Machine learning for prediction of sudden cardiac death in heart failure patients with low left ventricular ejection fraction: study protocol for a retroprospective multicentre registry in China.
ABSTRACT: INTRODUCTION:Left ventricular ejection fraction (LVEF) ?35%, as current significant implantable cardioverter-defibrillator (ICD) indication for primary prevention of sudden cardiac death (SCD) in heart failure (HF) patients, has been widely recognised to be inefficient. Improvement of patient selection for low LVEF (?35%) is needed to optimise deployment of ICD. Most of the existing prediction models are not appropriate to identify ICD candidates at high risk of SCD in HF patients with low LVEF. Compared with traditional statistical analysis, machine learning (ML) can employ computer algorithms to identify patterns in large datasets, analyse rules automatically and build both linear and non-linear models in order to make data-driven predictions. This study is aimed to develop and validate new models using ML to improve the prediction of SCD in HF patients with low LVEF. METHODS AND ANALYSIS:We will conduct a retroprospective, multicentre, observational registry of Chinese HF patients with low LVEF. The HF patients with LVEF ?35% after optimised medication at least 3 months will be enrolled in this study. The primary endpoints are all-cause death and SCD. The secondary endpoints are malignant arrhythmia, sudden cardiac arrest, cardiopulmonary resuscitation and rehospitalisation due to HF. The baseline demographic, clinical, biological, electrophysiological, social and psychological variables will be collected. Both ML and traditional multivariable Cox proportional hazards regression models will be developed and compared in the prediction of SCD. Moreover, the ML model will be validated in a prospective study. ETHICS AND DISSEMINATION:The study protocol has been approved by the Ethics Committee of the First Affiliated Hospital of Nanjing Medical University (2017-SR-06). All results of this study will be published in international peer-reviewed journals and presented at relevant conferences. TRIAL REGISTRATION NUMBER:ChiCTR-POC-17011842; Pre-results.
Project description:BACKGROUND:Low left ventricular ejection fraction (LVEF) increases risk for both sudden cardiac death (SCD) and for heart failure (HF) death; however, implantable cardioverter-defibrillators (ICDs) reduce the incidence of SCD, not HF death. Distinguishing individuals at risk for HF death (non-SCD) versus SCD could improve ICD patient selection. OBJECTIVE:This study evaluated whether electrocardiogram (ECG) quantification of myocardial infarction (MI) could discriminate risk for SCD versus non-SCD. METHODS:Selvester QRS scoring was performed on 995 MADIT-II trial subjects' ECGs to quantify MI size. MIs were categorized as small (0-3 QRS points), medium (4-7) or large (? 8). Mortality, SCD and non-SCD rates in the conventional medical therapy (CMT) arm and mortality and ventricular tachycardia/fibrillation (VT/VF) rates in the ICD arm were analyzed by QRS score group. Both arms were analyzed to determine ICD efficacy by QRS score group. RESULTS:In the CMT arm, mortality, SCD and non-SCD rates were similar across QRS score groups (P = 0.73, P = 0.92, and P = 0.77). The ICD arm showed similar rates of mortality (P = 0.17) and VT/VF (P = 0.24) across QRS score groups. ICD arm mortality was lower than CMT arm mortality across QRS score groups with greatest benefit in the large scar group. CONCLUSION:Recently, QRS score was shown to be predictive of VT/VF in the SCD-HeFT population consisting of both ischemic and nonischemic HF and having a maximum LVEF of 35% versus 30% for MADIT-II. Our study found that QRS score did not add prognostic value in the MADIT-II population exhibiting relatively more severe cardiac dysfunction.
