European Heart Journal (2014) 35, 2010–2020 CLINICAL RESEARCH

doi:10.1093/eurheartj/eht439 Myocardial disease

A novel clinical risk prediction model for sudden

cardiac death in hypertrophic cardiomyopathy
(HCM Risk-SCD)
Constantinos O’Mahony 1, Fatima Jichi 2, Menelaos Pavlou8, Lorenzo Monserrat 3,

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

Aristides Anastasakis 4, Claudio Rapezzi 5, Elena Biagini 5, Juan Ramon Gimeno 6,
Giuseppe Limongelli7, William J. McKenna1, Rumana Z. Omar2,8 and Perry M. Elliott1*,
for the Hypertrophic Cardiomyopathy Outcomes Investigators
1
The Inherited Cardiac Diseases Unit, The Heart Hospital/University College London, 16-18 Westmoreland St., London W1H 8PH, UK; 2Biostatistics Group, University College London
Hospitals/University College London Research Support Centre, University College London, Gower St., London WC1E 6BT, UK; 3Cardiology Department and Research Unit, A Coruña
University Hospital, Galician Health Service, Spain; 4Unit of Inherited Cardiovascular Diseases, 1st Department of Cardiology, University of Athens, 99 Michalakopoulou St, Athens 11527,
Greece; 5Institute of Cardiology, Department of Specialised, Experimental and Diagnostic Medicine, University of Bologna, Via Massarenti 9, Bologna 40138, Italy; 6Cardiac Department,
University Hospital Virgen Arrixaca, Murcia-Cartagena s/n. El Palmar, Murcia 30120, Spain; 7Monaldi Hospital, Second University of Naples, Via Leonardo Bianchi 1, Naples 80131, Italy; and
8
Department of Statistical Science, University College London, Gower St, London WC1E 6BT, UK

Received 26 July 2013; revised 17 September 2013; accepted 26 September 2013; online publish-ahead-of-print 14 October 2013

Aims Hypertrophic cardiomyopathy (HCM) is a leading cause of sudden cardiac death (SCD) in young adults. Current risk algo-
rithms provide only a crude estimate of risk and fail to account for the different effect size of individual risk factors. The aim
of this study was to develop and validate a new SCD risk prediction model that provides individualized risk estimates.
.....................................................................................................................................................................................
Methods The prognostic model was derived from a retrospective, multi-centre longitudinal cohort study. The model was devel-
and results oped from the entire data set using the Cox proportional hazards model and internally validated using bootstrapping. The
cohort consisted of 3675 consecutive patients from six centres. During a follow-up period of 24 313 patient-years
(median 5.7 years), 198 patients (5%) died suddenly or had an appropriate implantable cardioverter defibrillator
(ICD) shock. Of eight pre-specified predictors, age, maximal left ventricular wall thickness, left atrial diameter, left ven-
tricular outflow tract gradient, family history of SCD, non-sustained ventricular tachycardia, and unexplained syncope
were associated with SCD/appropriate ICD shock at the 15% significance level. These predictors were included in the
final model to estimate individual probabilities of SCD at 5 years. The calibration slope was 0.91 (95% CI: 0.74, 1.08),
C-index was 0.70 (95% CI: 0.68, 0.72), and D-statistic was 1.07 (95% CI: 0.81, 1.32). For every 16 ICDs implanted in
patients with ≥4% 5-year SCD risk, potentially 1 patient will be saved from SCD at 5 years. A second model with the
data set split into independent development and validation cohorts had very similar estimates of coefficients and perform-
ance when externally validated.
.....................................................................................................................................................................................
Conclusion This is the first validated SCD risk prediction model for patients with HCM and provides accurate individualized estimates
for the probability of SCD using readily collected clinical parameters.
-----------------------------------------------------------------------------------------------------------------------------------------------------------
Keywords Hypertrophic cardiomyopathy † Sudden cardiac death † Implantable cardioverter defibrillator † Risk prediction
model

in young adults.1,2 Patients at high risk of SCD need to be identified

Introduction so they can be offered lifesaving treatment with an implantable cardi-
Hypertrophic cardiomyopathy (HCM) is a common inherited heart overter defibrillator (ICD). Contemporary guidelines recommend
muscle disorder and a leading cause of sudden cardiac death (SCD) that the sudden death risk is assessed by evaluating clinical

* Corresponding author. Tel: +44 203 456 4801, Fax: +44 207 456 4901, Email: [email protected]
Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2013. For permissions please email: [email protected]
Risk prediction model for SCD in HCM 2011

parameters that reflect the severity of the underlying myocardial physicians). Deaths were assessed without knowledge of the candidate
disease. The presence or absence of these risk factors is then used predictor variables. The primary endpoint was SCD or an equivalent
to guide clinical decision-making with respect to prophylactic ICD event. Sudden cardiac death was defined as witnessed sudden death
implantation.1,2 Although observational cohort studies show that with or without documented ventricular fibrillation or death within 1 h
of new symptoms or nocturnal deaths with no antecedent history of wor-
this approach identifies patients with the greatest risk of SCD, valid-
sening symptoms.15 Aborted SCD during follow-up and appropriate ICD
ation of current algorithms suggests that they overestimate risk,
shock therapy were considered equivalent to SCD.16 – 21 Implantable car-
resulting in inappropriate prophylactic ICD implantation in a substan-
dioverter defibrillator shocks were considered appropriate if the treated
tial number of patients.3 The aim of this study was to derive and val- tachyarrhythmia was ventricular in origin in an identical manner to previ-
idate a new sudden death risk model that can be used to generate ous studies.16 – 21 All ICDs had the capacity to store intracardiac electro-
individualized risk estimates for SCD and improve the targeting of grams.
ICD therapy in patients with HCM.
Selection of predictor variables and coding

