New Concepts to 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 AEDs, 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.

Condensed Abstract:

Sudden cardiac arrest (SCA) is one of the largest causes of mortality with a dismal survival rate <10 %. We outline challenges in the areas of Respond, Understand and Predict, and Prevention for SCA. Response may be improved at the level of community response, AED access and by new technologies that match resources to victims. Understanding and predicting SCA may be improved by a new taxonomy along phenotypic/physiological axes of risk to identify less heterogeneous cohorts, enhanced by open-access data platforms and machine learning. Prevention of SCA, ultimately, requires greater awareness and societal debate, funding and focused public policy.

Introduction

Sudden cardiac death (SCD) is one of the largest causes of mortality and healthcare utilization in the world and, for many victims, their first and last contact with the healthcare system. Sudden cardiac arrest (SCA) affects 150,000 to 450,000 individuals per year in the U.S. alone (1), most of whom would not be considered at high-risk a priori (2), of whom approximately 90% suffer SCD (i.e. about 10 % survive SCA)(3). The causes for SCA are imprecisely defined, and evolving. Most cases still reflect ventricular arrhythmias, although increasing numbers reflect bradycardia and pulseless electrical activity(2) which have lower survival(4,5). It is unclear if this reflects trends for reduced mortality from coronary disease(4,5), but such trends have been more modest in SCA(2) despite advances in basic, translational and population science.

This state-of-the-art review identifies knowledge gaps in SCA, focused on unique avenues for therapy. We recognize the enormous strides made by past and current investigators but also the substantial challenges ahead. A central theme is the need for a broader taxonomy that recognizes SCA not as a disease entity but as the terminal event of several processes not limited to cardiovascular pathology (Figure 1). Current guidelines for SCA(6) identify patients mostly with cardiac phenotypes (7), yet non-cardiac comorbidities worsen outcome(8), and most victims have undefined pathology (9,10). Classifying SCA by organ-system comorbidities may also foster inter-disciplinary science and technology innovation. The second theme is creation of a digital infrastructure (Central Illustration) to translate scientific knowledge between multiple diverse specialty groups into unified management strategies for SCA. The digital network could use continuous biometric data and analytical tools including machine learning to better identify at-risk phenotypes, use these data and online platforms(11) to create actionable registries for allocating resources, then apply ubiquitous GPS devices to coordinate first response and emergency medical services (EMS) and advance resuscitation. Patient outcomes could then be used to reinforce or revise the network as appropriate. Such a framework may require substantial and novel public-private collaborations and funding initiatives.

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.

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.

The document is structured around 3 key concepts: Respond, Understand and Predict, and Prevent Sudden Cardiac Arrest.

Respond

Despite our best efforts, <20% of victims of out-of-hospital SCA have restoration of spontaneous circulation (ROSC) and, despite recent improvements(12), under 10% survive to hospital discharge (2). Since most SCA victims are unidentified until their collapse, improved resuscitation would have enormous impact. Potential opportunities include recognizing early warnings, electronic transmittal of event and location data to improve the rapidity and efficacy(13) of first responders and EMS networks, educational and policy initiatives, and refining the sciences of resuscitation and post-resuscitation care.

Symptoms and Warning Signs

It is increasingly recognized that SCA is preceded by warning symptoms and signs which, if promptly recognized, could greatly hasten first-response. SCA is often preceded by a non-specific physician encounter days or weeks before the event(14). However, until recently, immediately antecedent events were undefined due to difficulties in collecting data at an arrest. Muller et al(7) dispatched physicians with SCA emergency teams in Germany to interview survivors or bystanders for antecedent events or symptoms. They concluded that SCA is “not as sudden as typically defined”. Three quarters of victims had recognizable symptoms including angina pectoris and dyspnea (Table 1) of which 2/3rds lasted 1 hour or longer(7) of median duration 50 minutes in patients with asystole, 20 minutes before PEA, and 30 minutes before VF(7). In the Oregon Sudden Unexpected Death Study (SUDS), symptoms preceded SCA in half of victims yet only 21% contacted EMS (15). Those who did were more likely to have witnessed arrest, receive CPR, have a shockable rhythm and ultimately survive (32.1% versus 6.0%) (15). This favorable chain-of-survival was more likely in patients alerted to symptoms that were recognizable, such as continuous chest pain or who had a prior cardiac history. Other studies reiterate these themes (10). Symptom awareness is thus likely central to the emerging concept of "near-term prevention" of SCA(16) to bring the window for intervention forward in time.

Table 1.

Antecedent Symptoms For Out-Of-Hospital SCA (adapted from reference(7))

	All Patients with Known Symptoms (N=323)		Patients with Witnessed Arrest (N=274)		Patients with Symptom Duration <1h (N=116)
	N (%)	Duration, min	N (%)	Duration, min	N (%)
Angina Pectoris	88 (22)	120 (20, 630)	69 (25)	120 (15, 495)	38 (33)
Dyspnea	61 (15)	20 (10, 375)	47 (17)	10 (10, 180)	35 (30)
Nausea/vomiting	27 (7)	120 (20, 420)	19 (7)	90 (5, 240)	12 (10)
Dizziness/Syncope	21 (5)	10 (5, 60)	18 (7)	10 (5, 60)	17 (15)
Other	23 (6)	60 (10, 300)	23 (8)	60 (10, 270)	14 (12)
No Symptoms	103 (25)	N/A	71 (26)	N/A	N/A
Unknown	N/A	31 (11)	N/A	N/A	N/A

N/A indicates not available

Duration of symptoms is given in minutes as median and interpercentile range (25%, 75%)