Project description:OBJECTIVES:The aim of this study was to determine the incidence of sudden cardiac death (SCD) among persons living with human immunodeficiency virus infection (PHIV) with heart failure (HF), who were hospitalized for HF, and the risk factors associated with it. BACKGROUND:HF is associated with an increased risk for SCD. PHIV are at heightened risk for HF. METHODS:This was a retrospective study of 2,578 patients hospitalized with HF from a single academic center, of whom 344 were PHIV. The outcome of interest was SCD. Subgroup analyses were performed by strata of viral load (VL) and left ventricular ejection fraction (LVEF) <35%, 35% to 49%, and ?50%. RESULTS:Of 2,578 patients with HF, 2,149 (86%) did not have implantable cardioverter-defibrillators; of these, there were 344 PHIV and 1,805 uninfected control subjects. Among PHIV with HF, 313 (91%) were prescribed antiretroviral therapy and 64% were virally suppressed. There were 191 SCDs over a median follow-up period of 19 months. Compared with control subjects, PHIV had a 3-fold increase in SCD (21.0% vs. 6.4%; adjusted odds ratio: 3.0; 95% confidence interval: 1.78 to 4.24). Among PHIV, cocaine use, lower LVEF, absence of beta-blocker prescription, and VL were predictors of SCD. The SCD rate among PHIV with undetectable VL was similar to the rate among uninfected subjects. Similar findings were observed by LVEF strata. Among PHIV with HF without conventional indications for an implantable cardioverter-defibrillator, the rate of SCD was 10% per year. CONCLUSIONS:PHIV hospitalized with HF are at a markedly increased risk for SCD. SCD risk was increased in patients with lower LVEFs, lower CD4 counts, and higher VL.
Project description:Identification of patients with non-ischaemic dilated cardiomyopathy (NICM) who are at risk of sudden cardiac death (SCD) and could benefit from an implantable cardioverter defibrillator (ICD) is challenging. The study aims to systematically assess the prognostic value of left ventricular (LV) midwall late gadolinium enhancement (LGE) pattern in patients with NICM and further explore its value on predicting SCD events. The study was prospectively registered in PROPSERO (CRD42019138468). We systematically searched PubMed, Ovid Embase, Cochrane Library, Web of Science, and ClinicalTrials.gov to identify studies that evaluated the association between LV midwall LGE and clinical outcomes (all-cause mortality, cardiovascular mortality, and SCD or aborted SCD endpoint) in NICM patients. A meta-analysis was performed to determine pooled odds ratio (OR) for these adverse events. Seven studies including 1827 NICM patients over a mean follow-up duration of 36.1 ± 19.3 months were included. The presence of LV midwall LGE pattern was observed in 562 (30.8%) patients. The pooled OR was 3.37 [95% confidence intervals (CIs): 1.35-8.42] for all-cause mortality, 5.56 (95% CI: 1.23-25.22) for cardiovascular mortality, and 2.25 (95% CI: 1.16-3.16) for SCD or aborted SCD. In a subgroup analysis with mean ejection fraction cut-off point of 35%, the pooled OR for SCD or aborted SCD was 2.06 (95% CI: 1.32-3.22) for LV ejection fraction (LVEF) > 35% and 2.49 (95% CI: 1.48-4.20) for LVEF ? 35%. In addition, our study indicated that LV midwall LGE showed an excellent negative predictive value in identifying high-risk NICM patients and that the number needed to treat with ICD implantation in NICM patients with midwall LGE is 7. The presence of LV midwall on LGE is a significant prognosticator of adverse events in NICM patients. Additionally, patients with LV midwall LGE may be considered for ICD therapy irrespective of LVEF.
Project description:Background:Sudden cardiac death (SCD) risk stratification in dilated cardiomyopathy (DCM) has been based on left ventricular ejection fraction (LVEF), even though SCD may occur with LVEF?>?35%. Family history of unexplained SCD, especially in the young, raises concern about potential inheritable risk factors. It remains largely unknown how genetic tests can be integrated into clinical practice, particularly in the selection of implantable cardioverter defibrillator (ICD) candidates. We aimed to assess the diagnostic yield of genetic testing in DCM patients with a class I recommendation for ICD implantation, based on current guidelines. Methods:We included ambulatory stable adult patients with idiopathic or familial DCM with previously implanted ICD. Molecular analysis included 15 genes (LMNA, MYH7, MYBPC3, TNNT2, ACTC1, TPM1, CSRP3, TCAP, SGCD, PLN, MYL2, MYL3, TNNI3, TAZ, and LDB3) using next-generation sequencing. Results:We evaluated 21 patients, 12 (57%) males and 9 (43%) with familial DCM, including 3 (14%) with a family history of premature unexplained SCD. Mean age at DCM diagnosis was 40?±?2?years, and mean age at ICD implantation was 50?±?12?years. LVEF was 27?±?9%, and LV end-diastolic diameter was 65?±?7?mm. Genetic variants were found in six (29%) patients, occurring in 5 genes: TPM1, TNNT2, MYH7, PLN, and MYBPC3. The majority were classified as variants of uncertain significance. Family history of SCD was present in both patients with PLN variants. Conclusion:In patients with DCM and ICD, genetic variants could be identified in a significant proportion of patients in several genes, highlighting the potential role of genetics in DCM SCD risk stratification.