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

Methods Risk factors of SCD which were previously examined in multiple survival
studies were considered as candidate predictor variables following a
review of the literature completed in January 2010.22 Supplementary ma-
Study design and overview
terial online, Table S1 summarizes the predictor selection algorithm, the
The prognostic model was derived using data from a retrospective, multi-
relevant studies, and lists the candidate predictor variables. Briefly, clinical
centre longitudinal cohort study. The model presented in this article was
parameters were included as pre-specified predictor variables only when
developed using the entire data set, using the Cox proportional hazards
independently associated with SCD in at least one published multivari-
model, and internally validated using bootstrapping. A secondary model
able survival analysis, and were uniformly defined and collected in all par-
with external validation was developed de novo using a similar modelling
ticipating centres. The pre-specified predictors, their definitions, and
strategy with independent development and validation cohorts.
coding are summarized in Table 1.15,17,18,20,21,23 – 31 Even though both
The study conforms to the principles of the Helsinki declaration. The
left atrial (LA) size and atrial fibrillation (AF) are associated with
sponsors of this study did not have a role in study design, data collection,
SCD,31,32 only the former was included as a pre-specified predictor
analysis, and interpretation. C.O’M., R.O., F.J., and P.E. had access to all
since LA enlargement predisposes to AF33,34 and contained less
data and final responsibility to submit the article. The authors from
missing data. Left ventricular end-diastolic dimension24 was incorporated
each participating centre guarantee the integrity of data from their insti-
into fractional shortening to address the increasing concern of the risk of
tution. All investigators have agreed to the article as written.
SCD in HCM patients with impaired cardiac function.2 Echocardiographi-
cally derived ejection fraction and abnormal blood pressure response to
Study population and participating centres exercise were not selected as pre-specified predictors since they have
The study cohort consisted of all consecutively evaluated patients with not been independently associated with SCD in any multivariable survival
HCM, followed up at six participating European centres: (i) The Heart analyses.15,17,21,29,30
Hospital, London, UK, (ii) A Coruña University Hospital, A Coruña,
Spain, (iii) Unit of Inherited Cardiovascular diseases, 1st Department of
Sample size
Cardiology, University of Athens, Athens, Greece, (iv) Institute of Cardi-
ology, University of Bologna, Bologna, Italy, (v) University Hospital Virgen To ensure that the regression coefficients of the model were estimated
de la Arrixaca, Murcia, Spain, and (vi) Monaldi Hospital, Second Univer- with adequate precision, a minimum of 10 SCD/SCD equivalent events
sity of Naples, Naples, Italy. Patients from this study cohort are reported were required per coefficient estimated by the model.35 The 198 SCD/
in other recently published studies.3 – 12 SCD equivalent endpoints observed in this cohort allow the estimation
Only adult patients (≥16 years of age) with no prior ventricular fibril- of 19 regression coefficients, and were sufficient to develop the model
lation or sustained ventricular tachycardia were studied. Hypertrophic using the 8 pre-specified candidate predictors, perform sensitivity ana-
cardiomyopathy was defined as a maximum left ventricular wall thickness lyses, and develop a model based only on patients with complete data.
≥15 mm unexplained by abnormal loading conditions13 or in accordance
with published criteria for the diagnosis of disease in relatives of patients General statistical methods
with unequivocal disease.14 Patients with known metabolic diseases (e.g. All statistical analyses were carried out using STATA (version 12). Vari-
Anderson-Fabry disease) or syndromic causes of HCM (e.g. Noonan syn- ables are expressed as mean + standard deviation (SD), median and
drome) were excluded from the study. interquartile range (IQR) or counts and percentages as appropriate.
The follow-up time for each patient was calculated from the date of
Patient assessment and data collection their first evaluation to the date of reaching the study endpoint, or
All patients had planned clinical reviews every 6 – 12 months or earlier if death from another cause, or to the date of their most recent evaluation.
there was a change in symptoms. Patients underwent clinical assessment, The annual event rate was calculated by dividing the number of patients
pedigree analysis, physical examination, electrocardiography (resting and reaching the endpoint by the total follow-up period for that endpoint.
ambulatory), and transthoracic echocardiography. Left ventricular The cumulative probability for the occurrence of an outcome was esti-
outflow tract gradients were assessed at rest and on Valsalva. Data mated using the Kaplan– Meier method.
were independently collected at each participating centre using
uniform methodology. Missing data
To determine the degree of bias due to missing data, the characteristics of
Clinical outcomes patients with missing information were compared with those with com-
The cause of death was ascertained by experienced cardiologists at each plete information. Logistic regression was used to identify the predictors
centre using hospital and primary health care records, death certificates, of missingness. Data were assumed to be missing at random, and values
post-mortem reports, and interviews with witnesses (relatives and for the missing predictors were imputed using multiple imputation
2012 C. O’Mahony et al.

Table 1 Pre-specified predictor variables assessed at baseline evaluation

Predictor variable Definition Coding

...............................................................................................................................................................................
Age Age at evaluation.31 Continuous, years
Family history of SCD History of sudden cardiac death in 1 or more first degree relatives under 40 years of age or SCD in a Binary (yes ¼ 1/
first degree relative with confirmed HCM at any age (post- or ante-mortem diagnosis).20,23,29,30 no ¼ 0)
Maximal wall thickness The greatest thickness in the anterior septum, posterior septum, lateral wall, and posterior wall of Continuous, mm
the LV, measured at the level of the mitral valve, papillary muscles, and apex using parasternal
short-axis plane using 2-D echocardiography at time of evaluation15,17,21,24,25,27
Fractional shortening (LV end-diastolic dimension-LV end-systolic dimension)/ LV end-diastolic dimension measured by Continuous, %
M-Mode or 2D echocardiography at time of evaluation24
Left atrial diameter Left atrial diameter determined by M-Mode or 2D echocardiography in the parasternal long axis Continuous, mm

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

plane at time of evaluation31
Maximal left ventricular outflow The maximum LV outflow gradient determined at rest and with Valsalva provocation (irrespective Continuous, mmHg
tract gradients of concurrent medical treatment) using pulsed and continuous wave Doppler from the apical
three- and five-chamber views. Peak outflow tract gradients were determined using the
modified Bernouilli equation: Gradient ¼ 4V 2, where V is the peak aortic outflow
velocity18,20,23,28,29
Non-sustained ventricular ≥3 consecutive ventricular beats at a rate of ≥120 bpm and ,30 s in duration on Holter Binary (yes ¼ 1/
tachycardia monitoring (minimum duration 24 hours) at or prior to evaluation.17,20,29,30 no ¼ 0)
Unexplained syncope History of unexplained syncope at or prior to evaluation.20,21,26,29,31 Binary (yes ¼ 1/
no ¼ 0)

techniques based on chained equations.36 All predictors of missingness SCD.43,44 A value of 0.5 for C-index indicates no discrimination and a
were included in the multiple imputation model, together with the value equal to 1 indicates perfect discrimination. The D-statistic quanti-
outcome, all pre-specified predictors of the risk model, and the estimate fies the observed separation between subjects with low and high pre-
of the cumulative hazard function.37 A total of 25 imputed data sets were dicted risks as predicted by the model and can be interpreted as the
generated and the estimates were combined using Rubin’s rules.38 log hazard ratio for having SCD between the low and high risk groups
Patients with .50% missing predictors were excluded from model devel- of patients. A model with no discriminatory ability will produce a value
opment. of 0 for D-statistic, with increasing values indicating greater separation.

Model development Model presentation

Univariable Cox regression models were fitted for each continuous pre- The probability of SCD at 5 years for an individual patient can be calcu-
dictor to test the assumption of linearity with the outcome. To develop lated using the following equation, derived from the Cox proportional
the final risk model, multivariable Cox regression models were fitted with hazards model:
all predictors and quadratic terms for the continuous predictors where
non-linearity was found. The risk model was developed without centre P̂SCD at 5 years = 1 − S0 (t)exp (Prognostic Index) ,
as a predictor to allow the model to be used in other HCM populations.
To avoid problems related to variable selection when the data are not
where S0(t) is the average survival probability at time t (i.e. at 5 years), and
very large, a 15% significance level was used in the backward elimination
the prognostic index is the sum of the products of the predictors and their
procedure to select the predictors for the final risk model.39 The propor-
coefficients.
tional hazards assumption required by the Cox model was investigated
using Schoenfeld residuals.40 The risk model presented in this article
was developed using the entire cohort, rather than splitting it into Secondary model development with external
smaller development and validation data sets, making efficient use of validation
the data.41 For model development with external validation, the cohort was divided
into separate development and validation cohorts. The development
Model validation cohort consisted of all patients from five of the six participating
Bootstrapping was used to evaluate the performance of the model since centres: A Coruña University Hospital, 1st Department of Cardiology
this is the most efficient validation procedure as all aspects of the model of University of Athens, University of Bologna, University Hospital
development, including variable selection are validated.41 For this Virgen de la Arrixaca, and Monaldi Hospital (n ¼ 2082 with 109 SCD
purpose, 200 bootstrap samples were generated. endpoints). The model was fitted in the development cohort using pena-
The calibration slope was used to assess the degree of agreement lized Ridge Cox regression which adjusts the coefficients for model over-
between the observed and predicted hazards of SCD.42 A value close fitting when the number of events is small.39 The model derived from the
to 1 suggests good overall agreement. Graphical comparisons of the development cohort was then externally validated using The Heart Hos-
observed and predicted SCD at 5 years by risk groups (group cut-offs: pital patients (n ¼ 1593 with 89 SCD endpoints) who were excluded
0 – 2, 2 – 4, 4 – 6, and .6% 5-year risk of SCD) were performed. The from model development, and were thus independent. The same valid-
C-index (C-uno) and D-statistic were used to measure how well the ation measures as the ones used to validate the SCD risk prediction
model discriminated between patients with high and low risk of model derived from the entire data set were used.
Risk prediction model for SCD in HCM 2013

