The amount of late gadolinium enhancement outperforms current guideline-recommended criteria in the identification of patients with hypertrophic cardiomyopathy at risk of sudden cardiac death

This study was a multicenter analysis of patients with HCM undergoing CMR for diagnostic confirmation and/or risk stratification at Instituto do Coração (InCor), University of São Paulo (São Paulo, Brazil), Hospital da Luz (Lisboa, Portugal), Hospital de Santa Cruz (Lisboa, Portugal) and Hospital dos Lusíadas (Lisboa, Portugal). The initial study cohort consisted of 896 consecutive patients identified retrospectively in the databases maintained by the 4 laboratories. The Brazilian center included patients between 2003 and 2017 (n = 677) while the Portuguese centers included patients between 2009 and 2017 (n = 219). The diagnosis of HCM was made by experienced cardiologists using all available clinical data and based on typical features, with ventricular myocardial hypertrophy (left ventricular [LV] wall thickness ≥ 15 mm) occurring in the absence of any other disease responsible for hypertrophy [1]. Exclusion criteria comprised: 1) age < 16 years (n = 11); 2) CMR inconsistent with HCM or with an alternative diagnosis, such as athlete’s heart, Anderson-Fabry disease, cardiac amyloidosis or sarcoidosis (n = 53); 3) LV ejection fraction (LVEF) ≤ 35% by CMR (n = 5); 4) LGE pattern consistent with previous myocardial infarction (n = 3); 5) missing essential echocardiographic and/or 24 h Holter monitoring data (n = 313); 6) moderate or severe aortic/mitral disease (n = 13). Eighteen patients (2.0%) were lost to follow-up and were also excluded from the analysis. The study protocol was reviewed and approved by each institutions’ review board that waived the need for specific informed consent.

Data regarding clinical history and demographics were collected from patient chart review and electronic medical records. Familial history of SCD was defined as SCD in 1 or more first degree relatives under 40 years of age or SCD in a first degree relative with confirmed HCM at any age [4]. Information on genetic testing was available in 192 patients and was positive in 112 (58%).

Generally, all centers followed their patients at least once a year. Clinical status, annual echocardiography and Holter data were documented at the institution’s electronic medical records. The primary endpoint was a composite of SCD, aborted SCD, sustained ventricular tachycardia (VT), or appropriate ICD discharge. SCD was defined as witnessed sudden death with or without documented ventricular fibrillation (VF) or death within 1 h of new symptoms or nocturnal deaths with no antecedent history of worsening symptoms [18]. Event adjudication was performed by three Cardiologists who were blinded for the CMR data. Events were ascertained by reviewing electronic medical records including ICD electrograms. When death occurred outside of the hospital, the circumstances of death were determined by phone call interview with a family member. Any disagreement in event adjudication was discussed within this panel and solved by consensus.

Echocardiography, 24 h Holter monitoring and exercise test data within a 6-month window in relation to the CMR study were collected. If no left ventricular outflow tract (LVOT) obstruction was found, we assumed a gradient of 3 mmHg since it is the minimum accepted by the HCM Risk-SCD. Left atrial diameter was determined by M-Mode or 2D in the parasternal long-axis plane [4]. Non-sustained VT (NSVT) was defined as ≥ 3 consecutive ventricular beats at a rate of ≥ 120 beats/minute and < 30 s in duration on 24 h Holter monitoring [4]. Abnormal blood pressure response in exercise testing was defined as either a failure to increase by ≥ 20 mmHg or a drop of ≥ 20 mmHg during effort [2].

The 5-year HCM Risk-SCD was calculated for each patient [4] and categorized into 3 risk strata [1]: low risk (< 4%, ICD generally not indicated); intermediate risk (4–5.9%, ICD may be considered) and; high risk (≥ 6%, ICD should be considered). The ACCF/AHA guidelines consider that ICD implantation is reasonable if patients present any of the following major risk factors: family history of SCD in first degree relatives, LV wall thickness ≥ 30 mm, or recent unexplained syncope. Also, ICD implantation can be useful (minor risk factors) if NSVT is registered in Holter monitoring or abnormal blood pressure response occurs with exercise testing.

