2023 Volume 87 Issue 1 Pages 139-149
Background: Most of the factors and prediction models of sudden cardiac death (SCD) have been developed without considering the Asia population. The purpose of this study is to construct a point-based prediction model for the general population in Asia.
Methods and Results: Chin-Shan Community Cardiovascular Cohort (CCCC) is a community-based longitudinal cohort initiated between 1990 and 1991, enrolling participants aged ≥35 years and following them up until 2005. Participants with coronary artery disease (CAD) or a left ventricular ejection fraction (LVEF) of 35% were excluded from this study. The Framingham risk score function was used to derive a simple point-based prediction model. Based on bootstrapping, a novel model (CCCC-SCD-Score) was validated. A total of 2,105 participants were analyzed. The incidence rate of SCD was 0.406 per 1,000 person-years. The CCCC-SCD-Score score was calculated using age groups (maximal points=4), left ventricular hypertrophy, hypertension, left ventricular ejection fraction <40%, aortic flow rate >190 cm/s, and carotid plaque scores ≥5 (point=1 for each risk factor). The C-index of the CCCC-SCD-Score in predicting SCD risks was 0.888 (95% confidence interval: 0.807–0.969).
Conclusions: For the general Asian population without a history of CAD or a LVEF <35% and who are aged >35 years, the novel model-based scoring system effectively identifies the risk for SCD using the clinical factors, electrocardiographic and echocardiographic data.
Sudden cardiac death (SCD) is a non-traumatic and unexpected death resulting from sudden cardiac arrest.1 SCD occurs as a result of congenital heart defects in individuals aged <35 years; however, in the general population aged >35 years, it is most often caused by coronary artery disease (CAD).2 The current clinical guidelines emphasize primary prevention of SCD in the population with high-risk features of SCD risk, such as those with CAD and heart failure (HF).3 For example, individuals with left ventricular ejection fraction (LVEF) <35% are indicated for implantation of prophylactic implantable cardioverter-defibrillators (ICDs).4 Nevertheless, it remains a major challenge to prevent SCD in the general population.
Model-based risk prediction may be helpful in predicting the risk of SCD.5,6 Several imaging technologies (e.g., echocardiography) and electrocardiography (ECG) may be applied to improve the diagnosis and treatment of fetal cardiac disease.7 The echocardiography has been widely validated as a classifier to detect low LVEF and structural abnormalities in high-risk patients. In addition, the carotid artery duplex provided additional information for cardiovascular risk prediction because the plaque score reflects the severity of narrowing in the carotid artery.8
To the best of our knowledge, as most of the models and factors have been proposed beyond the Asia population, the risk factors should be assessed in relation to various ethnic groups.5,6,9 Although LV function acquired from echocardiography is currently the primary parameter for risk stratification for SCD, it is a poor marker with a low sensitivity and specificity. We will investigate the incidence rate of SCD in a Taiwanese community-based population, identify several risk factors for SCD, and construct a novel point-based prediction model of SCD for general populations in Asia.
The Chin-Shan Community Cardiovascular Cohort (CCCC) study is a community-based longitudinal cohort conducted between 1990 and 1991. A total of 3,602 residents (response rate: 82.8%, 47.3% men) aged ≥35 years were initially included for observation of cardiovascular events and related parameters. Participants without 12-lead ECG, echocardiography, and carotid artery duplex sonography data were excluded from this study (Figure 1). Finally, this study evaluated 2,105 participants (44.7% men) without a prior history of CAD and HF with reduced EF (HFrEF: LVEF <35%). Individual informed consents were obtained from each participant. This study was in accordance with the “Declaration of Helsinki” and the ethical standards of the responsible committee on human experimentation. The Institutional Review Board (IRB Number: 2011003001R) of the National Taiwan University Hospital approved this study according to Good Clinical Practice guidelines.
Study flow chart.
We re-assessed all cases biennially for cardiovascular risk factors, physical examinations, biochemical data, lipid profiles, and 12-lead ECGs (Figure 2). The baseline characteristics of the recruited study population have previously been reported (Supplementary Material).10,11 The baseline of systemic underlying diseases such as hypertension and diabetes mellitus were recorded if the patient had been diagnosed and treated for these conditions. Hypertension was defined as resting systolic blood pressure (SBP) ≥140 mmHg or diastolic blood pressure (DBP) ≥90 mmHg at baseline.12 Diabetes mellitus was defined as fasting blood glucose ≥126 mg/dL.13
Flow chart of data collection in the Chin-Shan Community Cardiovascular Cohort.