Project description:To test the value of Periodic Repolarization Dynamics (PRD), a recently validated electrocardiographic marker of sympathetic activity, as a novel approach to predict sudden cardiac death (SCD) and non-sudden cardiac death (N-SCD) and to improve identification of patients that profit from ICD-implantation.We included 856 post-infarction patients with left-ventricular ejection fraction (LVEF)??30% of the MADIT-II trial in sinus rhythm. Of these, 507 and 348 patients were randomized to ICD or conventional treatment. PRD was assessed from multipolar 10-min baseline ECGs. Primary and secondary endpoints were total mortality, SCD and N-SCD. Multivariable analyses included treatment group, QRS-duration, New York Heart Association classification, blood-urea nitrogen, diabetes mellitus, beta-blocker therapy and LVEF. During follow-up of 20.4?months, 119 patients died (53 SCD and 36 N-SCD). On multivariable analyses, increased PRD was a significant predictor of mortality (standardized coefficient 1.37[1.19-1.59]; P?<?0.001) and SCD (1.40 [1.13-1.75]; P?=?0.003) but also predicted N-SCD (1.41[1.10-1.81]; P?=?0.006). While increased PRD predicted SCD in conventionally treated patients (1.61[1.23-2.11]; P?<?0.001), it was predictive of N-SCD (1.63[1.28-2.09]; P?<?0.001) and adequate ICD-therapies (1.20[1.03-1.39]; P?=?0.017) in ICD-treated patients. ICD-treatment substantially reduced mortality in the lowest three PRD-quartiles by 53% (P?=?0.001). However, there was no effect in the highest PRD-quartile (mortality increase by 29%; P?=?0.412; P?<?0.001 for difference) as the reduction of SCD was compensated by an increase of N-SCD.In post-infarction patients with impaired LVEF, PRD is a significant predictor of SCD and N-SCD. Assessment of PRD is a promising tool to identify post-MI patients with reduced LVEF who might benefit from intensified treatment.
Project description:Heart failure patients with primary prevention implantable cardioverter-defibrillators (ICD) may experience an improvement in left ventricular ejection fraction (LVEF) over time. However, it is unclear how LVEF improvement affects subsequent risk for mortality and sudden cardiac death.This study sought to assess changes in LVEF after ICD implantation and the implication of these changes on subsequent mortality and ICD shocks.We conducted a prospective cohort study of 538 patients with repeated LVEF assessments after ICD implantation for primary prevention of sudden cardiac death. The primary endpoint was appropriate ICD shock defined as a shock for ventricular tachyarrhythmias. The secondary endpoint was all-cause mortality.Over a mean follow-up of 4.9 years, LVEF decreased in 13.0%, improved in 40.0%, and was unchanged in 47.0% of the patients. In the multivariate Cox models comparing patients with an improved LVEF with those with an unchanged LVEF, the hazard ratios were 0.33 (95% confidence interval: 0.18 to 0.59) for mortality and 0.29 (95% confidence interval: 0.11 to 0.78) for appropriate shock. During follow-up, 25% of patients showed an improvement in LVEF to >35% and their risk of appropriate shock decreased but was not eliminated.Among primary prevention ICD patients, 40.0% had an improved LVEF during follow-up and 25% had LVEF improved to >35%. Changes in LVEF were inversely associated with all-cause mortality and appropriate shocks for ventricular tachyarrhythmias. In patients whose follow-up LVEF improved to >35%, the risk of an appropriate shock remained but was markedly decreased.
