Hypertrophic cardiomyopathy (HCM) is the most common monogenic heart disease, with an estimated prevalence of 1 per 500 persons in the general population.1 The 2014 European Society of Cardiology (ESC) guidelines for HCM2 recommend a clinical risk prediction model for sudden cardiac death (SCD), based on the HCM Risk-SCD study.3 However, criticisms have been made regarding the ESC's SCD risk score, particularly its failure to include clinical signs of heart failure, electrocardiographic parameters or the presence and extent of myocardial fibrosis, an emerging tool for SCD risk stratification.4–6

The ESC risk prediction model has been assessed in different studies but the results have been conflicting. Although most reports and subsequent meta-analyses have validated the model,7–11 a single recent study suggested that it may misclassify risk.4 We used the Portuguese Registry of Hypertrophic Cardiomyopathy (PRo-HCM) to study the SCD risk of HCM patients and to develop a new SCD risk prediction model applicable to our population.

The PRo-HCM, run by the Portuguese Society of Cardiology, was a voluntary multicenter observational study with a two-year enrollment period (April 2013-April 2015), which was retrospective but included a prospective update. It was instituted to collect data on the epidemiology and outcomes of HCM patients seen in Portugal and the first overall results have been published.12 The criteria for enrollment were age >18 years at the time of enrollment and maximum left ventricular wall thickness ≥15 mm unexplained by abnormal loading conditions, in accordance with the 2014 ESC guidelines.2

For the purpose of the current study a cohort of 1022 patients enrolled in PRo-HCM was analyzed (20 patients were excluded due to missing data during follow-up). The follow-up time for each patient was calculated from the date of their first assessment to the study endpoint or death from another cause, or follow-up of five years. The study endpoint was defined as the composite of SCD or an equivalent event. SCD was defined as witnessed sudden death with or without documented ventricular fibrillation or death within one hour of new symptoms or nocturnal deaths with no previous history of worsening symptoms.13 Aborted SCD during follow-up and appropriate implantable cardioverter-defibrillator (ICD) shock therapy (ICD shock to treat sustained ventricular tachycardia or ventricular fibrillation) were considered equivalent events to SCD.

Baseline clinical, electrocardiographic, echocardiographic and cardiovascular magnetic resonance (CMR) characteristics, as well as medical treatment and follow-up data, were analyzed. Heart failure signs such as elevated jugular venous pressure, pulmonary crackles and peripheral edema were defined according to the most recent ESC guidelines.14 Electrocardiographic signs of myocardial fibrosis, such as Q waves and fragmented QRS complexes were analyzed. Fragmented QRS complexes were defined according to previous studies15: various RSR′ patterns with or without a Q wave, with >2 R waves (R′) or >2 notches in the R wave, or >2 notches in the downstroke or upstroke of the S wave, in two contiguous leads. Late gadolinium enhancement (LGE) on CMR was also recorded.

The characteristics of the patient cohort, including clinical, electrocardiographic, echocardiographic, CMR and treatment data, are summarized in Table 1. The mean age at diagnosis for the entire cohort was 53.2±16.4 years and 59.1% were male. During a median five-year follow-up, 19 of 1022 patients (1.9%) had SCD (n=7) or an equivalent event (n=12).

Variables identified with p<0.100 in univariate analysis were included in the multivariate Cox proportional hazards analysis. The model identified four categorical variables as potential predictors of SCD or equivalent (Table 2). The new proposed risk model (SHIFT model) is an acronym for these four variables: unexplained Syncope, Heart failure signs, Interventricular septum thickness ≥19 mm and FragmenTed QRS complex. The predicted five-year risk of SCD (PSCD) for an individual HCM patient can be calculated from the following equation: PSCD at five years=1-0.9695exp (prognostic index), where prognostic index=1.663211×Heart failure signs+1.281165×unexplained Syncope+1.289594×FragmenTed QRS complex+1.312358×Interventricular septum thickness ≥19 mm.

The ESC model was applied in a subgroup of patients for whom complete data were available on the eight risk factors used to calculate the ESC SCD risk score (total: 349 patients; low risk group: 331; intermediate risk group: 18; high risk group: 0), of whom eight patients (2.3%) had SCD or an equivalent event during the five-year follow-up.

The performance of the derived model was internally validated on the sample created with bootstrap resampling. The estimated beta coefficients and 95% CIs for the bootstrapped model are presented in Table 2. The calibration plot of the SHIFT model is shown in Figure 1, with calibration slope of 0.91 (95% CI: 0.74-1.07). The C-index was 0.81 (95% CI: 0.77-0.83) for SHIFT and 0.77 (95% CI: 0.73-0.81) for the ESC model (p=0.246, z: -1.160). The D-statistic was 2.38 (95% CI: 0.95-4.35) for the SHIFT and 1.97 (95% CI: 0.82-3.22) for the ESC model (Table 3). Through ROC curve analysis, a SHIFT score of 1.49 was the best cut-off above which the model recommends ICD implantation (a SHIFT score >1.49 has a sensitivity of 68.8% and a specificity of 83.1% in predicting SCD or equivalent events).

Figure 1.

Calibration plot between observed and predicted endpoints with linear regression fitting.