The data collection period for ECG, echocardiography, and carotid artery duplex sonography are summarized in Figure 2. Standard echocardiography was performed between 1992 and 1993 (first follow up) and 1994–1995 (second follow up). Carotid artery duplex sonography was performed once for study participants between 1994 and 1995. Qualified cardiologists measured M-mode echocardiography in accordance with the recommendations of the American Society of Echocardiography; measurements were repeated twice for calculating the average values. LVEF, LV mass, aortic flow rate (jet velocity), wall thickness, and any associated abnormalities were assessed using echocardiography. The values of intra-class correlation reliability for echocardiography were between 0.70 and 0.85 in the various measurements and had been reported previously.14 On the basis of echocardiography, significant LV systolic dysfunction was defined as LVEF <40%.15 LV mass index was defined as LV mass/body surface area. The aortic flow rate is a direct measurement of the highest antegrade systolic velocity signal across the aortic valve, and is defined as the highest velocity signal obtained from any window after a careful examination. A resting aortic flow rate of ≥260 cm/s (2.6 m/s) was associated with aortic stenosis.16 Both echocardiography and ECG were applied to define left ventricular hypertrophy (LVH) (echocardiography: LV mass index ≥132 g/m2 in men, LV mass index ≥109 g/m2 in women; ECG: S wave depth in V1+tallest R wave height in V5–V6 >35 mm [Sokolov-Lyon’s criteria]).17 In addition, the measurements of 12-lead ECG were corrected by a physician using a standard value or deviation, and the abnormalities detected by 12-lead ECG were based on 3 examinations during biannual follow ups (since 1994–1995).
The ECG risk score was calculated as:6
(Heart rate >80 beats/min) + (PR 220 ms) + (QRS >110 ms) + LVH + T wave inversion.
Duplex carotid artery sonography was used to detect the intima thickness near the bulb of the carotid artery, and to determine the carotid plaque score using a Hewlett-Packard SONO 1500 ultrasound system with a 7.5 MHz real-time B-mode scanner. Carotid plaque score was measured within the extracranial carotid bed based on the sum of sub-scores calculated from 10 segments (bilateral proximal/distal common carotid arteries, internal carotid arteries, external carotid arteries, and bulbs) using the Sutton’s scoring method:18 A grade was assigned to each chosen segment (Grade 0: normal or no observable plaque; Grade 1: one small plaque with diameter stenosis of 30%; Grade 2: for one medium plaque with 30–49% diameter stenosis or multiple small plaques; Grade 3: for one large plaque with 50–99% diameter stenosis or multiple plaques with at least one medium plaque; and Grade 4: for 100% occlusion). Reproducibility of carotid plaque score was good (kappa: 0.70).
The Atherosclerosis Risk in Communities (ARIC)-Framingham score was calculated as follows:5
(0.067*age) + (−1.262*male) + (0.008*cholesterol) + (0.444*lipid-lowing medication use) + (0.307*anti-hypertensive medication use) + (0.025*SBP) + (−0.024*DBP) + (0.617*current smoker) + (0.787*diabetes mellites) + (0.74*body mass index).
Follow-up Strategy and Outcome ConfirmationThe CCCC Study prospectively collected data about deaths from July 1, 1990 to February 28, 2005. We identified the mortality outcome based on the death certificates from the government, coupled with the interviews with the family members, witnesses, and the physician in charge regarding the cause and manner of death. The interviews were conducted within 1 month of the death and were reviewed by 3 investigating doctors, focusing particularly on the mode of death and the preceding symptoms and signs that corresponded to the definition of SCD.
SCD was defined as a sudden, unexpected, non-traumatic loss of heart function and vital signs without preceding complaints or illness, or occurring within 1 h of the onset of complaints. We included victims who were found dead but were seen alive and well within 24 h of the incident. Intoxicated or chronically ill patients experiencing circulatory arrest were excluded.