Sensitivity analyses seven of the eight predictors: maximal wall thickness in 11 (0.3%)
A model to adjust for centre effect was developed and validated using patients, fractional shortening in 286 (7.8%) patients, LA diameter
bootstrapping. An additional sensitivity analysis was carried out by devel- in 92 (2.5%) patients, left ventricular outflow gradient in 94 (2.5%)
oping a model using only patients with complete data, with and without patients, NSVT in 453 (12.3%) patients, family history of SCD in 7
centre. The same modelling strategy and validation measures as the ones (0.2%) patients, and unexplained syncope in 65 (1.8%) patients. Miss-
employed to develop and validate the SCD risk prediction model on the ingness was associated with New York Heart Association functional
entire data set were used. class III/IV, amiodarone treatment at baseline, ICD implantation
during follow-up, cardiovascular deaths not secondary to SCD/
Comparison with conventional risk factors heart failure/stroke, non-cardiovascular deaths, and the year of exit
Clinical practice guidelines consider severe hypertrophy (maximal wall from the study.
thickness ≥30 mm), non-sustained ventricular tachycardia (NSVT),
family history of SCD, and unexplained syncope as conventional risk
factors of SCD.1,2 To compare the risk model developed in this article Model development

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

with contemporary clinical practice, a risk score was constructed with The exploratory univariable analyses are shown in Table 4. Age,
a value of 0 (if no conventional risk factor was present), 1 (if one conven- maximal LV wall thickness (mm), LA diameter (mm), maximal left
tional risk factor was present), and 2 (if ≥2 conventional risk factors were ventricular outflow tract gradient (mmHg), family history of SCD,
present). This risk score was fitted as a continuous variable using the NSVT, and unexplained syncope were significantly associated with
entire data and validated using bootstrapping. The C-index and the cali-
SCD at the 15% significance level. Only maximal thickness was
bration slope were calculated.
observed to have a nonlinear association with SCD and hence a quad-
ratic term was included for this predictor. The risk of SCD associated
Clinical implications
with LV wall thickness tended to decrease with extreme hypertrophy
In patients with all the necessary data required to calculate the 5-year
(as assessed by maximal wall thickness). The association of maximal
SCD risk using the proposed model, the implications of ICD implantation
at different thresholds were compared with the observed events at 5
wall thickness and the SCD endpoint is shown in Supplementary ma-
years. Comparisons were carried out in three patient subgroups (0 risk terial online, Table S2. The data satisfied the assumption of propor-
factors, 1 risk factors, ≥2 risk factors). tional hazards. The estimates of the hazard ratios and the
corresponding confidence intervals for the prediction model are
shown in Table 5. The risk of SCD in 5 years for an individual HCM
Results patient can be calculated from the following equation:

Baseline clinical characteristics P̂SCD at 5 years = 1 − 0.998exp (Prognostic Index) ,

The study cohort consisted of 3675 patients. The baseline clinical
characteristics are shown in Table 2. Three thousand one hundred where Prognostic Index ¼ 0.15939858*Maximal wall thickness
twenty nine patients (85%) fulfilled conventional diagnostic criteria (mm) 2 0.00294271*Maximal wall thickness2 (mm2) + 0.0259082*
and 546 patients (15%) criteria for familial disease.14 During follow- Left atrial diameter (mm) + 0.00446131*Maximal left ventricular
up, 100 (3%) patients underwent alcohol septal ablation and 174 outflow tract gradient (mmHg) + 0.4583082*Family history
(5%) septal myectomy (8 patients (0.2%) had both procedures). SCD + 0.82639195*NSVT + 0.71650361*Unexplained syncope
During the study period, a total of 558 (15%) patients were treated 20.01799934*Age at clinical evaluation (years).
with an ICD.

SCD/SCD equivalent events during Model validation

follow-up Bootstrapping validation revealed a calibration slope of 0.91 (95% CI:
0.74, 1.08). Figure 1 illustrates a good agreement between the
During a follow-up period of 24 313 patient years [median 5.7 years;
observed and predicted risk of SCD at 5 years. The C-index was
IQR 2.8– 9.2 years; range 1 month (patient reached SCD endpoint)
0.70 (95% CI: 0.68, 0.72). The D-statistic was 1.07 (95% CI: 0.81,
to 33.6 years (patient censored)], 198 patients (5%) reached the
1.32) suggesting that the hazard of SCD is 2.91 times higher in the
SCD endpoint with an annual rate of 0.81% (95% CI: 0.71, 0.94),
high risk group compared with the hazard in the low risk group as pre-
and a 5-year cumulative incidence of 3.8% (95% CI: 3.1, 4.5). The
dicted by the model.
study outcome consisted of 53 appropriate ICD shocks (27%), 118
SCD (60%), and 27 aborted SCD (14%). Of the 198 SCD endpoints,
33 (17%) occurred within 6 months of baseline evaluation. The Secondary model development
follow-up characteristics by centre are summarized in Table 2. The with external validation
clinical characteristics of patients with and without the SCD endpoint The estimates of hazard ratios from the secondary model developed
are shown in Table 3. from patients excluding the Heart Hospital were similar to those
obtained using the entire data set (Supplementary material online,
Missing data Table S3). The results from external validation were also similar to
Seven hundred ninety six patients (21.7%) had at least one predictor those obtained from the bootstrapped model: calibration slope:
missing. Three patients were excluded from the analysis as they had 0.95 (95% CI: 0.67, 1.24); C-index: 0.67 (95% CI: 0.64, 0.70);
more than 50% of predictors missing. Missing data were present in D-statistic: 1.14 (95% CI: 0.75, 1.53).
 2014
Table 2 Cohort characteristics

All The Heart A Coruña 1St Department of Institute of University Hospital Monaldi
Hospital, UK University Cardiology, University Cardiology, Virgen de la Hospital,
Hospital, Spain of Athens, Greece Bologna, Italy Arrixaca, Spain Italy
.............................................................................................................................................................................................................................................

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

Baseline
Number of patients 3675a 1593 (43%) 590 (16%) 474 (13%) 456 (12%) 406 (11%) 156 (4%)
Male 2349 (64%) 1,018 (64%) 364 (62%) 340 (72%) 292 (64%) 243 (60%) 92 (59%)
Age; yearsb 48 + 17 43 + 15 57 + 15 47 + 16 50 + 17 53 + 17 44 + 16
NYHA III/IV 426 (12%) 136 (8.5%) 76 (13%) 77 (16%) 41 (9%) 82 (20%) 14 (9%)
Myectomy 34 (1%) 17 (1%) 6 (1%) 1 (0.2%) 5 (1%) 2 (0.5%) 3 (2%)
Alcohol septal ablation 10 (0.3%) 4 (0.3%) 2 (0.3%) 0 2 (0.4%) 2 (0.5%) 0
Amiodarone 468 (13%) 217 (14%) 79 (13%) 44 (9%) 78 (17%) 39 (10%) 11 (7%)
ICD 42 (1%) 14 (1%) 6 (1%) 1 (0.2%) 5 (1%) 8 (2%) 8 (5%)
Permanent /persistent AF 366 (10%) 98 (6%) 133 (23%) 39 (8%) 19 (4%) 65 (16%) 12 (8%)
NSVT 634 (17%) 300 (19%) 82 (14%) 56 (12%) 80 (18%) 85 (21%) 31 (20%)
LA diameter; mmb 44 + 8 44 + 8 45 + 8 44 + 6 46 + 9 44 + 8 45 + 8
LVOTGmax; mmHgb 12 (5– 49) 9 (5– 46) 10 (7 –45) 10 (4 –60) 20 (15– 50) 4 (4 –60) 4 (4– 40)
LVedd; mm 45 + 7 44 + 6 45 + 6 47 + 5 44 + 7 44 + 7 46 + 7
MWT; mmb 20 + 5 20 + 6 20 + 5 18 + 4 20 + 5 20 + 5 20 + 5
FS; %b,c 41 + 9 42 + 9 41 + 9 41 + 7 41 + 11 39 + 10 32 + 11
FHSCD 886 (24%) 482 (30%) 42 (7%) 147 (31%) 69 (15%) 110 (27%) 36 (23%)
Unexplained syncope 507 (14%) 274 (17%) 55 (9%) 90 (19%) 11 (2%) 55 (14%) 22 (14%)
.............................................................................................................................................................................................................................................
Follow-up
Enrolment period 1972– 2011 1988–2005 1983–2010 1993–2011 1972–2011 1983– 2009 2001–2010
End of follow-up period 2012 2010 2011 2012 2011 2011 2011
Total patient-years 24 313.4 11 779.4 4009.3 3709.8 2955.7 1274.2 584.6
Median follow-up; years 5.7 (2.8, 9.2) 6.6 (4.2, 10.0) 6.0 (3.5, 9.2) 7.4 (4.5, 11.1) 3.4 (1.2, 8.8) 2.5 (0.8,5.4) 3.8 (1.6, 5.4)
SCD endpoints 198 (5%) 89 (6%) 20 (3%) 25 (5%) 41 (9%) 20 (5%) 3 (2%)
5-year cumulative hazard; % (95% CI) 3.8% (3.1,4.5) 3.8% (2.9,5.0) 1.6% (0.8,3.3) 3.4% (2.1,5.6) 6.1% (3.8,9.7) 5.5% (3.3,9.0) 2.0% (0.7,6.2)