Improved Resuscitation Alert Systems

Since symptoms preceding SCA are often overlooked(15), there is a need for improved early-alert systems for symptoms or potentially signs of SCA. Public education is key and should include traditional efforts, focus groups and social media. Emerging digital technologies may be helpful. Using a smartphone software application (app) that transmits event and location data, lay responders were able to reach 58% of SCA victims, and initiate CPR prior to EMS arrival in 27% of cases (17). Medical or consumer wearable devices have potential to monitor biometrics for antecedent ‘signs’ of SCA, which may be transformational since many arrests are unwitnessed. Sensors could include ECG electrodes, photoplethysmography on limbs or the face (18) to detect pulse, or motion sensors (19) to detect lack of breathing or a sudden collapse. Sensors in smart houses (20), on smart mattresses (21) embedded in garment fabrics (22) or in other locations could network with portable devices to create an internet of things (IoT) for SCA. Although each independent data stream may be non-specific, their combination could be very effective particularly if personalized by adaptive tools including machine learning.

A comprehensive IoT for SCA would require debates on legal and ethical issues, on privacy and security, a consensus on shared communication protocols, and considerations of funding. Nevertheless, near-term focused IoTs are feasible, and include initiatives such as extending remote-management of cardiac implanted devices to include telephone call-back to verify status, provide instructions or to potentially to automatically activate EMS.

Improve Access to Automated External Defibrillators (AEDs)

The availability of AEDs significantly determines local outcome from SCA. Increasing AED numbers is important, but existing resources should also be optimally deployed. In a real-world study of AED availability in Toronto, Canada, Sun et al. found that half of AEDs closest to the location of an SCA were located in offices inaccessible to first-responders when needed(23). One solution for this place/time mismatch would be to locate community AEDs in lock boxes accessed by neighborhood codes. An ongoing community initiative in Progetto, Italy uses a volunteer network to deliver AEDs to the scene of SCA (24). Public policy could substantially improve outcomes but must be carefully designed. On one hand, AEDs could be allocated by residential or workplace density, and risk-stratified, taking precautions to cover less populous areas. Placing AEDs on police, firetruck, for-hire or ride-share vehicles would increase availability, but again will require logistical, ethical and legal discussions.

Rapid-response requires accurate data on AED location, yet this is often obscure. Recent apps provide comprehensive maps of AED locations, which may improve access (25,26) but have yet to be tested formally. Technology could incorporate traffic flow or weather at various times to dynamically optimize response.

Other technological innovations reported to improve AED availability include the use of drones to fly AEDs to sites of SCA(27), with remote audio and video instruction(28) to first-responders. Less exotic approaches include portable, battery operated pocket(29) or other less expensive designs to increase AED access.

Improving First Responder Rate

First-responder availability may be more critical than AED availability. SCA typically occurs in the presence of bystanders, yet under 1 in 7 perform resuscitation (7). In the Home Automated Defibrillator Trial (HAT), this reflected lack of training and factors including the psychological stress of performing CPR(30). Training and education of the public is thus a top priority. Resuscitation efficacy are enhanced by AEDs that provide instructions, or that network automatically to EMS personnel. Combining public CPR(7) with improved EMS access improves survival, but is underused nationally.

It is troubling and unclear why SCA survival rates vary so dramatically even between U.S. urban areas. In Seattle and King County, WA, survival rates after out-of-hospital cardiac arrest are ~19.9 %, while in Detroit the rate is ~3 %(31). Rates of bystander CPR vary with socioeconomic and racial characteristics of neighborhoods(32), which requires addressing public education, EMS resources and organizational limitations. First-response rates can also be affected by mundane issues such as road traffic, for which city planners already fast-track EMS vehicles.

Improving the Science of Resuscitation

The immediate goal of resuscitation is the return of spontaneous circulation (ROSC) via defibrillation and chest compression, which also improve coronary perfusion in animal and clinical(33) studies. Nevertheless, many patients even with ROSC(2,7) have poor survival to discharge, emphasizing the urgent need to identify additional mechanisms.

Targeted temperature management (TTM; therapeutic hypothermia) may reduce damage to the central nervous system as tracked by biomarkers(34), and also reduce metabolic demands or arrhythmias(35). Nevertheless, the role of TTM is under discussion(36) as it has been linked to events such as coronary (stent) thrombosis(37). Early coronary revascularization shows efficacy when combined with a network for rapid triage(38), but since studies question its benefits(39) this could be reserved for patients whose axes of risk include coronary disease (Figure 1).

Several devices have been proposed to improve resuscitation. Mechanical support devices can now be deployed by first responders in cases of SCA due to PEA or asystole, which are increasingly common(2), or when EMS is delayed. However, chest compression devices are controversial and may cause visceral damage (40). Other emerging solutions include electrical stimulation to activate chest wall muscles (41), and simplified methods for cardiopulmonary bypass or extracorporeal circulation.

One may anticipate a spectrum of device complexity for SCA, tailored to individual risk. Consumer wearable technology and apps could be recommended to the general public, while ‘medical-grade’ devices requiring FDA-clearance could be prescribed for higher risk patients (Central Illustration). The existing implanted base of devices may provide useful information in this regard, since even implantable monitors or pacemakers which cannot defibrillate are Bluetooth enabled. Such devices could identify SCA in real time via dedicated closed-loop systems, or networked with other sensors in an SCA focused internet of thing (IoT). This represents a new application of device-device interaction.

Acute resuscitation of the SCA victim does not end with ROSC, and a multidisciplinary team of cardiologists, neurologists, rehabilitation specialists, nurses and others are needed to optimize outcomes(42). Further research into post-ROSC derangements in cardiac, neurological, and metabolic systems, and designing therapy to address them, may improve ultimate survival.