Statistical AnalysesNormally distributed continuous variables were presented as mean±standard deviation and compared using the Student’s t-test. Continuous variables with a non-normal distribution were presented as medians and interquartile ranges (IQR). Categorical values were presented as absolute numbers (N) with percentages (%), and chi-squared tests were used for statistical comparisons. Incidence rates of events were calculated as the number of cases per 1,000 person-years (PYs) along with 95% confidence intervals (CIs).
The event-free survival curves were plotted using the Kaplan-Meier method, and the log-rank test was used to determine statistical significance. Because the SCD rate may change over time, competing risk models (cause-specific hazard [CSH] vs. sub-distribution hazard [Fine and Gray]) based on the Cox proportional hazard models were used to analyze the hazard ratios (HRs).19,20 For prognostic research such as estimating the absolute risk function, applying the sub-distribution hazard model is recommended.19,20
Due to the limited event rate and over-dispersion in the study data, we also applied the negative binomial regression model to generate the β coefficients for calculating the risk prediction scores,21 and compared them with the competing risk models (SCD as primary vs. other deaths from SCD). To perform the negative binomial regression model, the number of events were summarized by time groups (follow-up years: 0–5, 6–10, >10 years), baseline age group (<45, 45–54, 55–64, 65–74, ≥75 years), and other important risk groups based on the risk factors selected from the competing risk models under the weight of log (person-years) as an offset variable.
In an attempt to construct a simple point-based SCD prediction model, this study examined the incremental predictive values of adding these variables in the multivariable model-derived coefficients. As outlined in the Supplementary Material, the methods used for selecting the risk factors and Framingham risk score are summarized. After obtaining the β value, it would be applied to the Framingham risk score, and the probability of SCD over the following 10 years can be calculated.22 The novel model (CCCC-SCD-Score) has been validated internally by using the bootstrapping method.23 The training dataset was repeatedly resampled 100 times to produce 5 replicated bootstrap sample sets. Each the bootstrap sample size was the same as the training dataset (Figure 1).
Receiver operating characteristic curves and area under curves (AUCs) were used to summarize the prediction performance. The best cut-off value for predicting incident events was determined using the Youden index of the area under the receiver operating characteristic curve (AUC) (sensitivity+specificity−1). The CCCC-SCD-Score was compared to the ARIC-Framingham score and the ECG risk score using the DeLong’s test. We assessed the goodness-of-fit based on the Hosmer-Lemeshow test. A 2-tailed alpha level of <0.05 was used to determine the statistical significance of the data. The analyses were performed using SAS version 9.4.
The study enrolled a total of 2,105 participants (44.1% men) (Figure 1). The baseline characteristics are provided in Supplementary Table 1. During a median follow-up period of 16.4 years (IQR: 15.7–16.9), a total of 401 deaths (19.0%) were recorded. Among these, 13 were classified as SCD (3.24% of all deaths, 0.61% of total participants). The incidence of SCD was 0.406 per 1,000 person-per years (95% CI: 0.185–0.627). Among SCD victims, 23.1% were attributed to CAD, 7.69% to valvular heart disease, 7.69% to arrhythmia, and 61.5% to other causes.
Clinical History, ECG Patterns, and SCDIn this study, we analyzed several background factors associated with SCD, including: gender, age, body mass index, smoking status, alcohol drinking history, regular exercise, diabetes status, fasting blood glucose, triglyceride levels, total cholesterol levels, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, cholesterol levels, antihypertensive and hypoglycemic medications (Supplementary Tables 1,2). In comparison to the participants without SCD, SCD victims were only significantly associated with older age (sub-distribution HR: 1.09, 95% CI: 1.04–1.15) and hypertension (sub-distribution HR: 3.50, 95% CI: 1.01–10.9) after multivariable adjustment (Supplementary Table 2). In addition, the educational differences, marital status, and occupation were not significant risk factors of SCD in this study. Supplementary Table 3 summarized the baseline characteristics of ECGs. LVH documented by 12-lead ECG was the only independent risk factor of an abnormal ECG pattern for SCD (sub-distribution HR: 6.04, 95% CI: 1.47–24.9; Supplementary Table 4).
Associations Between Ultrasonographic Findings and SCDSupplementary Table 5 summarizes the findings of carotid artery duplex sonography. Using multivariable adjustment, carotid plaque scoring ≥5 (sub-distribution HR: 5.76, 95% CI: 1.15–28.7), aortic flow rate >190 cm/s (sub-distribution HR: 72.1, 95% CI: 12.4–418.8; cut-off point identified by the Youden index of the AUC), LV systolic dysfunction (LVEF <40%) (sub-distribution HR: 23.6, 95% CI: 2.35–237.5), and LVH based on the echocardiography (sub-distribution HR: 5.92, 95% CI: 1.37–25.7) were independent factors for SCD (Supplementary Table 6).
