Hypertrophic Cardiomyopathy With Left Ventricular Systolic Dysfunction

Supplemental Digital Content is available in the text.

H ypertrophic cardiomyopathy (HCM) is the most common inherited cardiac disease, with an estimated prevalence of 1 in 500 in the adult population. [1][2][3] Pathogenic variants in genes that encode cardiac sarcomeric proteins are responsible for causing disease in roughly 60% of familial and 20% to 30% of apparently sporadic HCM. 4 Left ventricular (LV) systolic function is characteristically normal to hyperdynamic in HCM. However, prior studies have suggested that 4% to 9% of patients develop systolic dysfunction, defined by LV ejection fraction (LVEF) <50% (herein referred to as HCM with LV systolic dysfunction [LVSD]). This complication of disease has previously been called "endstage" or "burnt-out" HCM. [5][6][7][8] HCM-LVSD is typically accompanied by diffuse myocardial fibrosis, although LV wall thinning and cavity enlargement may also be present. 6,9 The reported prognosis of patients with HCM-LVSD has been quite poor, with mortality as high as 11%/year. 6 However, because of its relative rarity, the natural history remains incompletely characterized. There is a clear need to better understand HCM-LVSD to improve risk stratification and to inform clinical management. In this study, we leverage the international SHaRe (Sarcomeric Human Cardiomyopathy Registry), 10 analyzing data on nearly 7000 patients with HCM, to better describe the prevalence and natural history of HCM-LVSD, to identify predictors of prognosis, and to identify features that predict incident development of systolic dysfunction.

METHODS
SHaRe is an international collaborative consortium of highvolume HCM centers that maintain longitudinal databases capturing phenotypic, genetic, and clinical outcomes data on patients with HCM and their families. The structure of the registry and initial findings have previously been described in detail. 10 Briefly, definitions for phenotypic features and clinical outcomes were harmonized, and site data are mapped to a secure, centralized data set. Historical events that occurred before the initial visits to SHaRe sites are carefully ascertained and vetted for accuracy through detailed patient history and medical record review. Prospective longitudinal data are captured by sites as they occur or during clinical encounters and uploaded quarterly from site databases. Institutional review and ethics approval were obtained in accordance with applicable site policies. This study analyzed data from 1960 through March 2019 from 11 different sites around the world (Brigham and Women's Hospital, Boston, MA; Boston Children's Hospital, MA; University of Michigan, Ann Arbor; Cincinnati Children's Hospital, OH; Yale-New Haven Hospital, CT; Stanford University, CA; Cardiomyopathy Unit, University of Florence, Italy; Erasmus Medical Center, Rotterdam, the Netherlands; Royal Brompton Hospital, London, United Kingdom; University of São Paulo, Brazil; University of Sydney, Australia). The data cannot be made available to other researchers for purposes of reproducing or replicating the procedure because of constraints related to human subjects research. Analytical methods can be made available on request.

Study Population
Patients were included in this study if they had at least 1 complete echocardiographic study at a SHaRe site and a site-designated diagnosis of HCM, defined as unexplained LV hypertrophy with a maximal LV wall thickness >15 mm or >13 mm in members of families with HCM (or equivalent LV wall thickness z score in pediatric patients). Extracardiac features, family history, and genotype were integrated to allow accurate and informed diagnosis by experienced clinicians. Genetic testing was performed at sites using different platforms available over time, focusing on the 8 sarcomeric genes definitively associated with HCM (MYBPC3, MYH7, TNNT2,  TNNI3, TPM1,

Outcomes
Patients were designated as having HCM-LVSD at the first documentation of an echocardiographic LVEF <50% on a clinically performed echocardiographic study. To identify predictors of prognosis in HCM-LVSD, patients were followed from recognition of HCM-LVSD until last follow-up or meeting the composite outcome of all-cause mortality, cardiac transplantation, or implantation of an LV assist device (LVAD). Development of atrial fibrillation and New York Heart Association (NYHA) functional class III/IV symptoms were also assessed. Analyses to identify factors associated with incident HCM-LVSD included genotyped patients with LVEF ≥50% and no history of septal reduction at or before their initial SHaRe visit. Patients with existing HCM-LVSD at their initial SHaRe evaluation were excluded from this model. Sensitivity analysis was performed to compare septal reduction therapies (SRTs) in patients diagnosed with HCM before and after January 2000 to limit analysis to patients who were diagnosed with HCM and underwent procedures in the current era.

