New Concepts in Sudden Cardiac Arrest to Address an Intractable Epidemic: JACC State-of-the-Art Review

Abstract

Sudden cardiac arrest (SCA) is one of the largest causes of mortality globally, with an out-of-hospital survival below 10% despite intense research. This document outlines challenges in addressing the epidemic of SCA, along the framework of respond, understand and predict, and prevent. Response could be improved by technology-assisted orchestration of community responder systems, access to automated external defibrillators, and innovations to match resuscitation resources to victims in place and time. Efforts to understand and predict SCA may be enhanced by refining taxonomy along phenotypical and pathophysiological "axes of risk," extending beyond cardiovascular pathology to identify less heterogeneous cohorts, facilitated by open-data platforms and analytics including machine learning to integrate discoveries across disciplines. Prevention of SCA must integrate these concepts, recognizing that all members of society are stakeholders. Ultimately, solutions to the public health challenge of SCA will require greater awareness, societal debate and focused public policy.

Keywords: ECG; acute coronary syndrome; cardiopulmonary resuscitation; heart failure; informatics; machine learning; sudden cardiac arrest.

Central Illustration.. An Emerging Digital Network Which Could Address Pathophysiological, Clinical and Infrastructural Gaps in SCA.

In the current model, there is an undefined yet often long period between an anticipatory medical visit, typically for non-specific complaints, and time zero of arrest (arrow). Resuscitation is often the first medical contact. In a potential future model, SCA care could be improved in 3 areas backwards in time from time zero of arrest. (A) Response may be coordinated by the digital infrastructure. Smartphone Apps can alert first responders to event and GPS location; wearable biometric sensors can sense antecedent warnings that occur in one half of victims; both may facilitate rapid dispatch of responders, AEDs and EMS. (B) Understand and Predict may be assisted by defining novel SCA phenotypes from next-generation registry databases, that upload data from wearable biometrics and existing devices (an internet of things for SCA) in the hours and minutes preceding time zero, and upstream and downstream clinical data. Such databases will enable analytics to improve prediction of high-risk individuals, who could be identified or treated e.g. at an anticipatory visit. (C) Prevent. Upstream efforts at prevention guided by registry data that prospectively includes clinical data plus non-invasive autopsy, genetic profiles and real-time-biometrics across organ systems. Emerging smart-analytics such as machine learning hold promise to reveal novel risk phenotypes and predictors as targets for prevention.

Figure 1.. Sudden Cardiac Arrest: Hypothesized Axes of Risk.

Applied to populations, axes plot the severity of each comorbidity, while dashed ellipses are stylized representations of the prevalence of SCA in populations with each overlapping combination of comorbidities. Applied for personal risk-stratification, areas represent the SCA risk for an individual with those specific comorbidities. (A) General risk along proposed major axes of prior resuscitated SCA, cardiac comorbidity, family history and non-cardiac comorbidity. (B) Detailed axes further dissect cardiac comorbidities into reduced LVEF and other structural disease, add granularity on non-cardiac comorbidities, and expand family history into inherited arrhythmia syndromes and less defined heritability. New at-risk phenotypes may be defined by new analytics, such as unsupervised cluster analysis and supervised machine learning in defined populations. The precise shapes of these plots remain to be defined in various populations.

Figure 1.. Sudden Cardiac Arrest: Hypothesized Axes of Risk.

Applied to populations, axes plot the severity of each comorbidity, while dashed ellipses are stylized representations of the prevalence of SCA in populations with each overlapping combination of comorbidities. Applied for personal risk-stratification, areas represent the SCA risk for an individual with those specific comorbidities. (A) General risk along proposed major axes of prior resuscitated SCA, cardiac comorbidity, family history and non-cardiac comorbidity. (B) Detailed axes further dissect cardiac comorbidities into reduced LVEF and other structural disease, add granularity on non-cardiac comorbidities, and expand family history into inherited arrhythmia syndromes and less defined heritability. New at-risk phenotypes may be defined by new analytics, such as unsupervised cluster analysis and supervised machine learning in defined populations. The precise shapes of these plots remain to be defined in various populations.

Figure 2.. Changing Incidence and Pathophysiology of Sudden Cardiac Arrest In Recent Decades.

(A) Falling Incidence of Sudden Cardiac Arrest (from New Engl J Med, "Declining Risk of Sudden Death in Heart Failure", Shen, L., et al. (2017). 377(18): 1794-1795 Copyright (c) 2017 Massachusetts Medical Society Reprinted with permission from(117)); (B) Changing Arrhythmia Presentations of SCA (with permission from (43)).

Figure 3.. Competing risk factors for mortality.

(A) Relative benefit of the ICD over conventional medical therapy based on number of risk factors: higher NYHA class, end-stage renal disease, atrial fibrillation, QRS widening and diabetes mellitus, confer a J-shaped benefit for the ICD. From the MADIT-2 trial (with permission from (8)). (B) Cox proportional hazards model of the ICD against Seattle Heart Failure Model (SHFM) predicted risk of SCA (as a continuous variable) using the Seattle Proportional Risk Model (SPRM). As SHFM predicted-risk increases, the ICD hazard ratio for death becomes more favorable (with permission from (65)).

Figure 4.. Disease Contributions for Forms of Heart Failure, Along Observed Phenotypic Axes of Risk.

Specific axes code for cardiac, non-cardiac pathology and demographics. The shape of the clinical area profiles in these plots represents individual patient phenotypes (with permission from (69)).