CCCC-SCD-Score Construction Using the Training DatasetThe CCCC-SCD-Score was developed for the purpose of estimating 10-year SCD risks for the general population in Asia after carefully selecting several risk factors associated with SCD occurrence. Two competing risk models (cause-specific approach and sub-distribution approach) and the multivariable negative binomial regression were fit to the selected risk factors for comparisons among various models (Table 1). Based on the training dataset, the points of the CCCC-SCD-Score calculated by the 3 regression models were identical. The points were assigned based on the 1-year increments of adjusted β coefficient change in age: <45 years: 0; 45–54 years: 1; 55–64 years: 2; 65–74 years: 3; and ≥75 years: 4 points. There were other clinical risk factors included in the simple point-based SCD prediction score, including: hypertension (point=1), LVH (ECG or echocardiography) (point=1), LVEF <40% (point=1), aortic (valve) flow rate >190 cm/s (point=1), and carotid plaque scores ≥5 (point=1) (Table 1). In Table 2 and Figure 3A, the CCCC-SCD-Score was used to illustrate the risk function that predicts 10-year SCD rates on the basis of the cause-specific approach, the sub-distribution approach, and the negative binomial model. The sub-distribution approach is more suitable for estimating the 10-year SCD rate in this study.
| Clinical risk factors | Distribution of the population, N=2,105 (mean or proportion (%)) |
Wi-j-Wi-ref | Cause-specific approach | Sub-distribution approach | Negative binomial model | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Estimated coefficient (βrisk factor) |
βrisk factor* (Wi-j-Wi-ref) |
Risk points | Estimated coefficient (βrisk factor) |
βrisk factor* (Wi-j-Wi-ref) |
Risk points | Estimated coefficient (βrisk factor) |
βrisk factor* (Wi-j-Wi-ref) |
Risk points | |||
| Age, +1 year | 53.9±11.7 | βage-1=0.251 | βage-10=2.510 | βage-1=0.191 | βage-10=1.910 | βage-1=0.304 | βage-10=3.044 | ||||
| <45 (reference) | 26.1 | 0 | 0.000 | 0 | 0.000 | 0 | 0.000 | 0 | |||
| 45–54 | 27.0 | 10 | 2.510 | 1 | 1.910 | 1 | 3.044 | 1 | |||
| 55–64 | 27.2 | 20 | 5.020 | 2 | 3.820 | 2 | 6.088 | 2 | |||
| 65–74 | 15.4 | 30 | 7.530 | 3 | 5.730 | 3 | 9.133 | 3 | |||
| ≥75 | 4.28 | 40 | 10.04 | 4 | 7.640 | 4 | 12.18 | 4 | |||
| Hypertension (SBP ≥140 mmHg or DBP ≥90 mmHg) |
27.6 | 1 | 1.606 | 1.606 | 1 | 1.105 | 1.105 | 1 | 1.929 | 1.929 | 1 |
| LV systolic dysfunction (LVEF <40%) |
0.52 | 1 | 3.595 | 3.595 | 1 | 1.966 | 1.966 | 1 | 2.964 | 2.964 | 1 |
| LVH (ECG or echocardiography) |
7.41 | 1 | 1.655 | 1.655 | 1 | 1.734 | 1.734 | 1 | 4.123 | 4.123 | 1 |
| Aortic flow >190 cm/s | 0.52 | 1 | 3.177 | 3.177 | 1 | 2.786 | 2.786 | 1 | 4.006 | 4.006 | 1 |
| Carotid plaque scores ≥5 | 3.28 | 1 | 1.916 | 1.916 | 1 | 1.867 | 1.867 | 1 | 3.125 | 3.125 | 1 |
(Wi-j-Wi-ref) represents the difference between each value of risk factor and its reference value; Risk points = βrisk factor*(Wi-j-Wi-ref) / βage-10. DBP, diastolic blood pressure; ECG, electrocardiography; LV, left ventricular; LVEF, left ventricular ejection fraction; LVH, left ventricular hypertrophy; SBP, systolic blood pressure.