Understand and Predict

Several opportunities exist to improve understanding of SCA. These include defining novel SCA phenotypes or their determinants (comorbid risk factors), identifying cellular-phenotype interactions, leveraging technology innovations such as continuous data collection to enrich data registries, and applying novel analytics to define novel phenotypes.

Epidemiology

SCA affects 180,000 to 450,000 individuals per year in the US alone (1), varying with age from 2.28 per 100,000 under age 35, to ~100 at age 50 and ~600 at age 75. Notably, its epidemiology is evolving and from 1980-2000 an increasing proportion of SCA cases presented as pulseless electrical activity or asystole than VF(4,5,43,44) (Figure 2). This may reflect increasing beta-blocker use(43) or better treatment of coronary disease (4,5).This impacts resuscitation, as survival is lower for non-VF compared to VF arrests (4,5), risk prediction and understanding.

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)).

Epidemiology has been our primary lens into sudden cardiac arrest (SCA), providing key clinical, pathophysiological and therapeutic insights including who is affected, what are their triggers and pathophysiologies, and what therapies work. However, epidemiological data on SCA have been limited in definitions, disease taxonomy, and reporting (ascertainment) that must be addressed to move the field forwards.

Refining Definition and Taxonomy for SCA

Cardiac arrest is defined as the sudden unexpected termination of cardiac activity associated with loss of consciousness, spontaneous breathing and circulation occurring within 24 hours after the onset of symptoms of cardiac origin. This definition is useful to target resuscitation efforts, and for public education campaigns, yet it describes a constellation of terminal symptoms rather than defining specific pathophysiology or patient phenotypes.

SCA is not a monolithic event but a family of mechanisms leading to common pathways of rhythm disturbances (VT, VF, asystole) or hemodynamic failure (as in PEA), as emphasized in recent studies (10). Accordingly, it may be useful to develop structured definitions that stratify cases of SCA by other means, such as by outcome or potentially by organ-system. Definitions should identify non-overlapping categories consistently, and avoid ambiguity ("GiGo", garbage in garbage out). This section discusses some emerging taxonomies which could translate into actionable clinical phenotypes for SCA.

One nascent approach classifies SCA along axes of risk (Figure 1) attempting to unify known comorbidities (Table 2) and mechanisms (Table 3) for SCA. While some SCA phenotypes are relatively well understood, such as VF following acute ST elevation MI, others such as the association of PEA with a history of syncope, lung disease, black race or female gender(2) are less understood – yet rising in prevalence(45). Conditions such as hypertension and diabetes are considered as risk factors for cardiovascular mortality, yet contribute to SCA (Table 3) and could be placed along axes of risk.

Table 2.

Potential Clinical Taxonomy for Sudden Cardiac Arrest

CLINICAL PHENOTYPE	COMMENTS
Coronary artery disease
Acute myocardial ischemia with or without prior infarct	Long duration post MI benefits from ICD. CABG reduces SCA risk in chronic ischemic cardiomyopathy(6).
Acute coronary occlusion (no prior CAD events)	Size of ischemic zone; genetic determinants. Molecular mechanisms relatively well understood(118).
Coronary artery spasm	Diagnosis may be occult. Vasodilator therapy. ICD need unclear. Prognosis worse if atherosclerotic CAD too.
Coronary artery dissection(119)	Female predominance. Arteritis and connective tissue disease such as fibromuscular dysplasia. Diagnosis challenging; IVUS, OCT may help*
Prior myocardial infarction, no acute ischemia	The prior infarct may be clinically evident or silent.
HFrEF (Ischemic cardiomyopathy)	For EF ≤ 35 an ICD is indicated after 40 days(6); further risk stratification may be warranted however. Beta-blockers, ACE-I, statins and other drugs.
HFmEF (ischemic)	Number of SCA may outnumber HFrEF though per patient risk lower; largely unstudied. Beta-blockers, ACE-I, statins and other drugs.
HFpEF (ischemic)	Statins; most pharmacologic trials have been neutral.
Dilated, idiopathic cardiomyopathy
HFrEF (non-ischemic)	For EF ≤ 35 an ICD is indicated after 90 days(6); further risk stratification may be warranted however. Beta-blockers, ACE-I, statins and other drugs.
HFmEF (non-ischemic)	ICD not indicated for primary prevention but SCA risk high; scar may be predictive(87).
HFpEF (non-ischemic)	Statins; most pharmacologic trials have been neutral. *
Other Cardiac
Inherited channelopathy	Sudden Arrhythmic Death Syndrome; Drug-induced QT prolongation.
Hypertrophic cardiomyopathy	Risk factors include severe hypertrophy (3.0 cm), NSVT, abnormal blood pressure response to exercise, LVOT obstruction, LV aneurysm, syncope, suspected VT, FH of SCD, extent of LGE(120).
Inherited cardiomyopathy	Desmosomal proteins (ARVC/ALVC/AC); lamin A/C; other proteins.
Congenital heart disease	Related to VT, VF or also AT with rapid ventricular response; bradyarrhythmias. Cyanotic lesions, Tetralogy, transposition of great vessels, univentricular hearts and Ebstein’s are common causes(6).
Sarcoidosis(121)	May cause AV block, monomorphic or polymorphic VT, atrial arrhythmias, right and/or left ventricular dysfunction. MRI or PET for diagnosis.
Mitral valve prolapse	Bileaflet prolapse; need not be severe(122); outflow and fascicular or papillary muscle ectopics; localized fibrosis near mitral annulus.
Aortic Disease	Aortic Dissection
Subsets at Risk for Non-VF Arrests
At-risk for PEA	PEA more common in women, African Americans, older patients, pulmonary disease, antipsychotic drug use, prior syncope; OHT recipients. Seizure patients with SUDEP may present with PEA and without antecedent seizure. Survival higher than asystole but lower than VF. Women survive PEA > men(45).
At-risk for primary asystole	Nonischemic causes of SCA; dialysis patients; OHT recipients; laryngospasm in SUDEP; pulmonary disease; some ethnicities, e.g. Asians
Other
OSA	Nocturnal SCA predominance; role of stretch, ischemia, autonomic changes, hypoxia.
Neuromuscular Disorders	Myotonic dystrophy; Emery-Dreifuss; limb-girdle; facio-scapulo-humeral; mitochondrial; dystrophinopathies (Duchenne, Becker, other)
Schizophrenia(123)	Antipsychotic drugs, coronary risk factors, DVT/PE, other.
Neurological catastrophe(10)	Intracranial hemorrhage, Sudden unexplained death in epilepsy, aneurysm rupture, acute stroke, other neurological event (e.g. Huntington disease)
Infection(10)	Pneumonia, sepsis, other infections
Metabolic(10)	Occult drug overdose (opiates and nonopiates), hypoglycemia, hyperglycemia, acute renal failure, acute alcohol withdrawal, hypothermia
Gastrointestinal(10)	GI hemorrhage, incarcerated/strangulated hernia, bowel obstruction, hepatorenal failure/pancreatitis; liver failure
Aspiration, Asphyxia(10)
Disseminated cancer(10)
Hypercoagulable States	Pulmonary Embolism