Statistical Analysis
Normally distributed data were expressed as mean±SD and compared with the Student t test. Nonnormally distributed data were expressed as median and interquartile range and compared with the Wilcoxon rank-sum test. Differences in categorical variables were calculated with the χ 2 test. The Kaplan-Meier method with log-rank test for significance was used to estimate the cumulative incidence and time to event of the end points of interest, starting from the time of the initial visit to a SHaRe site. To adjust for baseline characteristics, Cox proportional hazards models were developed, requiring a minimum of 10 events (incident HCM-LVSD or composite outcome) per covariate included. 12 Results of Cox regressions were reported as adjusted hazard ratios (HRs) with 95% CIs. Analyses were depicted in forest plots, and values of P˂0.05 were considered statistically significant. If patients were missing echocardiographic measures at the first SHaRe evaluation, data were imputed from subsequent echocardiographic studies if the echocardiogram was performed before the development of HCM-LVSD. Patients missing key echocardiographic data in all of their studies in SHaRe were omitted from Cox regression models. R version 3.5.2 was used for all analyses. Raw counts and HRs of all events experienced by patients with HCM-LVSD and HCM without LVSD are presented in Table 2. HRs are adjusted for age, sex, and follow-up time from the initial SHaRe evaluation. Patients with HCM-LVSD had a markedly higher prevalence of cardiac transplantation (11.4% versus 0.7%) and LVAD implantation (1.6% versus 0.1%; P<0.001 for all comparisons). Patients with HCM-LVSD were more likely to experience all-cause death (25.0% versus 6.7%), implantable cardioverter-defibrillator (ICD) implantation (54.4% versus 21.6%), and appropriate ICD firing (25.2% versus 12.1%; P<0.001). Atrial fibrillation (49.3% versus 20.9%) and stroke (8.4% versus 2.3%) were also more common in patients with HCM-LVSD.

Prevalence and Clinical Features of HCM-LVSD
A greater proportion of patients with HCM-LVSD underwent SRT compared with patients without LVSD (27.3% versus 17.5%; P<0.001; Table 3). The majority of procedures were myectomy (78% of SRT procedures). The proportion undergoing alcohol septal ablation did not differ between groups (4.0% versus 3.2%;

ORIGINAL RESEARCH ARTICLE
P=0.34), but patients with HCM-LVSD were more likely to have had both myectomy and alcohol ablation performed (1.6% versus 0.6%; P<0.001).

Natural History of HCM-LVSD
We characterized natural history by focusing on our cohort of 553 patients with HCM-LVSD. As Table 4 shows, patients with HCM-LVSD were a mean age of 35.6±19.2 years when diagnosed with HCM and 50.3±17.9 years when recognized to have LVEF <50%.
The mean LVEF at presentation with HCM-LVSD was 40±8%, and 27.1% of patients had an LVEF <35% at presentation. Thirty percent (165 of 553) had NYHA class III/IV symptoms at presentation with HCM-LVSD. Table 5 lists clinical events associated with the total HCM-LVSD cohort, both before developing systolic dysfunction and during follow-up. Overall, 74.   Figure 1A). Assessing death alone as an outcome showed that 138 patients (25%) with HCM-LVSD died, including 10 patients who died after cardiac transplantation or LVAD implantation. The median time to death was 11.4 years (95% CI, 9.3-14.9 years) after the development of systolic dysfunction. Patients with HCM-LVSD at the initial evaluation had a prognosis from the time of HCM-LVSD similar to those who developed incident HCM-LVSD during follow-up (data not shown). Clinical course varied substantially from patient to patient. As Figure 2 illustrates, many patients with HCM-LVSD across a broad spectrum of ages do not experience the composite outcome, whereas many others at similar ages and duration of HCM-LVSD experienced serious outcomes.

Incident Development of HCM-LVSD
A cohort of 5905 patients with longitudinal data and LVEF >50% at the initial evaluation was available to