Project description:Clinical trials of prophylactic implantable cardioverter-defibrillators (ICDs) have included a minority of patients with a left ventricular ejection fraction (LVEF) between 30% and 35%. Because a large number of ICDs in the United States are implanted in such patients, it is important to study survival associated with this therapy.To characterize patients with LVEF between 30% and 35% and compare the survival of those with and without ICDs.Retrospective cohort study of Medicare beneficiaries in the National Cardiovascular Data Registry ICD registry (January 1, 2006, through December 31, 2007) with an LVEF between 30% and 35% who received an ICD during a heart failure hospitalization and similar patients in the Get With The Guidelines-Heart Failure (GWTG-HF) database (January 1, 2005, through December 31, 2009) with no ICD. The analysis was repeated in patients with an LVEF less than 30%. There were 3120 patients with an LVEF between 30% and 35% (816 in matched cohorts) and 4578 with an LVEF less than 30% (2176 in matched cohorts). Propensity score matching and Cox models were applied.The primary outcome was all-cause mortality; data were obtained from Medicare claims through December 31, 2011.There were no significant differences in the baseline characteristics of the matched groups (n?=?408 for both groups). Among patients with an LVEF between 30% and 35%, there were 248 deaths in the ICD Registry group, within a median follow-up of 4.4 years (interquartile range, 2.7-4.9) and 249 deaths in the GWTG HF group, within a median follow-up of 2.9 years (interquartile range, 2.1-4.4). The risk of all-cause mortality in patients with an LVEF between 30% and 35% and an ICD was significantly lower than that in matched patients without an ICD (3-year mortality rates: 51.4% vs 55.0%; hazard ratio, 0.83 [95% CI, 0.69-0.99]; P?=?.04). Presence of an ICD also was associated with better survival in patients with an LVEF less than 30% (3-year mortality rates: 45.0% vs 57.6%; 634 and 660 total deaths; hazard ratio, 0.72 [95% CI, 0.65-0.81]; P?<?.001) (P?=?.20 for interaction).Among Medicare beneficiaries hospitalized for heart failure and with an LVEF between 30% and 35% and less than 30%, survival at 3 years was better in patients who received a prophylactic ICD than in comparable patients with no ICD. These findings support guideline recommendations to implant prophylactic ICDs in eligible patients with an LVEF of 35% or less.
Project description:Sudden cardiac death (SCD) from ventricular tachyarrhythmias accounts for approximately 450,000 annual deaths in the United States; many of these cases involve patients with chronic heart failure (HF). Prediction of which HF patients are most susceptible to SCD is difficult, and it is uncertain whether gene polymorphisms associated with HF outcomes are also linked to arrhythmic risk.We evaluated 485 patients with chronic HF to see whether the angiotensin receptor type 1 (AT1R) 1166A/C or angiotensin-converting enzyme insertion/deletion (ACE I/D) polymorphisms were associated with a higher rate of ventricular arrhythmias requiring implantable cardioverter defibrillator (ICD) therapies over a 5-year period. We assessed the correlation between polymorphisms and antitachycardia pacing (ATP) and/or ICD shocks.Patients with AT1R-1166CC genotype had an increased rate of all events: ATP plus ICD shocks (P = .02). There was no association between ACE I/D genotype and ICD therapies. Furthermore, circulating levels of microRNA-155 (miR-155), a microRNA known to posttranscriptionally regulate AT1R expression, were significantly decreased in the CC compared with the AC and AA genotypes and were associated with ICD events.Our study suggests that the AT1R-1166CC genotype is associated with increased ICD therapies in patients with chronic HF, and the level of circulating miR-155 may be a potential marker for arrhythmic risk. Although these findings are novel, they will need replication and validation in larger cohorts of chronic HF patients.
