robust models for both absolute and proportional SCD risk will be meaningful. These models will be critical for the design of sudden death clinical trials and thoughtful evaluation for ICD candidacy. Ultimately, we envision a future where patients and physicians can readily access tools that translate the concepts of absolute and proportional SCD risk into actionable information for decision-making at the bedside. Conflict of interest: W.C.L. is a consultant to Medtronic, Impulse Dynamics, PharmIN, clinical endpoint committee member for CardioMems trials (Abbott, Baim Institute) and EBR Systems Inc, steering committee member for ADMIRE ICD (GE Healthcare), Cadiac Dimensions, and Respircardia. The University of Washington CoMotion holds the copyrights for the Seattle Heart Failure Model and the Seattle Proportional Risk Model and has received license fees from various companies. N.A.C. has nothing to disclose.