All CMR scans were performed using a 1.5 T systems (Siemens Avanto® and Aera®, Siemens Healthineers, Erlangen, Germany; Toshiba Vantage Titan®, Toshiba Medical Systems, Tokyo, Japan; Phillips Achieva®, Phillips Healthcare, Best, the Netherlands; GE Signa CVi® General Electric Healthcare, Waukesha, Wisconsin, USA). Images were transferred to a core laboratory where an experienced cardiologist in CMR, blinded for clinical events, analyzed all CMR data using a dedicated software (Circle Cardiovascular Imaging® release 5.6.4, Calgary, Canada). LV volume, mass, and EF were measured by use of standard volumetric techniques. LV endocardial and epicardial borders on cine images were manually traced to define the myocardium, taking care to exclude papillary muscles and the intertrabecular blood pool. Maximal LV wall thickness was defined as the greatest dimension at any site within the LV myocardium.

In all centers, LGE images were acquired 10 min after the administration of 0.2 mmol/kg intravenous gadolinium chelate contrast agent with breath-hold 2-dimensional segmented inversion-recovery spoiled gradient echo sequence or phase-sensitive inversion-recovery sequences. Imaging was performed in short-axis views covering the LV from the mitral annular plane to the apex with 8 mm slice thickness and 2 mm gaps. The typical in-plane spatial resolution was 1.5 mm × 1.5 mm. Inversion time was optimized to null normal myocardial signal. The LV short-axis stack of LGE images was first assessed visually for the presence of LGE, followed by quantification when LGE was present. LGE was defined as areas of signal intensity ≥ 6 standard deviations from normal myocardium and was expressed as the percentage of total LV myocardial mass (LGE%). Any areas that were identified as LGE by the software, but deemed artifactual on visual analysis, were manually excluded. Finally, LGE% was categorized into four risk strata (0%, 0.1–10%; 10.1–19.9%; ≥ 20%) [10].

To evaluate interrater agreement, a second blinded cardiologist analyzed 50 random CMR studies. A strong correlation was achieved in the evaluation of LGE% between the two raters (Spearman’s Rho: 0.97; p <  0.001). A very good agreement was found between the two raters regarding LGE strata (weighted κ: 0.85; 95% CI: 0.73–0.96; p <  0.001).

Categorical variables are presented as frequency and percentage, and continuous variables are presented as mean ± standard deviation (normal distribution) or median and interquartile range (non-normal distribution). Student’s t-test, Mann-Whitney U and Fisher’s exact test were used for comparison where appropriate. The amount of LGE in different risk categories was compared with a Kruskal-Wallis test.

Weighted κ was used to assess interrater agreement between the three classifications (HCM Risk-SCD, ACCF/AHA and LGE% strata). Quadratic weights were applied where the penalties for disagreement are milder for small disagreements but grow harsher as the disagreements become larger. The strength of agreement between each classification was considered poor (κ <  0.2), fair (κ = 0.21–0.4); moderate (κ = 0.41–0.6), good (κ = 0.61–0.8) or very good (κ = 0.81–1.0).

The effect of HCM Risk-SCD, ACCF/AHA and LGE on overall survival was assessed using Cox proportional hazards models. A sensitivity analysis was conducted with censoring at 5-years of follow-up since this is the time frame considered by the HCM Risk-SCD tool. Clinically relevant variables and/or variables with a p-value < 0.10 on individual analysis were included in multivariate models. The discriminative ability was assessed by calculation of the c index. For binary outcomes, the c index is identical to the area under the receiver operating characteristic curve [19]. Kaplan-Meier survival curves were plotted for each risk stratification tool. The log-rank test was used to assess for significant differences in time to endpoint between the risk strata. Net reclassification index (categorical NRI) was used to ascertain if LGE improves the risk stratification strategies of the American and European guidelines. Statistical significance was set at p-value < 0.05 (two-sided). All analyses were performed using SPSS® 25.0 and MedCalc® 9.3.8.0.