Variables are expressed as mean + standard deviation (SD), median and interquartile range (IQR) or counts and percentages as appropriate.
NYHA, New York Heart Association; ICD, implantable cardioverter defibrillator; AF, atrial fibrillation; NSVT, non-sustained ventricular tachycardia; LA, left atrium; LVOTG, left ventricular outflow tract gradient at rest or Valsalva; LVedd, left
ventricular end diastolic dimension; MWT, maximal wall thickness; FS, fractional shortening; FHSCD, family history of sudden cardiac death; SCD, sudden cardiac death.
a
Three patients were excluded from model development because .50% of predictors were missing. One of the three excluded patients had suffered SCD/appropriate ICD shock.
b
Range: 1%; 99% centiles: age: 16.9; 81.4 years, LA diameter: 28; 67 mm, LVOTGmax: 2; 154 mmHg, MWT: 10; 36 mm, FS: 15; 64%.
c
207 patients (5%) with FS , 27%, 489 patients (13%) with FS.50%.

C. O’Mahony et al.
Risk prediction model for SCD in HCM 2015

Table 3 The clinical characteristics of patients with and

Sensitivity analyses
without the sudden cardiac death endpoint The estimates of hazard ratios from the model adjusting for centre
effects were similar to the developed model without centre and
Clinical Patients without Patients with SCD are shown in Table 5. The calibration slope for this model was
characteristic SCD endpoints, endpoints, n 5 198 0.876 (95% CI: 0.869, 0.883), C-index was 0.69 (95%: CI 0.68,
n 5 3477
................................................................................ 0.71), and D-statistic was 1.115 (95% CI: 1.065, 1.164). The SCD
Male 2207 (63%) 142 (72%) risk models developed using only patients with complete data, with
Age; years 49 + 17 43 + 15 and without centre were also similar to the model developed in
NYHA III/IV 395 (12%) 31 (17%) the primary analysis using bootstrapping (Supplementary material
Myectomy 202 (6%) 6 (3%) online, Tables S4 and S5).
Alcohol septal 96 (3%) 4 (2%)
ablation Comparison with conventional risk factors

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

Amiodarone 412 (12%) 47 (23%) The model with a continuous risk score based on the current practice
Permanent 339 (10%) 27 (13%) of using four conventional risk factors produced a C-index of
/persistent AF
0.54 (95% CI: 0.51, 0.56) and calibration slope of 1.02 (95% CI:
NSVT 572 (19%) 62 (31%)
0.73, 1.30).
LA diameter; mm 44 + 8 46 + 9
LVOTGmax; 11 (5 –49) 18 (6–58)
mmHg Clinical implications
LVedd; mm 45 + 6 45 + 8 The clinical implications of using the model were examined in 3066
MWT; mm 19 + 5 22 + 6 patients with the necessary data to calculate the 5-year SCD risk.
FS; % 41 + 9 41 + 10 Table 6 shows the simulated effect of using different thresholds of
FHSCD 813 (23%) 73 (37%) 5-year SCD risk to implant an ICD in patient subgroups defined
Unexplained 455 (13%) 52 (26%) using conventional risk factors. The 5-year SCD risk estimates that
syncope correctly identified the maximum number of SCD endpoints at
5 years and simultaneously minimized ICD implantation in patients
Variables are expressed as mean + standard deviation (SD), median and without the SCD endpoint were ≥6% for patients with ≥2 conven-
interquartile range (IQR) or counts and percentages as appropriate
NYHA, New York Heart Association; ICD, implantable cardioverter defibrillator; tional risk factors (Figure 2A) and ≥4% for patients with a single risk
AF, atrial fibrillation; NSVT, non-sustained ventricular tachycardia; LA, left atrium; factor (Figure 2B). In patients with no risk factors where contempor-
LVOTG, left ventricular outflow tract gradient at rest or Valsalva; LVedd, left ary guidelines do not recommend primary prophylaxis ICD,2 a ≥3%
ventricular end diastolic dimension; MWT, maximal wall thickness; FS, fractional
shortening; FHSCD, family history of sudden cardiac death; SCD, sudden cardiac 5-year SCD risk would have resulted in ICD implantation in 8 (32%)
death. of 25 patients with SCD endpoints and 264 (17%) patients without
SCD endpoints (Figure 2C).

Table 4 Summary of the characteristics of patients with sudden cardiac death endpoints and univariable Cox regression
models

Predictor variable SCD group characteristics Hazard ratio 95% confidence P-value
(n 5 198)a,b interval
...............................................................................................................................................................................
Age (years) 42.5 + 15 0.988 0.979, 0.997 0.007
Maximal wall thickness (mm) 21.5 + 6 1.048 1.025, 1.071 ,0.001
Fractional shortening (%) 41.0 + 10 0.992 0.977, 1.008 0.344
Left atrial diameter (mm) 46.2 + 9 1.035 1.018, 1.052 ,0.001
Left ventricular outflow gradient (mmHg) 18 (6 –58) 1.005 1.001, 1.008 0.005
Family history of sudden cardiac death 73 (37%) 1.760 1.318, 2.350 ,0.001
Non-sustained ventricular tachycardia 62 (31%) 2.533 1.849, 3.469 ,0.001
Unexplained syncope 52 (26%) 2.326 1.693, 3.195 ,0.001

Variables are expressed as mean + standard deviation (SD), median and interquartile range (IQR) or counts and percentages as appropriate.
a
Range of values (minimum; maximum) in SCD group: age: 16.3; 77.4 years, maximal wall thickness: 9; 37 mm, fractional shortening: 15; 62%, left atrial diameter: 28; 70 mm, maximal left
ventricular outflow tract gradient: 2;190 mmHg.
b
Missing data in SCD group: maximal wall thickness: 3%, fractional shortening: 11%, left atrial diameter: 6%, left ventricular outflow tract gradient: 3%, non-sustained ventricular
tachycardia: 19%, unexplained syncope: 2%.
2016 C. O’Mahony et al.