(0.03MB).

Decision curve analysis showed the value of the SHIFT model. The results are presented as a graph with the selected threshold probability between 0.01 (1%) and 0.26 (26%) plotted on the x axis and the net benefits of models (SHIFT model, treat all model or treat none model) on the y axis (Figure 2). The SHIFT model presents higher net benefit values between 0.01 (1%) and 0.06 (6%). Knowing the net benefit, we can also calculate the reduction in the number of ICDs implanted per 100 patients using the formula: 100× (net benefit of the SHIFT model-net benefit of treat all)/[threshold probability/(1-threshold probability)]. For example, considering a threshold probability of 5%, the reduction in the number of ICDs implanted is 54, meaning that there are 54 fewer false positives per 100 patients.

Figure 2.

Decision curve analysis of the SHIFT model in predicting sudden cardiac death (SCD) or equivalent. The dotted line represents the net benefit of an implantable cardioverter-defibrillator (ICD) for patients according to the SHIFT prediction model; the gray line represents the net benefit of providing an ICD for all patients, assuming that all patients would have SCD or equivalent; and the solid line represents the net benefit of an ICD to no patients, assuming that none would have SCD or equivalent.

(0.06MB).

The present study shows that combining clinical, electrocardiographic and echocardiographic characteristics in HCM patients enables prediction of SCD, aborted SCD or appropriate ICD shock therapy. The SHIFT model is easy to use and does not require complex measurements or calculations since unexplained syncope, heart failure signs, interventricular septum thickness and fragmented QRS complexes are routinely and easily collected variables.

Syncope in HCM is a complex entity since several mechanisms may be responsible for this symptom, including arrhythmias and a primary hemodynamic mechanism. Some of these mechanisms are directly linked to SCD, such as sustained ventricular tachycardia, complete heart block or left ventricular outflow tract obstruction.17 The prognostic significance of unexplained syncope has been widely investigated18,19 and has been identified as a risk factor of SCD in different guidelines.2,20,21 In our study, unexplained syncope was more prevalent in patients with SCD (13.4% vs. 33.3%); similar differences were observed in the HCM Risk-SCD study (13.0% vs. 26.0%).3

HCM patients may present heterogeneous clinical profiles, particularly in terms of signs and symptoms of heart failure, related to left ventricular outflow obstruction, atrial fibrillation, and diastolic and systolic dysfunction.22 Ismail et al.23 identified left ventricular dysfunction as an independent risk factor and a harbinger of increased SCD risk. In our cohort, 8.4% of the population had heart failure signs, which were clearly more prevalent in the SCD group (26.3%), and this variable was an independent predictor of SCD or equivalent.

In our analysis, maximum left ventricular wall thickness and interventricular septum (IVS) thickness were both found to be predictors of the endpoint in univariate analysis (p<0.100). Nevertheless, in multivariate analysis only IVS thickness was an independent predictor, the most predictive cut-off value being ≥19 mm (p=0.024). The degree of hypertrophy and left ventricular outflow tract obstruction are directly related to the risk of SCD and progression of heart failure, and are strong and independent predictors of prognosis.11,24–27

The SHIFT model also considers fragmented QRS as an independent predictor of SCD or equivalent. Fragmented QRS complexes have been associated with myocardial fibrosis,13,28,29 which leads to myocardial scar-related heart failure and arrhythmic events.30 Fragmented QRS has also been associated with HCM Risk-SCD score >4% and increased ventricular arrhythmias in HCM patients.31 Moreover, myocardial fibrosis identified by LGE-CMR is a known predictor of SCD in HCM patients.5,23,32 From this standpoint, the inclusion of fragmented QRS complexes in the proposed risk model provides additional information to assess SCD event risk among HCM patients, and indirectly signifies the presence of fibrosis. It should be noted that the small number of patients who underwent CMR in our study (n=473, 46.3%) may explain why the presence of LGE on CMR imaging was not an independent predictor of SCD.

Our proposed risk prediction model has good discrimination, with a C-index of 0.81 and a D-statistic of 2.38. It is interesting to note that the C-index for the ESC model (0.77) was similar to those obtained from HCM Risk-SCD validation (0.70 [95% CI: 0.68-0.72]).3 A direct comparison between the SHIFT and ESC models was not feasible, as only a small subgroup of patients had complete data for the eight risk factors used to calculate the ESC SCD risk score.

Based on ROC curve analysis and with a cut-off of 1.49, the SHIFT model has the potential to improve patient management by simultaneously minimizing the number of unnecessary and potentially harmful ICD implantations in patients who will not suffer SCD (specificity 83.1%) while accurately identifying the majority of those who will suffer SCD and are most likely to benefit from an ICD (sensitivity 68.8%).

On this basis, the SHIFT model may complement clinical decisions by providing additional and objective prognostic information. In particular, it can be applied to patients whose risk remains uncertain after consideration of conventional risk factors.

The SHIFT model appears to be a promising tool to assess the probability of SCD, aborted SCD or appropriate ICD shock therapy in HCM patients. It may potentially provide additional prognostic value and can be helpful when deciding to implant an ICD for primary prevention.

The authors have no conflicts of interest to declare.