Table 3.

Potential Mechanistic Axes for Sudden Cardiac Arrest

Mechanism	Comment	Reference
Electrophysiological
Reentry	Most common mechanism. Multiple subtypes. Myocardial reentry related to scar; bundle branch reentry; rotors	(118,124)
Repolarization prolongation	Pathophysiology likely related to temporo-spatial dispersion. Long QT; Acute ischemia; potassium channel blockade; enhanced late sodium current.	(118,125)
Repolarization dispersion	Measured as T peak to T end; phase 2 reentry in ischemia or Brugada syndrome; T-wave area dispersion; microvolt T-wave alternans (TWA).	(76,126,127)
Triggered -Early After Depolarizations (EADs)	Long QT syndromes; Heart failure-related Enhanced late inward Na current (INa-L)	(128)
Triggered -Delayed After Depolarizations (DADs)	CPVT; Idiopathic OT and annular VPDs and VT	(129)
Wavebreak	Transition from organized circuits, foci to fibrillation	(89,130)
Myocardial Ischemia
Myocardium	Myocardial infarction; Myocardial ischemia	(6)
Coronary	Atherosclerosis and collaterals; Coronary dissection; Coronary artery spasm; Myocardial- arterial bridge	(131)
Inflammation	Coronary events; QT modulation.	(132)
Mechanical
Stretch	LV dysfunction; mitral valve prolapse; sleep apnea; left bundle branch block	(91,93)
Fibrosis/scar	Post infarct; NICM; HCM. Facilitates reentry, other rhythms	(87)
Dyssynchrony	Bundle branch may cause cardiomyopathy; predispose to SCA	(133)
Mechanical Disruption	Commotio cordis, Aortic dissection	(134)
Neurologic
Sympathetic stimulation	Stellate ganglion activity may precede ventricular arrhythmias; takotsubo syndrome.	(135,136)
Neural sprouting	Infarction can cause sympathetic nerve sprouting and excessive regional sympathetic innervation.	(135)
Regional Myocardial inhomogeneity (MIBG)	Regionally inhomogeneously denervated myocardium is Arrhythmogenic.	(137)
Cerebral Events	Known causes of Sudden death	(10,49)
Genetic	Monogenic disorders (LQTS, CPVT, Brugada, HCM, ARVC, etc); familial pattern of death in acute MI	(47)

Even areas of mechanistic agreement are less clear than thought. While half of SCA victims have cardiac disease (7) by autopsy, chart review(46) and molecular autopsy(47) (Table 3), this often comprises modest LV dysfunction(46) or abnormalities that would be not considered critical a priori (48). Moreover, many SCA cases are unrelated to cardiac disease, but instead to gastrointestinal bleeds or cerebrovascular events(10,49). A taxonomy for SCA which recognizes the contribution of these entities (Table 2) provides a framework which may help advance the field.

Improve Ascertainment: Reportable Disease Status, Structured Data Resources

Reporting of SCA (ascertainment) is currently limited by variability in death certificate data(50), geographical variations in EMS reporting(2), non-uniform enrollment (e.g. age exclusions in some (51) in some studies) and other factors.

A “reportable disease” status for SCA may improve our meager understanding of its actual incidence and pathophysiology. Reporting requirements fuel funding, knowledge and attention in a positive feedback loop(52). One immediate benefit would be to increase awareness of SCA, which is often synonymous with “heart attack” by the lay press. SCA already has reportable status in Finland, where a legal mandate exists for victims to undergo autopsy (53). In the U.S., coordinated programs such as the Resuscitation Outcomes Consortium (ROC) and Cardiac Arrest Registry to Enhance Survival (CARES), have greatly advanced our understanding (3,54), and offer a partial blueprint of specific data that ought to be collected in expanded regional or national reporting databases(11), and may ultimately enable more nuanced research and personalized care.

A transformational opportunity currently exists to reinvent the reporting for SCA. This may take the form of uploading granular data, potentially including continuous data from wearables recorded in the hours and minutes leading up to SCA, updating databases semi-automatically to build registries, curating data using novel analytics to identify actionable trends, then sharing databases for specific clinical or logistic goals.