Table 5 Sudden cardiac death risk prediction model and sensitivity analysis for centre effect

Predictor variable SCD risk prediction model Sensitivity analysis: model with centre
................................................................. ..................................................................
Hazard ratio (95% confidence interval) P-value Hazard ratio (95% confidence interval) P-value
...............................................................................................................................................................................
Age (years) 0.98 (0.97, 0.99) 0.001 0.98 (0.97, 0.99) ,0.001
Maximal wall thickness (mm) 1.17 (1.01, 1.37) 0.042 1.15 (0.99, 1.35) 0.069
Maximal wall thickness2 (mm2) 0.997 (0.99, 1.0003) 0.078 0.997 (0.99, 1.001) 0.116
Left atrial diameter (mm) 1.03 (1.01, 1.05) 0.006 1.02 (1.01, 1.04) 0.008
LV outflow gradient (mmHg) 1.004 (1.001, 1.01) 0.021 1.004 (1.0004, 1.01) 0.031
Family history SCD 1.58 (1.18, 2.13) 0.002 1.62 (1.20, 2.19) 0.002
NSVT 2.29 (1.64, 3.18) ,0.001 2.14 (1.53, 2.99) ,0.001

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

Unexplained syncope 2.05 (1.48, 2.82) ,0.001 2.29 (1.64, 3.20) ,0.001
Centre
London — Baseline: 1 (– ) 0.0015
Athens — 0.999 (0.64, 1.57)
Bologna — 2.08 (1.39, 3.12)
Coruna — 1.08 (0.65, 1.79)
Murcia — 2.04 (1.22, 3.41)
Naples — 0.73 (0.23, 2.33)

LV, left ventricle; NSVT, non-sustained ventricular tachycardia.

A 5-year SCD risk of ≥4% identified 60 (71%) of 84 SCD end- The usefulness of this model lies in providing accurate prognostic
points with 896 (30%) ICD implants in patients without SCD at 5 information that aids clinical decision making. The model achieved
years (Figure 2D). For every 16 ICD implantations in patients with this by showing good agreement between the predicted and
≥4% 5-year SCD risk, 1 patient can potentially be saved from SCD observed hazards of SCD and by demonstrating the ability to separ-
at 5 years. Patients not reaching the SCD endpoint at 5 years ate patients with regard to their 5-year risk of SCD.42 The C-indices
(n ¼ 2982) had a mean predicted 5-year SCD risk of 3.7% (IQR: indicated that the proposed risk prediction model has superior dis-
1.8 –4.5%), while the corresponding figures for those reaching the crimination compared with the model of conventional risk factors
SCD endpoint (n ¼ 84) were 7.3% (IQR: 3.4 –10%). used in contemporary clinical practice. The sensitivity analysis
demonstrated that the relationship between the predictors and
SCD remains unchanged with the inclusion of centre in the model
and the risk model without centre is proposed for general clinical use.
Discussion The risk prediction model has the potential to improve the man-
This is the first validated risk prediction model for SCD in patients agement of patients with a solitary and multiple risk factors by simul-
with HCM. The model was derived from a large, diverse, and well- taneously reducing unnecessary and potentially harmful ICD
characterized population of patients followed at six different Euro- implants in patients who do not suffer SCD and correctly identifying
pean centres and provides accurate, individualized estimates for the majority of those who suffer SCD and are most likely to benefit
the probability of SCD using readily collected clinical parameters. from an ICD. Currently, patients without conventional risk factors
The broad patient inclusion criteria of the study mean that the are reassured and reassessed and are not routinely offered ICD
model can be used in the majority of adult patients with HCM, includ- therapy.1,2 However, approximately one-third of all SCD come
ing those with mild disease identified during family screening. from this subgroup of patients, and contemporary management strat-
Current clinical guidelines for HCM in the USA and Europe recom- egies fail to address this problem. The risk prediction model may help
mend SCD risk algorithms based on a simple summation of a limited identify a small proportion of SCD in this group which represents an
number of binary clinical parameters (NSVT, severe hypertrophy, un- improvement when compared with current clinical practice, but the
explained syncope, family history of SCD, and abnormal BPRE).1,2 performance of the model in this patient subgroup is not optimal.
Even though this approach has been used in clinical practice for The probability of SCD at 5 years is derived from a range of readily
more than a decade, it provides only a very crude estimate of relative available clinical parameters, each with a unique contribution to risk.
risk of SCD and fails to account for the different effect size of individ- For example, consider the management of two patients with NSVT,
ual risk factors.3 Moreover, some risk factors such as hypertrophy are maximal wall thickness of 23 mm, and LA diameter of 44 mm. One is
considered as binary variables when in fact they are associated with a aged 24 years with a resting LVOTGmax of 64 mmHg and the other is
continuous increase in SCD risk.25 As a result, existing algorithms 64 years old with an LVOTGmax of 36 mmHg. Current guidelines treat
have a low positive predictive accuracy for SCD that results in the un- these two patients identically as they each have a single risk factor
necessary treatment of patients who are at intrinsically low risk.3 (NSVT). By applying the clinical risk prediction model in this clinical
Risk prediction model for SCD in HCM 2017

this level of risk might have different implications in an otherwise well

20 year old compared to a 70-year-old patient with significant co-
morbidity. When deciding on device treatment, physicians and
patients have to balance the benefits of protection from SCD
against the potential hazards of therapy. Approximately one-third of
ICD recipients experience implant complications or inappropriate
shocks after 5 years, and while the majority of ICD-related adverse
events are not life threatening, they often require hospitalization and
additional invasive procedures.10 In addition, the impact of ICD
therapy on employment, driving, and recreational activities has to be
considered. Ultimately, the decision on treatment rests on the relative
weights of the risks and benefits of ICD therapy in individual patients.42

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

Future development of the model
Prognosis is the outcome of the interaction of host, disease, and en-
vironmental factors which can be considered as predictors in prog-
nostic models.42,47 The candidate predictors consisted primarily of
disease factors that reflect the severity of the cardiac phenotype.
Of the potential predictors relating to the host, only age was included
on the basis of previously published studies. When updating the
model, other factors, such as sex and ethnic background as well as en-
vironmental parameters such as social deprivation and access to
health care should be investigated to improve the prognostic per-
formance of the model. These factors may be responsible for the
observed centre effect in our study. The incremental predictive
value of genetic information and late gadolinium enhancement
imaging can also be examined in future iterations of the model. The
ultimate test of the usefulness of a prediction model is an impact
Figure 1 (A) Calibration by risk group. Vertical bars represent
study that determines whether the application of the model in
observed (black) and model-based predicted (grey) predicted
routine clinical practice improves decision making, patient outcomes,
probabilities of sudden cardiac death in 5 years. The four risk
groups (1 – 4) were created using model-based predicted probabil-
and cost effectiveness.45
ities (0 – 2, 2 – 4, 4 – 6, and .6% 5-year risk of SCD).These groups
are selected for the purposes of validation rather than clinical deci-
Limitations
sion making. (B) Plot showing the proportion surviving by risk, as No risk stratification strategy will ever be able to predict SCD with
predicted by the model. The graph compares the risk of SCD as pre- absolute certainty and the risk prediction model should be used by
dicted from the model with the observed SCD endpoints at 5-years physicians experienced in the management of the condition. The
(combining 5 random imputations). The horizontal axis labels indi- model should be used only in patients with similar characteristics
cate the risk groups, e.g. [0 – 1) indicates a predicted risk of 0 to ,1, to the study cohort. The model is not validated in paediatric patients
[1 – 2) is a risk of 1 to ,2, etc. (,16 years), elite athletes, or in individuals with metabolic diseases
(e.g. Anderson-Fabry disease) and syndromes (e.g. Noonan syn-
drome). Patients with a previous history of aborted SCD or sustained
scenario, the 5-year risk of SCD is estimated at 9.7% for the first ventricular arrhythmia were specifically excluded from the cohort
patient and at 4.3% for the latter. We envisage that an on-line risk and should continue to be treated with an ICD for secondary preven-
calculator can be used to estimate the SCD risk, and this information tion. The model does not account for patients with purely exercise
can then be shared and discussed with patients and their families induced left ventricular outflow tract obstruction48 or for the
(www.HCMRisk.org). effect of myectomy or alcohol septal ablation and thus should be
The aim of this prediction model, like any other prognostic tool, is used cautiously in such patients. Patients with drug refractory symp-
not to render physicians’ clinical judgment obsolete, but rather to tomatic left ventricular outflow tract should be considered for septal
complement clinical reasoning by providing objective individualized reduction therapy, irrespective of their risk of SCD. On the other
prognostic information.45,46 Guidelines recommend ICD therapy hand, asymptomatic patients with left ventricular outflow tract ob-
for the primary prevention of SCD in high risk patients,1,2 but there struction should be treated conservatively and the severity of ob-
is no international consensus on the absolute SCD risk that justifies struction should be used in risk stratification. In patients with
ICD therapy. The intention of the proposed SCD risk model is not exercise induced ventricular arrhythmias,20 strong consideration of
to categorize patients into simplistic high or low risk groups with pre- ICD implantation should be considered. The effect of amiodarone
defined therapeutic strategies, but to treat SCD risk as a continuum, and b-blocker treatment on the risk of SCD was not examined.
interpreted within each patient’s clinical context. So while a 5-year The SCD risk tended to fall in patients with extreme hypertrophy.
SCD risk of ≥4% identified 71% of patients that had a SCD endpoint, Patients with a maximal wall thickness≥ 35 mm had a very low SCD
2018 C. O’Mahony et al.