ORIGINAL RESEARCH ARTICLE
analyze incident development of HCM-LVSD. Of those, 350 (5.9%) developed HCM-LVSD during follow-up. The estimated incidence rates were 0.5%/y from initial evaluation, 1.7% (95% CI, 1.4%-2.2%) at 5 years, 4.5% (3.8%-5.3%) at 10 years, and 7.5% (6.5%-8.6%) at 15 years after the initial SHaRe visit (Figure 4). A Cox proportional hazards model was performed to identify predictors of developing HCM-LVSD. Of the 5905 patients with available follow-up, 3524 had genetic data available and no history of SRT at initial evaluation. Complete echocardiographic data were available in 2627 of those, and in 189 patients, data were imputed with a median time of 87 days (interquartile range, 39-380 days) between the initial evaluation and echocardiography. Hence, the model was based on a cohort of 2816 patients with HCM, with 170 outcomes of incident HCM-LVSD ( Figure 5). Greater LV cavity size, greater LV wall thickness, and the presence of a pathogenic/likely pathogenic sarcomeric variant were associated with increased risk of incident HCM-LVSD, with HRs ranging from 1.2 to 1.5. The presence of a variant in a thin filament gene (TNNT2, TNNI3, TPM1, ACTC) was associated with an HR of 2.5 (95% CI, 1.2-5.1). Borderline LVSD at the initial SHaRe evaluation was associated with a higher risk of incident HCM-LVSD (LVEF, 55%-59%: HR, 1.8 [95% CI, 1.2-2.8]; LVEF 50%-54%: HR, 2.8 [95% CI, 1.8-4.2]), likely identifying patients with incipient HCM-LVSD remodeling at presentation. In patients with a baseline LVEF between 50% and 59% who later developed HCM-LVSD, the median time to onset of HCM-LVSD was 3.4 years (2.32-4.23 years) if LVEF was 55% to 59% and 2.2 years (1.80-3.2 years) if LVEF was 50% to 54%.
Although atrial fibrillation was a predictor of prognosis and was more frequently seen in patients with HCM-LVSD (11.3% versus 4.3%), it was not an independent risk predictor for the development of incident HCM-LVSD remodeling (HR, 1.1 [95% CI, 0.8-1.5]). Patients with obstructive physiology (maximal LV outflow tract gradient >30 mm Hg) were less likely to develop HCM-LVSD (HR, 0.6 [95% CI, 0.4-0.9]), suggesting that obstructive physiology was not driving the development of systolic dysfunction in the majority of patients with HCM-LVSD.
A full list of multiple pathogenic/like pathogenic sarcomeric variants is provided in Table I    β-blocker nor calcium channel blocker use had a significant association with risk for developing incident HCM-LVSD or prognosis with HCM-LVSD once developed.
In a Cox regression model adjusted for genotype, age at diagnosis, cardiac magnetic resonance LVEF, and sex, the presence of LGE was associated with an HR of 2.3 (95% CI, 1.0-4.9; P=0.039) for developing incident HCM-LVSD.

Impact of SRT
Given the observation of a higher prevalence of SRT in patients with HCM-LVSD, we included SRT in the model

ORIGINAL RESEARCH ARTICLE
assessing incident development of HCM-LVSD ( Figure  5). An increased risk for developing incident HCM-LVSD was seen in patients who previously had a myectomy (n=40 of 465; HR, 2.7 [95% CI, 1.8-4.0]) or alcohol ablation (n=10 of 80; HR, 2.6 [95% CI, 1.3-5.2]). Further analyses were then pursued to better characterize the association between SRT and incident development of systolic dysfunction. Overall, 1221 patients in SHaRe underwent SRT, of whom 138 (11.4%) developed incident HCM-LVSD. These 138 patients were a mean 43±18 years of age at SRT and developed HCM-LVSD a median of 5.6 years (3.9-8.6 years) after SRT. A sensitivity analysis was performed that excluded patients who underwent SRT before the year 2000 (n=8) to limit analyses to patients whose procedures were performed in the current era. Results were similar to the full model (HR for SRT, 2.4 [95% CI, 1.7-3.6]), suggesting that the association between HCM-LVSD and myectomy was not driven primarily by patients who underwent procedures in the remote past. In addition, adjusting for participating site did not change the estimated risk (data not shown).
We compared baseline clinical characteristics between patients with HCM-LVSD who did and those who did not undergo SRT. Patients with HCM-LVSD with prior SRT were similar in age, sex, sarcomeric variant status, and presenting LV wall thickness and LV cavity size, but they had larger left atrial diameter and greater LVEF at baseline evaluation (Table III in the Data  Supplement). Patients with HCM-LVSD with prior SRT were also more likely to have NYHA class III/IV symptoms at presentation with HCM-LVSD.
We postulated that the development of HCM-LVSD after myectomy could be the result of procedurerelated left bundle-branch block, leading to a mild to moderate reduction in LVEF (LVEF, 35%-49%) but no other clinical sequelae (no atrial fibrillation, appropriate ICD therapy, or NYHA class III/IV symptoms, denoted uncomplicated HCM-LVSD). No significant difference in the proportion of uncomplicated HCM-LVSD was found between patients with and those without a history of SRT ( P=0.31), suggesting that systolic dysfunction in patients with prior myectomy is not purely a reflection of otherwise inconsequential bundle-branch block.
Once HCM-LVSD developed, the natural history in patients with prior SRT was not significantly different from that in patients without prior SRT (Table IV in the ORIGINAL RESEARCH ARTICLE Data Supplement). Among patients with HCM-LVSD with prior SRT, 26.8% developed LVEF <35% (versus 36.5% without prior SRT; P=0.051). Patients with prior SRT were more likely to have an ICD implanted but less likely to have appropriate firing. No significant differences were found for death, need for cardiac transplantation, LVAD, or atrial fibrillation between patients with HCM-LVSD with and those without prior SRT.