| Scores | Total proportion (%) |
Observed 10-year rate (%) |
Cause-specific approach |
Sub-distribution approach |
Negative binomial model |
|---|---|---|---|---|---|
| Estimated 10-year rate (%) |
Estimated 10-year rate (%) |
Estimated 10-year rate (%) |
|||
| 0 | 22.95 | 0.000 | 0.000 | 0.000 | 0.000 |
| 1 | 22.66 | 0.000 | 0.000 | 0.000 | 0.000 |
| 2 | 22.13 | 0.137 | 0.000 | 0.001 | 0.000 |
| 3 | 17.01 | 0.581 | 0.002 | 0.008 | 0.000 |
| 4 | 10.12 | 1.353 | 0.020 | 0.052 | 0.007 |
| 5 | 3.610 | 3.312 | 0.241 | 0.351 | 0.143 |
| 6 | 0.950 | 4.949 | 2.924 | 2.346 | 2.967 |
| 7 | 0.570 | 24.36 | 30.59 | 14.81 | 46.87 |
| 8 | NA | NA | 98.88 | 66.12 | 99.99 |
| 9 | NA | NA | 100.0 | 99.93 | 100.0 |
NA, not available.
(A) Risk function based on various scores and (B) prediction accuracy based on the Hosmer-Lemeshow chi-squared test (x-axis: observed probability of SCD; y-axis: predicted probability of SCD). SCD, sudden cardiac death.
In the training dataset, analyses of the receiver operating characteristic curve (ROC) demonstrated that the CCCC-SCD-Score had good predictive performance in predicting incident events of SCD (Table 3). The AUC was 0.881 (95% CI: 0.805–0.958; sensitivity: 0.923, specificity: 0.955; positive predictive value: 0.172; negative predictive value: 0.999; positive likelihood ratio: 80.1; negative likelihood ratio: 0.92). SCD events were predicted most accurately with a cut-off value of ≥6. A Kaplan-Meier survival plot demonstrated significant differences between the survival curves for patients stratified according to a CCCC-SCD-Score ≥6 or <6 (Log-rank test, P<0.001) (Supplementary Figure). In fitting the observed and predicted values, the CCCC-SCD-Score had good prediction accuracy (Chi-squared of Hosmer-Lemeshow test: 0.906; P value=0.82; Figure 3B).
| Step | Risk assessment | AUC | 95% CI of AUC | P value 1 | P value 2 |
|---|---|---|---|---|---|
| 1-0 | Taking history and clinical examination: Age groups, and history of hypertension |
0.842 | 0.772–0.912 | Reference | 0.11 |
| 2-0 | 12-lead ECG: Patterns of LVH | 0.694 | 0.539–0.848 | 0.040 | 0.72 |
| 2-1 | Step 1-0 plus Step 2-0 | 0.864 | 0.778–0.951 | 0.38 | 0.12 |
| 3-0 | Echocardiography data: LV mass index, LVEF, and aortic flow rate |
0.645 | 0.458–0.833 | 0.025 | 0.85 |
| 3-1 | Step 1-0 plus Step 3-0 | 0.871 | 0.798–0.965 | 0.14 | 0.052 |
| 4-0 | Carotid artery duplex: Carotid plaque score | 0.561 | 0.461–0.620 | <0.001 | 0.23 |
| 4-1 | Step 1-0 plus Step 4-0 | 0.833 | 0.770–0.915 | 0.86 | 0.10 |
| 5-0 | Step 1-0 plus Step 2-0 plus Step 3-0 | 0.908 | 0.834–0.983 | 0.035 | 0.027 |
| 6-0 | Step 1-0 plus Step 2-0 plus Step 3-0 plus Step 4-0 | 0.908 | 0.834–0.983 | 0.035 | 0.027 |
| 7-0 | CCCC-SCD-Score | 0.888 | 0.807–0.969 | 0.039 | 0.031 |
| 8-0 | ARIC-Framingham score | 0.662 | 0.469–0.914 | 0.027 | Reference |
| 9-0 | ECG risk score | 0.666 | 0.498–0.835 | 0.013 | 0.85 |
P value 1: each model was compared with the reference of Step 1-0. P value 2: each model was compared with the reference of Step 8-0. ARIC, Atherosclerosis Risk in Communities Study; AUC, area under the curve of receiver operating characteristic curve; CI, confidence interval. Other abbreviations as in Table 1.