Table 6 The simulated effect of using different thresholds of 5-year sudden cardiac death risk to implant an implantable
cardioverter defibrillator in patient groups defined using conventional risk factors as recommended in current treatment
guidelines

Patient groups defined using four Estimate for 5-year SCD risk Patients reaching SCD Patients without SCD
conventional risk factors used to implant ICD (%) endpointa endpointb
................................. ...................................
ICD No ICD No ICD ICD
implant Implant implant implant
...............................................................................................................................................................................
Two or more conventional risk factors: 412 patients ≥3 32 (100%) 0 9 (2%) 371 (98%)
with 32 SCD endpoints at 5 years ≥4 32 (100%) 0 32 (8%) 348 (92%)
≥5 32 (100%) 0 69 (18%) 311 (82%)
≥6 31 (97%) 1 (3%) 131 (34%) 249 (66%)

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

Single conventional risk factor: 1074 patients ≥3 27 (100%) 0 338 (32%) 709 (68%)
with 27 SCD-endpoints at 5 years ≥4 24 (89%) 3 (11%) 580 (66%) 467 (44%)
≥5 18 (67%) 9 (33%) 754 (62%) 293 (28%)
≥6 12 (44%) 15 (56%) 877 (84%) 170 (16%)
No conventional risk factors: 1580 patients with ≥2 18 (72%) 7 (28%) 822 (53%) 733 (47%)
25 SCD endpoints at 5 years ≥3 8 (32%) 17 (68%) 1291 (83%) 264 (17%)
≥4 4 (16%) 21 (84%) 1474 (95%) 81 (5%)
≥5 0 25 (100%) 1531 (99%) 24 (1%)
All patients: 3066 patients with 84 SCD ≥3 67 (80%) 17 (20%) 1638 (55%) 1344 (45%)
endpoints at 5 yearsc ≥4 60 (71%) 24 (29%) 2086 (70%) 896 (30%)
≥5 50 (60%) 34 (40%) 2354 (79%) 628 (21%)
≥6 43 (51%) 41 (49%) 2556 (86%) 426 (14%)

a
Denominator for %¼ number of SCD endpoints.
b
Denominator for %¼ number of patients without SCD endpoints.
c
During the study period of 1972 –2011, of the 3066 patients with the necessary data required to calculate the 5-year SCD risk, 49% with ≥2 risk factors, 19% with a single risk factor,
and 5% with no conventional risk factors had an ICD.

rate (Supplementary material online, Table S2) which is reflected in Hypertrophic Cardiomyopathy
the nonlinear relation between maximal wall thickness and 5-year
SCD risk. This subgroup was also small in size, and pending further Outcomes Investigators (www.
analysis, the model should also be used cautiously in patients with HCMRisk.org)
maximal wall thickness≥ 35 mm.
This work is not an epidemiological study exploring new associations The following were also investigators in the study: Martin Ortiz-
with SCD. Instead, previously described associations were used to Genga MD (A Coruña University Hospital, Spain), Xusto Fernandez
develop a prognostic model. Predictors were only included if shown MD (A Coruña University Hospital, Spain), Vassiliki Vlagouli MD
to be independently associated with SCD in published multivariable (University of Athens, Greece), Chris Stefanadis MD (University of
survival analyses. This strategy achieved a high event-to-predictor Athens, Greece), Fabio Coccolo MD (University of Bologna, Italy),
ratio which improves the generalizability and accuracy of predictions. Maria Jose Oliva Sandoval MD (Murcia, Spain), Giuseppe Pacileo
Although prospective cohort studies are desirable,46 the low rate MD (Monaldi Hospital, Italy), Daniele Masarone MD (Monaldi
of SCD in HCM made a prospective design impractical. All participat- Hospital, Italy), Antonios Pantazis MD (The Heart Hospital, UK),
ing centres assess HCM patients in an almost identical manner, which Maite Tome-Esteban MD (The Heart Hospital, UK), Shaughan Dickie
allowed the use of retrospectively collected data from all sites. The (The Heart Hospital, UK), Pier D Lambiase PhD (The Heart Hospital,
prediction model will be further improved through the incorporation UK), and Shafiq Rahman PhD (University College London, UK).
of additional cohorts, thus increasing the number of events that will
reduce model optimism.42 Authors’ contributions
C.O’M designed the study, collected and interpreted the data, carried
out the descriptive statistical analysis, and wrote the article. P.M.E.
Conclusions designed the study, interpreted the data, and wrote the article.
The risk prediction model proposed in this study provides accurate R.Z.O. was involved in study design, led the statistical aspects of
prognostic information regarding SCD in HCM. The model provides the risk modelling and wrote the article. F.J. carried out the statistical
the basis for a rational and informed approach to treatment that analysis. Dr Rahman evaluated some of the measures used for model
empowers patients by helping them to understand the relative validation as part of his PhD thesis with Prof Omar. M.P. carried out
merits of prophylactic therapies including implantable cardioverter the statistical analysis in relation to the external validation of the
defibrillators. model. LM., A.A., E.B., J.R.G., G.L., and C.R. collected and interpreted
 Risk prediction model for SCD in HCM
Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024
Figure 2 The impact of the risk prediction model on clinical decision making. Conventional risk factors (wall thickness ≥30 mm, NSVT, family history of SCD and unexplained syncope) were used to
classify patients in those with multiple (A), single (B), no risk factors (C). Each figure shows the implications of implanting an ICD in all (bottom bar) or none of the patients (top bar), while the rest of the bars
show the impact of using a threshold of ≥3, ≥4, ≥5, ≥6 5-year SCD risk to implant an ICD. The impact on the whole cohort is shown in (D). The raw data are shown in Table 6. Only patients with all data
necessary for the 5-year SCD risk calculation were included (n ¼ 3066).

