Integrated Assessment of Late Gadolinium Enhancement and Left Ventricular Global Longitudinal Strain for Prognostic Stratification in Hypertrophic Cardiomyopathy = 비후성 심근증에서 지연조영증강과 좌심실 종축 변형률의 통합 평가를 통한 예후 예측 분류

국문 초록 (Abstract)

배경: 심장 자기공명영상(cardiovascular magnetic resonance, CMR)을 이용한 지연조영증강(late gadolinium enhancement, LGE)과 심초음파를 이용한 좌심실 종축 변형률(left ventricular global longitudinal strain, LV-GLS)은 비후성 심근증(hypertrophic cardiomyopathy, HCM)에서 좌심실의 구조적 및 기능적 리모델링을 반영하는 지표이다. 본 연구는 이 두 영상 지표를 통합하여 HCM 환자의 예후 예측과 위험도 층화에 대한 임상적 가치를 평가하고자 하였다.
방법: 2008년부터 2020년까지 진단 시점에 심초음파와 CMR을 모두 시행한 비후성 심근증 환자 652명을 대상으로 한 2개 기관 공동연구에서, 지연조영증강 비율(LGE%)과 좌심실 종축 변형률을 중앙 판독실에서 정량화하였다. 좌심실 종축 변형률은 절대값으로 보고하였다. 1차 평가변수는 심혈관 사망(돌연심장사[sudden cardiac death, SCD] 포함), 소생된 SCD 또는 적절한 삽입형 제세동기(implantable cardioverter-defibrillator, ICD) 충격, 심부전 입원을 포함한 복합 사건으로 정의하였다. 2차 평가는 SCD 관련 사건이었다. LGE%와 LV-GLS의 예후 연관성은 다변량 Cox 회귀모형에서 제한된 입방 스플라인(restricted cubic spline, RCS) 곡선과 랜덤 생존 포레스트(random survival forest, RSF) 분석에서 도출된 부분의존도곡선(partial dependence plot)을 이용하여 평가하였다. 이로부터 자료 기반의 탐색적 역치값을 도출하고, 해당 역치를 기준으로 환자를 4개 군으로 분류하였다. 이후 Cox 분석, Kaplan–Meier 생존곡선, Harrell의 C-index, 순 재분류 개선 지수(net reclassification improvement, NRI), 통합 판별 개선 지수(integrated discrimination improvement, IDI)를 이용하여 예측력의 향상을 평가하였다. 또한 구조방정식모형(structural equation model, SEM)을 통해 LGE와 임상결과 간의 연관성이 LV-GLS를 매개로 이루어지는지를 분석하였다.
결과: 전체 652명의 환자(중앙값 연령 56세, 남자 74%)에서 LGE%의 중앙값은 4.2%, LV-GLS의 중앙값은 14.3%였다. 중앙 추적관찰 기간 7.4년(IQR 3.4–10.3년) 동안 59명(9.0%)에서 심혈관 사건이 발생하였다. LGE%의 증가와 LV-GLS의 감소는 심혈관 사건 위험 증가와 독립적으로 관련되었다(LGE% 1% 증가당 보정위험비[aHR] 1.04, 95% CI 1.01–1.06, p=0.005; LV-GLS 1% 감소당 aHR 1.08, 95% CI 1.00–1.15, p=0.038). RCS 곡선과 partial dependence plot 모두에서 LGE%와 LV-GLS는 심혈관 사건과 비선형적 연관성을 보였으며, LGE%의 경우 낮은 수치에서 위험 증가가 관찰되었고, LV-GLS는 약 15% 이하에서 위험이 급격히 증가하는 변곡점이 확인되었다. 이를 바탕으로 LGE% 3%와 15%, LV-GLS 15%가 예후적으로 의미 있는 탐색적 역치로 도출되었다. LGE% <=3%군 대비 3–15%군과 >15%군은 각각 약 3배 높은 심혈관 사건 위험을 보였으며(aHR 2.75, 95% CI 1.36–5.56, p=0.005; aHR 2.94, 95% CI 1.26–6.85, p=0.013), LV-GLS <15%는 >=15% 대비 약 2배 높은 위험을 보였다(aHR 2.00, 95% CI 1.04–3.85, p=0.037). LGE% 3%와 LV-GLS 15%를 기준으로 한 4분류 모델에서, 두 지표 모두 이상이 있는 군이 가장 높은 사건 위험을 보였다(aHR 4.11, 95% CI 1.50–11.24, p=0.006). 해당 분류를 기존 임상모델에 추가함으로써 예측력이 유의하게 향상되었다(C-index 0.808 vs. 0.777, bootstrap p=0.026; NRI 0.537, 95% CI 0.195–0.878, p=0.002; IDI 0.040, 95% CI 0.008–0.073, p=0.016). 매개분석 결과, LGE는 LV-GLS를 통한 유의한 간접효과를 통해 심혈관 사건과 연관되어 있음을 확인하였다. SCD 관련 사건 역시 LGE%가 높고 LV-GLS가 낮은 군에서 가장 빈번하게 발생하였다.
결론: 지연조영증강과 좌심실 종축 변형률은 비후성 심근증 환자에서 서로 독립적이면서도 상호보완적인 예후 정보를 제공한다. 두 지표를 통합함으로써 기존 임상모델 대비 예측력을 향상시킬 수 있으며, 구조적 섬유화와 기능적 장애가 공존하는 고위험군을 식별할 수 있다. 이러한 결과는 영상기반 위험도 평가를 통해 정밀한 예후 예측이 가능함을 보여주며, 비후성 심근증의 임상적 의사결정에 도움이 될 수 있는 근거를 제공한다.