Table 3 summarized the predictive performance of various models for risk assessment of SCD according to the flow chart of Figure 4 (refer to the Supplementary Material). Excellent predictive performance was shown when adding several parameters together. By comparing the CCCC-SCD-Score (Step 7-0) with the ARIC-Framingham score and ECG risk score (Step 8-0 and Step 9-0 in Table 3), the CCCC-SCD-Score effectively identifies the risk for SCD (Table 3).
Summary for applying electrocardiogram and medical ultrasound to detect the warning signs of sudden cardiac deaths based on the Chin-Shan Community Cardiovascular Cohort. DBP, diastolic blood pressure; ECG, electrocardiogram; LVMI, left ventricular mass index; LVEF, left ventricular ejection fraction; SBP, systolic blood pressure.
The distribution of the AUCs based on bootstrapping validation with 100 times re-sampling is shown in Figure 5A. ROC analyses showed that the CCCC-SCD-Score was still effective at predicting incidents of SCD (mean AUC: 0.880, 95% CI: 0.874–0.887; Figure 5B) when comparing with the ARIC-Framingham score and the ECG risk score. The AUCs for the ARIC-Framingham score (mean AUC: 0.731, 95% CI: 0.715–0.756; Figure 5C) and the ECG risk score (mean AUC: 0.650, 95% CI: 0.637–0.663; Figure 5D) were lower than the AUCs for the CCCC-SCD-Score.
Discrimination performance based on the area under the receiver operating characteristic (ROC) curves (AUCs) for the various SCD prediction scores: (A) Distribution of the AUCs based on the bootstrapping validation (the training dataset was repeatedly resampled 100 times), and histogram of the AUCs based on the frequency (number) of bootstrap sample sets for: (B) CCCC-SCD-Score, (C) ARIC-Framingham score, and (D) ECG risk score. AUC, area under the curve; CI, confidence interval; ECG, electrocardiogram; N, number; SCD, sudden cardiac death; StDev, standard deviation.
In this study, a novel model-based point scoring system was developed for the Asian general population without a history of CAD or a LVEF <35% and who were aged >35 years. The study revealed several relevant findings, including: (1) hypertension was significantly associated with a higher risk of SCD; (2) LVH documented by ECG or echocardiography was independently associated with SCD; (3) medical ultrasound findings of LV systolic dysfunction (LVEF <40%), increased aortic flow, and evidence of significant carotid plaque were associated with SCD; and (4) the newly developed CCCC-SCD-Score system in Asia effectively identifies the risk of SCD with a good predictive performance, even after bootstrapping validation with 100 times re-sampling.
Applying Electrocardiography and Medical Ultrasounds to Detect SCD RiskEchocardiography and ECG can be widely used to improve the diagnosis and treatment of fatal cardiac disease,6,7,24 and to identify subjects who are at risk of SCD.25 The 12-lead ECG is a widely available, inexpensive, non-invasive tool available to all physicians, which may provide definitive clues for establishing the diagnosis. The Fingesture study (N=5,869, 75% men) reported that abnormal ECG patterns were associated with myocardial fibrosis among SCD victims in Northern Finland and Lapland.26 The study by Holkeri et al demonstrated that ECG risk score, combining various patterns of abnormal ECG, may predict 10-year SCD risk in the general population in Finland (N=6,830; age 30–59 years, 45.5% men).6 In this study, we found that an abnormal ECG pattern of LVH was strongly associated with SCD, but other abnormal ECG patterns were not selected in the CCCC-SCD-Score. However, the CCCC-SCD-Score is more effective at predicting SCD risk than the ECG risk score.