2019
2020 C. O’Mahony et al.

the data and critically reviewed the manuscript. Ortiz-Genga, Fernan- 4. Barriales-Villa R, Centurion-Inda R, Fernandez-Fernandez X, Ortiz MF,
Perez-Alvarez L, Rodriguez GI, Hermida-Prieto M, Monserrat L. Severe cardiac con-
dez, Vlagouli, Sandoval, Pacileo, Pantazis, Tome-Esteban, Dickie,
duction disturbances and pacemaker implantation in patients with hypertrophic car-
Lambiase, Stefanadis, Coccolo, and Masarone were involved in data diomyopathy. Rev Esp Cardiol 2010;63:985 – 988.
collection and interpretation. W.J.M. was involved in the drafting of 5. Di Salvo G, Pacileo G, Limongelli G, Baldini L, Rea A, Verrengia M, D’Andrea A,
the article and revising it critically for important intellectual content. Russo MG, Calabro R. Non sustained ventricular tachycardia in hypertrophic cardio-
myopathy and new ultrasonic derived parameters. J Am Soc Echocardiogr 2010;23:
581 –590.
6. Oliva-Sandoval MJ, Ruiz-Espejo F, Monserrat L, Hermida-Prieto M, Sabater M,
Ethics approval Garcia-Molina E, Ortiz M, Rodriguez-Garcia MI, Nunez L, Gimeno JR,
Castro-Beiras A, Valdes M. Insights into genotype-phenotype correlation in hyper-
Patients at A Coruña University Hospital (Spain), Unit of Inherited trophic cardiomyopathy. findings from 18 Spanish families with a single mutation in
Cardiovascular diseases at the 1st Department of Cardiology MYBPC3. Heart 2010;96:1980 –1984.
(Athens), University Hospital Virgen de la Arrixaca (Spain), and 7. Monserrat L, Lopez B, Gonzalez A, Hermida M, Fernandez X, Ortiz M,
Barriales-Villa R, Castro-Beiras A, Diez J. Cardiotrophin-1 plasma levels are asso-
Monaldi Hospital (Italy) provided written informed consent. The ciated with the severity of hypertrophy in hypertrophic cardiomyopathy. Eur

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

ethics committees at Heart Hospital (UK), and Institute of Cardi- Heart J 2011;32:177–183.
ology at the University of Bologna (Italy) were informed, but approval 8. Pacileo G, Baldini L, Limongelli G, Di Salvo G, Iacomino M, Capogrosso C, Rea A,
D’Andrea A, Russo MG, Calabro R. Prolonged left ventricular twist in cardiomyop-
was not required under local research governance arrangements.
athies: a potential link between systolic and diastolic dysfunction. Eur J Echocardiogr
2011;12:841 –849.
9. O’Mahony C, Lambiase PD, Rahman SM, Cardona M, Calcagnino M, Quarta G,
Supplementary material Tsovolas K, Al Shaikh S, McKenna W, Elliott P. The relation of ventricular arrhythmia
electrophysiological characteristics to cardiac phenotype and circadian patterns in
Supplementary material is available at European Heart Journal online. hypertrophic cardiomyopathy. Europace 2012;14:724 – 733.
10. O’Mahony C, Lambiase PD, Quarta G, Cardona M, Calcagnino M, Tsovolas K, Al
Shaikh S, Rahman SM, Arnous S, Jones S, McKenna W, Elliott P. The long-term sur-
Acknowledgements vival and the risks and benefits of implantable cardioverter defibrillators in patients
The authors would like to thank Dr Gareth Ambler (University with hypertrophic cardiomyopathy. Heart 2012;98:116 – 125.
College London) for helpful discussions on risk modeling method- 11. Peteiro J, Bouzas-Mosquera A, Fernandez X, Monserrat L, Pazos P,
Estevez-Loureiro R, Castro-Beiras A. Prognostic value of exercise echocardiog-
ology. raphy in patients with hypertrophic cardiomyopathy. J Am Soc Echocardiogr 2012;
25:182 – 189.
Funding 12. Biagini E, Lorenzini M, Olivotto I, Rocchi G, Lovato L, Lai F, Rosmini S, Pazzi C,
Pasquale F, Reggiani ML, Fattori R, Rapezzi C. Effects of myocardial fibrosis assessed
This work was undertaken at University College London Hospitals /Uni- by MRI on dynamic left ventricular outflow tract obstruction in patients with hyper-
versity College London who received a proportion of funding from the trophic cardiomyopathy: a retrospective database analysis. BMJ Open 2012;2:
United Kingdom Department of Health’s National Institute for Health e001267. doi:10.1136/bmjopen-2012-001267.
Research Biomedical Research Centres funding scheme. L.M. received 13. Elliott P, Andersson B, Arbustini E, Bilinska Z, Cecchi F, Charron P, Dubourg O,
Kuhl U, Maisch B, McKenna WJ, Monserrat L, Pankuweit S, Rapezzi C, Seferovic P,
a Carlos III Health Institute grant (PI11/0260). G.L. received support
Tavazzi L, Keren A. Classification of the cardiomyopathies: a position statement
from a Progetto ministeriale (MIUR) PRIN 2008. J.G. received support from the European Society of Cardiology Working Group on Myocardial and Peri-
from the Cardiovascular Research Network (RECAVA) from the cardial Diseases. Eur Heart J 2008;29:270 –276.
Carlos III Health Institute (C03/01,RD06/0014/0017). 14. McKenna WJ, Spirito P, Desnos M, Dubourg O, Komajda M. Experience from clinical
genetics in hypertrophic cardiomyopathy: proposal for new diagnostic criteria in
Conflict of interest: L.M. is shareholder in Health in Code SL. All other adult members of affected families. Heart 1997;77:130 –132.
authors have no conflicts of interest to declare. 15. Elliott PM, Poloniecki J, Dickie S, Sharma S, Monserrat L, Varnava A, Mahon NG,
McKenna WJ. Sudden death in hypertrophic cardiomyopathy: identification of
high risk patients. J Am Coll Cardiol 2000;36:2212 –2218.
References 16. Olivotto I, Gistri R, Petrone P, Pedemonte E, Vargiu D, Cecchi F. Maximum left ven-
1. Maron BJ, McKenna WJ, Danielson GK, Kappenberger LJ, Kuhn HJ, Seidman CE, tricular thickness and risk of sudden death in patients with hypertrophic cardiomy-
Shah PM, Spencer WH III, Spirito P, Ten Cate FJ, Wigle ED, ACCF Task Force on Clin- opathy. J Am Coll Cardiol 2003;41:315 –321.
ical Expert Consensus Documents Members, Vogel RA, Abrams J, Bates ER, 17. Monserrat L, Elliott PM, Gimeno JR, Sharma S, Penas-Lado M, McKenna WJ. Non-
Brodie BR, Danias PG, Gregoratos G, Hlatky MA, Hochman JS, Kaul S, sustained ventricular tachycardia in hypertrophic cardiomyopathy: an independent
Lichtenberg RC, Lindner JR, O’Rourke RA, Pohost GM, Schofield RS, Tracy CM, marker of sudden death risk in young patients. J Am Coll Cardiol 2003;42:873 –879.
Winters WL Jr, ESC Committee for Practice Guidelines Members, Klein WW, 18. Maron MS, Olivotto I, Betocchi S, Casey SA, Lesser JR, Losi MA, Cecchi F, Maron BJ.
Priori SG, Alonso-Garcia A, Blomstrom-Lundqvist C, De Backer G, Deckers J, Effect of left ventricular outflow tract obstruction on clinical outcome in hyper-
Flather M, Hradec J, Oto A, Parkhomenko A, Silber S, Torbicki A. American trophic cardiomyopathy. N Engl J Med 2003;348:295 – 303.
College of Cardiology/European Society of Cardiology Clinical Expert Consensus 19. Adabag AS, Casey SA, Kuskowski MA, Zenovich AG, Maron BJ. Spectrum and prog-
Document on Hypertrophic Cardiomyopathy: a report of the American College nostic significance of arrhythmias on ambulatory Holter electrocardiogram in hyper-
of Cardiology Foundation Task Force on Clinical Expert Consensus Documents trophic cardiomyopathy. J Am Coll Cardiol 2005;45:697–704.
and the European Society of Cardiology Committee for Practice Guidelines. Eur 20. Gimeno JR, Tome-Esteban M, Lofiego C, Hurtado J, Pantazis A, Mist B, Lambiase P,
Heart J 2003;24:1965 –1991. McKenna WJ, Elliott PM. Exercise-induced ventricular arrhythmias and risk of
2. Gersh BJ, Maron BJ, Bonow RO, Dearani JA, Fifer MA, Link MS, Naidu SS, sudden cardiac death in patients with hypertrophic cardiomyopathy. Eur Heart J
Nishimura RA, Ommen SR, Rakowski H, Seidman CE, Towbin JA, Udelson JE,
2009;30:2599 –2605.
Yancy CW. 2011 ACCF/AHA Guideline for the Diagnosis and Treatment of Hyper-
21. Efthimiadis GK, Parcharidou DG, Giannakoulas G, Pagourelias ED, Charalampidis P,
trophic Cardiomyopathy: Executive Summary: A Report of the American College of
Savvopoulos G, Ziakas A, Karvounis H, Styliadis IH, Parcharidis GE. Left ventricular
Cardiology Foundation/American Heart Association Task Force on Practice Guide-
outflow tract obstruction as a risk factor for sudden cardiac death in hypertrophic
lines. Circulation 2011;124:2761 – 2796.
cardiomyopathy. Am J Cardiol 2009;104:695 –699.
3. O’Mahony C, Esteban MTT, Lambiase PD, Pantazis AA, Dickie S, McKenna WJ,
22. O’Mahony C, Elliott P, McKenna W. Sudden cardiac death in hypertrophic cardiomy-
Elliott PM. A validation study of the 2003 American College of Cardiology/European
Society of Cardiology and 2011 American College of Cardiology Foundation/Ameri- opathy. Circ Arrhythm Electrophysiol 2012;6:443 –451.
can Heart Association risk stratification and treatment algorithms for sudden cardiac 23. Maki S, Ikeda H, Muro A, Yoshida N, Shibata A, Koga Y, Imaizumi T. Predictors of
death in patients with hypertrophic cardiomyopathy. Heart 2013;99:534 –541. sudden cardiac death in hypertrophic cardiomyopathy. Am J Cardiol 1998;82:774–778.
Risk prediction model for SCD in HCM 2020a

