Risk stratification for sudden death (SD) is an essential component of hypertrophic cardiomyopathy (HC) management, given the proven effectiveness of implantable cardioverter-defibrillators (ICD) for preventing SD. Although highly effective in identifying high-risk patients, current stratification algorithms remain incomplete and novel strategies are encouraged. In this regard, reliability of the statistical model to predict SD risk in HC, as recommended by the recent European Society of Cardiology (ESC) guidelines, was retrospectively tested in an independent cohort of 1,629 consecutive patients with HC aged ≥16 years. Of the 1,629 patients, 35 incurred SD events, but only 4 of these (11%) had high predictive risk scores >6%/5 years consistent with an ICD recommendation, and most (60%; n = 21) had scores <4%/5 years that would not justify ICDs. Of 46 high-risk patients with appropriate ICD interventions for ventricular fibrillation/tachycardia, 27 (59%) had low SD risk scores of <4%/5 years, regarded by ESC as insufficient to recommend ICDs, and only 12 (26%) had scores >6%/5 years, considered an ICD indication; 11 of these 12 had already met conventional criteria warranting implantation with 2 to 3 risk markers. Of 414 patients with ICDs but without appropriate interventions, 258 (62%) had low risk scores (<4%/5 years) that would argue against implant. In conclusion, primary risk stratification using the ESC prognostic score applied retrospectively to a large independent HC cohort proved unreliable for prediction of future SD events. Most patients with HC with SD or appropriate ICD interventions were misclassified with low risk scores and therefore would have remained unprotected from arrhythmic SD without ICDs.
Sudden death (SD) is a highly visible and devastating consequence of hypertrophic cardiomyopathy (HC). Availability of implantable cardioverter-defibrillators (ICD) for high-risk patients has substantially reduced the frequency of arrhythmic SD. However, efficacy of the ICD strategy for SD prevention is highly dependent on reliable identification of those patients within the heterogeneous disease spectrum who are at highest risk and would benefit most from primary prevention device therapy. Current risk stratification strategy uses generally accepted major noninvasive risk factors to guide selection of patients for ICDs, on the basis of previous observational and retrospective cohort studies and incorporated into expert consensus guidelines.
Recently, the European Society of Cardiology (ESC) has advanced a novel strategy to identify patients with high risk HC, using a mathematical and statistical modeling technique and prognostic score which can be accessed with an online calculator. This approach has been advanced to provide an individualized risk score aimed at identifying patients most likely to benefit from ICDs and heavily promoted as a strategy to replace current evidence-based risk stratification methods. Given the recent visibility of the HC–Sudden Death risk score in the ESC guidelines and throughout Europe and its clinical implications for management of many patients with HC, we considered it a priority to test this statistical modeling strategy against an independent and diverse cohort of patients with HC.
Methods
Case records of the Hypertrophic Cardiomyopathy Centers of the Minneapolis Heart Institute Foundation and Tufts Medical Center were accessed, and 1,629 consecutively evaluated and prospectively followed patients with HC aged ≥16 years at the first evaluation (October 1992 to May 2014) were identified, with a total follow-up period of 13,274 patient-years, to November 2014. This project was reviewed and approved by the institutional review boards at each center.
The study population of 1,629 patients consisted of (1) 944 surviving without ICDs or events; (2) 190 who died from a variety of causes: non-cardiac, or cardiac but non-HC, or HC-related but not sudden; (3) 460 judged at increased SD risk using conventional risk factor analysis, that is, ≥1 major marker within the clinical profile of individual patients, consistent with our practice and experience with HC, and the published U.S. guidelines ; these patients were implanted with primary prevention ICDs at age 43 ± 15 years; and (4) 35 patients who incurred arrhythmic SD events (23 SDs and 12 with resuscitated cardiac arrest). These 35 patients included 13 who declined an ICD despite a recommendation based on standard clinical criteria, 13 assessed in the era before the general application of primary prevention ICDs to HC (i.e., before 2000), and 8 without conventional risk factors sufficient to recommend an ICD. Only 1 of these patients had an ICD in place which failed to convert ventricular fibrillation because of a mechanical/electrical defect. Thirty-six patients with secondary prevention ICDs were excluded from this study. Overall, 81 of our 1629 patients (5%) reached an SD event end-point with an annual rate of 0.61%, similar to that of O’Mahony et al.