Aging is associated with increased vascular stiffness and aortic valve flow rate. Hypertension could lead to hypertensive heart disease with clinical manifestations of LVH and diastolic dysfunction.27 The association between LVH and SCD was reported,28 especially in the presence of myocardial ischemia, fibrosis and scar tissue.27 In fact, SCD can be caused not only by ischemic heart disease, but also by genetic channelopathies (e.g., hypertrophic cardiomyopathy,29 Brugada syndrome,30 or arrhythmogenic right ventricular dysplasia). The inherited heart conditions are related to genetic mutations and result in cardiomyopathy, but the incidences of inherited heart conditions are relatively rare in the general population who are aged <35 years, and these people often have genetic mutations that results in cardiomyopathy.31 Approximately 60% of patients with a family history (30% of patients without a family history) of hypertrophic cardiomyopathy will have a positive genetic result; however, genetic screening tests are extremely labor-intensive and expensive. In the absence of abnormal conditions or other causes of LVH (e.g., hypertension or valvular heart disease), standard 2D echocardiography is the first-line imaging tool for identifying LVH.29 In this study, participants aged <35 years were excluded from enrollment initially. The inherited heart conditions and family history were rarely reported in the CCCC cohort.
It has been reported by Sutton-Tyrrell et al that an elevated aortic pulse is a marker of arterial stiffness, which is predictive of cardiovascular and death events.32 Except for LVH, sudden death in patients with severe aortic stenosis is a clinically important issue. The severity of aortic stenosis is determined by aortic jet velocity and mean gradient (mild: aortic jet velocity ranging from 260 to 300 cm/s (2.6–3 m/s); moderate: aortic jet velocity ranging from 300 to 400 cm/s (3–4 m/s); severe: aortic jet velocity >400 cm/s [4 m/s]).16 As reported by Alcón et al, increased aortic jet velocity (flow rate) ranging from 150 to 200 cm/s (1.5–2 m/s) was significantly associated with increased cardiovascular and mortality outcomes, as found by analyzing 5,994 adults without/with aortic stenosis.33 In our study, we demonstrated similar findings that increased aortic flow rate >190 cm/s (1.9 m/s, cut-off point identifying by the Youden index of the AUC) was an independent risk factor of SCD, even in patients with normal flow. The possible mechanism could be: (1) increased aortic flow rate or aortic stenosis may lead to arterial hypotension, stimulation of LV baroreceptors may cause a fall in venous return and consequent bradycardia (abnormal Betzold-Jarisch reflex); or (2) inappropriate hypotension and a low cardiac output, which provokes coronary hypoperfusion in patients who already have a predisposition through LVH, which may lead to VT.
Reduced EF is the most widely used marker for increased SCD risk in patients with either ischemic heart disease or non-ischemic cardiomyopathy.25 For risk assessment of SCD, echocardiography is highly recommended to assess the structural and functional state of the heart.25 The ARIC Study and the Cardiovascular Health Study found that echocardiography-derived variables for predicting 10-year SCD provided incremental value for risk stratification after adjustment for Framingham risk scores and renal function in the general population.34 In agreement with our findings, the echocardiographic evidence of reduced LVEF (<40%) and LVH were significantly related to incident SCD events in the general population.
In addition, a previous study conducted in Japan demonstrated that carotid plaque scores were associated with cardiovascular deaths in the elderly who have a low cardiovascular risk.8 The ARIC Study and the Cardiovascular Health Study (CHS) reported that the presence of carotid plaque was associated with SCD risk (HR: 1.37, 95% CI: 1.13–1.67).35 In order to detect subclinical carotid atherosclerosis, patients without obvious symptoms of cardiovascular events may benefit from carotid artery duplex sonography. Several mechanisms have been proposed regarding the link between early carotid atherosclerosis and SCD. Subclinical atherosclerosis may lead to ischemic events and inflammatory status, which may result in incident SCD. Atherosclerosis may remodel the LV myocardium in a chronic, subclinical manner, leading to cardiomyopathy and may contribute to fatal arrhythmia and lead to incident SCD consequently.36–38
Model-Based Risk Prediction Score and Traditional Risk FactorsTraditional risk factors include physical factors such as age, gender, obesity, and race, as well as underlying diseases such as CAD, HF, atrial fibrillation, hypertension, diabetes mellitus, and renal dysfunction, which may present a risk for SCD.39,40 Given a number of identifiable risk factors, a model-based risk prediction score could be helpful for risk stratification of SCD. Bogle et al once developed a simple 10-year risk prediction score based on the ARIC Study (N=11,335) and the Framingham cohort (N=5,626) in the United States (patients were aged 45–65 years, 47.6% men).5 This risk score was derived from the following factors: age, sex, total cholesterol, use of lipid-lowering and hypertension medications, blood pressure, smoking status, diabetes, and body mass index, with a C-index of 0.75. The risk of SCD was also higher in Black people than in White people in each risk strata.5 In contrast, the CCCC-SCD-Score did not include diabetes, smoking, lipid profiles, body mass index, renal dysfunction, or cardiac arrhythmias. Nevertheless, the prediction performance of the present score was significantly higher than the ARIC-Framingham score, suggesting that different scoring systems should be applied to risk stratification for different ethnicities.