Registries to Better Understand SCA

Death certificate statistics likely overestimate SCA events. Conversely, phenotyping SCA survivors, while enlightening, represents the <10% of individuals who survive and thus may or may not represent the population in general. Unfortunately, it is equally challenging to learn from those who succumb to SCD in an era of few autopsies. Large well-organized registries are needed that include data uploaded at the time of SCA as well as data upstream and downstream from the event.

Registries may be community based, such as the Oregon Sudden Unexpected Death Study (SUDS) (46) or San Francisco Postmortem Systematic Investigation of Sudden Cardiac Death Study (POST SCD) (10), or enriched for specific populations such as prior MI and preserved EF, diabetes, less understood populations like African Americans with a higher risk of SCA(32) or women without structural disease(46). Registries should be prospective, include in-depth post-mortem and genetic material, and should be stratified by cardiopulmonary and neurologic outcome to guide actionable analyses.

Arrest-related data elements should include rhythm, antecedent symptoms, witness status, location type (home, public area), bystander CPR or AED usage and could be uploaded automatically. Downstream data should focus on STEMI/ NSTEMI diagnosis, genetic panels, evolution of the ECG and biomarkers over time, coronary angiography, imaging, cardiac or non-cardiac diagnosis. Upstream elements should include antecedent symptoms (7), increasingly available antecedent biometrics from wearable devices, prior cardiac imaging or stress testing, and medications such as beta-blockers which may contribute to PEA (43). Upstream data are particularly relevant for SCA from asystole, which in an unknown but possibly majority of instances may have started with VF.

In SCD victims, MR or CT imaging can extend the traditional autopsy via the concept of post-mortem imaging (“non-invasive autopsy”) to detect coronary or structural heart disease, subtle infarction(55), mechanical complications (cardiac rupture), and non-cardiac causes of arrest such as cerebral hemorrhage (10).

The challenges of organizing and populating such data platforms, then making such data secure and readily accessible are formidable. Registries should leverage technology trends to address this challenge, using novel predictive algorithms to rapidly and efficiently curate data using ‘big data’ methods. Nevertheless, due to the importance and complexity of the data, these authors believe in the continued need for expert data adjudication and validation.

Professional societies, academic institutions and other stakeholders may accelerate efforts at creating digital databases for SCA(56). The American Heart Association recently launched an open digital platform to enable precision medicine for cardiovascular disease (11).

Known Cardiovascular Risk Phenotypes in SCA

Several systems have been proposed to score risk for SCA and other causes of mortality (Table 4). All systems recognize the risk continuum of Myerburg (57) that individuals at highest risk are the sickest and easiest to identify, yet contribute fewer overall cases than individuals at lower risk who are more prevalent.

Table 4.

Current Risk scores for SCA.

Risk Score	Target population	Variables	Reference
Duke Sudden Cardiac Death risk score	Coronary artery disease (> 75% stenosis)	LVEF, Number of coronary vessels, HF, tobacco, DM, HTN, cerebrovascular disease	(138)
Seattle Heart Failure Model	Mild to severe HF	age, gender, ischemic etiology, NYHA, ejection fraction, systolic blood pressure, K-sparing diuretic use, statin use, allopurinol use, hemoglobin, % lymphocyte count, uric acid, sodium, cholesterol, and diuretic dose/kg	(139)
MUSIC (MUerte Subita en Insuficiencia Cardiaca) study	Mild-moderate HF	NYHA class, LVEF, T-wave alternans, T-peak-to-end restitution and T-wave morphology restitution	(140)
ESC HCM SCD Risk Calculator	Hypertrophic cardiomyopathy	Age, maximal wall thickness, LA size, LVOT gradient, family history SCA, NSVT, syncope	(141)
Atherosclerosis Risk in Communities (ARIC) Study	Community population 45-64 years old	age, sex, total cholesterol, lipid-lowering and hypertension medication use, blood pressure, smoking status, diabetes, and body mass index	(142)
eMust Study (Paris)	Prehospital STEMI	younger age, absence of obesity, absence of diabetes mellitus, shortness of breath, and a short delay between pain onset and call to emergency medical services	(143)
Atherosclerosis Risk in Communities (ARIC) Study	Community population 45-64 years old	age, male sex, black race, current smoking, systolic blood pressure, use of antihypertensive medication, diabetes mellitus, serum potassium, serum albumin, high-density lipoprotein, estimated glomerular filtration rate, and QTc interval.	(144)
MUSTT Study	Post-MI, EF ≤ 40	Lower LVEF (<40%), inducible VT, inpatient status, LBBB or IVCD, NSVT >10 days after CABG/no CABG, HF	(62)
REFINE study (prediction of cardiac mortality or SCA)	Post-MI, EF ≤50	Heart rate turbulence (HRT) and T-wave alternans (TWA) at > 8 weeks	(79)

Key: CABG, coronary artery bypass grafting; DM, diabetes mellitus; HF, heart failure; HTN, hypertension; IVCD, intraventricular conduction defect; LA, left atrial; LBBB, left bundle branch block; LVOT, Left ventricular outflow tract; NSVT, non-sustained ventricular tachycardia; NYHA, New York Heart Failure Association class; QTc, corrected QT interval

The highest-risk group comprises SCA survivors, in whom the case for an ICD is strong barring specific limited reversible causes (58). Even with modern shock deferring therapy, their 1-year VT/VF event rate is near 20%(59), and even those with a “reversible cause” face a mortality rate of 18% at 2-years (60).