According to the ACCF/AHA criteria, ICD would ‘not be recommended’ in 57%, ‘could be useful’ in 13.8%, and would be ‘reasonable’ in 29.2%. According to the HCM Risk-SCD criteria, ICD would not be ‘indicated’ in 73.4%, ‘could be considered’ in 13.4%, and ‘should be considered’ in 13.2%. LGE was present in 79.3% of patients, with a median LGE% of 2.9% (IQR 0.4–8.4%). Amongst patients with LGE, the distribution was the following: 0.1 to 10.0% of the LV mass (n = 285, 72.9%), 10.1 to 19.9% (n = 63, 16.1%), and ≥ 20% (n = 43, 11.0%).

The concordance between risk assessment by these methods was relatively weak (Fig. 1). Weighted κ analysis revealed moderate agreement between the HCM Risk-SCD and the ACCF/AHA classification [κ = 0.51 (95% CI: 0.44–0.58); p <  0.001]. Poor agreement was found between ACCF/AHA and LGE classifications [κ = 0.19 (95% CI: 0.11–0.27); p <  0.001] and between HCM Risk-SCD and LGE [κ = 0.17 (95% CI: 0.07–0.26); p <  0.001].

The median total duration of follow-up was 3.4 years (IQR 1.5–6.8 years), during which a total of 23 events occurred (12 SCD, 6 appropriate ICD discharges and 5 sustained VT – Additional file 2: Table S1). An additional 4 cardiovascular deaths and 5 non-cardiac deaths were recorded. Throughout this period 42 ICDs were implanted for primary prevention. The distribution of events across each of the three studied classifications is presented in Table 2. A significant proportion of the patients who experienced an event were considered low-risk according to the clinical scores (35 and 48% for the ACCF/AHA and HCM Risk-SCD, respectively). Conversely, amongst those who were considered high-risk by the clinical scores, less than 10% experienced an event.

The unadjusted primary endpoint incidence increased in direct relation to the extent of LGE (Fig. 2) ranging from 0 events per 1000 person-years (95% CI: 0–0.4) in patients without LGE to 41 events per 1000 person-years (95% CI: 14–68) in patients with LGE% ≥ 20%.

The amount of LGE showed greater discriminative power (C-statistic 0.84; 95% CI: 0.76–0.91) than the ACCF/AHA (C-statistic 0.61; 95% CI: 0.49–0.72; p for comparison < 0.001) and the HCM Risk-SCD strategies (C-statistic 0.68; 95% CI: 0.59–0.78; p for comparison = 0.006). Kaplan-Meier survival curves according to the pre-defined LGE% strata, HCM Risk-SCD strata and ACCF/AHA algorithm are depicted in Fig. 3. No statistically significant differences in SCD-event free survival were found between the risk strata of the HCM Risk-SCD and ACCF/AHA criteria (log-rank p = 0.109 and log-rank p = 0.101, respectively). In contrast, SCD-event free survival was significantly different between LGE% risk strata (log-rank p <  0.001).

Univariate analysis showed an association between SCD events and LGE%, LVEF, LA diameter, NSVT and known AF (Table 3). In multivariate analysis, the amount of LGE was the only independent predictor of outcome (adjusted HR: 1.08; 95% CI: 1.04–1.12; p <  0.001). To assess the incremental value of LGE%, two additional analyses were conducted, forcing the HCM Risk-SCD and ACCF/AHA into the models. Neither of these risk stratification tools modified the independent prognostic value of LGE%.

The ability of LGE% to reclassify risk was further assessed by calculating the net reclassification index. The amount of LGE yielded an overall NRI of 0.43 (p = 0.011) when added to the HCM Risk-SCD, and an overall NRI of 0.36 (p = 0.021) when added to the ACCF/AHA algorithm – Table 4.

Since HCM Risk-SCD was derived to predict 5-year risk of SCD, we conducted a sensitivity analysis where follow-up was censored at 5 years. During this period, a total of 19 events were recorded (10 SCD, 5 appropriate ICD discharges and 4 sustained VT). Overall, findings remained similar (Additional file 1: Figure S1; and Additional file 3: Table S2, Additional file 4: Table S3, Additional file 5: Table S4, Additional file 6: Table S5).