DISCUSSION
This study leverages a large, international cohort to better characterize the prevalence and natural history of systolic dysfunction in HCM. The major findings are the following: First, ≈8% of patients with HCM develop systolic dysfunction with LVEF <50% (7.5% incidence over 15 years). Second, the natural history of HCM-LVSD was variable, but ≈75% of patients experienced clinically relevant adverse events, including 35% experiencing death, cardiac transplantation, or LVAD implantation an estimated median of 8.4 years after developing systolic dysfunction. Third, atrial fibrillation, LVEF <35%, and the presence of multiple sarcomeric variants were independently associated with poor prognosis. Last, sarcomeric variants (particularly in thin filament genes), borderline low baseline LVEF (50%-60%), and LGE on cardiac magnetic resonance were independently associated with incident development of HCM-LVSD.

Prevalence and Prognosis of HCM-LVSD
Earlier, smaller studies on HCM-LVSD (n=20-156) have reported a prevalence ranging from 4% to 9% and a malignant natural history, with ≈60% of patients with HCM-LVSD experiencing death or cardiac transplantation over a mean time of only 2.7 to 5 years from recognition of systolic dysfunction. [6][7][8][13][14][15] In our cohort of 6793 patients with HCM with a median follow-up of >3 years at HCM specialty centers, 553 developed systolic dysfunction, representing a prevalence of 8.1%. Our results confirm that systolic dysfunction confers adverse outcomes. Seventy-five percent of patients with HCM-LVSD experienced adverse sequelae, and they were at least twice as likely as patients without systolic dysfunction to have NYHA class III/IV symptoms, appropriate ICD therapy, atrial fibrillation, and stroke. One-third of patients with HCM-LVSD developed severe LVSD (LVEF <35%), and 35% experienced a death equivalent as reflected by the composite outcome of death, cardiac transplantation, or LVAD implantation. Compared with patients with HCM without systolic dysfunction, mortality during follow-up was >2-fold higher in patients with HCM-LVSD; the need for cardiac transplantation or LVAD was >11-and 26-fold higher, respectively. However, our findings indicate that the time frame is less precipitous for most individuals with HCM-LVSD than previously described. On average, systolic dysfunction developed a median of 15 years after initial diagnosis of HCM. The median time to death, cardiac transplantation, or LVAD implantation was 8.4 years after recognition of systolic dysfunction and 11.4 years to death alone. Atrial fibrillation (HR, 2.6 [95% CI, 1.7-3.8]), LVEF <35% (HR, 2.0 [95% CI, 1.3-2.8]), and the presence of multiple pathogenic sarcomeric variants (HR, 5.6 [95% CI, 2.1-13.5]) were independent predictors of poor prognosis, as assessed by the composite outcome of death, cardiac transplantation, or LVAD implantation.
Moreover, the experience of patients with HCM-LVSD was quite broad, with many free from serious events for many years after the initial decline in LVEF was documented. Thus, end stage, the previously used nomenclature, does not accurately reflect the majority of patients with HCM-LVSD. Further investigation is needed to more fully understand this heterogeneity and to characterize the underlying factors that drive progressive adverse remodeling associated with decreased LV systolic function in HCM.  16 in 2010. They also identified a greater extent of myocardial fibrosis associated with declining LVEF. 16 Similarly, our study identified the presence of late gadolinium and LVEF of 50% to 60% as independent predictors for incident systolic dysfunction. Although this LVEF would be considered within normal range for other populations, it appears to identify patients with HCM with a nearly 3-fold increased risk of developing HCM-LVSD who would likely benefit from closer management.