The next highest risk group comprises those with reduced LVEF without prior SCA. Such patients qualify for primary prevention ICDs, yet their rate of appropriate therapy at 1 year is <5-10%(61) with modern programming. Further risk stratification of this group has proven elusive. Electrophysiological study moderately predicts events in patients with ICDs implanted for nonsustained VT, prior MI and an LVEF ≤ 40%(62), but is less predictive in those with lower LVEF (≤ 30%)(63). Several non-invasive risk factors have been studied (Table 4) but few are predictive enough to guide therapy. Comorbidities such as higher NYHA class, lower LVEF, atrial fibrillation and nonsustained VT certainly confer higher risk for arrhythmias (64), yet also increase the proportion of deaths due causes other than to arrhythmias (8) (Figure 3). These and other competing mortality risks limit the effectiveness of ICDs. ICD effectiveness could be improved through the use of risk scores allowing targeted ICD use for patients with the highest proportionate risk of arrhythmic death and the greatest potential mortality benefit (65) (Figure 3B).

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)).

The next risk group, individuals with preserved LVEF, contribute more SCA events yet their low incidence makes it difficult to devise actionable strategies (44,66). In a meta-analysis of 48,286 patients(67) with preserved LVEF enriched by prior NSTEMI, SCA contributed a third of cardiovascular deaths yet its rate was only 2.37% over 30 months. Revascularization reduced risk by 25%, suggesting an ischemic role, yet none of the comorbidities associated with SCA had sufficient hazard ratios to guide therapy a priori. SCA accounted for one fifth of deaths in 1767 patients with HFpEF (LVEF>45%) in the Aldosterone Antagonist Therapy for Adults With Heart Failure and Preserved Systolic Function (TOPCAT) trial, with modest additional risk conferred by male sex and presence of insulin-treated diabetes mellitus (68).

An emerging risk stratification strategy for SCA (66) is to quantify risk along concurrent ‘axes of comorbidity’ such as age, coronary ischemia and diabetes mellitus, for a population or individual. Figure 1 shows a hypothesized schematic approach. Applied to a population, axes represent worsening of each comorbidity and areas represent numbers of individuals with each combination of comorbidities at risk. Applied to an individual, the area represents the likelihood for SCA given his or her combination of comorbidities.

Figure 4 shows such comorbidity plots for heart failure with reduced ejection fraction (HFrEF, LVEF <40%), preserved ejection fraction (HFpEF, LVEF ≥50%) and moderately reduced (or “mid-range”) ejection fraction (HFmEF; LVEF 40-50%) (69). HFmEF includes individuals with or without coronary disease, potentially transitioning from preserved to reduced(70) LVEF. Such novel analyses extend risk stratification by LVEF alone, and resulting phenotypes may form populations for novel clinical trials.

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)).

Table 2 summarizes other defined SCA populations. Athletes may have increased risk for SCA; it must be reconciled why acute exercise is associated with SCA risk while long-term exercise reduces risk. SCA in competitive athletes is highly publicized yet uncommon, of incidence 0.11 to 33.3 per 100,000 (71). Arrhythmogenic ventricular cardiomyopathy (ARVC/D) is a unique entity where exercise may be pathogenic and acutely trigger an arrhythmia (72). Other notable at-risk populations include recreational users of drugs(73,74) from cocaine to agents such as the anti-diarrheal loperamide (75). Mechanisms for SCA from drugs include overdose, QT prolongation and torsades de pointes, coronary vasospasm and dissection, accelerated atherosclerotic coronary disease, myocardial infarction, myocarditis and seizures.

Mechanistically, current models have had modest success in explaining SCA (Table 3). ECG-based fluctuations in repolarization (microvolt T-wave alternans) (76), slow ventricular conduction (77) and autonomic imbalance measured by heart rate variability, baroreflex sensitivity or heart rate turbulence are predictive in some populations (78), yet may confer only modest value beyond clinical judgement. Initiatives suggest combining (79) or improving (78) such metrics, but alternative paradigms may be needed for major advances.

Novel Cardiovascular and Non-Cardiovascular Phenotypes for SCA

Possible mechanisms for SCA in Table 3 include inflammation, mechanical factors, neurological and metabolic comorbidities and genetic factors, as well as better-studied indices arrhythmic mechanisms of abnormal ventricular repolarization, conduction or autonomic innervation.

Inflammation is increasingly associated with SCA and its comorbidities. This may be mediated by coronary inflammation or modulation of potassium currents (despite only subtle ECG QT prolongation), or autonomic perturbations (80). Statins have anti-inflammatory effects yet were not associated with reduced SCA in a randomized trial of at-risk patients with HF (81).

Comorbidities for SCA are increasingly described. Diabetes mellitus is an important risk through protean mechanisms including fibrosis, inflammation(82), ventricular hypertrophy, dysautonomia or modulation of the renin-angiotensin system(83). Sleep apnea is associated with SCA(84) through potential mechanisms of autonomic modulation, hypoxia, hypercapnia and cardiac stretch from intrathoracic pressure swings (85).

Mechanical factors may explain some cases of SCA, via ventricular hypertrophy particularly if eccentric(86), fibrosis and scar(87). Cardiac MRI abnormalities better predict monomorphic VT than VF(88), which may reflect the current inability of clinical MRI to detect microfibrosis related to VF(89). Stretch (mechano-electric feedback) is a plausible arrhythmic mechanism in patients with LV dysfunction (90), may induce hypertrophy and explain extra-systoles from acute volume loading (91) and arrhythmic risk with chronic heart failure exacerbation. Mitral valve prolapse, long-associated with SCA(92), may reflect inferobasal stretch and fibrosis. Left bundle branch block, a risk factor for SCA(78) reflects slow conduction which induces regional stretch on the LV(93).

Premature ventricular beats and/or nonsustained VT are associated with SCA in ischemic and non-ischemic cardiomyopathy (78). Suppression with antiarrhythmic drugs may cause drug-related pro-arrhythmia(94), yet it is unclear if ablation to reduce electrical or mechanical ventricular heterogeneity or autonomic modulation will prolong survival.