Diagnosis of HC was made at initial evaluation using the standard definition based on the identification by 2-dimensional echocardiography or cardiovascular magnetic resonance (CMR) of a hypertrophied nondilated left ventricle (LV) in the absence of another cardiac or systemic disease capable of producing the magnitude of hypertrophy evident.
For the present analysis, we replicated the method and study design presented in O’Mahony et al, essentially the basis of the ESC guidelines for risk stratification. The ESC risk model introduces a set of 7 binary or continuous potentially prognostic variables into a formula. An online calculator generates a prognostic index score promoted as predictive of SD risk projected over 5 years. On the basis of the score, patients are stratified into 3 groups for ICD recommendations: <4%/5 years (ICD generally not considered); 4% to 6%/5 years (ICD can be considered), and >6%/5 years (ICD should be considered).
In accord with O’Mahony et al, follow-up duration was >1 month after clinical evaluation (and risk scoring) with the end points of either the first appropriate ICD intervention or SD/resuscitated cardiac arrest, or clinical status at the last follow-up. Duration of follow-up for the overall study group was 6.4 ± 5.3 years.
The ESC–HC prediction formula for SD is as follows:
Probability SCD at 5 years = 1 − 0.998 exp ( Prognostic index ) ;
ESC risk scores were calculated at initial evaluation and compared to the specified end points for individual patients within 5 subsets judged at different levels of risk based on conventional risk markers and guidelines, that is, (1) SD, (2 and 3) ICD with or without appropriate intervention, (4) survivors without ICDs/events, and (5) other cardiac or noncardiac deaths ( Table 1 ).
| Variable | Survivors without ICDs/events (n=944) | Other deaths ∗ (n=190) | ICD | Sudden death (n=35) | |
|---|---|---|---|---|---|
| With appropriate intervention (n=46) | Without appropriate intervention (n=414) | ||||
| Age at initial evaluation (years) | 47 ± 17 | 65 ± 16 | 39 ±14 | 40 ± 14 | 38 ± 16 |
| Male | 660 (70%) | 80 (42%) | 21 (46%) | 277 (67%) | 29 (83%) |
| Follow-up (years) † | 6.0 ± 5.2 | 6.5 ± 5.6 | 4.9 ± 5.3 | 7.6 ± 5.3 | 4.9 ± 3.5 |
| Risk score, median (25th, 75th percentile) | 2.1 (1.5, 3.0) ‡ | 2.1 (1.5, 3.1) ‡ | 3.8 (2.9, 6.2) | 3.3 (2.3, 5.0) | 3.2 (2.4, 5.2) |
| Range (minimum, maximum) | 18.6 (0.1, 18.8) | 16.0 (0.5, 16.6) | 13.9 (1.4, 15.3) | 21.7 (0.5, 22.1) | 8.2 (1.2, 9.4) |
| Risk %/year † | 0.42%/year | 0.44%/year | 0.76%/year | 0.66%/year | 0.62%/year |
| ICD recommendation | |||||
| Low risk <4% (generally not considered) | 820 (86.9%) § | 159 (83.7%) § | 27 (58.7%) | 258 (62.3%) | 21 (60.0%) |
| Moderate risk 4-6% (can be considered) | 82 (8.7%) | 18 (9.5%) | 7 (15.2%) | 95 (22.9%) | 10 (27.8%) |
| High risk >6% (should be considered) | 42 (4.5%) | 13 (6.8%) | 12 (26.1%) | 61 (14.7%) | 4 (11.1%) |
∗ Includes non-cardiac (n = 103); cardiac related, but non-HC (n = 59); HC-related heart failure including transplant (n = 14); embolic stroke (n = 8); postoperative (n = 6).
† To the first appropriate ICD intervention; sudden or non-sudden deaths; or most recent follow-up visit or telephone contact.
‡ Lower risk scores compared to ICD and SD patients (p ≤0.001).