다국어 초록 (Multilingual Abstract)

Background: Late gadolinium enhancement (LGE) assessed with cardiovascular magnetic resonance (CMR) and left ventricular global longitudinal strain (LV-GLS) measured by echocardiography are structural and functional markers of left ventricular (LV) remodeling in hypertrophic cardiomyopathy (HCM). This thesis aimed to investigate the prognostic value of integrating these two markers for risk stratification in HCM.
Methods: In a two-center cohort of 652 consecutive patients with HCM who underwent echocardiography and CMR at diagnosis between 2008 and 2020, the percentage of LGE (LGE%) and LV-GLS were quantified in a central core laboratory. LV-GLS value was reported as an absolute value. The primary outcome was a composite of cardiovascular (CV) events, including CV death (which includes sudden cardiac death [SCD]), aborted SCD or appropriate implantable cardioverter-defibrillator shock, and heart failure hospitalization. The secondary outcome was SCD-related events alone. The relationships of LGE% and LV-GLS with composite CV events were evaluated using restricted cubic spline (RCS) curves from multivariable Cox models and partial dependence plots derived from the random survival forest (RSF) analysis. Data-driven exploratory thresholds for both markers were identified, and patients were subsequently stratified into four groups based on these cutoffs. The prognostic value of this classification was assessed using Cox analyses, Kaplan–Meier curves, and measures of discrimination improvement—including Harrell’s C-index, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Structural equation modeling (SEM) was performed to evaluate whether the association between LGE and outcomes was mediated by LV-GLS.
Results: Among 652 patients with HCM (median age, 56 years; 481/652 [74%] male), median LGE% and LV-GLS were 4.2% and 14.3%, respectively. During the median 7.4-year follow-up (IQR: 3.4–10.3 years), 59 patients (9.0%) experienced CV events. Higher LGE% and lower LV-GLS were associated with increased CV events (per 1% LGE% increase, adjusted hazard ratio [aHR]: 1.04, 95% CI: 1.01–1.06, p=0.005; per 1% LV-GLS decrease, aHR: 1.08, 95% CI: 1.00–1.15, p=0.038). Both the RCS curves and partial dependence plots demonstrated similar non-linear associations between LGE%, LV-GLS, and composite CV events, with an early increase in risk at low levels of LGE% and an inflection point indicating higher risk at lower LV-GLS. Based on these curves, data-driven exploratory thresholds were identified at 3% and 15% for LGE% and 15% for LV-GLS. Compared with patients with LGE% <=3%, those with LGE% 3–15% and >15% had approximately threefold higher risk of CV events (aHR: 2.75, 95% CI: 1.36–5.56, p=0.005; aHR: 2.94, 95% CI: 1.26–6.85, p=0.013). LV-GLS <15% was associated with a twofold higher risk compared with LV-GLS >=15% (aHR: 2.00, 95% CI: 1.04–3.85, p=0.037). When stratified into four groups using the thresholds of LGE% 3% and LV-GLS 15%, patients with both elevated LGE% and impaired LV-GLS had the highest risk of CV events (aHR: 4.11, 95% CI: 1.50–11.24, p=0.006). Incorporating this four-group classification improved risk discrimination beyond the baseline clinical model (C-index 0.808 vs. 0.777, bootstrap p for comparison=0.026; NRI 0.537, 95% CI: 0.195–0.878, p=0.002; IDI 0.040, 95% CI: 0.008–0.073, p=0.016). Mediation analysis confirmed a significant indirect effect of LGE on CV events via LV-GLS, suggesting a mechanistic link between myocardial fibrosis and systolic dysfunction. For SCD-related outcomes, events were most frequent in patients with both high LGE% and low LV-GLS.
Conclusions: This thesis demonstrated that LGE and LV-GLS provide independent and complementary prognostic information in HCM. Integrating both markers enhances risk prediction beyond conventional models, identifying a high-risk subgroup with concurrent structural fibrosis and functional impairment. These findings support an imaging-based framework for refined risk stratification and improved clinical decision-making in HCM management.