Genetic Contributions to SCA Risk

Younger victims of SCA (under 35 years) may have structural cardiac disease, yet a third who remain elusive despite detailed histopathologic and toxicological studies (95) are said to suffer from sudden arrhythmic death syndrome (SADS) (95). SADS may reflect inherited arrhythmias such as LQTS, catecholaminergic polymorphic ventricular tachycardia (CPVT) and Brugada syndrome (BrS). While exome studies have demonstrated genetic abnormalities (96), recent studies combining chart review with molecular autopsy of patients and surviving relatives found genetic etiologies in only half of cases(47). These studies re-emphasize the need to explore genetic, non-genetic, cardiac and non-cardiac etiologies for SCA at all ages (Figure 1).

Genetics also contributes to SCA in older individuals, in a more complex fashion(97). In a case control study post STEMI (98), the only factors associated with acute VF were a family history of SCA (OR 2.72; 1.84-4.03) and ECG markers of infarct size. Genes may explain VF risk after acute MI (99), although such studies are limited in that VF survivors may not represent the wider population. Indeed, attempts to use common genetic variations as SCA risk scores have been disappointing, although they used GWAS and specific SNPs which could ultimately be improved by next generation whole genome sequencing (100).

Impact of Race, Ethnicity and Gender

Novel SCA taxonomies should be able to explain differences in SCA based on racial and demographic factors, and hence reduce disparities in care. SCA is less likely in women than men (42 vs 58% of victims) (2), and female victims exhibit less structural heart disease than men(101). The risk for SCA in African Americans is nearly double(32) that in Caucasians, and occurs at a younger age(102), only partly explained by known factors such as socioeconomic status(32) or poorer residential area and lower bystander CPR rates (103). Disparities may be partly mitigated by novel methods such as patient-centric video instruction (104).

Prevent

Prevention of SCA is our highest goal, yet poses challenges for pathophysiological understanding, clinical guidelines, and technology development, each impacted by societal priorities and funding. This section focuses on Research methodologies and tools, Epidemiological and Health Services Research and Public Policy Research to improve SCA prevention.

Scientific Pathways and Tools

Prevention should identify populations and individuals at risk for SCA then, using increasingly granular information on where such individuals live and work, and what their likely environmental and personal triggers may be, orchestrate delivery of resuscitation resources to them. Novel discoveries and prevention pathways for SCA will require attracting truly interdisciplinary researchers to the field.

Digital approaches to collecting registry data for prevention, as discussed in the section Understand and Predict, could facilitate research innovations such as consent via mobile devices(56), automatic updates and curation of data for analysis. Challenges to such approaches include ensuring data quality, integrity and security.

Analytical approaches to SCA must be developed in parallel with other technological innovations. Properly implemented, databases will contain oceans of data yet much may not inform clinical judgement. Data must be analyzed with a view to change the course of clinical endpoints including incident SCA, presentation with shockable rhythm, survival to hospital discharge, or full functional recovery. Analytics must enable actionable clinical pathways or at least risk scores (Tables 3,4). Population studies to date have used methods such as stepwise multivariate regression to identify variables associated with each endpoint, then tested in trials. This may need to be revised. Firstly, this approach has provided limited insights thus far, albeit on less granular data. Secondly, traditional biostatistics have limitations when applied to large volumes of data. Thirdly, our limited current level of understanding limits the hypotheses that can be posed a priori.

Machine learning is an emerging science which can adaptively learn, classify and predict patterns in complex multidimensional data by combining computer science, domain-specific pattern recognition and statistics(105). Machine learning has had great success in many fields including voice recognition and image processing, yet its applications in cardiology and for SCA are at an early stage(106). Studies should test whether adaptive learning applied to granular data can separate individuals with and without SCA, type of presenting rhythm, or other clinical endpoints. This could be attempted using networks trained on specific data inputs (supervised learning), or via systems that identify novel data clusters (unsupervised analysis) for future testing and validation.

Ramirez et al (107) used support vector machines to analyze ECG dispersion of repolarization restitution (Deltaalpha), which quantifies T-wave alternans and heart rate turbulence, to identify patients who may suffer SCA versus death from heart failure. Lee et al (108) recently trained artificial neural networks on traditional ECG parameters to predict onset of sustained VT with sensitivity 88%, specificity 82%, and c-statistic 93%. Aro et al. reported on 14 ECG parameters to better predict SCA in large cohorts (109).

Limitations of this approach include our current inability to explain results from machine learning, the fact that identified parameters are often known with previously modest results, and the fact that most machine learning is validated against experts who are limited. Nonetheless, there are high expectations that machine learning results will eventually be interpretable, and reveal clinical associations which are currently unanticipated, yet these challenges for AI remain.

A separate challenge is the need to move potentially useful tools rapidly to clinical trials(56). This has been a major obstacle. For instance, the first randomized trial testing SAECG to predict SCA was a MUSTT substudy (77) that confirmed its value 1-2 decades after its introduction. From a systems perspective, approaches to accelerate studies would be very useful, and may be facilitated by open source resources. In this regard, the AHA recently set up an open source forum for hosting precision medicine based data, code and other resources on Amazon Web Services (11). From a trialist perspective, novel clinical study and statistical design methods are needed to demonstrate clinical effectiveness of interventions in large clinical studies at a sufficiently low cost to be feasible.