24. Spirito P, Bellone P, Harris KM, Bernabo P, Bruzzi P, Maron BJ. Magnitude of left ven- 34. Olivotto I, Cecchi F, Casey SA, Dolara A, Traverse JH, Maron BJ. Impact of atrial fib-
tricular hypertrophy and risk of sudden death in hypertrophic cardiomyopathy. N rillation on the clinical course of hypertrophic cardiomyopathy. Circulation 2001;104:
Engl J Med 2000;342:1778 –1785. 2517 –2524.
25. Elliott PM, Gimeno B Jr, Mahon NG, Poloniecki JD, McKenna WJ. Relation between 35. Peduzzi P, Concato J, Feinstein AR, Holford TR. Importance of events per independ-
severity of left-ventricular hypertrophy and prognosis in patients with hypertrophic ent variable in proportional hazards regression analysis. II. Accuracy and precision of
cardiomyopathy. Lancet 2001;357:420–424. regression estimates. J Clin Epidemiol 1995;48:1503 –1510.
26. Kofflard MJM, Ten Cate FJ, van der Lee C, van Domburg RT. Hypertrophic cardio- 36. van Buuren S, Boshuizen HC, Knook DL. Multiple imputation of missing blood pres-
myopathy in a large community-based population: clinical outcome and identifica- sure covariates in survival analysis. Stat Med 1999;18:681 – 694.
tion of risk factors for sudden cardiac death and clinical deterioration. J Am Coll 37. White IR, Royston P, Wood AM. Multiple imputation using chained equations: issues
Cardiol 2003;41:987 –993. and guidance for practice. Stat Med 2011;30:377 –399.
27. Autore C, Bernabo P, Barilla CS, Bruzzi P, Spirito P. The prognostic importance of left 38. Rubin D. Multiple Imputation for Nonresponse in Surveys. New York: John Wiley and
ventricular outflow obstruction in hypertrophic cardiomyopathy varies in relation to Sons; 1987.
the severity of symptoms. J Am Coll Cardiol 2005;45:1076 –1080. 39. Ambler G, Seaman S, Omar RZ. An evaluation of penalised survival methods for
developing prognostic models with rare events. Stat Med 2012;31:1150 –1161.
28. Olivotto I, Maron MS, Adabag AS, Casey SA, Vargiu D, Link MS, Udelson JE, Cecchi F,
40. Schoenfeld DAVI. Partial residuals for the proportional hazards regression model.
Maron BJ. Gender-related differences in the clinical presentation and outcome of
Biometrika 1982;69:239 – 241.
hypertrophic cardiomyopathy. J Am Coll Cardiol 2005;46:480 –487.

Downloaded from https://academic.oup.com/eurheartj/article/35/30/2010/467191 by guest on 01 October 2024

41. Harrell FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic Re-
29. Elliott PM, Gimeno JR, Tome MT, Shah J, Ward D, Thaman R, Mogensen J,
gression, and Survival Analysis, 1 edn. Springer; 2001.
McKenna WJ. Left ventricular outflow tract obstruction and sudden death
42. Steyerberg EW. Clinical Prediction Models. A Practical Approach to Development, Valid-
risk in patients with hypertrophic cardiomyopathy. Eur Heart J 2006;27:
ation and Updating, 1st edn. New York: Springer Science+Business Media; 2009.
1933 – 1941.
43. Royston P, Sauerbrei W. A new measure of prognostic separation in survival data.
30. D’Andrea A, Caso P, Severino S, Cuomo S, Capozzi G, Calabro P, Cice G, Ascione L, Stat Med 2004;23:723 –748.
Scherillo M, Calabro R. Prognostic value of intra-left ventricular electromechanical 44. Gonen M, Heller G. Concordance probability and discriminatory power in propor-
asynchrony in patients with hypertrophic cardiomyopathy. Eur Heart J 2006;27: tional hazards regression. Biometrika 2005;92:965 –970.
1311– 1318. 45. Moons KG, Altman DG, Vergouwe Y, Royston P. Prognosis and prognostic research:
31. Spirito P, Autore C, Rapezzi C, Bernabo P, Badagliacca R, Maron MS, Bongioanni S, application and impact of prognostic models in clinical practice. BMJ 2009;338:b606.
Coccolo F, Estes NA, Barilla CS, Biagini E, Quarta G, Conte MR, Bruzzi P, Maron BJ. 46. Moons KG, Royston P, Vergouwe Y, Grobbee DE, Altman DG. Prognosis and prog-
Syncope and risk of sudden death in hypertrophic cardiomyopathy. Circulation 2009; nostic research: what, why, and how? BMJ 2009;338:b375.
119:1703 –1710. 47. Hemingway H. Prognosis research: why is Dr. Lydgate still waiting? J Clin Epidemiol
32. Sorajja P, Ommen SR, Nishimura RA, Gersh BJ, Berger PB, Tajik AJ. Adverse progno- 2006;59:1229 – 1238.
sis of patients with hypertrophic cardiomyopathy who have epicardial coronary 48. Shah JS, Esteban MTT, Thaman R, Sharma R, Mist B, Pantazis A, Ward D, Kohli SK,
artery disease. Circulation 2003;108:2342 –2348. Page SP, Demetrescu C, Sevdalis E, Keren A, Pellerin D, McKenna WJ, Elliott PM.
33. Cecchi F, Olivotto I, Montereggi A, Santoro G, Dolara A, Maron BJ. Hypertrophic Prevalence of exercise-induced left ventricular outflow tract obstruction in symp-
cardiomyopathy in tuscany: clinical course and outcome in an unselected regional tomatic patients with non-obstructive hypertrophic cardiomyopathy. Heart 2008;
population. J Am Coll Cardiol 1995;26:1529 –1536. 94:1288 –1294.