목차 (Table of Contents)

• ABSTRACT I
• ACKNOWLEDGEMENTS IV
• TABLE OF CONTENTS V
• LIST OF TABLES VII
• LIST OF FIGURES VIII
• LIST OF ABBREVIATIONS IX
• CHAPTER 1. INTRODUCTION 1
• 1.1 Study Background 1
• 1.1.1 Significant Cardiovascular Risk Associated with Hypertrophic Cardiomyopathy 1
• 1.1.2 Cardiovascular Magnetic Resonance in Risk Stratification of HCM 1
• 1.1.3 Myocardial Strain Assessment for Detecting Subclinical Systolic Dysfunction 2
• 1.2 Statement of the Problem 3
• 1.3 Research Hypothesis 3
• 1.4 Purpose of the Study 4
• CHAPTER 2. METHODOLOGY 6
• 2.1 Study Design and Population 6
• 2.1.1 Study Design of the Current Study 6
• 2.1.2 Cohort Characteristics and Variable Definitions 7
• 2.2 Echocardiography 8
• 2.2.1 Standard Transthoracic Echocardiography 8
• 2.2.2 LV-GLS Measurements 9
• 2.3 CMR 10
• 2.3.1 CMR Protocol 10
• 2.3.2 CMR Image Quantification 11
• 2.4 Definitions of Clinical Outcomes 14
• 2.5 Statistical Analysis 15
• 2.5.1 Baseline Characteristics Assessment 15
• 2.5.2 Conventional Statistical Methods for Prognostic Relationship of LGE and LV-GLS 15
• 2.5.3 Random Survival Forest Analysis to Explore the Relationship Between LGE, LV-GLS, and Adverse Events 16
• 2.5.4 Risk stratification by both LGE and LV-GLS 17
• 2.5.5 Mediation Analysis 18
• 2.5.6 Statistical Significance 20
• 2.6 Ethical Statement 20
• CHAPTER 3. RESULTS 21
• 3.1 Distribution and Relationship between LGE and LV-GLS 21
• 3.1.1 LV Structural and Functional Characteristics Across the Study Population 21
• 3.1.2 Comparison of Characteristics Between the Study Centers 24
• 3.2 Prognostic Role of LGE and LV-GLS: Conventional Cox Analysis 27
• 3.2.1 Clinical Outcomes and Univariable Cox Associations 27
• 3.2.2 Multivariable Cox analysis and Median based thresholds 29
• 3.2.3 RCS curves Depicting the Relationship Between Imaging markers and CV Events 30
• 3.3 Random Survival Forest Model Approach to Identify Nonlinear Associations 31
• 3.3.1 RSF Model Development and Performance 31
• 3.3.2 Non-linear Association of LGE% and LV-GLS with CV Events in the Partial Dependency Plot 33
• 3.3.3 Evaluation of Exploratory Thresholds of LGE% and LV-GLS 34
• 3.4 Combined Risk Stratification Using Both LGE% and LV-GLS 36
• 3.4.1 Characteristics According to the Four Groups Stratified by LGE% and LV-GLS. 36
• 3.4.2 Prognosis According to the Four Groups Stratified by LGE% and LV-GLS. 38
• 3.5 Mediation Analysis Between LGE, LV-GLS, and CV Events 44
• 3.6 Implications for SCD-related Event 45
• CHAPTER 4. DISCUSSION 49
• 4.1 Adverse Cardiac Remodeling in HCM: Fibrosis and Myocardial Strain 49
• 4.2 LGE and Prognostic Association in HCM 50
• 4.3 LV-GLS and Prognostic Association in HCM 51
• 4.4 Mediating Effect of LV-GLS in the Relationship Between LGE and CV events 52
• 4.5 Complementary Prognostic Information Provided by LV-GLS and LGE 53
• CHAPTER 5. CONCLUSION 54
• BIBLIOGRAPHY 55
• 국문 초록 63