Prevention will ultimately require steps to alter the natural history of disease phenotypes. However, this is not always intuitive. Reverse remodeling in cardiomyopathy that normalizes LVEF may decrease overall mortality while not reducing arrhythmic risk or ICD firing rates (110). Other steps to reduce arrhythmic substrate such as reducing fibrosis may have more success. Future therapies may include systemic oral agents or small molecules that reduce myocardial fibrosis, although their cardiac selectivity may be difficult to achieve. Techniques such as treating abnormalities in the Purkinje network, which may be important in VF genesis, may reduce SCA risk in selected individuals. Development of non-invasive computational techniques to map sites that might promote ventricular arrhythmias, then ablating these regions non-invasively by external beam irradiation (111) might someday permit a non-invasive solution in high-risk individuals. Gene and stem cell therapies may in the future have a role in reversing genetic causes of arrhythmias leading to SCA.

Epidemiologic and Health Services Research

Epidemiologic and health services research must focus on closing gaps in treatment by ensuring guideline adherence, and help develop new guidelines. Many accepted level I guidelines have poor adherence or are implemented ad hoc.

Current SCA screening, for instance, lacks an accepted and systematic approach to identify candidate patients for guideline-based ICD therapy. Notwithstanding the limitations of current guidelines based largely on LVEF, society based screening systems have not developed to identify asymptomatic individuals meeting this criterion. One approach is to use improve organizational quality to close identified specific gaps. Another is to develop systems across health care systems, which in some cases are countrywide, that are scaleable and able to evolve with guidelines. Another approach is use systems to identify gaps in which to pursue innovations. For instance, since current imaging technology can accurately identifies reduced LVEF, one system-based solution is to develop a low-cost screening tool with a low false positive rate. Similarly, current technologies can identify coronary artery disease and with computer modeling predict coronary events, but simpler, more cost-effective tools are needed for widespread screening.

Public Policy Research

Public policy research is critical to define better ways to prioritize SCA in the sea of demands within society, and can shape health care prioritization. Improving outcomes from SCA is one of the most pressing contemporary public health concerns, yet education of many stakeholders will still be necessary for them to recognize its importance.

One of our greatest challenges in SCA is the difficulty making financial and policy decisions related to screening for SCA. Societies may, on one hand institute regulations requiring dissemination of AEDs, yet in parallel find it difficult to reach consensus on requiring an ECG for sports-related screening (112). Funding such initiatives will inevitably prove costly, and neither industry nor governmental funding alone may be sufficient for this purpose. Reducing regulatory and reimbursement barriers may promote development of new avenues for intervention. Notably, studies in the U.S. have shown that the public may be open to novel concepts such as crowdsourcing of resuscitation for out of hospital SCA (113), and public policy can successfully enable risk screening and prevention of SCA (114).

Such studies demonstrate that paradigms outlined in this document are potentially feasible. We must avoid the real prospect that even the creation of cost-effective technologies that could plausibly save lives might be difficult to fund or implement. Educational efforts and societal commitment to these strategies for SCA are critical to their success. A call to action by major public and lay organizations will have a major impact on developing both governmental and societal priorities for both implementing strategies to reduce SCA as well as supporting necessary scientific investigation.

Funding Directions

Research in SCA has long-been funded by traditional agencies and/or industry. Research in SCA spans clinical, bioengineering and cellular scales, which are likely to converge with industry. Novel funding opportunities that span these stakeholders are thus needed. Public support is critically important to create a mandate for the increases in government supported research necessary to advance SCA outcomes. Compared to the resources and breadth of investigations for all cancers, for example, NIH funding for SCA research is modest. Foundations and large funding agencies such as the American Heart Association have a large role to play in funding seed grants to open innovative approaches to research for which traditional funding may be less suitable.

One major goal must be to fund studies which move the first point of contact from the time of SCA to tens of minutes or hours earlier. Promising approaches have been discussed and include developing a digital infrastructure to optimize resource allocation to at-risk individuals, wearable devices transmitting data to detect ‘early warnings’, alerting first responders and EMS networks. While machine learning currently provides only modest accuracy, for instance in detecting arrhythmia from ambulatory ECGs(115), such studies show at least that it is feasible to automatically and continuously analyze biometric data to inform clinical decisions. A glimpse into the wealth of biometric sensors which could provide continuous data is provided by reports using fluctuations in facial skin reflectance (18) or in weight (116) to indicate heart rate.

Legal and ethical concerns on the application of technologies to acquire and analyze personal, medical, genetic and other data must be addressed but are likely surmountable. Concerns may be mitigated if appropriate security measures are introduced by thoughtful design involving professional organizations, academia, patients, regulatory bodies and industry, with appropriate guidelines(56). These discussions continue to encroach new territory.

Conclusions

Improvements in managing SCA will require a concerted effort across scientific disciplines, clinician groups and industry to combine technological innovation and rigorous scientific studies, and focused public policy initiatives. Advances can be considered in the categories of improving first-response, improving our pathophysiological understanding, and using insights to prevent SCA in at-risk populations and individuals. These initiatives are diverse, and some are more well defined than others. Nevertheless, the enormity of the problem mandates urgent action at scientific, professional and society-wide levels.

Abbreviations

ACS

Acute coronary syndrome

AED

Automated External Defibrillator

AF

Atrial Fibrillation

CPR

Cardiopulmonary Resuscitation

DM

Diabetes mellitus

EMS

Emergency medical services

HFmEF

Heart Failure with moderately reduced ejection fraction

HFpEF

Heart Failure with preserved ejection fraction

HFrEF

Heart Failure with reduced ejection fraction

LVEF

Left ventricular ejection fraction

NSTEMI

Non-ST segment elevation myocardial infarction

PEA

Pulseless Electrical Activity

ROSC

Restoration of spontaneous circulation

SCA

Sudden cardiac arrest

SCD

Sudden cardiac death

STEMI

ST-segment elevation myocardial infarction

VF

Ventricular Fibrillation

VT

Ventricular Tachycardia