Incident Development of Systolic Dysfunction
Our data identified a potential association between invasive SRT and future development of HCM-LVSD. This signal (HR >2.6 for surgical myectomy and alcohol ablation) persisted even after considering LV morphology, function, and obstructive physiology before the procedure, atrial fibrillation, participating site, and limiting the analysis to procedures performed in the current era (since 2000). In our cohort, HCM-LVSD developed in 11% of patients who underwent SRT ≈6 years after the procedure was performed. Patients with HCM-LVSD who underwent prior SRT were more likely to have an ICD implanted than those without prior SRT, presumably related to procedure-associated conduction block, but appropriate ICD therapies did not differ significantly. Prognosis and clinical outcomes appeared similar in patients with HCM-LVSD with and without prior SRT.
Septal reduction therapies have been studied extensively for safety and efficacy of symptom relief. In particular, surgical myectomy has been shown to be associated with low morbidity and mortality in experienced centers such as SHaRe sites. [17][18][19] SRT plays an important role in the clinical management of HCM, providing highly effective relief of symptoms resulting from obstructive physiology. However, prior studies on SRT were not designed to identify a potential association between myectomy and future development of HCM-LVSD. Thus, they would not have captured the development of systolic dysfunction or death-equivalent outcomes of cardiac transplantation or LVAD implantation (30 patients in our cohort). Nonetheless, 2 smaller studies also identified an increased risk of developing impaired LV function after SRT 20 and emphasized that the duration of follow-up needs to be relatively long, at least 8 to 10 years, to capture this adverse remodeling. 21 This time frame is supported by our current findings. Similarly, all but 1 published study focusing on HCM-LVSD systematically excluded patients who underwent SRT. 5,13,16 The study by Harris et al 6 included 89 patients with a myectomy, 7 (8%) of whom developed HCM-LVSD, compared with 3% of patients who did not undergo myectomy. A value of P=0.84 was reported, without adjustment for age, sex, time to event, or disease severity. 6 We caution that our findings reflect a relatively small number of patients who underwent SRT before developing HCM-LVSD. Although potential causal mechanisms underlying this association cannot be inferred from this retrospective analysis, we have considered several possibilities that could be explored in future studies. First, we speculate that intraventricular conduction delay/ dyssynchrony related to the procedure may play a role, in which case early resynchronization therapy might be

ORIGINAL RESEARCH ARTICLE
considered in certain patients (eg, LVEF <55%). Second, the HCM-LVSD cohort with SRT may be confounded by some degree of selection bias because these patients may have been referred for a higher degree of symptoms as a result of more severe underlying adverse remodeling. Last, it is possible that preexisting pressure overload could lead to more advanced remodeling in a subset of patients, despite adequate relief of LV outflow tract obstruction. Our data suggest that patients undergoing SRT require longitudinal monitoring of cardiac function after the procedure, particularly younger patients with sarcomere gene mutations. Further study is needed to better characterize the associations between SRT and HCM-LVSD, to elucidate broader effects of SRT on long-term cardiac remodeling, and to fully understand the clinical implications.

Limitations
Although SHaRe incorporates curated, longitudinal data, it is subject to the limitations inherent to all observational and partially retrospective registry-based studies. Cardiac magnetic resonance data are derived from site-based clinical reports, which do not routinely capture data on quantification and location of LGE. On the basis of recent literature, we anticipate that <4% of patients with HCM will have extensive LGE (>15% involvement). 22 Because participating sites are highvolume HCM specialty centers, our study could be influenced by referral bias, resulting in an overestimation of the prevalence of HCM-LVSD and adverse prognosis of HCM-LVSD. Conversely, the registry is also subject to survival bias because patients must survive until being seen at a SHaRe site. It is notable that the patients with incident systolic dysfunction had longer follow-up compared with the group who did not develop systolic dysfunction, suggesting that the prevalence is likely to be higher with longer follow-up. Continued multicenter study in adequately powered cohorts is needed for more definitive understanding of HCM-LVSD.

Conclusions
LVSD develops in ≈8% of patients with HCM and carries important clinical implications in that ≈75% of patients with HCM-LVSD will experience atrial fibrillation, stroke, advanced heart failure (LVEF <35% or NYHA class III/IV), appropriate ICD therapy, cardiac transplantation, LVAD implantation, or death. However, the natural history appears to evolve over a number of years, and individual patient experience is variable. Genetic substrate appears to play a role in both the risk for developing systolic dysfunction and prognosis after systolic dysfunction is present. In addition, patients with HCM with an LVEF between 50% and 60% had a nearly 3-fold increased risk of developing systolic dysfunction and may benefit from closer clinical surveillance. Further study is required to refine clinical predictors of progressive adverse remodeling, especially in patients with prior SRT, and to more fully characterize the determinants of clinical outcomes.