Project description:<h4>Aims</h4>There is an urgent need to extend sudden cardiac death (SCD) risk stratification beyond the left ventricular ejection fraction (LVEF). We evaluated whether a cumulative electrocardiogram (ECG) risk score would improve identification of individuals at high risk of SCD.<h4>Methods and results</h4>In the community-based Oregon Sudden Unexpected Death Study (catchment population ?1 million), 522 SCD cases with archived 12-lead ECG available (65.3?±?14.5 years, 66% male) were compared with 736 geographical controls to assess the incremental value of multiple ECG parameters in SCD prediction. Heart rate, LV hypertrophy, QRS transition zone, QRS-T angle, QTc, and Tpeak-to-Tend interval remained significant in the final model, which was externally validated in the Atherosclerosis Risk in Communities (ARIC) Study. Sixteen percent of cases and 3% of controls had ?4 abnormal ECG markers. After adjusting for clinical factors and LVEF, increasing ECG risk score was associated with progressively greater odds of SCD. Overall, subjects with ?4 ECG abnormalities had an odds ratio (OR) of 21.2 for SCD [95% confidence interval (CI) 9.4-47.7; P?<?0.001]. In the LVEF >35% subgroup, the OR was 26.1 (95% CI 9.9-68.5; P?<?0.001). The ECG risk score increased the C-statistic from 0.625 to 0.753 (P?<?0.001), with net reclassification improvement of 0.319 (P?<?0.001). In the ARIC cohort validation, risk of SCD associated with ?4 ECG abnormalities remained significant after multivariable adjustment (hazard ratio 4.84; 95% CI 2.34-9.99; P?<?0.001; C-statistic improvement 0.759-0.774; P?=?0.019).<h4>Conclusion</h4>This novel cumulative ECG risk score was independently associated with SCD and was particularly effective for LVEF >35% where risk stratification is currently unavailable. These findings warrant further evaluation in prospective clinical investigations.
Project description:The role of implantable cardioverter-defibrillator (ICD) placement in the primary prevention of sudden cardiac death (SCD) in all consecutive patients with left ventricular ejection fraction (LVEF) ? 35% is still a matter of hot debate due to the fact that the population of these patients is highly heterogeneous in terms of the SCD risk. Nevertheless, reduced LVEF is still the only established criterion during qualification of patients for ICD implantation in the primary prevention of SCD, therefore identification of persons with particularly high risk among patients with LVEF ?35% is currently of lesser importance. More important seems to be the selection of individuals with relatively low risk of SCD in whom ICD implantation can be safely postponed. The aim of the study was to determine whether well-known, non-invasive parameters, such as microvolt T-wave alternans (MTWA), baroreflex sensitivity (BRS) and short-term heart rate variability (HRV), can be helpful in the identification of low-arrhythmic risk patients with ischemic left ventricular systolic dysfunction.In 141 patients with coronary artery disease and LVEF ? 35%, MTWA testing, as well as BRS and short-term HRV parameters, were analysed. During 34 ± 13 months of follow-up 37 patients had arrhythmic episode (EVENT): SCD, non-fatal sustained ventricular arrhythmia (ventricular tachycardia [VT] or ventricular fibrillation [VF]), or adequate high-voltage ICD intervention (shock) due to a rapid ventricular arrhythmia ?200/min. LVEF, non-negative MTWA (MTWA_non-neg), BRS and low frequency power in normalized units (LFnu) turned out to be associated with the incidence of EVENT in univariate Cox analysis. The cut-off values for BRS and LFnu that most accurately distinguished between patients with and without EVENT were 3 ms/mmHg and 23, respectively. The only variable that provided 100% negative predictive value (NPV) for EVENT was negative MTWA result (MTWA_neg), but solely for initial 12 months of the follow-up; the NPVs for other potential predictors of the EVENT were lower. The cut-off values for BRS and LFnu that provide 100% NPV for EVENT during 12 and 24 months were higher: 6.0 ms/mmHg and 73 respectively, but the gain in the NPV occurred at an expense of the number of identified patients. However, the number of identified non-risk patients turned out to be higher when the predictive model included MTWA_neg and the lower cut-off values for ANS parameters: 100% NPV for 12 and 24 months of follow-up was obtained for combination MTWA_neg and BRS ? 3 ms/mmHg, for combination MTWA_neg and LFnu ? 23 100% NPV was obtained for 12 months.Well-known, non-invasive parameters, such as MTWA, BRS and short-term HRV indices may be helpful in the identification of individuals with a relatively low risk of malignant ventricular arrhythmias among patients with ischemic left ventricular systolic dysfunction; in such persons, implantation of ICD could be safely postponed.