Clinical ImplicationsSCD can be caused by structural (e.g., CAD, non-ischemic cardiomyopathies, and valvular heart disease) and non-structural etiologies (e.g., arrhythmic causes).2 After excluding prior histories of CAD and LVEF <35%, 23.1% of the victims in this study were likely to have a cause of death from CAD. The cause of >76.9% of SCD in this study was not attributed to CAD. Identifying the risk factors of SCD that extend beyond CAD is essential for preventive medicine. Considering the risk assessment of SCD, the CCCC-SCD-Score is recommended to be re-assessed at least annually due to the status of disease varying with time for timely risk stratification and treatment. In fact, risk assessment of SCD should be initiated with a clinical interview or blood testing (e.g., lipid profile or other biological data). Just like the CHA2DS2-VASc score is used to assess stroke risk in patients with atrial fibrillation, a population-appropriate risk calculator is an easy way to begin assessing cardiovascular risks. In addition to clinical interview, excellent predictive performance to detect SCD was shown in this study when adding ECG and echocardiography data together. However, the measurements of 12-lead ECG or medical ultrasonography shall be corrected using a standard value as per the clinical guidelines to control the diagnostic accuracy.
The CCCC-SCD-Score represents an integrated point-based scoring system as an initial step in developing routine screening for SCD. As the consequence of SCD is severe, even if the positive predictive value is not so high, patients who are grouped in the high-risk group shall be followed routinely, and earlier diagnosis and proper prognostic stratification may reduce disease-related mortality by promoting advanced examination (e.g., genetic screening or cardiac magnetic resonance imaging) and timely treatment.
Study LimitationsThere are several limitations to this study. First, the single measurement at baseline of biomarkers, hemodynamic information, and medical ultrasound data were obtained for the study. The abnormalities detected by 12-lead ECG were based on 3 examinations during biannual follow ups. Time-varying covariance may occur when a covariate changes over time during the follow-up period, which is a common phenomenon in clinical research. Such a variable can be analyzed with the Cox regression model to estimate its effect on survival time. However, the CCCC-SCD-Score represents an integrated point-based scoring system based on binary data of abnormalities or diseases as an initial step in developing routine screening for SCD. Second, this study was based on the data from a general Chinese population. Interpretations of data for various races should be made with caution, and external validation should be conducted in the future. Third, genetic data and family history of inherited heart conditions were not collected in this study. Genetic screening test is extremely labor-intensive and expensive, but the inherited heart conditions and family history were rare in the CCCC cohort.
There are several strengths of our study as well. First, CCCC is a valuable study with a large sample size and a long-term follow-up period in Asia. In addition, a community-based population could reduce the possibility of selection bias compared to a hospital-based cohort. Second, we established a comprehensive strategy for identifying risk factors of SCD and ensuring subjects were followed up. Integrated analyses based on the ECGs and imaging technology may be applied to improve the diagnosis and treatment of fetal cardiac disease. A proper population-based risk score system is helpful in predicting the risk of SCD in Asia.
The newly constructed clinical model-based point scoring system is useful in identifying the SCD risks among the Asian general population who are aged at least 35 years. In addition to higher age, LV systolic dysfunction, and hypertension, there are significant risk factors associated with SCD, such as LVH, increased aortic flow, and higher carotid plaque score. Early diagnosis using screening tools such as electrocardiography, echocardiography, and carotid artery duplex sonography is important for the primary prevention of SCD.
We acknowledge the support from the Ministry of Science and Technology of Taiwan (MOST 110-2314-B-A49A-541-MY3); a grant from the TVGH (C19-027); a grant from the TVGH and NTUH (VN111-05); and the Research Foundation of Cardiovascular Medicine (110-02-006).
The Institutional Review Board (IRB Number: 2011003001R) of the National Taiwan University Hospital approved this study.
The deidentified participant data will not be shared.
Please find supplementary file(s);
https://doi.org/10.1253/circj.CJ-22-0322