§ Distribution of patients by risk level significantly different compared to ICD and SD patients.
Also, to examine how individual patients would be classified into different risk categories by the ESC-HC SD prediction formula, we tabulated the projected 5-year probabilities of SD and appropriate ICD interventions for a range of selected clinically relevant covariates, as follows:
First, we examined 5-year projected probabilities with respect to single conventional risk factors (i.e., unexplained recent syncope, family history of SD, and NSVT) across a range of patient ages and maximal wall thicknesses. Mean values for left atrial size and LV outflow gradient from the derivation cohort (44 mm and 12 mm Hg, respectively) were used in these simulations. Similarly, we examined those patients with 2 traditional risk factors across a range of ages and maximal LV wall thickness.
In addition, we examined projected 5-year probabilities in the absence of traditional risk factors. We chose 3 representative ages (20, 40, and 60 years) and 2 representative LV wall thicknesses (20 mm and 25 mm) and assessed a range of LV outflow gradients and left atrial sizes. We then examined patients with different ages and left atrial sizes with respect to 2 representative LV wall thicknesses (20 mm and 25 mm). Finally, we examined 5-year probabilities for different ages and maximal LV wall thicknesses, using the mean left atrial size of the derivation cohort.
Single or dual-chamber ICDs with capacity for antitachycardia and antibradycardia pacing were implanted in 460 patients for primary prevention. Selection of patients for ICDs was according to the risk stratification models previously advanced for HC: American College of Cardiology/American Heart Association guidelines, American College of Cardiology/ESC expert consensus panel, and prevailing practice patterns.
Stored intracardiac electrograms were analyzed for arrhythmias responsible for precipitating defibrillator discharges (shocks or antitachycardia pacing) according to previous definitions in populations with HC. Defibrillator discharges were considered appropriate in 46 patients when triggered by ventricular fibrillation (n = 18) or sustained rapid monomorphic ventricular tachycardia (VT; n = 28; rate >200 minutes), as documented by interrogation of stored electrocardiographic data and interpreted by expert managing electrophysiologists at each institution, and also consistent with O’Mahony et al. Rate cutoffs for arrhythmia detection were programmed and antitachycardia pacing activated at the discretion of the responsible electrophysiologist.
Descriptive statistics are displayed as mean and standard deviation or median (twenty-fifth, seventy-fifth percentile) for continuous variables and number and percentage for categorical variables. Fisher’s exact or the chi-square tests were used to assess the statistical significance of categorical variables between groups, and analysis of variance and the Kruskal–Wallis tests were used for continuous variables. p Values comparing multiple groups were adjusted using the Bonferroni method. A p value of ≤0.05 was considered statistically significant, and reported p values are 2 sided where appropriate. All statistical calculations were performed in Stata 11.2 (StataCorp, Release 11, College Station, Texas).
Results
Of the 1,629 total study patients, at initial evaluation, 901 (55%) were asymptomatic (New York Heart Association class I), 443 (27%) were mildly symptomatic (class II), and 285 (17%) were severely symptomatic (classes III/IV); 1,067 (66%) patients were men. Echocardiographic measurements were: maximum LV wall thickness (usually ventricular septum), 21 ± 6 mm with 135 patients (8%) ≥30 mm; LV end-diastolic dimension, 34 ± 5 mm; left atrial dimension, 42 ± 8 mm; ejection fraction, 64 ± 7%.
In 35 patients, SD events occurred at 43 ± 15 years of age (range 19 to 76), 4.7 ± 3.6 years after initial evaluation (and risk scoring). Notably, of the 35 patients, only 4 (11%) were judged to be at high-risk with an ESC score >6%/5 years, sufficient to recommend an ICD; also, 3 of these 4 patients had 2 or 3 conventional risk factors (each with massive LV hypertrophy) and would be regarded at high risk independent of the calculated risk score ( Table 1 and Figure 1 ). In 10 patients (28%), risk scores were intermediate (4% to 6%/5 years), but the majority, that is, 21 patients (60%), were judged at lowest risk (<4%/5 years) inconsistent with an ICD recommendation and with an ESC score in which an ICD is generally not considered .
